<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Juan Carlos Saravia on Medium]]></title>
        <description><![CDATA[Stories by Juan Carlos Saravia on Medium]]></description>
        <link>https://medium.com/@jsaraviadrago?source=rss-468fc5c36a75------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*mSE2Mq7oUCjiKZhPwryPbg.jpeg</url>
            <title>Stories by Juan Carlos Saravia on Medium</title>
            <link>https://medium.com/@jsaraviadrago?source=rss-468fc5c36a75------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 20:07:25 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@jsaraviadrago/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[El juego de mi abuela estaba en peligro de extinción, así que lo escribí en código]]></title>
            <link>https://jsaraviadrago.medium.com/el-juego-de-mi-abuela-estaba-en-peligro-de-extinci%C3%B3n-as%C3%AD-que-lo-escrib%C3%AD-en-c%C3%B3digo-d4bab6ee64c6?source=rss-468fc5c36a75------2</link>
            <guid isPermaLink="false">https://medium.com/p/d4bab6ee64c6</guid>
            <category><![CDATA[ai-for-good]]></category>
            <category><![CDATA[rust]]></category>
            <category><![CDATA[backgammon]]></category>
            <category><![CDATA[tradiciones]]></category>
            <dc:creator><![CDATA[Juan Carlos Saravia]]></dc:creator>
            <pubDate>Sat, 03 Oct 2026 12:22:18 GMT</pubDate>
            <atom:updated>2026-10-03T12:22:18.217Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Cómo convertí un chaquete familiar en un motor de reglas en Rust para que no se pierda nunca</em></p><p>¿Sabías que el backgammon también se llama <strong>chaquete</strong>?</p><p>Yo no tenía ni idea. Para mí, el chaquete era simplemente el juego que me enseñó mi abuela cuando era chico. Recuerdo que lo jugábamos con unas reglas muy suyas, distintas a las del backgammon que uno encuentra en cualquier lado. Años después, cuando me puse a buscar en internet, descubrí que lo que ella me había enseñado era una variación familiar. Un chaquete con sello propio.</p><h3>Cuando las reglas viven solo en tu cabeza</h3><p>Mi abuela falleció y, con el tiempo, le enseñé el juego a un amigo. Jugamos un buen rato, hasta que él se fue a vivir fuera del Perú. Y sin con quién jugar, yo también dejé de hacerlo.</p><p>Pasaron los años. Un día se me ocurrió preguntarle a mi padrino, que también sabía jugar. Me dijo algo que me dejó pensando: había reglas que ya no se acordaba.</p><p>Ahí caí en la cuenta. El chaquete de mi abuela, que luego descubrí que era una variación del <strong>Jacquet</strong>, un juego muy popular en Europa hacia 1800, vivía únicamente en mi memoria. Y mi memoria, como la de cualquiera, no es un sistema de respaldo muy confiable.</p><p>Quiero enseñárselo a mis hijos. Todavía son chiquitos, pero seguro llegará el día. Y mientras lo pensaba, me di cuenta de que esto me suena a algo que veo seguido en mi trabajo: <strong>el conocimiento que solo está en la cabeza de alguien se pierde.</strong> Pasa con la documentación, pasa con las reglas de negocio de una empresa. Si nadie las escribe, un día alguien se va, se olvida, y adiós.</p><p>En mi caso era peor, porque ese juego, exactamente así, no existe en internet.</p><h3>Entonces hice lo que sé hacer: programarlo</h3><p>Así que tomé una decisión medio literal: <strong>volcar mi memoria al código.</strong></p><p>Escribí el juego completo, tal como lo llamaban mi abuela y mi padrino. Las reglas, la lógica, todo. La idea era crear una versión digital que guardara el juego tal cual lo conocí.</p><p>Se juega con <strong>tres dados</strong>, y cada número tenía su nombre. El 1 era <strong>AS</strong>, el 2 era <strong>don</strong>, el 3 era <strong>tren</strong>, y así sucesivamente. Y cuando tirabas los tres dados y salía, por ejemplo, un 4–4–2, había que decirlo de una forma específica, ni una palabra más ni una menos. Toda esa nomenclatura la dejé escrita en la documentación del repositorio, porque también es parte del juego.</p><p>Pero no me quedé en la documentación:</p><ul><li><strong>Rust</strong> para construir todo el motor de reglas.</li><li><strong>JavaScript</strong> para el front.</li><li><strong>Un servidor</strong> para el multijugador, de modo que dos personas puedan jugar desde distintos celulares o computadoras.</li></ul><p>Todo está en este repositorio: <a href="https://github.com/jsaraviadrago/rusty-walk"><strong>https://github.com/jsaraviadrago/rusty-walk</strong></a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/768/1*NQKuVOOz3CiylKniUPuoiA.png" /></figure><h3>Una segunda vida para los juegos de mesa</h3><p>Soy de los que piensan que la magia del chaquete está en sentarse frente a un tablero de verdad, con los dados en la mano. Eso no se reemplaza.</p><p>Pero también me puse a pensar en lo que pasó con el ajedrez y plataformas como chess.com. Cada vez más juegos de mesa pueden jugarse también de forma digital, y eso les da una segunda vida.</p><p>Ahora el Jacquet, o chaquete, ya no depende de mi memoria. Hay un lugar donde están las reglas, donde se puede jugar y donde el juego puede perdurar.</p><p>Y mi misión cambió. Ya no es solo enseñarles a mis hijos. Es que ellos, yo y quien se anime le enseñemos a más personas, para que el juego siga pasando de mano en mano. Y ojalá no solo online, sino también en físico, en vivo, con un tablero y tres dados de por medio.</p><h3>Tecnología para cuidar lo que importa</h3><p>En NTT DATA Europe &amp; Latam hacemos proyectos increíbles y, más allá del impacto que generamos en nuestros clientes, el ambiente nos inspira a usar la tecnología en otros espacios. Esta historia nació de ahí: de creer que la tecnología también sirve para hacer cosas buenas por el mundo y para mantener vivas las tradiciones.</p><p>Si tienes un juego, una receta o una costumbre familiar que vive solo en la memoria de alguien, quizá sea buen momento de escribirla. O de programarla.</p><p>#Rust, #JavaScript, #AI for Good, #Backgammon, #Jacquet, #Chaquete, #Tradiciones</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d4bab6ee64c6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Who wins the Champions League?]]></title>
            <link>https://jsaraviadrago.medium.com/who-wins-the-champions-league-8dd30c832d20?source=rss-468fc5c36a75------2</link>
            <guid isPermaLink="false">https://medium.com/p/8dd30c832d20</guid>
            <category><![CDATA[soccer]]></category>
            <category><![CDATA[soccer-prediction]]></category>
            <category><![CDATA[soccer-analytics]]></category>
            <dc:creator><![CDATA[Juan Carlos Saravia]]></dc:creator>
            <pubDate>Fri, 27 May 2022 16:13:13 GMT</pubDate>
            <atom:updated>2022-05-27T16:13:13.443Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/855/1*3YNhJimON8Fwq6iqsU6TKQ.png" /><figcaption>RpcTV</figcaption></figure><p>That wonderful moment of the year has arrived, the UEFA Champions League final, for many Europeans and Latin Americans one of the most anticipated football events of the year. But what does this final await us?</p><p>Well, in truth, too many things await us, <strong>plenty of history, some drama, football and different tactics but more similar than we think</strong>. Oh obvious! Real Madrid versus Liverpool and always, of course, the data. But enough of introductions let’s go to what we came, the analysis of the game and what we can expect from this amazing game.</p><p><strong>Lots of History!</strong></p><p>This Saturday, May 28, two teams that are among the three teams with the most Champions League titles will face each other. To give us an idea, between them they add 23 different endings. Real Madrid has participated in 16 Champions League finals, the most successful team with 13, and two of those finals were precisely against Liverpool, who to their credit have 9 finals, of which they were champions 6 times.</p><p>Among them? In the 1980–81 season, Liverpool became champions by beating Madrid 1–0, while in the 2017–18 season they met again and this time, Madrid took revenge. He won 3–1 and with this he was proclaimed champion for the last time.</p><p>Do you remember that game? Well, that was the match where Sergio Ramos’ cursed man hurt Salah’s shoulder and took him out of the match. But it was also the game where Gareth Bale scored a goal from Chalaca (yes, it is Chalaca) and Karius had a blooper where he practically threw the ball at Benzema’s leg and apart from that he did not catch Bale’s shot from outside the area that entered and put the final 3–1.</p><p>Some media say that Ramos caused Karius a concussion on the head in a crash and that is why he had such a bad game. Beyond the bloopers, Bale’s goal is one to remember.</p><p>A new final is coming, and Mané is still the author of Liverpool’s transitory draw and my friend Karius is already very far from Anfield after he was loaned out a couple of times and his contract is almost expected to expire? And Ramos? Well we already know that story, he went to Paris Saint Germain to win French leagues but not be a force in the UEFA Champions League.</p><p><strong>Drama?</strong></p><p>Yes, to reach this final, Real Madrid had to go through drama more than Liverpool. How what drama? Yes, if you don’t remember it all started in the round of 16 when Madrid lost in Paris against Lionel Messi’s PSG with a great goal from Mbappé, the player in current controversy over his renewal with the French team.</p><p>But not content with that, Madrid in Spain started by losing the game with another goal from Mbappé until pressure came from Madrid. But we already know that story, Benzema putting pressure on Donnarumma and converting 3 goals in 17 minutes that literally gave him a pass to the quarter-finals and a date with Chelsea, current champions.</p><p>It seemed that this was going to be Madrid’s feat in this Champions League, but although they first beat Chelsea 3–1 at Stamford Bridge, Chelsea at the Bernabéu turned the game around. Yes, at one point the English team won 3–0 and left Madrid out. But then Rodrygo appeared…</p><p>Who? Rodrygo, the 21-year-old Brazilian striker who would become Madrid’s saving weapon. The carioca in the 80th minute scored the discount and although Chelsea was still winning 3–1, he gave life to Madrid because with this they went to supplementary.</p><p>The rest is history, Benzema in perhaps one of the best scoring streaks of his career, at least in the champions he converted again and in the 96th minute he gave the aggregate 5–4 to Madrid, which would be the final result . Benzema had 4 goals in two games, he dressed as a hero again but before that Rodrygo had rescued Madrid from being eliminated after volleying a brutal pass with the outside of Croatian Luca Modric’s right foot.</p><p>Around here in Peru, they say that Modric and Christian Cueva get along because they both like to give assists with the outside of their feet. Benzema is in a devastating moment but Rodrygo was only the beginning of what was to come, the game against Manchester City.</p><p>Again faithful to the drama, Madrid lost 4–3 in the first round of the semifinals and for a moment lost 3–1 but in the end, a penalty goal from Benzema (yes, the Frenchman again and with this he added 6 goals in 3 games) put things 4–3. Madrid had lost, but it was close, as they say, a stone’s throw away and only one goal behind to go to the Bernabéu.</p><p>What happened at the Bernabéu speaks of two things: on the one hand, City is a team that needs a psychologist rather than greater football power by hiring stars. The issue was already coming, he escaped from Atlético de Madrid in the quarterfinals because in reality the mattress team did not have a very clear one. But it is absolutely inexplicable how cracks of that nature simply stop playing when they feel they have the advantage. But City’s analysis could go to another post.</p><p>How so? Very simple, in the second leg until the 70th minute the game continued with City qualifying with a difference of one goal. But Mahrez put a distance of 2 goals when he scored in the 73rd minute, it seemed all done. And this is when the feat happened.</p><p>Rodrygo, dressed as a hero again as against Chelsea he scored the discount goal and the equalizer in a span of one minute and sent the match into extra time. A City totally confused by what happened, committed a quick penalty to Benzema, who would score his seventh goal in 4 games to send Madrid to the final.</p><p>Merit of Madrid? Yes, of course, but City also has a lot of responsibility, you can’t stop playing a game, the game ends when the final whistle blows. City stopped playing, but why do I say this? Beyond the goals is their form.</p><p>At the end of the match, Benzema receives a cross that throws Rodrygo backwards, who is the <strong>ONLY</strong> one who enters the ball. The City players jog, Rodrygo runs. They didn’t see her. But the second goal is even worse because not only does Carvajal cross with space, but in the box almost <strong>NO</strong> City players jump and Laporte does it by lowering his head a bit. The only one who jumped and attacked the trajectory of the ball well to connect it was logically Rodrygo, the author of the goal. City in the video simply seem to have stopped playing.</p><p>Well, you already told me a lot about Madrid! and Liverpool? The truth is that Liverpool does not have much drama to tell. Liverpool have been a locomotive, they only lost against Inter 1–0 but after having defeated them 2–0 in the Round of 16.</p><p>Perhaps the most dramatic moment of the Anfield team was in the semifinals when after beating Villareal 2–0 in Spain they were losing 2–0 and it seemed that a similar feat was coming. Recall that Villareal dramatically beat Bayern Munich in aggregate with a goal two minutes from time to maintain their 1–0 lead over Spain with a draw in Munich.</p><p>But the same thing was not going to happen with Liverpool, although a share of drama seemed to occur when at minute 41 Villareal had tied the series 2–2 after conceding two goals to the Reds, this quickly vanished in the second time.</p><p>Liverpool in just 12 minutes scored 3 goals against Villareal and went from having a tied series 2–2 to winning the series 5–2. Jurgen Klopp’s Liverpool is the engine of this tournament. And to be honest I don’t know if Madrid has enough to stop him.</p><p>But this is Mathematically possible so if there is no data we really leave the spirit that characterizes us. Even for the drama there is data, and Real Madrid has scored 14 goals between the Round of 16 and the Semifinals in those 6 games. He made 3 to PSG, 5 to Chelsea and 6 to Manchester City. But not only that, 6 of those goals were from the 80th minute onwards. He has even only scored 3 goals in the first half, clearly Real Madrid in this Champions League has not lowered its arms regardless of the number of goals it has been down.</p><p>One more fact! Real Madrid in the elimination phase has lost a game in all the series. They lost 1–0 against PSG in Paris, 3–2 with Chelsea in Madrid and lost 4–3 against City in Manchester. He was even down on the scoreboard at some point in 5 of 6 games, but the most important thing is the addition, it is here when things can get complicated.</p><p>Because although on aggregate it turned the result around in the three elimination series, which is an enormous feat, the final is a single match, Madrid cannot afford to be behind on the scoreboard, which has been its tendency in this phase of elimination of Champions.</p><p>But what about Liverpool? In this aspect the Anfield team is quite the opposite. As I say, in this champions it has been a locomotive. First in only 2 of the 6 games was he down on the scoreboard at some point. They only lost against Inter in the round of 16 1–0 after having beaten them 2–0 in England. Has he allowed goals? Yes, of course, he allowed 7 goals but even in the most dramatic moment, which was the match against Villareal, which I already mentioned that he was losing by 2 goals, he recovered and turned the game around. The offensive arsenal is powerful, they have scored 13 goals in these games.</p><p>Was there drama? Almost no, Liverpool were hardly down and when they were in trouble against Villareal they pulled themselves together and pulled it off. The interesting thing is the form, he has only scored 2 goals from minute 80 onwards and only 3 of his 13 goals in the first half.</p><p>But also, and thinking about how Real got here, Liverpool only allowed 1 goal after the 80th minute and allowed themselves to tie against Benfica after they won 3–1, after they had already won in the first leg 2–0. In other words, they let themselves draw when they already had a wide four-goal lead, but still only conceded 2 of the four that the Portuguese team lacked at that time.</p><p>The amazing thing is that we talk about Liverpool being a locomotive but Madrid have even scored one more goal than the Reds in the knockout stages. The white team has maintained the offensive production in terms of goals as it did in the group stage (if anyone wants to remember the data, they can go here and <a href="https://matematicamenteposible.net/?p=619">here</a> outputs everything). On the other hand, Liverpool scored 4 goals less than in the group stage but has still been much more dominant than its rival.</p><p>Let’s remember something super important, and this I think should be an interesting key. Real Madrid has allowed 11 goals in this knockout phase and in no match did their goal remain at 0. Liverpool had 2 games in which their goal was not beaten and allowed 7 goals in 6 games. Even that figure may be somewhat inflated by the two he conceded to Benfica when the series was almost closed. Equal! Although it was only once, the fact that the Reds have allowed him to return to the Portuguese team opens a door for Real Madrid, which has been seen as a resilient team.</p><p><strong>Tactics? Numbers?</strong></p><p>As we already know, Real Madrid is led by the great Carlo Ancelotti and Liverpool by the fabulous Jurgen Klopp. What do we talk about when we talk about tactics? Simple, Ancelotti’s Madrid, despite the fact that the coach is Italian, is more offensive than it seems. Since he is a disciple of Arrigo Sacchi, Ancelotti’s teams have played for a long time with 4 in the background and without these ultra-defensive and petty liberos that Italian teams have played for a great time.</p><p>But the coach when he started his career used to be, like Sacchi, a faithful practitioner of the 4–4–2 formation with two forwards and four midfielders and defenders. What Ancelotti does have as an Italian and quite strongly is that he uses a lot of counterattack. eye! It is not the typical counterattack where a ball is thrown from the area by a defender or a midfielder towards the forward, Madrid can even make counters with the ball on the ground with long passes to the ground from even the other striker who is Benzema.</p><p>Obviously, it has the Modric variant with that brilliance in the passes, but Madrid beyond the argument with the Croatian or the passes of the German Toni Kroos has these exits where Benzema makes Vinicius jr run so that he kicks or I returned it to the French for him to accompany. Crosses back after a long pass, either from above or low to the ground, is a weapon that Madrid uses a lot.</p><p>In general, the Madrid team has played with a fairly common 4–3–3 playing with Vinicius Jr, Benzema and Valverde as the top three players. Modric, Kroos are a little further behind to create the game and also support Casemiro somewhat in the brand. But Madrid can also sometimes push the full-backs to the top, remember that Carvajal sent the cross to Rodrygo for the headed goal against Manchester City.</p><p>But what do the numbers say? In attack according to <a href="https://www.driblab.com/analysis-team/liverpool-vs-real-madrid-all-the-data-on-the-champions-league-final/?utm_medium=email&amp;amp%C2%A0;_hsmi=214389188&amp;amp;_hsenc=p2ANqtz-_02JPFm-4_eJo3Q3ewfU9ideZyj8FR890a9aVHy6Ux9GEhLlRBFMQ_XzrRr_dng5QuroQqlQBSzQ-7-ebdjO9ofq0Lig&amp;amp;utm_content=214389188&amp;amp;utm_source=hs_email">Driblab</a> Real Madrid has averaged 2 goals without counting those he has converted from penalties and has shot almost 15 shots per game. In addition, in general he has divided possession of the ball with his rival (51.6%) and on average he has had 30.4 touches in the rival area.</p><p>Having 15 shots beyond that are on goal is a large volume. An issue that may have to do with this is that Madrid also recovers in the opposite field, on average it has 13.2 recoveries in the opposite field that although the recovery is a defensive event, if it does it in the opposite field it becomes an attack very easily . For example, the pressure that Benzema generated on Donnarumma at Madrid PSG generated a recovery that turned into a goal. It is a defensive act to recover the ball but with a lot of speed it can become an attack.</p><p>In simple terms, Madrid is offensive and as we will see later with Liverpool’s numbers it is not that far from the English team. Not only is it that both have scored almost the same number of goals but, in their style, they have had a very similar volume of attack.</p><p>But Real’s problem isn’t attacking but rather defense, again with data from <a href="https://www.driblab.com/analysis-team/liverpool-vs-real-madrid-all-%20the-data-on-the-champions-league-final/?utm_medium=email&amp;amp;_hsmi=214389188&amp;amp;_hsenc=p2ANqtz-_02JPFm-4_eJo3Q3ewfU9ideZyj8FR890a9aVHy6Ux9GEhLlRBFMQ_XzrRr_dng5QuroQqlQBSzQ-7-ebdjO9ofq0Lig&amp;amp;utm_content=214389188&amp;amp;utm_source=hs_email">Driblab</a>. He has conceded an average of 15 shots to his rival, but also, as we already know, they have converted 11 goals, although he maintains a positive difference because he scored 14 goals, the difference is not that big.</p><p>But beyond the shots, Madrid hasn’t been very prone to winning in the aerial game either, on average it has won 54.7% of the duels it has had. Although there are vital aerial duels such as Rodrygo’s header against City, failing here in a pivot by a Liverpool midfielder can generate a dangerous play so that, for example, Mané entered to connect the ball in the area of ​​​​Courtois, the Madrid goalkeeper.</p><p>What is very interesting is that Madrid has been shot quite a lot at the goal also because by design it has waited for its rival. Madrid has allowed 21.2 passes per defensive action (PPDA). This means that he allowed an average of 21 actions before trying to take the ball from the rival even with a foul. This speech shows an attitude of waiting a bit for his opponent before launching into pressure and having a more conservative way of defending. Who knows if he will push harder? The question is if Modric, Kroos and Casemiro give them the air to do it too. Pressing more is very exhausting. In these the Madrid team and Liverpool are diametrically different.</p><p>In the case of Liverpool, the issue is somewhat different, unlike Madrid, Liverpool plays absolutely modern football, with very high pressure on opposing defenders that pushes the entire team forward. If we see in the analysis of pressure from the field, in general, the presence of the Liverpool players is in the opposite field and well up pressing.</p><p>Why is this important? Because although it seems that the reds play with the same Ancelotti formation with 4–3–3 with Mané, Salah, Diego Jota or Luis Diaz as the three forwards in attack, this is not so much the case. First of all, Klopp has a choice between all these spectacular strikers which makes it an extremely flexible team to have variants in attack.</p><p>But Liverpool don’t just attack with their strikers. He actually attacks with the whole team and that’s what makes him very dangerous. It’s a super vertical game, what do I mean by this? They give a lot of forward passes and do not usually give many passes to the sides (as Manchester City does, for example). Precisely to calculate the verticality in football analytics, the angle and the number of passes you send forward beyond the distance are used.</p><p>But additionally, Trent Alexander-Arnold, although he is a winger, is a marker that goes up a lot, at times he even seems like a midfielder for everything that goes up. Again, if you review the position he is in, he is usually above midfield in attack, which shows that the team is very high.</p><p>Salah and Mané, although they are forwards and can play very well in the area, their position is from the sides to the inside. They cover the whole band and it’s not that they stay on top, especially Mané. This makes them also more dangerous because they can receive the ball from behind. Unlike Madrid who give these long passes where the ball travels great distances, in Liverpool they make a succession of very fast forward passes and then Mané, Salah or Diego Jota run with the ball. The addition of Luis Diaz has given the team in attack that shot from outside the area where days goes from the left to the middle and connects from outside the area.</p><p>Overall, Liverpool’s attacking numbers are still impressive, again according to data from <a href="https://www.driblab.com/analysis-team/liverpool-vs-real-madrid-all-%20the-data-on-the-champions-league-final/?utm_medium=email&amp;amp;_hsmi=214389188&amp;amp;_hsenc=p2ANqtz-_02JPFm-4_eJo3Q3ewfU9ideZyj8FR890a9aVHy6Ux9GEhLlRBFMQ_XzrRr_dng5QuroQqlQBSzQ-7-ebdjO9ofq0Lig&amp;amp;utm_content=214389188&amp;amp;utm_source=hs_email">Driblab</a> averages 2.25 goals excluding penalties and generates almost 17 shots on goal. He has an average possession of 63% despite having a very vertical game. But also, given the vertical and the pressure they exert, they recover almost 20 balls in the rival field, which is well above what Madrid does.</p><p>Winning it on the opponent’s court can be turned into a super fast attack with Salah and Mané up. With such talented players who can attack and with such a vertical game, it is interesting that the number of touches he has on average per game in the rival area is very similar to Madrid. It barely has 2 more than Madrid with 32.5 which shows that both are vertical and attack a lot but in different ways.</p><p>But in defense it’s not that he’s far behind. Virgil Van Dijk who is probably the best defender in the world, has a great ability to come out with closures and calculates very well when to leave and when to stay. On TV and YouTube we can always find videos of him frustrating counterattacks practically alone because he knows when to go out and challenge the ball and when to stay. But that is just one example of his great skill as a central defender, among the many he has.</p><p>In defense, Liverpool barely allowed 7 shots on goal, in other words, half of the shots allowed by Madrid. In addition, he wins 66.1% of aerial duels, which is well above what Madrid wins. Logically, having Van Dijk to play those aerial duels is a great help for this.</p><p>But what really sets these two teams apart is that the Reds have a PPDA of 7.77, which means that they don’t let their rival breathe and allow him barely 7 passes before going out to try to take possession of the ball. It is possible that this contributes to them having so many recoveries in the opposite field. Is it good to press? Yes as long as it works for you. With this system of pressing, Liverpool allows few shots on goal and recovers many balls, in other words, pressing works for him, while Madrid is more cautious.</p><p>But if we talk about pressing, it also has its pros and cons. We have already seen numerically the pros that Liverpool gives pressure but the big con is logically if the player escapes you. In Madrid versus City, Vinicius jr’s goal was because Fernandinho practically went out into the rival field so much that when Vinicius Jr lost, he was left with an open field to be able to make the 50-meter race to the Moraes goal. Trent Alexander-Arnold tends to come out a lot because in general pressing makes you advance all the lines with the big risk that you leave a big space behind, but it can be very risky Vinicius clearly showed us that on TV.</p><p>If you want more data, including players for this final, I recommend you visit the link <a href="https://www.driblab.com/analysis-team/liverpool-vs-real-madrid-all-the-%20data-on-the-champions-league-final/?utm_medium=email&amp;amp;_hsmi=214389188&amp;amp;_hsenc=p2ANqtz-_02JPFm-4_eJo3Q3ewfU9ideZyj8FR890a9aVHy6Ux9GEhLlRBFMQ_XzrRr_dng5QuroQqlQBSzQ-7-ebdjO9ofq0Lig&amp;amp;utm_content=214389188&amp;amp;utm_source=hs_email">Driblab</a>.</p><p><strong>Predictions!</strong></p><p>According to the <a href="https://www.sportingnews.com/ar/futbol/news/real-madrid-liverpool-betting-favorites-predictions-final-champions-league-2022/gcy6zx6vpyxoraso5gb9a5ru">Sportingnews</a> Liverpool wins the game and is theoretically favored by 1.61 and Real Madrid by 2.3. In other words, if you bet 100 you win 161 if Liverpool wins and if you bet 100 you win 230 if you put your chips on Real Madrid.</p><p>For me, Liverpool wins 3–1 and don’t be scared if the vast majority of goals come in the second half and the safest thing is that we will have a wonderful game.</p><p>If you liked my content, leave a message and follow me for more articles. Be encouraged to leave a comment with your opinion!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8dd30c832d20" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How making my own open source package saved me tones of time]]></title>
            <link>https://medium.com/analytics-vidhya/how-making-my-own-open-source-package-saved-me-tones-of-time-848a77f77b82?source=rss-468fc5c36a75------2</link>
            <guid isPermaLink="false">https://medium.com/p/848a77f77b82</guid>
            <category><![CDATA[coding]]></category>
            <category><![CDATA[open-source]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[r]]></category>
            <dc:creator><![CDATA[Juan Carlos Saravia]]></dc:creator>
            <pubDate>Thu, 05 Aug 2021 14:16:45 GMT</pubDate>
            <atom:updated>2021-08-08T17:14:16.667Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/444/1*GaIHNsJJsUeHtRvkvaJT4w.png" /><figcaption>Own source</figcaption></figure><p>Packages are core in many open source languages. For example, in CRAN in R you can find 17898 packages already published and more than 200,000 packages in Python.</p><p>These numbers doesn&#39;t count all of the packages that are in github that are very useful. I spend a lot of time programming in R and Python but there are some specific projects that I favor using R.</p><p>Almost all of my psychometric projects are coded in R because I feel it is a bit more developed than Python in that topic and I am also very used to it.</p><p>One of the big pains that I always have is that sometimes I can&#39;t find very specific analysis that are becoming popular and highly. One of them is the Composite reliability (CR) index and the other is pretty common average variance extracted (AVE) in a Confirmatory factor analysis (CFA) framework (Kline, 2016).</p><p>Since both analysis are in many books I thought that I could find a very simple function that could execute those two analyses. Well I was dead wrong!</p><p>Since I used to use the lavaan package in R for all analysis around the family of structural equation modelling (SEM) I started calculating the CR and AVE manually.</p><p><strong>How did I do it?</strong></p><p>I created a simple function in R and run it inside the script I was working on. Something like this:</p><pre># Manually had to hard code the values <br>cargas1 &lt;- abs(c(0.67,0.66,0.69,0.59))<br>errorv1 &lt;- abs(c(0.55,0.57,0.52,0.66))</pre><pre># Created the functions for composite reliability </pre><pre>comp.rel &lt;- function(x,y) {  operacion &lt;- (sum(x) * sum(x))/(sum(x) * sum(x) + sum(y))  return(operacion)} </pre><pre>comp.rel(cargas1,errorv1)</pre><pre># Manually had to hard code the values</pre><pre>cargas2 &lt;- abs(c(0.30,0.54,0.54))<br>errorv2 &lt;- abs(c(0.91,0.71,0.71))</pre><pre># Created the functions for AVE</pre><pre>ave &lt;- function(x) {  operacion &lt;- sum(x^2)/length(x)  return(operacion)}</pre><pre>ave(cargas1)</pre><p>This solution was suboptimal in MANY ways:</p><ol><li>It was hard coded therefore if in any future I wanted to deploy this it could cost massive headache to the team.</li><li>It was messy because I had to copy paste the newly created functions and sometimes I didn&#39;t find them and I had to rewrite them.</li><li>It was difficult to read for me and the person that check my code because it was tough to understand the numbers without a good explanation or a readme.</li></ol><p>And the list goes on…</p><p><strong>Introducing the </strong><a href="https://github.com/jsaraviadrago/bluegrafir"><strong>bluegrafir</strong></a><strong> package</strong></p><p>This happened some years ago and one day I just became fed up with the situation. So I decided to make my own package and put it in github. The bluegrafir was named after my 2 cats (blue and grafito) and aimed to make a simple way to execute these and other functions.</p><p>Overall, it is an implementation of tidy tables, linear transformations of continuous variables, average variance extracted and composite reliability for Confirmatory factor analysis (CFA).</p><p>Besides that was super fun to make my own package it REALLY helped make my work much more efficient. In order to make the package I had to really go deep into the inner workings of the lavaan package, construct the functions, test them and make documentation. This really helped me solve some of my main problems and make my work more efficient. For example:</p><ol><li>I found that all the values needed to calculate CR and AVE where inside a a deep lavaan object therefore I could extract it from there instead of hard coding them.</li><li>To open the function I didn&#39;t need to create it again I just had to call it through the <a href="https://github.com/jsaraviadrago/bluegrafir">bluegrafir</a> package.</li><li>I created a pretty detailed Readme so if somebody wanted to know what in the world the functions and the package did they could read it there.</li></ol><p>Furthermore…</p><p>4. My team started to use the package so we weren&#39;t &quot;passing functions&quot; around and also everybody knew what the analysis was about and how it was done. So it was much easier to make reproducible results because all of us executed the same code.</p><p>5. My team started to have more time because we were available to reduce time on manual tasks and use it more on creative things like for example business problem solving.</p><p>To make the package organized there are some blue_something functions and other graph_functions. All blue_ functions are focused on linear transformations and grafi_ in graphs and tables. I decided to make it that way so there is a coherent line of thinking between the name of the package and it&#39;s functions. CR and AVE had different name because they were made to calculate different things. Packages like “testthat”, “devtools” and “roxygen” were fundamental to create, test and put the package to use.</p><p>If you are curious to check on the package you can do it by clicking here (<a href="https://github.com/jsaraviadrago/bluegrafir">bluegrafir</a>).</p><p><strong>Just a quick recap</strong></p><ol><li>It is pretty fun to make your own package, you learn a lot about the ins and outs of making package and forces you to learn good coding practices.</li><li>Creating my own package helped me become much more efficient with my work. It helped reduce hard coding, inefficient coding and communicating coding with team members.</li><li>Creating a package helped all the team reduce manual tasks, spend it in more valuable things and increased the probability of reproducible results and deployment.</li></ol><p>That&#39;s it! Hope this post motivated you to make your own package! Who knows? Maybe you can personalise it and give it a great name and logo.</p><p>Happy coding!</p><p><strong>Reference</strong></p><p>Kline, R.(2016). <em>Principles and practice of structural equation modeling. Forth edition</em>. NY: Guilford Press.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=848a77f77b82" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/how-making-my-own-open-source-package-saved-me-tones-of-time-848a77f77b82">How making my own open source package saved me tones of time</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The story that has helped develop my Data Science career]]></title>
            <link>https://medium.com/analytics-vidhya/the-story-that-has-helped-develop-my-data-science-career-9cc6dbae16dc?source=rss-468fc5c36a75------2</link>
            <guid isPermaLink="false">https://medium.com/p/9cc6dbae16dc</guid>
            <category><![CDATA[advice-and-opinion]]></category>
            <category><![CDATA[storytelling]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Juan Carlos Saravia]]></dc:creator>
            <pubDate>Thu, 22 Jul 2021 14:27:35 GMT</pubDate>
            <atom:updated>2021-08-01T14:37:13.237Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fpTLDEJ9ZfZDw-y5WVBmLA.jpeg" /><figcaption>source: uptake.com</figcaption></figure><p>Having a career in data is a hefty challenge. You have to know many stuff about a lot of different topics: math, coding, machine learning, cloud, just to name a few. The list is huge and it becomes bigger by the day.</p><p>If you search the web for &quot;advice for a data scientist&quot; you could get 128,000,000 hits in less than 1 second. You go to Linkedin and there are a bunch of people posting articles about what does a Data Science needs to do or the phases of a good data science project. Sure, I am guilty on that as well and probably this post also has some things that have already been said there.</p><p>But in this case, I wanna tell you a story my thesis supervisor told me that has sticked with me ever since I heard it and has been vital in my Data Science career.</p><p><strong>The story goes like this…</strong></p><p>We were finishing the multivariate techniques course which by the way was called &quot;From problem to analysis&quot; (a really great way to emphasize what really data science is all about) and he told us that a very important lesson for the course was not regarding the technical stuff but soemthing else. And he went something like this:</p><blockquote>&quot;Imagine that you have to make an important presentation which airs on the Sunday news. It will be the National Channel and everyone is going to hear about your work. In your audience you will have my grandmother who is a mathematician and my auntie who is a writer and hates maths.</blockquote><blockquote>Therefore, your work needs to target all audiences. You have to be as <strong>PRECISE</strong> as possible so my grandmother doesn&#39;t get mad at you and at the same time as <strong>CLEAR</strong> as possible so my auntie really understands what you are saying.</blockquote><blockquote>Your work will be very important and you should focus on ensuring that nobody in the country is left out&quot;</blockquote><blockquote>- Tacq</blockquote><p>At the time I thought the story was nice and made lots of sense. As time passed I realized that it was more useful that I could imagine. It talked about many important skills a Data Scientist should have to have to be successful. After giving a big deep thought about it I made a list of actions to get those skills for my own use that now I am sharing with you.</p><p><strong>The skills to be PRECISE and CLEAR on the Sunday news…</strong></p><ol><li>Learn the math so well that you can be <strong>PRECISE</strong> when you explain what your work is about to stakeholders or fellow data professionals.</li><li>Be sure you understand the math well, so you can be crystal <strong>CLEAR</strong> when you explain to stakeholders what your work is about and it&#39;s impact.</li><li>Get really deep into your own coding skills, so you can do tidy and <strong>CLEAR </strong>code that can be explained to a colleague or a team member and put it into production.</li><li>Understand the problem well, so you can be <strong>PRECISE </strong>and can tackle the problem with the proper analytics solution (From problem to analysis, remember?).</li><li>Learn how to be a good storyteller so you can be <strong>PRECISE</strong> and <strong>CLEAR</strong>. Telling a great story with Data is difficult it requires communication skills, preparation, a profound understanding of your work and great problem solving. But if you manage to do it you can make a great impact and bring lots of value.</li></ol><p><strong>…and the list goes on.</strong></p><p>6. Have fun.</p><p>7. Always be prepared to be amazed…</p><p>Probably all these points aren&#39;t new to you. Everyone talks about them and you have certainly came across them a thousand times. In reality and in my humble experience, it is very easy to talk about them but difficult to put them into practice.</p><p>Daily work can easily shift your focus from some important stuff or maybe experience may let you think you are proficient on one or more of these skills. Even though this could be true, for me, thinking that one day I will have to be on the Sunday news is a good reminder that it is always good to revisit and keep developing these skills.</p><p>If your work makes sense to stakeholders, develop into actions from other company members and translate into great results then you&#39;re doing well. But for that to happen, you and your team has to do great analytics and make every stakeholder know about it.</p><p>And for that to happen you have to be <strong>CLEAR</strong> and <strong>PRECISE.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9cc6dbae16dc" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/the-story-that-has-helped-develop-my-data-science-career-9cc6dbae16dc">The story that has helped develop my Data Science career</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The use and misuse of Net promoter score (NPS) metric]]></title>
            <link>https://medium.com/analytics-vidhya/the-use-and-misuse-of-net-promoter-score-nps-metric-39d5448331f7?source=rss-468fc5c36a75------2</link>
            <guid isPermaLink="false">https://medium.com/p/39d5448331f7</guid>
            <category><![CDATA[product]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[marketing]]></category>
            <category><![CDATA[metrics]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Juan Carlos Saravia]]></dc:creator>
            <pubDate>Wed, 30 Jun 2021 22:48:05 GMT</pubDate>
            <atom:updated>2021-07-05T06:44:15.625Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NLX6lrUdH7HUSPzFBtuWuQ.png" /><figcaption>retrieved from: trustmary.com</figcaption></figure><p>Net promoter score (NPS) is a widely used score that measures the likelihood that clients would recommend your product (Reichheld, 2003). Since it was introduced it has wreak havoc in many industries pushing many metrics aside and becoming the go to metric for customer loyalty.</p><p>But as a result of it&#39;s great popularity some dangers have also arise with it. Some of them are a result of how the NPS metric works and others on how people tend to use it. Just to make myself perfectly clear, NPS is a metric that can give you valuable information when it is used properly and in the right conditions. The thing is that this is not always the case.</p><p>Therefore, the purpose of this article is to share with you some ideas about the proper use and misuse of the NPS metric. These ideas are based on simple research methods, psychometrics and statistics.</p><p>In this case, in order to use it properly you should follow this guidelines:</p><ol><li>Understand the business problem and judge if the NPS really answers the question you are asking.</li><li>Avoid using it alone, always with more metrics</li><li>Divide the metric into it&#39;s pieces and use the overall metric AND their pieces.</li></ol><p>Before I go into detail I will also share with you the dangers of this metric:</p><ol><li>Neglecting to ask the business questions and therefore use the metric as a jack of all trades.</li></ol><p>2. Have a lack of information as a result of collapsing the values to calculate the metric.</p><p>3. Gather measurement error as a result of the difficulty customers could have discriminating between the numbers of the question.</p><p><strong>Ok! very nice but how do I do all this and avoid the dangers?</strong></p><p>The great thing is that when you follow good practices at the same time you avoid the dangers (a bit obvious but important to point it out).</p><p>In this case,<strong> I would say that the best practices is to really understand the problem and make the research question. </strong>NPS aims to measure customer likelihood to recommend a product. In that sense, if you want to use it you should be perfectly clear that you truly want to measure customer likelihood to recommend and not another thing.</p><p><strong>Understanding the problem</strong></p><p>NPS doesn&#39;t measure customer satisfaction directly. Even though it can be a proxy of it, it&#39;s not exactly the same. So be careful not to think that you could equate likelihood of recommending a product with general customer satisfaction. People could have a different logic when they answer both questions. You could be satisfied with a product but not recommend it because you think it isn&#39;t for everyone.</p><p>The important idea here is, use the metric when the problem really suits it. NPS isn´t a jack of all trades for customer satisfaction, likelihood of buying a product or not leaving a mobile app.</p><p><strong>Avoid using the metric alone</strong></p><p>If you want to understand overall satisfaction or your customers use many questions besides the NPS. There is no silver bullet, therefore with one question it is not posible to gather all the information you need. Human brain is pretty complex and psychological constructs such as customer satisfaction and likelihood to recommend is better to measure it with multiple questions.</p><p>After you measure it, you could also make some nice psychometric analysis to check for measurement reliability and evidence of validity. This is possible by measuring your construct with multiple questions.</p><p><strong>Divide the metric so you can avoid information loss</strong></p><p>NPS is measured in a scale from 0 to 10. But when you calculate the metric you really use, 9, 10 (promoters) and 0 to 6 (detractors). Therefore, 7 and 8 isn&#39;t used in the calculation. In spite of this, the person that has to answer the question has to think on it and therefore draws his/her attention and cognitive energy. This wastes valuable seconds that can be used for example to plug another question that could give you even more information.</p><p>Furthermore, a scale of 0 to 10 is pretty large and for a person it could be very difficult to discriminate between a 5 and 6 or a 4 and 3. All of these cases are detractors. In an extreme example, a 0 and a 6 are also tagged the same. But one is pretty near to the passive side and the other is totally unwilling to recommend your product. This is pretty valuable information you are losing as a result of collapsing the numbers.</p><p>In that sense, a great thing to do would be to reduce the amount of options you are using for the NPS so every option could have a more meaningful value for the person. In other words, the person is able to discriminate between one number and the other.</p><p>If you really don&#39;t want to tweak the real metric a bit you could get two measurements from it. 1) the real NPS with your group of detractors and promoters and 2) the amount of people that answered each value so you could zoom in the detractors, promoters and even passive group.</p><p>It is possible that you have lots of people near the edges of another group (for example many people answered 6) and are just outisde of going out from detractors. Also, it would be wise to know how many people really would recommend your product (10) and also the amount of people that by no means will do it (0 and 1 maybe).</p><p><strong>Just a quick recap!</strong></p><ul><li>Understand the problem, check if the NPS is really the right metric for the business problem you want to solve.</li><li>Don&#39;t use NPS alone, make a nice questionnaire where you could gather a greater variety of features and test how feasible that information is.</li><li>Reduce the number of option in the NPS. For many people detecting the difference between a 4 and a 5 could be very difficult. Also, use the real NPS and also the whole scale. A detractor that answers 6 is quite different from one that answers 0, that is valuable information.</li></ul><p>Well that’s it! Hope you like it! Happy measuring!</p><p><strong>Reference</strong></p><p>Reichheld, F. (2003). The one number you need to grow. <em>Harvard Business Review.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=39d5448331f7" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/the-use-and-misuse-of-net-promoter-score-nps-metric-39d5448331f7">The use and misuse of Net promoter score (NPS) metric</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Market segmentation is not enough? Try Multilevel modelling afterwards]]></title>
            <link>https://medium.com/analytics-vidhya/market-segmentation-is-not-enough-try-multilevel-modelling-afterwards-eee67d11954f?source=rss-468fc5c36a75------2</link>
            <guid isPermaLink="false">https://medium.com/p/eee67d11954f</guid>
            <category><![CDATA[data]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[multilevel]]></category>
            <dc:creator><![CDATA[Juan Carlos Saravia]]></dc:creator>
            <pubDate>Thu, 03 Jun 2021 18:49:37 GMT</pubDate>
            <atom:updated>2021-06-09T06:25:03.710Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/825/1*Lxa0OQ_Nr1O5lbp7Y9gzFw.jpeg" /><figcaption>Picture: Eric Paquette brookmark blog post</figcaption></figure><p>One of the most common strategies in analytics projects is marketing segmentation. In analytics, there are a wide variety of techniques that can help you make sense of unlabeled data and cluster your market. Unsupervised learning techniques are very nice, they let you reveal hidden patterns and make sense of data that you didn’t have a clear idea how it was behaving. Remember! As Airbnb says, data captures the people’s voices.</p><p>The thing is that when you execute a clustering technique such as K means clustering or K nearest neighbours you end up with multiple clusters. Afterwards, Marketing and sales teams tend to label those groups and based on the characteristics of each grou put them a very nice name.</p><p>The trick here is that your broad clusters even though they are supposed to be homogenous, they still have some hidden behaviors inside those big clusters.</p><p>This is were multilevel modelling comes into play. With this technique you can go one step further and check for specific effects on specific groups and take into account more variables that can affect your market segmentation.</p><p>Therefore, the purpose of this post is to tell you about how Multilevel analysis can give you an edge over other data scientist and companies that only settle on market segmentation. In that case, this post is aimed for people that do analytics in a daily basis and also marketing teams that propose different market strategies based on a market segmentation.</p><p>Just for the sake of discussing the same terms, when I talk about market segmentation I am referring to the process that “divides a market into well-defined slices” (Kotler &amp; Keller, 2016, p. 268).</p><p>Ok, I knew that one! But, what in the world is Multilvel modeling? As Joop Hox said, “In multilevel research the data structure in the population is hierarchical, and the sample data are viewed as a multistage sample from this hierarchical population” (Hox, 2010, p. 1). Sorry…I don’t have a clue about what you are talking about.</p><p>In simple terms, is the technique which models variables that are related to a broader level such as location, state or even type of industry. Therefore, multilevel modelling can capture the different effects that the context has in specific groups.</p><p><strong>Ok, very nice but market segmentation has done wonders for me why change?</strong></p><p>Just to be clear here, I’m NOT suggesting that you drop your marketing segmentation or cluster analysis all together. What I’m implying is that you go a step further. AFTER doing the market segmentation apply multilevel modeling to get further inisights.</p><p><strong>Why should you do it?</strong></p><ol><li>One: Market segmentation is a grouping technique and assumes that inside every cluster behaviors are homogeneous. If you put the clusters in a model (for example a regression) to predict sales, you are assuming that the effects of broader variables (such as weather) is the same for every group. Which can be false.</li><li>Two: Multilevel modelling can help you check if the effect of a higher variable differ in your outcome variable. For example, if you execute multilevel linear regression to predict sales, since this model separates characteristics of people (age, sex, etc) and context features (weather, altitude, etc), you can check if the effect of a higher variable such as weather differ in sales inside each cluster.</li><li>Three: Marketing teams can make even more personalized promotions because they have more information on the mechanisms that are working between the segments and the context.</li></ol><p>Ok, this sounds a bit abstract. Let’s go make an example, shall we?</p><p><strong>Quick example</strong></p><p>Imagine there is an electronics web store named “techsports” which sells all kinds of electronic gadgets for sports.</p><p>Now, imagine there is a marketing team that wants to understand what “types” of clients are shopping in there web page, based on time of purchase, age, sex and even location. So they ask the analytics teams if they are so kind to make the analysis. The team has all the data in the cloud so they use a bit of SQL to retrieve the data and make there master table.</p><p>Afterwards they perform the clustering analysis, flag each client and put an excel file using a FTP so the marketing team can analyze the result with a Dashboard. Pretty cool, huh?</p><p>The marketing team finds out that there is a group of clients that are from Europe, female, around 35 years old that like to buy headphones in the morning. Like that group, they find 5 other different groups. Based on that information they do a marketing strategy but afterwards, find out that the promotions they are implementing are not resulting in any marginal revenue. They feel puzzled so they go back again to the analytics team to think together what is happening.</p><p>During the discussion the whole team finds that even though the group described before likes to buy headphones, at the time of the promotion in some places this is not the case. So they start doing some problem solving and hypothesis testing and find out that something has to do with the weather.</p><p>So the analytics team starts to do some multilevel analysis and tries to find out if the place people lived and the climate has something to do with client behavior. So, in level 1 they put people&#39;s variables (such as the segments) and in level 2 they put cities features such as weather.</p><p>After the setup, the analytics team predicts sales with a regression model and in general the type of weather is related to greater sales. But they find out that this effect of weather is reversed in rainy cities. In other words, people buys less headphones in those cities.</p><p>These results were revealing because it helped the Marketing understand that it wasn&#39;t such a good idea to take only the segments and that even though the weather was a nice indicator of greater sales this didn&#39;t hold up in all cases.</p><p>In that case, they shifted their promotion strategy and took into account the segments and the context variables. At the end, the marketing team was very happy because they were able to refocus and the analytics team were satisfied of being able to help them redirect the marketing campaign.</p><p><strong>Very interesting! Are there any software for this?</strong></p><p>Of course! In the open source community you can find libraries such as lm4 or multilevel for R. If you are not an R user and you prefer Python there is no problem! The statsmodels package has also a lmm function that helps you make these kind of models.</p><p>If you are a Random forest fan you can even make mixed effects random forests in Python. <a href="https://towardsdatascience.com/mixed-effects-random-forests-6ecbb85cb177">Here</a> I leave you a great post about it.</p><p><strong>Just a quick recap!</strong></p><ul><li>Market segmentation is a great tool for marketing teams to understand their costumers and make marketing campaigns. But, with the inclusion of multilevel analysis AFTER your segmentation you could end up with much better results that yields greater value.</li><li>Multilevel analysis takes into account not only people&#39;s characteristics but also contexts features.</li><li>Multilevel analysis could be very helpful to reveal if the effect of a certain context variable vary between different places. In the literature, this is called a random effects model (Hox, 2010)</li></ul><p>Well that’s it! Hope you like it! Happy coding!</p><p><strong>Reference</strong></p><p>Kotler, P., &amp; Keller, K. L. (2016). <em>Marketing Management</em> (15th ed.). Pearson.</p><p>Hox, J. (2010). Multilevel analysis. Techniques and Applications. Quantitative Methdology Series. Second Edition. Routledge.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=eee67d11954f" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/market-segmentation-is-not-enough-try-multilevel-modelling-afterwards-eee67d11954f">Market segmentation is not enough? Try Multilevel modelling afterwards</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why I use Psychometrics in my analytics projects and you should too]]></title>
            <link>https://medium.com/analytics-vidhya/why-i-use-psychometrics-in-my-analytics-projects-and-you-should-too-369b1233ff91?source=rss-468fc5c36a75------2</link>
            <guid isPermaLink="false">https://medium.com/p/369b1233ff91</guid>
            <category><![CDATA[data]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[psychometrics]]></category>
            <dc:creator><![CDATA[Juan Carlos Saravia]]></dc:creator>
            <pubDate>Sat, 22 May 2021 16:06:47 GMT</pubDate>
            <atom:updated>2021-06-09T03:11:34.606Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/698/1*4b8aIkKwXMFK0m88toEa3A.jpeg" /></figure><p>In the last couple of years, data analytics has been in a roll. Since big powerful companies such as Google, Twitter, Airbnb and Facebook started using data, every other industry has been trying to get on the analytics-side of things.</p><p>Some of these companies have managed to make it work and others -not so much. Even though there are a variety of reasons why this happens, I am going to focus on the day-to-day use of analytics.</p><p>So, this post is aimed for people that do analytics in a daily basis (as a job, for fun, on premise or on cloud) and why you should use psychometrics in your projects. The answer is very simple: you will get better inputs to introduce in your models and probably much better results. Why is that? Let&#39;s find out!</p><p>First of all, what in the world is &quot;Psychometrics&quot;? Psychometrics is a discipline focused on the quantitative measurement practices in social sciences and education. Or, as Galton referred more than a century ago: Psychometrics is “The art of imposing measurement and number upon operations of the mind” (Galton, 1879, p149). So in simple terms: psychometrics attempts to measure the mind with questionnaires and other tools.</p><p>Psychometrics is mainly used in academics and educational settings such as in big standardized assessment tests. But it is also used in enterprises by marketing teams who apply questionnaires in order to collect information about customers. With the digital world and e-commerce in full speed, questionnaires are less and less paper-based and more web-based. We now use tools such as Survey Monkey and Google Docs -to name a few. So if you think psychometrics is only for educational settings, you might be wrong! I would say that this is a misjudgment of the potential of the Psychometric tools.</p><p><strong>Why should you use it?</strong></p><ol><li>One: It is quite simple: many analytics models tend to use information about people’s thoughts, perceptions and feelings. Ok, I know there are many techniques that can also be used for that -such as sentiment analysis or even tweets mining… but the nice thing of making questionnaires is that you can control EXACTLY what information you want to gather.</li><li>Two: This one is very important, you can handle the garbage-in and garbage-out problem. If you test for test-structure and reliability before you insert the results of the questionnaire in the model, you could further understand if the answers people give you are useful and a truly measure of what you are asking. In that case, you could first check if the measurements are reliable before putting them into a model (statistics or machine learning model). If the measurement is not reliable you don&#39;t add noise to your model. That said: a lesser noise means better predictions or a better precision in your model.</li><li>You reduce the danger of making wrong assumptions on the effect people&#39;s perceptions have in the model.</li></ol><p><strong>Let&#39;s give a quick example</strong></p><p>You make an index based on 10 questions regarding clients satisfaction on the experience of a web page. Maybe in your questionnaire you had 3 questions regarding the font type, and other questions regarding the look and feel of the web page, how easily you can navigate on it and find what you are looking for.</p><p>You create your index with the 10 questions and you assume that font type has something to do with client&#39;s satisfaction. You start doing your exploratory data analysis, you do some correlations and realize that there is a relationship between client satisfaction and revenue but it&#39;s not that strong. So you feel happy but after I while you start to wonder… Why is the relationship so low?</p><p>You go back to your questionnaire do some factor analysis and realize that your 3 questions regarding font type has no relationship to the whole measurement of client&#39;s satisfaction. Therefore, you drop those items and recalculate your index, execute the same correlation once again and the effect has become much stronger. Then you can feel more confident on introducing your index to a more complex model to maybe predict the website&#39;s sales.</p><p><strong>Are there available tools to do this?</strong></p><p>In the open source community in languages such as R and Python (to name a few) there are many packages and libraries that can help you with this. For example, in R you can use Mirt or Lavaan. In Python, you can use girth or psychometrics. Therefore, the language you use should not be a problem to execute psychometric analysis.</p><p><strong>Ok, very nice but I&#39;m a big and powerful enterprise that puts models in production and I don&#39;t play games, is this going to help me?</strong></p><p>Of course! You can even add the psychometric analysis as a part of the data prep with a nice ordered, scalable and not hardcoded pipeline. You can even automatize the times the questionnaires needs to be applied and those answers could go to your cloud or you could even manually put it inside your cloud storage if you wish. You could pick the information digitally, clean it, make it part of your data prep and then put it in your final model in production.</p><p><strong>Just a quick recap!</strong></p><ul><li>In this article I suggested that as a good practice you should use psychometric tools to check if your questionnaires results are relaible.</li><li>Using questionnaires gives you the power on what you can ask, handle the garbage-in, garbage-out problem and reduce the danger of wrong assumptions.</li><li>There are lots of open source tools that can help you with this such as mirt and lavaan in R and girth and psychometrics in Python.</li><li>Psychometrics can be put into your pipeline for model production without your data engineer suffering from excruciating pain in the process.</li></ul><p>Well that&#39;s it! Hope you like it! Happy coding!</p><p><strong>Reference</strong></p><p>Galton, F. (1879). Psychometric experiments. <em>Brain: A Journal of Neurology, 11</em>, 149–162.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=369b1233ff91" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/why-i-use-psychometrics-in-my-analytics-projects-and-you-should-too-369b1233ff91">Why I use Psychometrics in my analytics projects and you should too</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[¿Qué nos trae el Super Bowl 2021 (Mahomes vs Brady)?]]></title>
            <link>https://jsaraviadrago.medium.com/qu%C3%A9-nos-trae-el-super-bowl-2021-mahomes-vs-brady-278ee155621a?source=rss-468fc5c36a75------2</link>
            <guid isPermaLink="false">https://medium.com/p/278ee155621a</guid>
            <category><![CDATA[super-bowl]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[nfl]]></category>
            <dc:creator><![CDATA[Juan Carlos Saravia]]></dc:creator>
            <pubDate>Sat, 06 Feb 2021 19:52:35 GMT</pubDate>
            <atom:updated>2021-02-06T19:52:35.878Z</atom:updated>
            <content:encoded><![CDATA[<p>Este Domingo es el Super Bowl y para los fanáticos del fútbol americano es el evento más importante del año. El partido será entre los Tampa Bay Buccaneers vs Kansas City Chiefs. La estrella que pareciera eterna, Tom Brady participa en su décimo Super Bowl versus la estrella emergente Patrick Mahomes que busca coronarse por segunda vez como campeón. Entonces, sin más que decir les traigo algunos datos para que tengan mayor contexto del partido que se viene el Domingo 7 de Febrero.</p><p>El Super Bowl suele ser en teoría un partido donde el mejor equipo de la AFC (Conferencia americana) se enfrenta al mejor equipo de la NFC (Conferencia nacional), dado que son los mejores en teoría debería ser un partido muy reñido y parejo. Si bien este suele ser el caso esto no siempre ha sido así.</p><p>Márgenes de victoria en Super Bowl y quién fue el <strong>equipo ganador</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3HnKsMH-MtL5-1PM1BEXLQ.jpeg" /></figure><p>En esta gráfica vemos por años el margen de victoria que tuvo el equipo que se coronó campeón. Como se puede ver existen partidos sumamente parejos donde Los NY Giants ganaron por 1 punto a los Buffalo Bills (1991) 20–19. Sin embargo, han habido otros partidos donde las victorias han sido sumamente holgadas, Los San Francisco 49ers en 1990 ganaron el Super Bowl venciendo a los Denver Broncos 55–10, la victoria más aplastante de la historia.</p><p>Márgenes de victoria en Super Bowl: y quién fue el <strong>equipo perdedor</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7eyveZVKHoI-9J51Vtn_Hw.jpeg" /></figure><p>Naturalmente, dentro de estas gráficas se encuentran tanto los equipos de Tom Brady que participaron en el Super Bowl así como el equipo de Patrick Mahomes quien se coronó campeón el año pasado venciendo a los San Francisco 49ers 31–20. Entonces el diferencial de puntos a lo largo de los años ha variado entre 1 y 45 puntos de diferencia.</p><p><strong>¿Pero más allá de la historia y los márgenes que particularidades trae este Super Bowl?</strong></p><p>Es la primera vez que Tom Brady no participa en un Super Bowl como miembro de los New England Patriots con quién participó en 9 ocasiones. De esas 9, ganó 6 y perdió 3 y esos partidos fueron decididos por un margen entre 3 y 10 puntos como máximo. Lo curioso de las victorias de Brady es que ha perdido con dos equipos (Los NY Giants y los Philadelphia Eagles) de la misma división (NFC del este) a pesar de ser favorito en esos partidos. Adicionalmente, ganaron el Super Bowl del 2015 deteniendo a los Seattle Seahawks en la última jugada del partido a menos de 5 yardas que anoten los Seahawks para ponerse arriba en el marcador.</p><p>Esta no es la primera vez que Tom Brady se enfrenta a una ofensiva potente como la de Kansas City y Patrick Mahomes. En el 2002 los New England Patriots con Brady ganaron el Super Bowl, derrotando a los Rams de Kurt Warner deteniendo &quot;El show más grande en el gramado&quot;, que era la ofensiva de St Louis (ahora Los Ángeles), con jugadores fabulosos como Kurt Warner, Marshal Faulk, Isaac Bruce y Torry Holt. Además, en el 2017 New England volvió 27 puntos abajo contra la mejor ofensiva de ese año liderada por Matt Ryan de los Atlanta Falcons quien se coronó como el jugador más valioso de la liga ese año.</p><p>¿Pero con esto quiero decir que los Bucs de Brady van a ganar el partido? Ni mucho menos, es más Los Kansas City Chiefs según las predicciones de Las Vegas son los favoritos para ganar el partido. Además, la última vez que el equipo de los Chiefs no fue favorito en un partido esta temporada fue la tercera semana de la temporada cuando se enfrentó contra los Baltimore Ravens. A pesar de ello, Kansas City ganó ese partido 34–20 y llega al Super Bowl tan solo con 2 derrotas en la temporada (Contra Los Ángeles Raiders y Los Ángeles Chargers). Y si hablamos de esta temporada, Los Chiefs y los Bucs ya se han enfrentado una vez (Semana 12) donde Kansas City le ganó a Tampa Bay 27–24. En ese partido Patrick Mahomes, destrozó a la defensa secundaria (defensa contra el pase generalmente) de los Bucs, lanzando para 462 yardas con 3 pases de anotación y comenzando ganando el partido 17–0.</p><p>Sin embargo, esto no quiere decir que el partido está definido. Si pensamos en la historia de estos jugadores, en el 2007 Brady también se reencontró con un equipo que venció (38–35) en la temporada regular (Los NY Giants) y llegó al Super Bowl con un record de 16 ganado y 0 perdidos y la mejor ofensiva de la temporada. Sin embargo, en la final los New England Patriots fueron derrotados por Eli Manning y los NY Giants 17–14, que sería el primer partido entre estos equipos en la era Brady en el Super Bowl. En la revancha, Brady tampoco lograría vencer a Manning estando 0–2 en el Super Bowl contra su equipo. Pero si hablamos de revanchas, el que también buscará vencer a Brady en el Super Bowl es Andy Reid el entrenador de los Kansas City Chiefs quién buscará este 2020 conseguir lo que no pude en el 2005 como entrenador de los Philadelphia Eagles, vencer a Brady.</p><p>Lo que es seguró es que será un partido emocionante y habla de la grandeza de ambos equipos y sus mariscales de campo. Por un lado, Brady participará en su décimo Super Bowl siendo el jugador con más apariciones en la contienda. Por otro lado, Mahomes intentará coronarse como campeón en años consecutivos proeza que no se ha logrado, desde que lo hicieron los Patriots el 2003 y 2004. Además, Mahomes ya tiene 2 apariciones en la final en sus 3 temporadas como titular. Si aparece nuevamente en el Super Bowl en el futuro cercano sería 1 de 3 mariscales de campo (conjunto con Tom Brady y Ben Roethlisberger de Pitsburgh) que han llegado al Super Bowl 3 veces en sus primeras 7 temporadas como profesional.</p><p>No sabemos quién ganará este partido, lo que si tenemos seguridad es que será un partido emocionante, lo más seguro que reñido y que muestra por un lado el pasado y presente en Brady y el presente y futuro en Mahomes de la NFL. ¡Así que abrochémoos los cinturones y a disfrutar del Super Bowl del 2021!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=278ee155621a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[¿Qué nos enseño Gareca y la selección peruana en su camino hacia el mundial?]]></title>
            <link>https://jsaraviadrago.medium.com/qu%C3%A9-nos-ense%C3%B1o-gareca-y-la-selecci%C3%B3n-peruana-en-su-camino-hacia-el-mundial-c849dc50fe1?source=rss-468fc5c36a75------2</link>
            <guid isPermaLink="false">https://medium.com/p/c849dc50fe1</guid>
            <category><![CDATA[peru]]></category>
            <category><![CDATA[soccer]]></category>
            <category><![CDATA[conmebol]]></category>
            <dc:creator><![CDATA[Juan Carlos Saravia]]></dc:creator>
            <pubDate>Wed, 30 Sep 2020 01:01:26 GMT</pubDate>
            <atom:updated>2020-09-30T02:33:55.571Z</atom:updated>
            <content:encoded><![CDATA[<p>Como ya la mayoría de peruanos sabe, de la mano de su entrendor Ricardo Gareca Perú finalmente logró participar en un mundial (Rusia 2018) después de más de 30 años de no poder hacerlo. Esta fue una proeza por parte del equipo, no solo porque rompió una racha de varios años, sino también porque lo logró en el grupo de Clasificatorias Latinoamericanas (CONMEBOL) que es considerado como uno de los más dificiles para entrar al mundial debido al gran nivel que presentan todos los equipos. Además, este torneo es de largo aliento son casi 2 años de competencia y muchísimas cosas pueden ocurrir.</p><p>¿Qué ocurrió con Perú durante este torneo que le permitió ir al Mundial? ¿Qué nos enseño Gareca y el equipo? Hubieron varios cambios pero enfoquémonos en uno. Perú cambió la forma de jugar, re ajustó la cantidad de pases que daba con el pasar del torneo y eso lo ayudó bastante para poder meter más goles. Gareca nos enseñó que en un torneo tan complicado si no cuestionas tus ideas, te adaptas y piensas fuera de la caja, romper las rachas largas y duras es sumamente complicado.</p><p>Comencemos por un poco de contexto. Un dato interesante es que en promedio a lo largo de todos los partidos se encontró una relación inversa entre la cantidad de goles y pases dados por Perú. En otras palabras, en promedio mientras más pases daba menos goles metía (<em>r</em> = -.22, para los interesados en el coeficiente numérico de la relación).</p><p>Pero esa es una parte general de la historia si hacemos zoom al torneo podemos encontrar hallazgos mucho más interesantes. Perú cuestionó su esquema y cambió a lo largo de los 18 partidos. En los primeros 6 partidos donde sumó 4 de 18 puntos posibles, mientras más pases daba menos goles metía (<em>r</em> = -.23). En otras palabras, era un equipo muy ineficiente donde se compartía mucho la pelota entre los jugadores pero no lograba concretar goles. Perú en promedio dio más de 350 pases y logró solo 6 goles y en 3 partidos no consiguió meter ni uno, parecía que el toque peruano no traía beneficios.</p><p>Al obtener resultados así, Gareca decidió no solo replantear su equipo y su grupo de jugadores, sino también ensayar nuevas opciones tácticas lo cual le trajo ciertos resultados. Del partido 7 hasta el 12, Perú sumó 10 de 18 puntos posibles. Mientrás más pases daba Perú más goles metía (<em>r</em> = .18), la parte interesante es que también Perú se fue al otro extremo. En comparación a los primeros 6 partidos, bajó en promedio casi 100 pases y en promedio en este tramo daba un poco menos de 250 pases por partido y estos se traducían en 12 goles, 9 anotados en partidos y 3 adicionales porque Bolivia rompió una regla. En otras palabras, se convirtió en un equipo más eficiente con la pelota donde dar pases era una forma de atacar en lugar de solo tener la posesión de la pelota. El cambio fue drástico, pero Perú aún estaba a 4 posiciones de la clasificación directa entonces aún había mucho camino que remontar. En los últimos 6 partidos, que fue el tramo donde a Perú le fue mejor (sumó 12 de 18 puntos) la relación lineal entre pases y goles desapareció (r = -0.07) pero la cantidad de pases en promedio se incrementó a casi 300 pases por partido. En un esfuerzo por optimizar el juego, Perú volvió a ajustar la cantidad de pases y esto le trajo aún mejores resultados que le permitió conseguir el quinto puesto y la opción de jugar 2 partidos adicionales contra Nueva Zelanda por el boleto al mundial.</p><p>¿Pero qué significa esto de dar pases, cuántos son pocos o muchos? Para poner una referencia, la figura 1 nos muestra el promedio de pases por equipo durante las eliminatorias en estos tramos del torneo.</p><p><strong>Figura 1: </strong>pases y goles por equipo a lo largo del torneo</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cEfxteIzahlJdrYZxcXiDg.png" /></figure><p>Después de revisar la figura 1 podemos ver cómo Perú fue aprendiendo y adaptándose a lo que la situación ameritaba. En los primeros 6 partidos era el equipo que más pases daba después de Brasil y muy parecido a Argentina. La diferencia era que no tenía la contundencia de Brasil o Chile (Perú en promedio metía casi 1 gol por partido y Brasil y Chile alrededor de 2). Además, dar muchos pases tampoco le traían los resultados de Argentina que daba una similar cantidad de pases con un promedio de goles parecido (Perú estaba 8 de 10 (22% de los puntos posible) mientras que Argentina 3 de 10). La idea del susodicho toque peruano y que somos un ¨pequeño¨ Brasil nos traían resultados fatales.</p><p>Luego, entre el partido 7 y el 12, como ya mencionamos Perú cambió drásticamente su forma de jugar, la cantidad de pases fue más similar a Bolivia y Paraguay pero con la efectividad de Uruguay en la cantidad de goles (promedió alrededor de 2 goles por partido). Si bien esto le trajo buenos resultados (55.6% de los puntos obtenidos), aún esto no se vio reflejado en los puestos ya que se mantuvo en el puesto 8 en la fecha 12, justo detrás de Paraguay y adelante de Bolivia. Perú se convirtió en un equipo más eficiente, donde con menos pases podía meter más goles</p><p>Entre los partidos 13 al 18 Perú bajó el promedio de goles por partido (un poco menos de 2 goles por partido), aumentó la cantidad de pases a un promedio muy parecido a Uruguay, Colombia y Ecuador pero con ligeramente más contundencia ofensiva que estos equipos. Además, Perú consiguió el 66% de los puntos disputados a pesar de en promedio meter menos goles. Si analizamos, quizás podemos ver que Perú encontró un mejor balance en el juego, si bien aumentó la cantidad de pases, la cantidad de goles convertidos se redujo y también la cantidad de recuperaciones. Perú tuvo 195 recuperaciones en los primeros 6 partidos y 199 en los partidos 6 siguientes (del 7 al 12 partido) y en los últimos partidos las recuperaciones bajaron a 119 en total. Al comienzo de las eliminatorias no había relación entre la cantidad de pases y recuperaciones (<em>r</em> = 0.01). Sin embargo, en los siguientes 12 partidos tener la posesión de la pelota mediante pases sí se manifestaba como una forma de defender. Del partido 7 al 12, mientrás más pasaba el balón Perú menos recuperaciones necesitaba hacer (<em>r</em> = -0.61), está relación se mantuvo (<em>r</em> = -0.31) en los últimos 6 partidos. Sin embargo, dado que Perú aumentó la cantidad de pases no tenía tanta necesidad de recuperar tanto la pelota. En otras palabras, al final en los últimos partidos Gareca logró encontrar un mejor balance en la cantidad de pases para defenderse mejor sin desgastar tanto el equipo y ser lo suficientemente eficiente para meter los goles necesarios para conseguir los resultados esperados. Como vemos, a lo largo de los partidos Gareca replanteó la táctica de varias maneras, si bien los resultados toman tiempo eventualmente retar las ideas pasadas pueden traer grandes beneficios.</p><p>La figura 1 no solo habla mucho de la cantidad de pases de Perú y la relación con los goles sino también de la capacidad de replantear estrategias por parte de Gareca. Ninguno de los otros 9 equipos tuvo un replanteamiento táctico en cuanto pases tan grande como Perú. En general, en 18 partidos, Perú pasó de ser uno de los equipos que más pases daba, a uno de los que que daba menos pases hasta que se situó en un grupo del medio con un aproximado de 300 pases, lugar donde mejores resultados pudo obtener. Uruguay pasó ser de los equipos que menos pasaban la pelota en los primeros 12 partidos a dar mayor cantidad de pases pero sin llegar a dar tantos pases como los equipos que más lo hicieron.</p><p>Brasil, Argentina y Chile fueron los equipos que más pases dieron y esta tendencia se mantuvo a lo largo de los 18 partidos. Colombia y Ecuador se mantuvieron en un grupo intermedio en todas las clasificatorias. Venezuela, Bolivia y Paraguay fueron los equipos que menos pases dieron en promedio en todo el torneo. Perú logró una gran hazaña al romper una racha de más de 30 años para poder llegar al mundial. Gareca y el equipo nos enseñaron varias cosas y sobretodo que para romper rachas y obtener los resultados hay que retar las ideas previas, tener capacidad de adaptación y buscar el camino óptimo para cumplir las metas. Si seguimos haciendo lo mismo, hay una alta probabilidad que los resultados no varíen mucho, pero si somos creativos incluso haciendo algo totalmente diferente a todos los demás, quizás podamos fallar pero también podamos lograr resultados increíbles como romper una racha de más de 30 años y llegar al mundial.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c849dc50fe1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[La edad y el COVID una reflexión del caso peruano]]></title>
            <link>https://jsaraviadrago.medium.com/la-edad-y-el-covid-una-reflexi%C3%B3n-del-caso-peruano-bf0e71944a7f?source=rss-468fc5c36a75------2</link>
            <guid isPermaLink="false">https://medium.com/p/bf0e71944a7f</guid>
            <category><![CDATA[peru]]></category>
            <category><![CDATA[covid19]]></category>
            <dc:creator><![CDATA[Juan Carlos Saravia]]></dc:creator>
            <pubDate>Thu, 20 Aug 2020 14:44:04 GMT</pubDate>
            <atom:updated>2020-08-20T14:44:04.765Z</atom:updated>
            <content:encoded><![CDATA[<p>El COVID-19 ha tenido un fuerte impacto a nivel mundial y en el caso de Perú no es la excepción. A la fecha (27 de Julio) el Perú es uno de los 10 países con más confirmados por COVID-19 a nivel mundial. Según el Ministerio de Salud (MINSA) más de 375 mil peruanos han sido contagiados por este virus con una edad promedio de 43 años.</p><p>Para esta pandemia, el gobierno ha decidido tomar algunas medidas para poder enfrentar esta pandemia, algunas relacionadas a la edad. Por ejemplo, las personas de 65 años en adelante no deberían en la medida de lo posible salir de su casa. Si bien este artículo no tiene como objetivo criticar las reglas que ha propuesto el estado, sí busca analizar un poco más estos cortes de edad que fueron impuestos. Utilizando la data de fallecidos por COVID del MINSA, se harán algunas reflexiones sobre estos cortes. Tienen sentido los cortes impuesto por el estado? Se deberían cambiar? Por qué el estado pone la restricciones con adolescentes y niños? La respuesta ha estas preguntas se harán sobre la perspectiva del análisis de fallecidos, seguramente hay mucho más miradas y opiniones. El objetivo específico es dar un argumento objetivo con datos y basado en evidencia.</p><h3>¿Por qué el gobierno pone el límite de 65 años?</h3><p>Según la data de contagios, 12% de las personas que contrajeron COVID tienen 65 años a más. Sin embargo, de esa cantidad de personas más de la mitad (56%) han fallecido por COVID. Si bien el porcentaje de fallecidos es relativamente alto y responde un poco a los factores de riesgo que un adulto mayor puede tener, es bastante curioso el punto de corte de 65 años. De dónde sale el número mágico? Es posible que este criterio salga de las AFPs, donde una persona puede retirar su fondo de inversión cuando llegue a esa edad o más. Si bien esto tiene sentido y responde a la edad de adultos mayores, en el Perú y según la Organización Mundial de la salud, legalmente un adulto mayor es una persona que tiene 60 años a más.</p><p>Si volvemos a hacer el análisis descriptivo cambiando el corte a 60 años podemos ver que 18% de personas de 60 años a más han contraído COVID. Si bien la diferencia en puntos porcentuales no parece ser muy importante, cuando nos enfocamos en la cantidad de fallecidos es interesante ver que 69% de las personas fallecidas por COVID tenían 60 años a más. En este caso, sí podemos ver que la diferencia de 5 años es importante ya que la diferencia es de 13 puntos porcentuales entre 65 años a más y 60 años a más. Esto quiere decir que él estado debería cambiar este primer punto de corte? Quizás! De todas formas, sería importante poner también ha discusión si es que los adultos mayores deberían estar en cuarentena absoluta pero ese es tema para debatir en otro momento. El punto importante acá es que sería interesante poner los puntos de cortes o cambiarlos sobre la base de la evidencia. Como se ve, claramente esos 5 años de diferencia muestran una diferencia importante con respecto a las personas que podrían estar en riesgo a causa del COVID. En conclusión, sería interesante que los tomadores de decisiones reflexionen sobre el punto de corte basándose en evidencia.</p><p>Estos datos no quieren decir que el único factor de riesgo para fallecer de COVID es la edad, situaciones como la exposición a la enfermedad, el lugar donde vive o por donde transita la persona o si tiene una enfermedad pre existente son factores que incluso pueden pesar más que la edad en sí. Por otro lado, en algunas casos hay factores que vienen con la edad (enfermedades pre existentes) y en otros depende de la rutina de cada persona (el nivel de exposición a la enfermedad).</p><p>Es por ello que la toma de decisiones de restricciones de edad por ejemplo, debería estar basado en una multiplicidad de factores recogidos de diferentes formas. La buena noticia es que existen cada vez más formas de recoger información, por ejemplo es posible obtener información de la web en google maps y otros de la locación de las personas. Asimismo, también comienzan a salir apps y páginas web como Hospitapp que te permiten registrar síntomas, locación o enfermedades de pacientes. Incluso, el app de las farmacias nos permiten conocer los medicamentos que las personas toman y conocer más sobre sus conductas diarias.</p><p>Si bien los apps de los celulares aún no tienen la capacidad de abarcar a toda la población peruana y la conectividad es un factor que aún no logra abarcar todos las zonas del país está en constantemente mejora, además es necesario empezar por algún lado. Quizás una forma adecuada sería incluir diversas fuentes de información, apps, registros en centros y postas médicas y comenzar a recoger información sobre registro de rutas de transporte público entre otros. Si bien esto no busca interferir en el quehacer diario de la persona ni irrumpir en su privacidad, mientras más información se pueda recoger y de mayor cantidad de personas en el Perú se podrán tomar acciones más informadas para poder sobrepasar no solo el COVID sino cualquier obstáculo. Las decisiones con data nos dan mucho poder y nos permiten tomar decisiones más informadas y que sean más exactas a lo que la realidad nos demuestra.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bf0e71944a7f" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>