<?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 Oleksandr Koval on Medium]]></title>
        <description><![CDATA[Stories by Oleksandr Koval on Medium]]></description>
        <link>https://medium.com/@Oleksandr1?source=rss-843a0a58a7c7------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*WjNURwyaaDuz54T2S3MY4Q.png</url>
            <title>Stories by Oleksandr Koval on Medium</title>
            <link>https://medium.com/@Oleksandr1?source=rss-843a0a58a7c7------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 10:55:00 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@Oleksandr1/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[How I was doing 270k requests to Steam and made money from this]]></title>
            <link>https://medium.com/@Oleksandr1/how-i-was-doing-270k-requests-to-steam-and-made-money-from-this-fe329b8bf393?source=rss-843a0a58a7c7------2</link>
            <guid isPermaLink="false">https://medium.com/p/fe329b8bf393</guid>
            <category><![CDATA[go]]></category>
            <category><![CDATA[steam]]></category>
            <category><![CDATA[counter-strike]]></category>
            <category><![CDATA[automatic-trading]]></category>
            <category><![CDATA[valve]]></category>
            <dc:creator><![CDATA[Oleksandr Koval]]></dc:creator>
            <pubDate>Fri, 15 Mar 2024 11:30:00 GMT</pubDate>
            <atom:updated>2024-03-15T12:01:35.418Z</atom:updated>
            <content:encoded><![CDATA[<p>Greetings! I am Oleksandr. In this article, I am excited to share with you the story of how my friend and I developed a highly efficient trading bot for the Steam market using Go. I will also discuss how we overcame various challenges such as the Steam request rate limiting. Additionally, I will delve into our project’s architecture that can handle 270k requests each day, and provide insights into our project’s profit, statistics, and lifecycle from the start to the present day.</p><h3>Disclaimer</h3><p>Please keep in mind that the actions outlined in the article are for storytelling purposes only and should not be copied. If you do repeat these actions, there is a risk of breaking the Steam Subscriber Agreement, terms of service, or other legal agreements with Steam. This could result in negative consequences, such as a trade ban on your account!!!</p><h3>The Backstory</h3><p>For a while, I had been encouraging my friend to create software that would automate item trading on the Steam marketplace. I found the idea intriguing, as I enjoy gaming and have been fascinated by my friend’s stories of his trades, profits, and all the intricacies of the CSGO items market. On March 25, 2023, we began planning and developing the software.</p><h3>What our bot is doing?</h3><p>Our bot has been designed to search for specific items on the Steam community market and compare prices based on a formula. The logic behind the bot was based on friend’s market strategy, and my role was to implement it while also considering the architecture, libraries, and hardware used in the final stage of the project. If the bot flow and formula meet the necessary conditions, it provides a link to the item on the community marketplace. Initially, we planned to send the link to our private telegram group for manual purchase, but the final version of the bot can automatically purchase the item, with the telegram notification serving only as a history for debugging or curiosity.</p><h3>Development</h3><p>The following libraries were utilized in the project:</p><ul><li>Log library for writing logs in a file with a customized Logger</li><li>net/http and io/ioutil for handling requests and responses</li><li>encoding/json for reading and writing JSON from files and responses</li><li><a href="https://github.com/spf13/viper">Viper </a>for configuring settings in a file</li><li><a href="https://github.com/go-telegram-bot-api/telegram-bot-api">TelegramBotAPI </a>for interacting with Telegram bots, writing messages, and manipulating bots in real-time with Telegram commands</li><li><a href="http://github.com/chromedp/chromedp">chromedp </a>for automating Chrome through dev tools</li></ul><p>I was considering using the colly library to parse Steam pages, but upon investigation, I discovered that Steam has its own endpoints that populate its pages with the data we need. For instance, when you search for “AK-47” in the CSGO category on the community marketplace and tick the “Include descriptions in search” box, Steam makes a request at <a href="https://steamcommunity.com/market/search?q=AK-47&amp;descriptions=1&amp;category_730_ItemSet%5B%5D=any&amp;category_730_ProPlayer%5B%5D=any&amp;category_730_StickerCapsule%5B%5D=any&amp;category_730_TournamentTeam%5B%5D=any&amp;category_730_Weapon%5B%5D=any&amp;appid=730">https://steamcommunity.com/market/search?</a><a href="https://steamcommunity.com/market/search/render/?q=AK-47&amp;descriptions=1&amp;category_730_ItemSet%5B%5D=any&amp;category_730_ProPlayer%5B%5D=any&amp;category_730_StickerCapsule%5B%5D=any&amp;category_730_TournamentTeam%5B%5D=any&amp;category_730_Weapon%5B%5D=any&amp;appid=730">query</a><a href="https://steamcommunity.com/market/search?q=AK-47&amp;descriptions=1&amp;category_730_ItemSet%5B%5D=any&amp;category_730_ProPlayer%5B%5D=any&amp;category_730_StickerCapsule%5B%5D=any&amp;category_730_TournamentTeam%5B%5D=any&amp;category_730_Weapon%5B%5D=any&amp;appid=730">=AK-47&amp;descriptions=1&amp;category_730_ItemSet%5B%5D=any&amp;category_730_ProPlayer%5B%5D=any&amp;category_730_StickerCapsule%5B%5D=any&amp;category_730_TournamentTeam%5B%5D=any&amp;category_730_Weapon%5B%5D=any&amp;appid=730</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ruycPM_1iHl161Q-H2J45Q.png" /></figure><p>However, the response is in HTML code, which isn’t ideal. By adding “render” to our link after market/search, we can obtain JSON with more data at <a href="https://steamcommunity.com/market/search/render/?q=AK-47&amp;descriptions=1&amp;category_730_ItemSet%5B%5D=any&amp;category_730_ProPlayer%5B%5D=any&amp;category_730_StickerCapsule%5B%5D=any&amp;category_730_TournamentTeam%5B%5D=any&amp;category_730_Weapon%5B%5D=any&amp;appid=730">https://steamcommunity.com/market/search/render/?query=AK-47&amp;descriptions=1&amp;category_730_ItemSet%5B%5D=any&amp;category_730_ProPlayer%5B%5D=any&amp;category_730_StickerCapsule%5B%5D=any&amp;category_730_TournamentTeam%5B%5D=any&amp;category_730_Weapon%5B%5D=any&amp;appid=730</a></p><p>Although the results are still in HTML under the “results_html” key, we can get beautifully structured JSON by adding norender=1 to our request at <a href="https://steamcommunity.com/market/search/render/?query=%22AK-47%22&amp;norender=1&amp;descriptions=1&amp;category_730_ItemSet%5B%5D=any&amp;category_730_ProPlayer%5B%5D=any&amp;category_730_StickerCapsule%5B%5D=any&amp;category_730_TournamentTeam%5B%5D=any&amp;category_730_Weapon%5B%5D=any&amp;appid=730">https://steamcommunity.com/market/search/render/?query=%22AK-47%22&amp;norender=1&amp;descriptions=1&amp;category_730_ItemSet%5B%5D=any&amp;category_730_ProPlayer%5B%5D=any&amp;category_730_StickerCapsule%5B%5D=any&amp;category_730_TournamentTeam%5B%5D=any&amp;category_730_Weapon%5B%5D=any&amp;appid=730</a>, even with additional data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MmDLTRnngmmKxdf2rMc0XA.png" /></figure><p>If you need to convert JSON data into a Go structure declaration, you can use the helpful tool at <a href="https://mholt.github.io/json-to-go/">mholt.github.io/json-to-go</a>. Keep in mind that there are many other parameters for this query, and there are plenty of information and browser extensions for Steam that use these links for data gathering.</p><p>The same approach can be applied to individual items, item graphs, and all data you see on Steam pages. You can find these endpoints through Dev Tools, the Network tab, or by searching on Google or Bing.</p><h3>Rate limits</h3><p>Our next task was to go through the necessary queries, but Steam has a limit on the number of requests we can make. If we exceed this limit, we will receive responses without any data. When you visit a market page in the browser, this is what it looks like:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1016/1*ZrpwUYti4YbMPJd-HLr4Og.jpeg" /></figure><p>We used delays between requests to extend the block’s delay. Initially, we tried using some free proxies, but it didn’t work. I spent countless hours researching how to integrate OpenVPN or something with an API or library on Go, but it seemed too complicated. We didn’t want to risk spending weeks trying to fix the issue and potentially losing profits. Additionally, we noticed that something in our flow would break when attempting to integrate possible solutions. Eventually, we found a fast and reliable solution — a VPN. We downloaded a free (product placement: Proton) VPN with 100+ servers in Europe and used each IP to send requests to Steam for 30–40 minutes. However, since the VPN only had a GUI version, we had to manually switch IPs once we received a notification in Telegram that we got blocked. I discovered CLI VPNs as a way to streamline the process, and after researching, I found one (Product placement: Mullvad) that only costs 5€ per month, offers CLI, can be downloaded on Windows, and allows for installation on up to 5 devices per account. I wrote my own integration for it using CLI commands and an external library for reliability. At the start of the project, we needed 36 IPs to run our bot for 24 hours straight, but after some optimizations, we managed to use an average of 20–25 IPs. More details will be provided in the following sections.</p><h3>Automatic purchase of items</h3><p>It can be time-consuming to sit and wait for a profitable skin to be sent by a bot in a Telegram group. There is a risk of missing out on a good deal if someone else buys it before you. Automating the purchasing process is faster and more efficient, but it requires a lot of work. I used the chromedp library to automate the purchase process since I was already familiar with it. It involved opening Chrome, following a link sent to the Telegram group, and completing all the steps to buy an item. It was a challenging task that took us four days of work, five hours each day. The main function for automatic purchase is about 180 lines long, and we had to account for Steam features and bugs. We used custom JS code and created a flow with nine steps to ensure everything was working correctly. If something unexpected happened, the rest of the steps would be useless. It took us around 20 hours to develop the automation process, but we eventually succeeded.</p><h3>Architecture</h3><p>In order to keep up with the constant activity of the Steam community marketplace and not miss out on profitable items, we needed to increase the frequency of our checks. However, we discovered that go routines won’t help, since making too many requests concurrently from the same IP address would result in a sooner block, around after 5 minutes, requiring us to switch to a different IP through a VPN.</p><p>After researching and considering various options, including virtual networking in Go and traffic routing, we ultimately decided to use hardware-level virtualization instead. We tested different virtual machine options and found that Windows own Hyper-V was a good fit for our needs. We installed a lightweight Windows distro on my second PC, which has an i7–4770k processor and 24GB of RAM, and set up virtual machines with dynamic RAM and 2 virtual CPUs each.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7xhWzNSW6T51hvzKaitdvQ.png" /></figure><p>We also used friend’s old notebook for the task. We imported a hard drive for each virtual machine, which made the setup process easy after configuring one VM and exporting it.</p><p>Each virtual machine typically uses only 0–2% of CPU and 1300MB of RAM while searching for items, with usage increasing when Chrome is used for automatic purchase. We have a total of 9 working bots across multiple VMs and physical devices, with an additional slot in our VPN account. All of this ensures that we can check the Steam marketplace frequently and efficiently without encountering blocks or other issues.</p><h3>A little bit about telegram bot functionality</h3><p>To ensure the efficient functioning of our Telegram bots, we created multiple bots and added them to our groups, as each bot requires a unique Telegram key that cannot be used by multiple instances at the same time. These bots are designed to send found items to a specific Telegram group, with the aim of distinguishing between manually purchased and automatic items for better tracking. Moreover, all bots have a list of 14 commands that enable us to view statistics and make changes in real time without having to restart the bot. This allows for easy management of the bot from anywhere using our phones. Our data shows that each bot makes approximately 30k requests every 24 hours, although this number may vary based on settings, Steam servers status, and other factors. In total, our 9 bots make 270k requests each day.</p><h3>Financial part</h3><p>Please be aware that our finances consist of the total account balances and the sum of all sell orders, minus the 15% Steam fee. Our selling prices may vary depending on various factors, including our profit margin on each item. Some items may sell quickly, while others may take longer or require price adjustments. We have two accounts — one for manual purchases and one for automated purchases. The manual account started with $140 and reached a balance of $420 on May 4th. We transferred $150 to the automated account as we shifted towards mostly automated purchases. This coincided with a CSGO major in Paris, causing a significant drop in the CSGO items market.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/862/1*0mbHrMlf5kN6yc-tpOUaYA.png" /></figure><p>As a result, our balance plus edited sell listings sum to around $230 instead of $270, but we are still profitable overall. The automated account started with $35 and received the $150 transfer, bringing the total to $185. As of writing this article, the automated account has a balance of $680, bringing our total balance to around $910. However, we do have expenses, including monthly fees of around $11 for each account and electricity bills. Additionally, withdrawing Steam balance to real money involves third-party websites with lower prices than the Steam market and fees of 3–7% for selling items. For example, if we buy a “★ Flip Knife | Forest DDPAT (Minimal Wear)” on Steam for around $220, we will sell it for $200 on a third-party website (product placement: like Dmarket), and after a 7% fee, we will receive $186, and minus 2% more if we withdraw to Payoneer and then to a credit card, leaving us with $182 in real money. However, we are optimistic that the summer and the upcoming release of CS 2 will increase the CSGO items market, leading to more volume and better deals for our bot.</p><p>Please note that the following metrics and history only cover the first two months of our journey. The trade history for the manually purchased skins account is included below, but please be aware that the dates are incorrect — all data is from the current year of 2023.</p><p>The amount sold on the first day refers to the funds received in the account through CSGO items as part of the starting budget. The currency for this account is UAH.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*76igjDdQ4Jgg7Ij2eUJ3uw.png" /><figcaption>Please note that the extension is displaying the wrong year due to a bug. All transactions occurred within the 2023–2024 timeframe.</figcaption></figure><p>The overall bought/sold count and trading volume show a negative difference due to friend withdrawing a few cool skins for 1300 UAH to his account and 150$ worth of skins transferred to the second account.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/431/1*wp1MXknAavIhCLkIO-g71w.png" /></figure><p>If we consider the exchange rate as 37 UAH for 1$, the total trading volume is approximately ~1564$.</p><p>Additionally, here is the data for the second account with automated purchasing. Again, please note that the dates are incorrect and all data is from the current year of 2023. The currency for this account is USD.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2l7BtufYRf7s84UmikOINA.png" /><figcaption>Please note that the extension is displaying the wrong year due to a bug. All transactions occurred within the 2023–2024 timeframe.</figcaption></figure><p>The overall bought/sold count and trading volume are shown below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/436/1*4EWiMqwJW2ORZ6LsJyWl9w.png" /></figure><p>It’s worth mentioning that April 25th was the most profitable day, as seen on the graph. We bought a lot of good skins and placed sell listings, resulting in a net profit of ~$59.</p><h3>Fun facts</h3><ul><li>We have one frequent buyer, who already bought 85 items from us. He resells these items with an even bigger margin of 50–100%+.</li><li>The second account with automatic purchase bought 3 items from our 1st manual account due to they met our own conditions in bot logic.</li><li>There were cases when we bought an item from someone, placed a sell listing with our margin, and after a couple of days, the same accounts were buying them back at a higher price.</li><li>We encountered multiple other bot accounts, especially Chinese that sold items on Steam marketplace, and yes we bought some from them since they met our conditions. Also, we encountered many weird bot accounts that were holding unusual sets of items</li><li>Steam and market/items backend are quite interesting and they have their workarounds and secrets but we won’t disclose them due to security and other reasons.</li></ul><h3>Conclusion</h3><p>Our project took about a month to develop and included 1258 lines of code. Despite encountering numerous issues and bugs(not only on our side), we successfully completed it after two months of production. This project has been one of the most enjoyable ones I have ever worked on. Not only did we create a 99% passive source of income, but I also discovered many new things along the way, with friend help and company. This is just the beginning, and we hope to improve it further. Thank you for reading this article, and I’ll update you on our progress soon. Please take a moment to consider supporting Ukraine during the ongoing Russian invasion by donating to <a href="https://u24.gov.ua/">United24</a> or <a href="https://savelife.in.ua/en/donate-en/">SaveLife fund</a>. Your contribution can make a difference in the lives of those affected by the war.</p><h3>March 2024 update</h3><p>Improvments: We’ve done multiple improvments and updates to the system and code including</p><ul><li>VPN CLI integration updates</li><li>Additional validations and comparison in auto buy processes</li><li>Steam search improvments</li></ul><p>Architecture changes: we replaced my friend notebook with PC we’ve built from used parts(4770 + 32Gb Ram). It cost us 180$ and thanks to that we scaled to 15 active bots.</p><p>Updated stat graph for main USD account:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3EH66pVM_CRWcPO7MCPpiA.png" /><figcaption>Spikes on graph represent items that we purchased as part of 400$ investment or skins for withdrawal on 3rd party markets</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/440/1*kus4kmIkCAOcaxVphxNPHg.png" /></figure><p>As you can see we are selling leftovers lately, since we are closing operations. We made this decision based on last CS 2 update and overall Steam back end support didn’t match our expectation and this niche is overflown with other competitors and low volume of valuable items. Our total gross profit is 1320$. More calculation details can be found in recent YT video I’ve made: <a href="https://youtu.be/6rSgwuhIQIY">https://youtu.be/6rSgwuhIQIY</a></p><p>Thank you for your attention, hope this article was interesting and gave you some perspective on Steam, Items Market and other aspects.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fe329b8bf393" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>