5 Most Profitable Trading Bots on Polymarket in August

I analyzed more than 1,000,000 executions from five of the most profitable trading bots on Polymarket in August using Claude to understand exactly how they build positions in Up/Down markets and what drives their consistent capital growth.

At first glance, their trading looks similar. But once you reconstruct their inventory, it becomes clear that similar activity is being generated by very different systems.

Some build nearly symmetrical positions through many small executions. Some spread capital across multiple assets and timeframes. Others use the opposite side mainly as a dynamic hedge around a core directional position.

But the most interesting part is that these bots are not simply trading direction. They are trading the microstructure of Up and Down markets based on entry price, the size of each addition, the cost of accumulated inventory, the balance between both sides, order book liquidity, and the timing of the next execution.

Below is a breakdown of each of the five systems.

1. Dynamic Two-Sided Inventory Rotation

Account: 0xce25e214d5cfe4f459cf67f08df581885aae7fdc

0xcE25E214… builds its position as a constantly changing inventory of Up and Down. The bot trades Bitcoin, Ethereum, Solana, and XRP across 5-minute and 15-minute markets. In 93% of cases, it accumulates both sides.

But this system cannot simply be described as standard Up + Down arbitrage. Across all markets where the bot bought both sides, the median combined average price of Up and Down is around $0.975. At the same time, it falls below $1 in only about 57% of those markets.

The bot clearly does not require every completed pair to immediately provide an edge. It allows inventory to temporarily become more expensive if the position structure and market movement provide an opportunity to restructure it later.

A typical cycle looks roughly like this:

• When Up becomes undervalued relative to its model, the bot starts increasing Up
• If the underlying asset price changes, it starts buying Down
• Part of the two positions becomes an arbitrage pair
• Contracts left beyond the paired amount form a directional bias

The weak point of this system appears during a strong move without a retracement.

If the bot has already accumulated one side but the market no longer provides an attractive opportunity to add the opposite side, the existing inventory can remain too expensive or too directional.

2. Fixed-Lot Directional Rotation

Account: 0xAAAAA

It trades 5-minute Bitcoin markets using a system built around base trading blocks of 60 and 120 contracts. Within each five-minute interval, the bot can switch between Up and Down several times, adding new 120-contract blocks as the market changes.

For example, the position can develop like this:

• 60 Down for the initial exposure
• Then 120 Up
• Another 120 Up after the move continues
• On a reversal, 120 Down
• After another move, one more block on the side that becomes stronger again

At first glance, this behavior can easily look like temporal arbitrage. But the numbers suggest otherwise. The bot bought both sides in 96.3% of markets, while the median sum of its average Up and Down prices is around $1.06. Only about 30% of two-sided markets had this value below $1. This means that building a cheap symmetrical pair is not the core objective of the algorithm.

In the median market, the smaller side of the position is only about 64% of the larger side. So even after all the switching, a substantial directional skew remains. Its system looks more like serial redistribution of directional inventory using fixed-size lots.

The opposite side is used for more than hedging. It allows the algorithm to change the structure of the position without constantly closing the inventory it already holds.

A 120-contract block is very convenient for automation, but the market does not always move in discrete steps of equal strength. If the signal changes too late, the next standard block can either create unnecessary exposure or make the hedge too expensive.

3. Balanced Micro-Fill Accumulation

Account: x-MoneyForWhiskas

x-MoneyForWhiskas is also fully focused on Bitcoin 5m, but its execution logic is completely different. While 0xAAAAA operates with large standardized blocks, Whiskas splits its position much more aggressively. The most common size is 50 contracts, but these blocks are surrounded by a large number of smaller fills.

In 98.7% of cases, it bought both sides. The smaller side is about 84% of the larger side at the median, while the average combined cost of Up + Down is around $0.99. The structure resembles temporal relative-value trading.

The bot actively uses price changes within each five-minute interval:

• buys one side;
• waits for relative value to change;
• starts accumulating the opposite side;
• returns to the first side again;
• gradually brings most of the inventory toward similar contract amounts.

Within a single market, it switches between Up and Down around 10 times at the median. This is no longer a single directional idea with a hedge layered on top. It is continuous interaction with two separate instruments. Out of these five bots, Whiskas is the closest to pure balanced temporal accumulation.

But this logic also depends on the price path.

If the first side is fully executed while the opposite side never reaches the required price, the system is left with an incomplete structure and has to either maintain directional exposure or worsen the average cost of the position.

4. Multi-Timeframe Regime Switching

Account: bosona

It trades 5m, 15m, 1h, 4h, and daily Up/Down markets across Bitcoin, Ethereum, Solana, Dogecoin, BNB, and XRP. Its behavior changes noticeably depending on the timeframe. Both sides are present in only 43% of markets. But those markets account for around 79% of total trading volume.

This means the two-sided structure is primarily used when the bot deploys more significant capital. This is especially clear on 5m. In two-sided five-minute markets, the median combined average price of Up + Down is around $0.80, and in almost 74% of those markets it is below $1.

Get Retro Valix’s stories in your inbox

Join Medium for free to get updates from this writer.

Remember me for faster sign in

On longer timeframes, the result changes. On 15m, pair cost is already around $1, while on 1h the median value is even slightly above it. So the same account is effectively using different position management regimes.

On shorter intervals:

• it can quickly take one side;
• capture a significant change in relative price;
• add the opposite side;
• turn part of the directional inventory into a cheap two-sided structure.

On longer horizons, the bot more often keeps a one-sided position and does not necessarily try to build a pair. Its main risk is not a specific entry, but choosing the wrong regime.

Logic that works extremely well on five-minute BTC does not necessarily fit hourly ETH or four-hour BNB. The algorithm has to simultaneously choose direction, sizing, hedge level, and the appropriate inventory management structure for each type of market.

5. Directional Core With a Hedge Overlay

Account: mo-money

Mo-money visually resembles bosona: it also trades BTC, ETH, SOL, DOGE, BNB, and XRP across multiple timeframes. But its position structure is even more heavily tilted toward directional trading. Both sides appear in around 41% of markets. At the same time, just like with bosona, those markets account for almost 80% of total volume.

So hedging is regularly used in larger positions, but the bot clearly does not try to make every market neutral. This is especially visible on 5m, where most of the capital is concentrated. In a two-sided five-minute position, the smaller side is only around 58% of the larger side at the median.

For example:

• 250 Up
• 145 Down

145 contracts on each side can be treated as the paired part of the inventory. But the remaining 105 Up contracts still create significant directional exposure. This structure describes mo-money well. First, the core directional leg is formed. Then, if the opposite side becomes cheap enough, the bot starts adding it.

With a favorable price path, this simultaneously:

• reduces net exposure;
• improves the average-price structure;
• creates a paired portion of the inventory;
• leaves additional size on the side of the primary signal.

On 5m, this structure is often quite efficient: the median combined average price of both sides is around $0.81, and in roughly 71% of two-sided five-minute markets it is below $1. But the bot does not turn the entire inventory into a pair. The directional remainder remains an important part of its edge.

The main vulnerability of this model is incorrect hedge sizing. An opposite position that is too small provides very little protection to the core inventory during a sharp reversal. But if it is added too late and at too high a price, the bot reduces directional risk at the cost of worsening the overall average price.

What These Five Trading Bots Have in Common

Despite using different algorithms, several very distinct patterns repeatedly appear in their trading:

1. Up and Down Are Used as Two Parts of the Same Inventory

For these systems, the opposite side can perform several functions at once:

> reduce net exposure
> improve the overall cost basis
> form the paired portion of the inventory
> create a new directional leg after the market changes
> allow the position to be restructured without constantly exiting the previous one

2. They Trade the Price Path, Not a Single Snapshot

The price of Up at the beginning of a five-minute interval and the price of Up 30 seconds before the end represent two completely different trading situations. These systems use the movements that happen in between.

If Up starts at 70¢ and later falls to 35¢ while Down moves in the opposite direction, the algorithm gets several different points where it can restructure its inventory. That is why the sequence of executions matters more than any single entry.

3. Sizing Is Part of the Algorithm Itself

Position size does not appear random here. Each bot has its own execution pattern.

0xAAAAA is built almost entirely around 120-contract blocks.

x-MoneyForWhiskas more often uses 50-contract blocks with fractional fills around them.

Bosona and mo-money use a much wider range, from micro-fills to 250-contract blocks.

0xcE25E214… sits somewhere between these models and dynamically adjusts size based on its current inventory.

So sizing is not only used for risk control. It determines how quickly the algorithm can change the balance between Up and Down after the next market move.

Key Differences

Although all five bots actively manage both sides of the market, their approaches to trading are different.

> x-MoneyForWhiskas is the most balanced. Its inventory is usually close to symmetrical, while the combined cost of Up + Down stays around $1. This is the cleanest example of relative-value trading between the two sides.

> 0xcE25E214 uses a similar idea, but much more flexibly. It trades multiple assets, allows a larger directional remainder, and does not require a cheap pair in every market.

> 0xAAAAA looks two-sided, but mathematically it is much more directional. Its defining feature is fixed 120-contract lots and the constant redistribution of those blocks between Up and Down.

> Bosona changes its trading behavior depending on the timeframe. On 5m, it actively uses cheap two-sided structures, while on longer intervals it is much more likely to maintain one-sided exposure.

> Mo-money maintains a stronger directional core than the other two multi-asset bots. The hedge is not the final objective, but an additional layer around the core position.

Conclusion

The August sample reveals something interesting. The five most profitable bots do not share one directional indicator, the same sizing model, or a single entry formula. Even buying both sides represents a different strategy for each one.

For some, it is a way to bring the combined cost closer to $1. For others, it is a way to reduce exposure. And for others, it is a way to change the direction of the position without closing the existing inventory. That is why analyzing these systems only through individual BUY records is practically useless.

You need to reconstruct the full position and look at:

> how many contracts have accumulated on each side
> the average price at which they were accumulated
> when the switches occur
> how much of the position can be matched into pairs
> how much directional remainder is left
> how sizing changes after each market move
> how many separate fills actually belong to a single limit order

The main edge of these trading bots does not come from a single entry point. It comes from a sequence of decisions that continuously reshapes the position as the market moves. And it is this ability to manage inventory within a short price interval that separates a complete automated trading system from a simple directional algorithm.

Thanks for reading! If you liked this article, follow me on medium:

https://medium.com/@RetroValix