A Polymarket market-making bot has drawn attention after a transaction-level breakdown showed more than $1.13 million in cumulative profit from roughly $67,000 in deployed capital. During a sampled 100-minute peak window, the address executed 3,379 trades, or about 34 per minute, while operating around the clock.
The address was not trading like a directional speculator. Data in the analysis shows it maintained 100 positions at the same time across BTC, ETH, SOL, and XRP, spanning timeframes from 5 minutes to 4 hours. One detail stood out immediately: 100% of its recorded trades were BUY orders, with no SELLs.
Order-book maintenance, not signal chasing
Trade distribution points to a system focused on constantly refreshing quotes rather than entering only when a market signal appears. ETH 1h was the busiest product with 737 trades in the 100-minute sample, while BTC 15m followed with 595 trades. XRP 5m showed no activity at all, suggesting the bot skipped markets where liquidity conditions did not justify participation.
The timing pattern was equally revealing. In the 1-hour markets, 11.6% of entries happened in the opening 0% to 10% of the candle, while 15.6% came in the closing 90% to 100% window. For 4-hour markets, 36% of trades were concentrated in the first 10% of the candle. That pattern fits a queue-priority strategy: the edge comes from being earlier in the book, not simply from being right on direction.
Why a bot with only BUY orders can still capture spread
Polymarket contracts are built around two tokens, UP (YES) and DOWN (NO). A market maker can buy both sides at the same time. In the example cited in the source material, buying UP at $0.48 and DOWN at $0.48 creates a combined cost of $0.96. At settlement, whichever side wins redeems for $1.00, leaving a theoretical profit of $0.04 per pair.
That is why the on-chain history can show only BUY activity while the strategy still makes money. The source also highlighted three structural advantages: 0% maker fees, maker rebates tied to trading volume, and queue priority from constant high-frequency quoting. The last rebate recorded for this bot was $145.
BTC was the weak spot, and losses could hit fast
The analysis makes clear that market making was not uniformly profitable across all instruments. BTC was the most problematic area. One BTC 1h candle produced a $1,988 loss, with the source attributing the damage to an oversized position of 20,138 shares that left the bot exposed during sharp price swings.
Performance also varied by timeframe. The 4-hour segment posted the best margin and was described as 100% profitable in the sample, yet it represented only 5.7% of total trading frequency. That trade-off matters. Higher-frequency markets bring more volume but also tougher competition, while lower-frequency markets can be cleaner but harder to scale.
The strategy added directional bias on top of market making
The most interesting finding was that the bot did not operate as a purely neutral two-sided market maker. After comparing minute-by-minute order flow against Binance price moves, the analysis concluded that the bot adjusted its UP and DOWN buying ratio using short-term price signals from Binance. The reported validation figures were 74% directional match rate, 77% directional accuracy, 100% allocation toward the more expensive side, 71% against 1h momentum, and 74% in line with 15m momentum.
That changes the picture. The bot was collecting spread, but it was also leaning into what it believed was the more likely winning side. The benefit was a higher potential return. The cost was larger one-sided exposure when that bias was wrong, which helps explain why BTC volatility caused outsized damage.
It recovered from drawdown, then shut down
The source gave one clear example of drawdown risk. On the night of February 21, a large BTC 1h move pushed the bot’s overall PnL from positive territory down to -$960. By the next morning, total PnL had recovered to +$947. The analysis described this as a self-repair feature of market making, where normal spread capture in later candles can absorb earlier losses if the pricing structure remains intact.
As of March 6, the bot had stopped operating. Its final activity was a batch of 130 REDEEM transactions, indicating a broad exit and settlement. Leaderboard data still showed all-time profit above $1.13 million. The source mentioned several possible reasons for the shutdown, including a strategy change, a move to a new address, or changes to Polymarket latency rules, but no definitive explanation was visible from on-chain data alone.