A fill is not a win: inside Polexbot’s new edge-capped orders
The bot used to give an order about eight seconds to fill. Now it asks a better question first: does the live price still leave enough edge to make the trade worth taking? If yes, it can take the ask. If no, it waits at a disciplined price — or walks away.
The short version
Polexbot now separates finding a trade from deciding how much to pay for it. Every initial entry gets a maximum acceptable price before the order is signed. The engine then reads Polymarket’s live order book and chooses one of two routes:
- » The best ask is inside the cap. The bot submits a marketable limit order at that ask, aiming for an immediate fill without offering to pay anything higher.
- » The ask is outside the cap. The bot does not chase. It rests at the signal price, or lower if the cap requires it, for up to 30 seconds.
- » The order fills or expires. Full and partial fills are tracked in the background. Any unfilled remainder is cancelled; a missed entry is not booked as a losing trade.
There is nothing new to configure. Your strategy, sizing and entry settings stay where they are.
Eight seconds was not a strategy. It was a coin flip.
Polexbot evaluates 5-minute markets using the midpoint between the best bid and best ask. That is useful for estimating the market’s center, but a buy submitted at the midpoint is not necessarily executable. If the midpoint is 41¢ and the best ask is 43¢, a 41¢ buy simply joins the queue.
The old engine waited roughly eight seconds and cancelled. The dashboard showed no position, users saw a bot that was running but apparently doing nothing, and a valid signal vanished before the order had a realistic chance to trade.
The obvious alternative — always pay the ask — would have created a different problem. A few extra cents on every entry can quietly erase the edge the strategy found. More fills would look busier, but busy is not the same as better.
A ceiling built from the trade’s edge
On AI Strat, the ceiling comes from the same calibrated model that approved the signal. In plain language:
maximum entry price = estimated win rate − required safety edge
Your configured entry ceiling still applies; the engine uses whichever limit is stricter. On the other strategies, where there is no calibrated win-rate estimate, the cap is the stricter of your entry ceiling and the signal price plus a small 2¢ slippage allowance.
- » Limit protection either way. Even the “take the ask” route is a limit order at the observed ask, not an open invitation to pay any price.
- » Tick-aware pricing. Passive prices are rounded down to the exchange’s valid price increment, so rounding can never push an order above the cap.
- » No blocking. While a passive order waits, the engine keeps scanning and managing every user rather than sleeping beside one order.
The 52% setup, the 41¢ midpoint and the 55¢ ask
Suppose AI Strat estimates a 52% chance of winning and requires a 3-point safety edge. The most it can pay while preserving that edge is 49¢. The signal midpoint is 41¢.
- » If the live ask is 42¢, the bot can take it. The remaining estimated edge is still 10 points.
- » If the live ask is 48¢, it can still take it, although the trade is less attractive than it was at 41¢.
- » If the live ask is 55¢, the estimated edge would be negative. The bot refuses to chase and instead rests at 41¢ for up to 30 seconds.
If five shares fill before the rest expires, Polexbot opens a position for those five shares at their actual cost and cancels the remainder. If nothing fills, it records a missed entry and moves on. No phantom position, no imaginary loss and no order left behind to fill after the signal is stale.
The goal is not more trades. It is better trades.
Winning at the order book starts before a position ever resolves. Two bots can see the same signal and get very different long-term results simply because one repeatedly pays more to enter.
There is also a catch to passive orders: they can fill when the market moves against you and miss when it moves in your favor. Waiting longer does not magically remove that risk. That is why the new path can take liquidity when the numbers still work, but will not sacrifice the model’s safety edge merely to make the trade counter rise.
No execution system can guarantee a profit. The model can be wrong, markets can gap, and a well-priced position can still lose. This update is about removing a preventable disadvantage: paying so much for a fill that the original idea no longer makes sense.
No new setting, no change to your strategy
You do not need to edit anything in the UI. AI Strat still uses its model and balance-aware sizing; Constellation and the directional strategies keep their existing signal rules. The update applies to initial entries; DCA behavior is unchanged.
The dashboard will now tell you when an order is resting and show its price cap. If it expires, you will see a missed-entry note rather than an alarming failed trade. When it fills, the position is booked from the filled share count and actual order cost, including partial fills.
Fill when the edge survives. Wait when it might return. Pass when it is gone.
That is the entire idea. The engine is no longer forced to choose between cancelling almost immediately and overpaying just to get in. It has a price ceiling, a live view of the ask, a short window to work the order, and the discipline to leave a bad entry alone.
Polexbot is an independent tool with no affiliation to Polymarket. Prediction markets carry real risk and past performance does not guarantee future results. Nothing here is financial advice.
Already active for new initial entries. Nothing to turn on.