Polymarket Fill Simulation Explained

An order is sized against the frame's ladder: first executable level, then deeper levels until the size is consumed, with any remainder priced where a crossing order would meet it. The execute-only-what-existed rule applies at every frame, including one-sided and crossed states.

That rule is why the model is honest on this venue: 16.9% of 5-minute snapshots are one-sided, so a fill model that invents the missing side is a model of a market that never existed.

Against a two-sided book, the fill walks levels and the cost is cumulative-consumption slippage. Against a one-sided book, the fill is the resting side at its price — entry is the bound, and there is no other price to cross to.

For crossed frames, the model prices at the executable side as recorded; the crossed flag is preserved so a simulation exposes, rather than hides, the moments of genuine disarray.

The model requires the full ladder (not top-of-book), the tape for consumption evidence, and the outcome column for settlement. All three are on the archive's 250ms grid — the same structure the replay and backtest recipes use.

The honest calibration is the reconstruction test: does the simulated fills history reproduce what the tape shows the market actually printed? The archive makes that check routine.

Fill simulation prices a strategy, it does not bless it: the honest result is the fill price series plus the states it traded through, and the honest habit is re-running the same fills at the opposite assumption to see how much the conclusion leans on the model.

• Slippage data
• Backtest methodology
• Strategy replay