The fill model · Polymarket
Backtest fills that match live
A backtest is only as real as its fills. Replayed against live fills from 5 to 16 October 2026, on matches held out from the fit, ours filled 1.00 times what the venue did on the ladder, 1.02 times on the first order in, and 94.3% of the first order’s contracts came back on the very same orders.
Updated
- The ladder after it
- 1.00×
- The first order in
- 1.02×
- Contracts filled again, order for order
- 94.3%
- Groups within 20%, 5 to 14 October 2026
- 8 of 8
In detail
Every group, not just the total
A total can match while the groups underneath are wrong. Ours holds group by group, by how much size sat ahead of the order and by how long it rested.
0.97× the venue’s fills in total · 8 of 8 groups within 20%
Queue ahead when it joined
- Empty level1.18×
- Up to 10 ahead0.97×
- 10 to 50 ahead0.92×
- Over 50 ahead0.94×
Time in the book
- Under 10 s0.86×
- 10 to 60 s0.92×
- 1 to 10 min0.95×
- Over 10 min1.04×
Use it on your orders
It runs on our millisecond record of the Polymarket book and every fill on the Polygon chain. Three ways in, all sold as the Calibrated Fill Model: run your backtest’s resting orders through it, fit it to your own live fills, or license it for your own engine. Your code stays with you.
Fifteen minutes on screen, then the check.
- 1
A 15-minute call
We show you a market like yours from our record. Nothing to prepare, nothing to sign.
- 2
A one-page letter
Scope, price and our commitments in writing before you share anything.
- 3
The check
One strategy’s order log in, the verdict out in 5 working days. $3,000 per strategy.
Questions
The part of a backtest that decides whether a resting order would have been filled, how much of it, and when. For a strategy that rests orders, it decides most of the result: a model that fills too much books trades the venue never gave you, and one that fills the wrong orders shows an edge that was really the price moving against you.
Replayed against live fills from 5 to 16 October 2026, on matches held out from the fit, it filled 1.00 times the contracts the venue filled on the ladder and 1.02 times on the first order in, and 94.3% of the venue’s contracts on those first orders came back on the very same orders. In the earlier replay, 5 to 14 October 2026, it filled 0.97 times the venue’s contracts on entry orders, and all 8 groups by queue depth and time in the book landed within 20%. 99.8% of live fills match a trade on our tape.
It was fitted on Polymarket. For your markets and order types, we check the fit on your own orders before you rely on it, and we tell you on the first call what is covered.
Yes, under licence, for your own engine. It starts with a 15-minute call, then the check on your own orders.