Skip to content

The fill & slippage model

When you submit a trade, the price you get isn't the headline market price: it's a fill built from several components. Ultraprop shows you that build before you commit, every time. This page documents each piece, because a slippage model you can't see is a slippage model you can't trust.

The components of a fill

  oracle mid price     market reference price
+ size slippage        bigger order, thinner book = worse
+ house tilt (+2 bps)  fixed, against you, always shown
± taker fee  (0.045%)  charged on the fill
─────────────────────
= your fill price

Oracle mid price

The live reference price for the market, sourced from real mainnet data, not a testnet feed and not a synthetic random walk. See Prices & oracles.

Size slippage

Real fills move against you as your order grows relative to available liquidity. The model accounts for order size and book depth, so a large order in a thin market fills worse than a small order in a deep one, exactly as it would on a live venue.

The house tilt

A fixed +2 basis points against the trader on every fill. This is the model's signature, and it's deliberate:

  • It's always disclosed. It appears in the fill preview every time. It is not a hidden spread.
  • It's identical for everyone. Same tilt, every participant, every fill.
  • It's conservative on purpose. The simulated fill is always a touch worse than the median real fill on a live aggregator. The point: a strategy that passes here would also pass on mainnet. The model never flatters you.

Taker fee

A taker fee is charged on every fill, including the fill that closes a position via a bracket. The reference rate is 0.045%. Fees feed your equity math, which means fees alone can move you toward your drawdown floor. A heavily traded account is paying its way down even between wins.

Why "show the math" is a principle, not a feature

The model is calibrated so that paper-trading is indistinguishable from live execution. If it weren't:

  • Traders couldn't trust the evaluation.
  • The pass rate would be meaningless as a skill signal.
  • Funded results in v2 would diverge from beta results, and everything learned in the beta would be worthless.

So the model is shown in full, calibrated against real historical swap data, and tilted conservatively. The whole credibility of the platform rests on this number being honest, which is why it's never hidden.