Hedge slippage: comparing what the book promised with what you got

When the bridge decides to hedge, it has a price in mind, usually the composite price for that size. The fill comes back at some other price. The difference is slippage, and it's the most useful number you can feed back into LP scoring and depth weighting, provided you measure it carefully.

The first question is what it's measured against. You can use the composite when the decision was made, when the order was sent, or when the LP acknowledged it. They answer different questions. Decision to send is mostly your own delay. Send to acknowledgement is the LP and the network. Pick one as the headline number and keep the others for working out where the cost came from.

Fix the sign convention early, too. Say positive means worse for you, count price improvement as negative, and use the same rule everywhere. A report where half the readers take the sign the other way is worse than no report.

A single average slippage figure hides most of what's going on. Cost tends to concentrate in particular sizes, particular pairs, and particular conditions such as wide spreads or the minutes after a big move. Bucket by size, symbol and some measure of volatility, and the picture can look very different from the overall average.

Compare LPs against each other in the same moments, not against zero. If everyone slipped during a fast market, scoring one LP down for it is just noise.

An LP with tiny slippage and a high reject rate isn't cheap. The rejected orders go somewhere else, often at a worse price because the market moved while you waited. Track what price you eventually got after a reject and compare it with the original reference. Add that to the cost of the LP that rejected.

Fold slippage into the same scorecard as rejects and quote age, so an LP can't look good on one number while it's bad on the others.

The part that's hard to settle is attribution. With one fill at a time you can't cleanly separate what the market did from what the LP did. Looking at several LPs in the same window helps, but you rarely trade with all of them at once, so some of the answer will always be an estimate. It's fine to say so in the report.