Last look and why the price on screen can be withdrawn

[Shadow Trader]

Last look is a rule some liquidity providers use that gives them a short window, often a few milliseconds up to a couple hundred, to accept or reject a trade after it has been sent. If the market moved against them in that time, they can reject. If it moved in their favor, they accept.

For a scalper this has a clear consequence. Trades that would have been profitable are rejected more often than trades that would have lost. Over hundreds of fills that asymmetry adds up to a measurable drag, and it shows up as a "requote" or a mysterious delay of a few hundred milliseconds.

Not every broker routes to last-look providers. Some publish their execution policy and declare they only use firm liquidity. Others stay vague.

There's a simple test. Log the time between sending a market order and the fill confirmation, and separate the trades by whether price moved in your favor during the delay. If the average delay is noticeably longer on favorable moves, it's the pattern. Not proof, but a flag.

It's also why scalpers who chase the last tick of a fast move often get filled worse than the quote they clicked. The market moved in the time it took to process.

Brokers rarely volunteer how they treat this. Some describe their liquidity as firm, meaning quotes are honored once accepted. Others mention "last look" in the execution policy, sometimes with a maximum hold time such as 10 or 50 milliseconds, sometimes without any figure.

A short hold time with a symmetrical rejection policy is a mild version of the practice. A long hold time with rejections that favor the provider is a heavier one. Without data, there's no way to tell which one applies, which is why the delay test matters.

It's also worth noticing the rejection messages. "Off quotes" and "requote" at the exact moment price moves in the client's favor are more informative than the occasional random one.

Has a broker ever been clear about whether their liquidity uses last look, or did the answer come out vague?

[Amanhasnoname]

I asked two brokers directly last year and got two very different answers. One sent me a link to their execution policy, which stated that they used last look with a maximum hold of a few dozen milliseconds and that rejections applied in both directions. The other replied with a long paragraph about "best execution" that never actually answered the question, which told me more than a yes would have.

I ran something close to your delay test on the second broker over roughly two months. My fills on trades where price had moved in my favour during the wait took visibly longer on average, and almost all of my requotes came on those trades too. It wasn't dramatic on any single trade, but it added up to a noticeable cost on a strategy aiming for small targets.

I moved most of my volume to the first broker afterwards. The spreads looked slightly worse on paper, but the average result per trade improved, mostly because I stopped losing the good fills.