Advertisement IC Markets

Delay filters and the speed bump

Compare ECN/Raw spread brokers, analyze execution speeds, report slippage, and evaluate commission structures for high-frequency traders.
Post Reply
Fairman
Posts: 6955
Joined: Tue Jul 21, 2026 7:11 am
Location: Abuja

Delay filters and the speed bump

Post by Fairman »

Delay filters and the speed bump

Some liquidity providers and brokers use a deliberate delay on certain flows. The idea is to protect against latency arbitrage: traders with fast feeds who exploit stale prices.

A stale quote is one that has not updated to reflect a move elsewhere. A fast participant sees the new price and hits the old quote before it can refresh. The market maker loses.

Delay filters, sometimes called speed bumps, slow incoming orders for a few milliseconds. During that time, the quote refreshes. If the price moved, the order is rejected or filled at the new price.

For most retail traders, the delay is invisible. For scalpers with fast setups, it is felt as slightly worse fills, mostly on entries that coincide with sharp moves.

The distinction between latency arbitrage and legitimate scalping is not always made carefully. A strategy that trades on short-term momentum looks similar to an arbitrage one from the broker's side.

Some brokers apply the delay only to accounts flagged by behaviour. Others apply it to everyone. A few advertise that they do not use any.

Signs worth checking are fill delays that exceed the usual by a consistent margin, particularly on winning trades, and differences between demo and live behaviour.

Clear information from the broker's execution policy is the starting point. Test results from small live trades are the proof.

An informal experiment is possible with two accounts: one placed on a fast VPS and one on a typical home line. If the faster account sees consistently worse fills than the slower one on identical orders, the broker may be penalising speed. That result sounds backwards and does occur. It suggests the broker's filter targets fast arrivals specifically, which is useful to know before investing money in lower latency.

What does the distribution of fill delays look like, and does it differ between trades that moved quickly and those that did not?
It’s Fairman :geek:
Recommended broker for automated trading & scalping IC Markets
FXS Prop Desk
Posts: 1689
Joined: Wed Oct 07, 2026 9:52 am

Re: Delay filters and the speed bump

Post by FXS Prop Desk »

Fairman wrote: It suggests the broker's filter targets fast arrivals specifically, which is useful to know before investing money in lower latency.
The speed bump is a fair tool for providers. For normal scalpers, it can add a few milliseconds to every fill, which matters most on very short targets. Comparing fill times before and after any announced change in execution helps measure its effect.
FXS Official Team | The Idea Lab & Strategy
Fairman
Posts: 6955
Joined: Tue Jul 21, 2026 7:11 am
Location: Abuja

Re: Delay filters and the speed bump

Post by Fairman »

FXS Prop Desk wrote:Comparing fill times before and after any announced change in execution helps measure its effect.
Good idea, though in my experience the change is rarely announced, so I keep a running log instead. Every trade records the time the order left the terminal and the server fill time, plus how far price moved in the 500 ms before entry. That last column is what exposes a filter.

When I ran it over about 300 trades at one broker, quiet entries filled in a median of 45 ms. Entries where price had already moved 1.5 pips or more in that half second filled in a median of 210 ms, and roughly one in five came back requoted or 0.4 to 0.8 pips worse. On a 3 pip target at 1 lot that slippage is 4 to 8 dollars a trade, a big slice of a 30 dollar winner.

Nothing I did with a faster VPS changed those numbers, which told me the delay was on their side. If anyone wants the column layout, I can post it.
It’s Fairman :geek:
Amanhasnoname
Posts: 2060
Joined: Fri Oct 30, 2026 9:45 am

Re: Delay filters and the speed bump

Post by Amanhasnoname »

That log layout is the right way to approach it, because the key signal is not slow fills in general but slow fills that depend on what price was doing just before the order arrived. A delay that is the same on every trade is probably network or bridge latency. A delay that grows only when the market has just moved in your favour looks much more like a deliberate filter, and that asymmetry is what you want the data to show or rule out.

It is worth adding a column for the direction of the pre-entry move relative to your order. If fills are slower when price moved toward your entry, but not when it moved away, that points strongly at a last look or speed bump check rather than general load on the server. Tracking the slippage sign as well as the size helps too, since a fair execution model should show some positive slippage, not only negative.

The main pitfall is small samples within each bucket. Fast-move entries are a minority of trades, so a few hundred trades overall can leave only a few dozen in the bucket that matters. Reading the execution policy and asking support directly whether any minimum hold or delay applies is a cheap check alongside the numbers.
Post Reply