Measuring Execution Quality: A Short List of Tests
Marketing pages talk about fast execution. You can measure it yourself with a few simple tests.
First, speed. Many platforms show the time between sending an order and receiving a confirmation. Record it for twenty or thirty orders at different times. Look at the median and the worst values.
Second, slippage, as covered earlier. Log the requested versus filled price, in both directions, for market orders and stop exits.
Third, rejection and requote rates. Count how often orders are rejected, requoted, or delayed. Frequent problems during normal conditions are a warning.
Fourth, spread behavior. Record spreads at set times: quiet hours, the London open, a news release, and rollover. Compare with advertised averages.
Fifth, stop and limit order fills. Check whether limit orders fill when price touches the level, or only when it trades through, and compare with other feeds. Check for unexpected stop-outs during spread spikes.
Sixth, consistency between platforms. Compare the broker's quotes with an independent price source during the same moments.
Run the tests on a small live account, since demo execution often differs.
Document results with screenshots and timestamps. If conditions are poor, you have evidence for a complaint or a reason to switch.
Repeat the tests every few months, since execution can change.
Practical step: log twenty market orders with request time, fill time, and slippage, and compute the average.
Measuring Execution Quality: A Short List of Tests
Measuring Execution Quality: A Short List of Tests
It’s Fairman 
Re: Measuring Execution Quality: A Short List of Tests
Doing this on a small live account matters. Demo execution can be very different, and that's sometimes on purpose. Demo servers often fill every order instantly at the quoted price with no slippage, which makes strategies look better than they are. A live account at minimum size gives you the real picture for the cost of a few spreads. Run the tests over a couple of weeks and at the times you actually trade. For most of us that means the London open, the New York data window and maybe the evening. Execution that looks fine at 10am can be very different at 1:30pm when US numbers come out.
It’s Fairman 
Re: Measuring Execution Quality: A Short List of Tests
Worst values are more useful than averages here. A broker can have fine average execution and still hit you with terrible fills during news. If ninety percent of your orders fill within a pip but the stop outs during releases slip by ten or fifteen pips, the average looks fine while the real damage happens on exactly the trades that matter. I'd record the worst three fills each month and compare them with another broker or a second data source. If they line up with normal market chaos, fine. If the other feed shows price never got that far, that's a pattern worth raising with the broker.
It’s Fairman 
-
FXS Prop Desk
- Posts: 1689
- Joined: Wed Oct 07, 2026 9:52 am
Re: Measuring Execution Quality: A Short List of Tests
A good test to add is the comparison between demo and live execution at the same broker. Running the same small orders on both accounts in parallel shows whether the live account receives the same prices and speed. Large differences suggest the demo results will overestimate real performance.
Running the tests again after a few months shows whether the broker's execution has changed, which happens more often than many traders expect.
Running the tests again after a few months shows whether the broker's execution has changed, which happens more often than many traders expect.
FXS Official Team | The Idea Lab & Strategy
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Measuring Execution Quality: A Short List of Tests
Direction of slippage is the more telling number, since the size alone can look normal. In a fair market, slippage on market orders should land on both sides, sometimes better and sometimes worse than the requested price.
Say you log 30 market orders and 15 fill at the quoted price, 4 fill better and 11 fill worse. A split that lopsided does not prove anything on its own, but it is the pattern worth taking to a second data source and testing again a month later.
Order size matters for the speed test. Running the same check at 0.01 lots and at your normal size can give different results, because some execution setups treat small tickets differently from larger ones.
For the timing figure, the median and the slowest few percent tell different stories. A median of 40 milliseconds with a slowest tenth above 800 milliseconds means the trades during busy moments, the ones a scalper cares about, are the slow ones.
Does anyone here test limit orders by placing one exactly at a recent high and noting whether it fills on the touch? Doing that a dozen times gives a clearer answer than reading the order execution policy.
Say you log 30 market orders and 15 fill at the quoted price, 4 fill better and 11 fill worse. A split that lopsided does not prove anything on its own, but it is the pattern worth taking to a second data source and testing again a month later.
Order size matters for the speed test. Running the same check at 0.01 lots and at your normal size can give different results, because some execution setups treat small tickets differently from larger ones.
For the timing figure, the median and the slowest few percent tell different stories. A median of 40 milliseconds with a slowest tenth above 800 milliseconds means the trades during busy moments, the ones a scalper cares about, are the slow ones.
Does anyone here test limit orders by placing one exactly at a recent high and noting whether it fills on the touch? Doing that a dozen times gives a clearer answer than reading the order execution policy.
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Measuring Execution Quality: A Short List of Tests
Slippage is easier to judge once it is converted into R rather than pips. Say the log shows market orders filling an average of 0.3 pips worse than requested, which sounds trivial.
Against a 6 pip stop it costs 0.05R per entry, and if exits slip by the same amount it doubles to 0.1R per round trip. Over 200 trades a year that is about 20R, more than many small edges produce in total.
Where the orders are sent from belongs in the test too. A home connection 120 milliseconds from the trade server and a VPS a couple of milliseconds away can show quite different fills on the same account, and that difference says more about the route than the broker.
So it is worth running the same batch of orders from both places on a quiet morning. If the fills match, the broker is the variable, and if they do not, the fix may be cheaper than switching.
Stop orders around releases deserve their own column. Recording the price at the moment of the trigger alongside the fill shows whether the slippage came from a genuinely fast market or from a delay before the order was handled.
Against a 6 pip stop it costs 0.05R per entry, and if exits slip by the same amount it doubles to 0.1R per round trip. Over 200 trades a year that is about 20R, more than many small edges produce in total.
Where the orders are sent from belongs in the test too. A home connection 120 milliseconds from the trade server and a VPS a couple of milliseconds away can show quite different fills on the same account, and that difference says more about the route than the broker.
So it is worth running the same batch of orders from both places on a quiet morning. If the fills match, the broker is the variable, and if they do not, the fix may be cheaper than switching.
Stop orders around releases deserve their own column. Recording the price at the moment of the trigger alongside the fill shows whether the slippage came from a genuinely fast market or from a delay before the order was handled.
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Measuring Execution Quality: A Short List of Tests
Professional liquidity providers evaluate every trade by looking at where the market moved shortly afterwards. This measure is called a markout: the difference between the fill price and a reference mid price some interval later, such as one second or one minute.
For a provider, consistently negative markouts mean clients are trading just before price moves in their favour, so the provider is being picked off. Consistently positive markouts mean the spread is being earned. Pricing, limits and even client acceptance decisions are adjusted on this evidence.
The same tool is available to an individual trader. Recording the mid price at fill time and at fixed intervals afterwards produces a profile of how entries behave. A scalper whose markouts decay quickly after the fill may be capturing short-lived imbalances; one whose markouts are flat or negative within seconds may be entering after the move has already been completed.
Markouts also expose execution costs that spread figures miss. If price tends to move away from entries during the first moments after fills, slippage or latency is probably involved. Collecting a few hundred observations turns a vague impression about execution quality into a measured profile that can be compared across sessions, instruments and providers.
For a provider, consistently negative markouts mean clients are trading just before price moves in their favour, so the provider is being picked off. Consistently positive markouts mean the spread is being earned. Pricing, limits and even client acceptance decisions are adjusted on this evidence.
The same tool is available to an individual trader. Recording the mid price at fill time and at fixed intervals afterwards produces a profile of how entries behave. A scalper whose markouts decay quickly after the fill may be capturing short-lived imbalances; one whose markouts are flat or negative within seconds may be entering after the move has already been completed.
Markouts also expose execution costs that spread figures miss. If price tends to move away from entries during the first moments after fills, slippage or latency is probably involved. Collecting a few hundred observations turns a vague impression about execution quality into a measured profile that can be compared across sessions, instruments and providers.