Running the same strategy on demo and live for two weeks to see what execution actually costs
everyone says demo fills are fake and live is different, but I wanted to know by how much for my own style. So I ran the exact same M1 and M5 scalping rules on a demo and a small live account side by side for two weeks, same pair, same entries, same stops, taking each trade on both.
What I compared was entry price, exit price and the spread at the time of entry. On calm entries the difference was tiny. On entries within a minute or two of the open of a session, or near any news, the live fill was usually a bit worse and the spread was wider than the demo showed. Over a lot of trades, that gap is basically the difference between a strategy that works and one that barely scrapes by.
The takeaway for me wasn't that live is rigged, it's that demo is optimistic and I shouldn't trust any backtest or demo result for a scalp style without taking a haircut off it. I now assume each trade costs me an extra fraction of a pip over what the demo showed.
Running the same strategy on demo and live for two weeks to see what execution actually costs
Re: Running the same strategy on demo and live for two weeks to see what execution actually costs
For anyone who wants to repeat the test, the easiest way I found to record it was a simple spreadsheet with one row per trade and columns for the demo fill, the live fill, the spread on each and the time. I wrote the numbers down right after each entry, before looking at the result, so I would not fudge anything. After two weeks I just averaged the difference in fills. It was boring to log but much better than relying on a feeling that live was worse.
It’s Fairman 
Re: Running the same strategy on demo and live for two weeks to see what execution actually costs
A fair question someone asked was whether the difference would shrink on bigger size or smaller size. I only tested at a small size, so I cannot say for sure. I would expect bigger orders to see a bit more slippage on fast entries, not less. If you are planning to scale up, it is worth running the same comparison again at the new size for a week or two, because the haircut I mentioned might not stay the same as you grow.
It’s Fairman 
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Running the same strategy on demo and live for two weeks to see what execution actually costs
Splitting the difference by entry time is where this test pays off. Say the trades are 70 percent calm entries costing an extra 0.1 pips live, 20 percent taken within two minutes of a session open costing 0.6, and 10 percent near news costing 1.1.
The average then comes out at 0.3 pips per trade, which on a 4 pip stop is 0.075R. The chart (illustration) shows where it comes from, since the open and news entries are 30 percent of trades but about 77 percent of the extra cost.
So the haircut does not have to be applied evenly. Skipping entries in the first two minutes of a session and near releases might cost a few good trades but removes most of the gap.
Measuring both fills against the signal price, rather than live against demo, is also worth doing. Demo fills are not always the clean reference they seem, and a demo that sometimes fills better than the quoted price would hide part of the real cost.
On the size question above, partial fills are the other thing to watch. A larger order at a thin moment may fill in pieces at different prices, so the average fill is the number to write down.
The average then comes out at 0.3 pips per trade, which on a 4 pip stop is 0.075R. The chart (illustration) shows where it comes from, since the open and news entries are 30 percent of trades but about 77 percent of the extra cost.
So the haircut does not have to be applied evenly. Skipping entries in the first two minutes of a session and near releases might cost a few good trades but removes most of the gap.
Measuring both fills against the signal price, rather than live against demo, is also worth doing. Demo fills are not always the clean reference they seem, and a demo that sometimes fills better than the quoted price would hide part of the real cost.
On the size question above, partial fills are the other thing to watch. A larger order at a thin moment may fill in pieces at different prices, so the average fill is the number to write down.
-
FXS Prop Desk
- Posts: 1689
- Joined: Wed Oct 07, 2026 9:52 am
Re: Running the same strategy on demo and live for two weeks to see what execution actually costs
What was the average gap per trade in pips? Even 0.2 pips per trade adds up quickly on an M1 strategy that takes many trades.
Another angle for the second run: note the clock time of each fill. Live slippage is often concentrated in the first minutes after the London open and around data releases, while mid-session fills look almost like demo. If most of the cost sits in those windows, simply sitting out the first five minutes of the session might recover a good part of it.
Another angle for the second run: note the clock time of each fill. Live slippage is often concentrated in the first minutes after the London open and around data releases, while mid-session fills look almost like demo. If most of the cost sits in those windows, simply sitting out the first five minutes of the session might recover a good part of it.
FXS Official Team | The Idea Lab & Strategy
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Running the same strategy on demo and live for two weeks to see what execution actually costs
Sample size decides whether the average gap is real. Say each trade's live minus demo difference has a standard deviation of about 0.5 pips and two weeks gives 80 paired trades. The standard error is then 0.5 divided by the square root of 80, about 0.06 pips.
So an average of 0.3 pips is clearly not noise, while an average of 0.05 could easily be zero. Splitting the sample by entry time shrinks each group, so a news group with only a handful of trades needs a much wider margin.
Rejected orders are a cost that a fill comparison misses. A demo account almost never refuses an order, but live can requote or reject during a fast move, and those are often the breakout entries that would have run furthest.
Say 1 live order in 25 is refused and those trades average plus 2R on demo. That works out to 0.08R per trade, slightly more than the 0.3 pip slippage figure from earlier in the thread costs on a 4 pip stop.
Positive slippage is the other half worth recording. Is the live account passing on better fills on limit orders, or only the worse ones on market orders?
So an average of 0.3 pips is clearly not noise, while an average of 0.05 could easily be zero. Splitting the sample by entry time shrinks each group, so a news group with only a handful of trades needs a much wider margin.
Rejected orders are a cost that a fill comparison misses. A demo account almost never refuses an order, but live can requote or reject during a fast move, and those are often the breakout entries that would have run furthest.
Say 1 live order in 25 is refused and those trades average plus 2R on demo. That works out to 0.08R per trade, slightly more than the 0.3 pip slippage figure from earlier in the thread costs on a 4 pip stop.
Positive slippage is the other half worth recording. Is the live account passing on better fills on limit orders, or only the worse ones on market orders?
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Running the same strategy on demo and live for two weeks to see what execution actually costs
Cent accounts denominate balances in cents, so small deposits behave like larger ones. They allow live testing of execution quality, slippage, and platform behavior with minimal financial risk.
Demo accounts often differ from live ones in fills, spreads, and rejection patterns, since demo servers may not route to the same liquidity. A cent account runs on the live feed and execution engine, producing results that more closely match real conditions.
For scalpers, a cent account is a way to validate a strategy before scaling up. Running the intended system for several weeks and logging fills, slippage, and rejections gives real data. The amounts at stake remain small, while the information gained is valuable. A limitation is that some brokers apply different conditions to cent accounts, such as wider spreads or restricted strategies. Confirming that cent conditions match standard accounts, or at least understanding the differences, ensures that conclusions transfer. The goal is information, so the results should be analyzed carefully.
Demo accounts often differ from live ones in fills, spreads, and rejection patterns, since demo servers may not route to the same liquidity. A cent account runs on the live feed and execution engine, producing results that more closely match real conditions.
For scalpers, a cent account is a way to validate a strategy before scaling up. Running the intended system for several weeks and logging fills, slippage, and rejections gives real data. The amounts at stake remain small, while the information gained is valuable. A limitation is that some brokers apply different conditions to cent accounts, such as wider spreads or restricted strategies. Confirming that cent conditions match standard accounts, or at least understanding the differences, ensures that conclusions transfer. The goal is information, so the results should be analyzed carefully.