Advertisement IC Markets

Backtest honesty for EURUSD M1: modeling spread and slippage

Discuss 1-minute to 15-minute price action setups, fading intraday momentum, key support/resistance zones, and proven short-term trading methodologies.
Post Reply
LondonScalper
Posts: 965
Joined: Sat Sep 05, 2026 7:54 am

Backtest honesty for EURUSD M1: modeling spread and slippage

Post by LondonScalper »

Backtest honesty for EURUSD M1: modelling spread and slippage.

An M1 EURUSD backtest that uses a fixed optimistic spread is fiction. I want variable spread by session, slippage stress on entries, and rejection of trades when spread exceeds a cap -- because that is how I trade live. Curve-fit entries die when costs wake up.

Minimum honesty
1. Session-aware costs.
2. Slippage scenarios, not zero.
3. Out-of-sample and walk-forward, not one magic decade.
4. Compare to a simple baseline (e.g. time-of-day filter only).

If the edge vanishes with realistic costs, it was never an edge.

How do you stress costs in M1 tests without making the model pure guesswork?

Publish assumptions with any equity curve you share: spread model, slip, and whether news trades were filtered. Curves without cost footnotes are entertainment. Honesty shrinks the curve and grows the chance you will trust live results.

I am interested in how others on this desk handle the same problem without turning it into folklore. Concrete rules and log fields beat slogans. If you have a version of "Backtest honesty for EURUSD M1: modeling spread and slippage" that survived contact with live sessions, what does the rule look like on a sticky note -- and what did you try that failed?
Recommended broker for automated trading & scalping IC Markets
PropScalpDesk
Posts: 521
Joined: Sat Sep 19, 2026 7:50 pm

Re: Backtest honesty for EURUSD M1: modeling spread and slippage

Post by PropScalpDesk »

For the spread side, the least guesswork I've found is MT5's "every tick based on real ticks" mode. It replays the broker's recorded ticks with their actual bid and ask, so session-aware spread comes for free, including the ugly minutes after the daily rollover. The catch is that the history only goes back as far as the broker kept ticks, and it's that one broker's feed, so I test on the account I trade.

Slippage I can't take from history, so I stress it on purpose. One run with zero, one with a fixed amount per side that matches what my live fills show on average, and one with double that. If the M1 edge only survives the zero-slippage run, I treat it as gone. I'd rather throw away a marginal system than discover it live.

The spread cap from your list belongs in the code itself, not as a filter after the test. If the EA refuses entries above the cap, the backtest also shows how many trades the cap removes, which is often a surprise.
Fairman
Posts: 6955
Joined: Tue Jul 21, 2026 7:11 am
Location: Abuja

Re: Backtest honesty for EURUSD M1: modeling spread and slippage

Post by Fairman »

PropScalpDesk wrote: One run with zero, one with a fixed amount per side that matches what my live fills show on average, and one with double that.
Is the fixed amount the same for entries and exits? In most live logs the average hides a big difference by order type. Market and limit entries tend to slip a little, while stop-loss exits are the ones that carry the tail, because they trigger exactly when price is moving fast against you. If the backtest applies the average to everything, it can overcharge the entries and undercharge the stops, which flatters a system that gets stopped out often. Do you set slippage separately per order type in the tester, or has the double-slippage run been close enough when you compared it with live results?
It’s Fairman :geek:
Shadow Trader
Posts: 5594
Joined: Wed Sep 09, 2026 5:19 pm

Re: Backtest honesty for EURUSD M1: modeling spread and slippage

Post by Shadow Trader »

Fairman's split matches my own log closely. Over my last 400 live EURUSD M1 trades, market entries slipped 0.1 pip on average, limit entries and limit targets filled at price, and stop loss exits averaged 0.6 pip.

Pooled across all 800 fills that comes to about 0.17 pip, which looks harmless. The chart puts the four order types next to that pooled figure, and the stop exits are the only bar well above it.
img_12445_r1.png
img_12445_r1.png (55.53 KiB) Viewed 139 times
The pooled number does its damage when comparing variants. In the live version 180 of the 400 trades ended at the stop, so 45 percent, and a variant stopped out 60 percent of the time pays about 0.36 pip per trade in stop slippage alone against 0.27, yet a flat per fill cost charges both the same.

The MT5 tester will not apply slippage by order type, so I export the deal list and add it afterwards in a spreadsheet. It takes about ten minutes and has changed which parameter set I picked twice.

One edge case for the tail. My worst stop fills all came in the two minutes after scheduled releases, so if the EA has a news blackout, the stop figure should come only from trades outside it.
Shadow Trader
Posts: 5594
Joined: Wed Sep 09, 2026 5:19 pm

Re: Backtest honesty for EURUSD M1: modeling spread and slippage

Post by Shadow Trader »

Since my last post I looked at the stop exits on their own, because a 0.6 pip average said nothing about how it was made up. Of the 180 stop exits in that sample, 64 filled at the stop price exactly and another 58 slipped half a pip or less.
img_12445_r2.png
img_12445_r2.png (77.92 KiB) Viewed 136 times
The chart splits all 180 into five slippage bands and shows each band's share of the total slippage. The 10 exits that slipped more than 2 pips make up about 37 percent of it on their own, which is why a flat 0.6 added to every stop misrepresents the risk.

For drawdown work I now resample instead. Each simulated trade that ends at the stop draws its slippage at random from those 180 real values, and the deal list is replayed 1,000 times in shuffled order.

The average result barely moves compared with the flat figure. The 95th percentile drawdown came out about 1R deeper, because a few large stop slips landing in the same week is exactly what a flat cost can never produce.

This only works if the slippage set matches the rules being tested. A variant with a different stop distance or session filter needs its own live sample before those numbers mean much.
Shadow Trader
Posts: 5594
Joined: Wed Sep 09, 2026 5:19 pm

Re: Backtest honesty for EURUSD M1: modeling spread and slippage

Post by Shadow Trader »

Many backtests assume a fixed spread, often the minimum advertised value. Real spreads vary through the day, widening around rollover, news, and low-liquidity hours. A model that ignores this variation overstates profitability, especially for strategies that trade during unfavorable times.

A more honest approach uses recorded spread data from the intended broker. Where tick history includes both bid and ask, the real spread at each moment can be replayed. Where it does not, a spread profile by time of day, built from live observation, can be applied.

Including a conservative buffer is also wise. Adding a fraction of a pip to every trade accounts for slippage and unmodeled costs. Testing how results change as the assumed spread increases shows the strategy's sensitivity. If a small increase in cost turns profit into loss, the edge is too thin to survive real conditions. Robust systems keep a positive expectancy even under pessimistic cost assumptions.
Post Reply