Most backtests fill at the touch — buy at the ask, sell at the bid, done. That is the most optimistic assumption you can make, and it is the default in almost every framework.
We wanted to know how much of an edge survives when you stop making it. So we took a finished result, left every entry and exit decision untouched, and only changed what we assumed the fill cost.
One thing to be clear about before the numbers: 44% is not a general property of slippage. It is what we measured on one specific strategy over one specific set of 2,084 trades. A strategy with a larger average trade or a longer holding period would decay far more slowly, and one with a smaller edge per trade would decay faster. What generalises here is the method and the shape of the curve, not the coefficient.
Edge remaining, baseline = 100
Baseline100
1 tick56
2 ticks13
5 bp75
10 bp50
20 bp0
1. What we changed and what we did not
The trade list is identical in every run. Same signals, same entries, same exits, same 2,084 trades. The only thing that moves is the price each fill is booked at:
- Baseline — buy at the ask, sell at the bid.
- Tick slippage — the fill walks one or two ticks against you on both sides.
- Basis-point slippage. The fill moves a fixed fraction against you on each side.
This isolates cost. Nothing about the strategy is being re-optimised to cope.
2. Result, normalised
Cost assumption
Edge remaining
Win rate
Profit factor
Baseline (touch)
100
64.1%
2.29
1 tick round trip
56
62.5%
1.82
2 ticks round trip
13
58.7%
1.44
5 bp one way
75
63.0%
2.04
10 bp one way
50
62.0%
1.80
20 bp one way
0
60.3%
1.42
We publish the edge column normalised to the baseline rather than as absolute returns, because the number that matters here is the decay, not the level.
3. Why the win rate barely moves while the edge collapses
Look at the two ticks row: the edge is down to 13% of baseline, but the win rate only fell from 64.1% to 58.7% and the profit factor is still above 1.4. Every headline statistic still looks acceptable. The result is nearly gone anyway.
That is the shape of a cost-sensitive strategy. Slippage does not remove your winners; it shaves every trade by the same small amount. With a short holding period and a modest average gain, that shave is a large fraction of the whole.
The corollary is uncomfortable: win rate and profit factor are poor early-warning signals for cost fragility. You have to re-price the fills.
4. What this implies for a short-holding strategy
If your average trade is small, the spread is not a detail of execution — it is most of your P&L. Before you tune entry logic, price the fills honestly and see whether anything is left. It is cheaper to find out at this stage than after funding an account.
The useful takeaway is not our coefficient. It is that you can run this test on your own result in an afternoon: keep the trade list, change only the fill price, and read the decay curve.
Conclusion
Test your cost assumptions before you test anything else. Here, one tick of round-trip slippage took 44% of the edge and two ticks took 87%, while the win rate moved by five points and the profit factor still read above 1.4. Your own numbers will differ. The point is that the summary statistics will not warn you.
Limitations
The 44% and 87% figures are specific to this strategy and this 2,084-trade set. They are not a general slippage rule and should not be quoted as one.
One strategy on US equities over one 75-session out-of-sample window. Decay depends entirely on average trade size and holding period; a strategy holding for days would barely notice these costs.
The costs are applied as uniform assumptions, not as a simulated order book. Real slippage is state-dependent. Worse when you are least likely to want it.
Fees are already inside the baseline. These figures are the additional effect of slippage on top of them.
We publish relative decay, not absolute performance. The absolute figures are retained internally as the source of these ratios.
Re-pricing a fixed trade list assumes the strategy would have taken the same trades under worse fills. A live system with a spread filter would have skipped some of them, which cuts both ways.
Originally published by TraderWe on August 11, 2026. You may quote and link to this page. Republishing the full text without a link back to the original is not permitted.
morning read with the coffee and honestly this one stung a bit. I've been fill-at-the-touch in every test I've ever run and just never questioned it. love that they held the trade list fixed and only changed the cost side, that's the part I can actually copy this week.
The bit that jumps out to me is what this does downstream of the P&L. If you size by anything Kelly-flavoured, the optimal fraction scales roughly with your edge, and the growth rate scales with edge squared over variance — so knocking 44% off the edge doesn't cost you 44% of your compounding, it costs closer to 70%. Cost assumptions are a sizing input, not just an accounting line.
What's the average gross profit per trade in ticks for this set? That number plus the tick size basically tells you the whole shape of your curve before you run it.