Rolling Expectancy on Twenty Trades

[Fairman]

Rolling Expectancy on Twenty Trades

A trader's last twenty trades in R: plus 2, minus 1, minus 1, plus 1.5, minus 1, minus 1, plus 3, minus 1, minus 1, plus 1, minus 1, minus 1, plus 2, minus 1, minus 1, minus 1, plus 1, minus 1, minus 1, minus 1.

Compute the rolling ten-trade average for three windows. Trades 1 to 10: the sum is 2 minus 1 minus 1 plus 1.5 minus 1 minus 1 plus 3 minus 1 minus 1 plus 1, which is plus 1.5. Average 0.15R.

Trades 6 to 15: minus 1, plus 3, minus 1, minus 1, plus 1, minus 1, minus 1, plus 2, minus 1, minus 1 sum to minus 1. Average minus 0.10R.

Trades 11 to 20: minus 1, minus 1, plus 2, minus 1, minus 1, minus 1, plus 1, minus 1, minus 1, minus 1 sum to minus 5. Average minus 0.50R.

The sequence of averages: 0.15, minus 0.10, minus 0.50. A downward trend.

What does it mean? With ten trades per window and a standard deviation of 1.5R, the standard error is 0.47R. A drop from 0.15 to minus 0.50 is 0.65R, roughly one to one and a half standard errors, interesting but not alarming. It is exactly the kind of pattern that warrants a review, not an overreaction.

The review questions: did the market regime change? Are the rules being followed? Were the last ten losses clustered in a particular session or pair?

Compare the win rate: windows show 3 of 10, 2 of 10 and 2 of 10 wins, a small decline.

Exercise: compute rolling averages for your own last thirty trades with a ten-trade window.

[FXS HongKong Scalper]

Nice example of why ten trades is too short a window. The numbers move from plus 0.15 to minus 0.10 to minus 0.50, but with a win rate around 30 percent and wins of 1 to 3R, the noise in a ten-trade average is larger than those differences.

We use rolling expectancy with a longer window, 50 trades, and also look at the win rate and average win separately. If the win rate drops while average wins stay the same, it may be market conditions. If the average win shrinks, it's often the trader cutting winners early after a losing run, which is fixable.

[Fairman]

A small correction on the win counts. Trades 1 to 10 contain four wins, plus 2, plus 1.5, plus 3, and plus 1, so the first window is 4 of 10, not 3. Trades 6 to 15 contain three wins, plus 3, plus 1, and plus 2. Trades 11 to 20 contain two, plus 2 and plus 1. So the sequence is 4, 3, 2, a steadier decline than the post suggests. The sums and averages in the post are correct, at plus 1.5, minus 1, and minus 5. The win rate trend reinforces the same message as the expectancy: something has weakened over these twenty trades.

[Amanhasnoname]

I tracked a rolling average like this for a stretch on a 1 minute scalping system and the first thing I learned was how much a ten trade window lies. My win rate was around 35 percent with a few 3R winners, and I had a window as bad as minus 0.6R followed three weeks later by one at plus 0.4R with no change in how I traded. The drop in your example, from 0.15 to minus 0.50, is well inside that kind of swing.

What made the number useful was widening the window to forty trades and plotting it next to the count of rule breaks I had logged. When both moved together, the fix was discipline. When the average sagged while my rule breaks stayed at zero, I treated it as a slow market and simply cut my size in half until it recovered.

I would also split the last twenty into winners and losers separately. Here the losses are all exactly minus 1R, so the stop discipline looks fine and the weakness is only that fewer winners show up, which points at conditions or entry timing, not execution.

[Fairman]

The statistical caution is right. The first and last windows don't overlap, so their difference of 0.65R can be compared with a standard error of about 0.47 times the square root of 2, roughly 0.67R. That's about one standard error, well within what chance can produce. The middle window overlaps both, so it doesn't add independent information. With only twenty trades, the decline could easily be noise. But as the post says, it's the kind of pattern that warrants a review, since checking the causes costs little and might catch a real problem early.

[Fairman]

The review questions are well chosen. A change in market regime, such as a shift from trending to ranging conditions, can hurt a trend strategy for a while without anything being wrong with it. A decline in rule following is a behavioural issue with a different fix. Clustering by session or pair might reveal that one part of the strategy is struggling while the rest is fine. Each answer leads to a different action, which is why the review is more useful than reacting to the number itself. Writing the answers down creates a record for the next review.

[Fairman]

It's worth noticing that all the wins in the last ten trades were small, plus 2 and plus 1, while the first window included a plus 3. Fewer large winners, combined with a falling win rate, suggests the strategy may be in a choppy period where trends don't extend. If the strategy relies on occasional big wins, a few weeks without them can produce exactly this pattern. Has anyone here used rolling expectancy to spot a regime change in their own trading? I'd be interested in how long the dip lasted and whether the strategy recovered on its own.