Day-of-the-Week Results and the Danger of Small Cells

[Fairman]

Day-of-the-Week Results and the Danger of Small Cells

A trader's results by weekday over 60 trades: Monday, 10 trades, minus 3R. Tuesday, 14 trades, plus 6R. Wednesday, 13 trades, plus 4R. Thursday, 12 trades, plus 2R. Friday, 11 trades, minus 5R.

Averages: Monday minus 0.30R, Tuesday plus 0.43R, Wednesday plus 0.31R, Thursday plus 0.17R, Friday minus 0.45R. Total plus 4R.

The temptation: skip Mondays and Fridays. That would remove 21 trades with a combined result of minus 8R, giving plus 12R from 39 trades.

The caution: each cell holds about a dozen trades. The standard error per cell is roughly 1.5 divided by the square root of 12, which is 0.43R. Friday's average of minus 0.45R is about one standard error below zero. Nothing is statistically distinguished.

Is there a reason behind the pattern? Mondays can be slow after the weekend gap, and Friday afternoons see position squaring and thin liquidity, so there are plausible structural explanations. That raises the credibility compared with a pattern with no story.

A sensible response is a trial. Reduce size by half on Mondays and Fridays for the next 60 trades, then recheck. This limits the cost of being wrong, and gives more data.

Alternatively, combine Monday and Friday into one group and compare with midweek, which gives larger cells.

Don't mine many splits. The more ways you slice the data, the likelier a false pattern.

Exercise: compare midweek with Monday plus Friday in your own results.