Sorting your journal results by day of the week

[Shadow Trader]

Pull a month of spread data by day of week and the averages are rarely flat. Tuesday through Thursday tend to have the tightest spreads and the deepest books. Monday mornings can be sluggish. Friday afternoons often thin out, particularly after London closes, and the spread creeps wider into the weekend.

Several forces are at work. Mid-week sees the most data releases and the most institutional participation. Friday afternoons carry position squaring and the reluctance of providers to take risk into a weekend. Monday carries the aftermath of the gap and a slower return of liquidity.

The patterns are small compared with the news or rollover effects, but they're consistent, and a scalper whose edge is thin can feel them.

A decent exercise is breaking performance down by day of week. Some strategies perform consistently better on certain days, and a few perform badly on one day in particular. With enough trades, the differences are statistically meaningful.

Dropping the worst weekday entirely sometimes improves the equity curve more than any parameter change.

Which day of the week produces the best results, and which one quietly costs money?

[Fairman]

The days around month end might deserve their own look, separate from the day of the week. The last trading day of the month can fall on any weekday, and flows around it can make that day behave differently. If one of your weekdays looks bad, check whether a lot of its trades happened to fall on month end days. Tagging month end separately and filtering it out of the weekday comparison can make a weekday effect disappear, or make it clearer. Either way, it's one more source of noise to rule out before drawing a conclusion.

[Fairman]

Do you see a difference between the first and second half of the week in your own trades? Splitting into two groups, Monday to Wednesday and Thursday to Friday, gives more trades per group than five separate days. Many big data releases cluster on Thursdays and Fridays, especially in the US, which can change how those days trade. If the second half of the week is clearly different for you, that's a more robust finding than a single bad weekday, since it's based on twice as many trades and has an obvious possible cause.

[Fairman]

The "tired on Fridays or rusty on Mondays" line made me check my own numbers. I had about 220 trades over three months, which is around 44 per day, so very thin as you say. Fridays looked awful at -0.28R average, and I nearly banned myself from trading them. Then I saw that two of those Fridays contained revenge sequences of four trades each, and without them the day was roughly flat.

So I tag the cause now instead of reacting to the number. If it's a market thing, like an NFP Friday, the response is to avoid that event, not the whole day. If it's me, I cut size.

On a 10,000 dollar account with 0.5 percent risk, a day that averages -0.28R over 8 trades costs about 110 dollars a week, which is worth fixing, but only once it holds up over a few months.

You mentioned a later start as a response. Have you tried delaying the first trade by an hour, say waiting until 9:30 rather than 8:30? And did the pivot table show anything for Wednesdays in your data?