Some arguments on this forum come round every few months. I can't settle them for everyone, but I can settle them for my own trading, because I have eleven years of tagged tickets in a database. Here are three, with my numbers. Yours may differ.
"London open is the best time to scalp." For me, partly true. My best hour by average net pips per trade is 07:00 to 08:00, but my best hour by total profit is 08:00 to 09:00, simply because I take more trades then. The very first minutes after 07:00 are actually below average once spread is counted.
"Avoid Mondays." My Monday results are slightly below my average, not dramatically. The real outlier is Monday before 08:00, where my net result over the years is negative. After 09:00 Monday looks like any other day.
"The London-New York overlap is where the money is." Not for me. My overlap results (13:00 to 16:00) are positive but with much higher variance than the morning. I make roughly the same per trade but have twice as many large losers, mostly around 13:30 and 15:00 releases.
The one I can't settle is "Friday afternoon is dangerous". I simply have too few Friday afternoon trades, because I stopped trading them years ago. Which is itself a kind of answer, I suppose.
If you have a decent sample of your own, I'd be interested in where your numbers disagree with mine.
Advertisement
Session mastery: common debates settled with data
-
LondonScalper
- Posts: 965
- Joined: Sat Sep 05, 2026 7:54 am
Recommended broker for automated trading & scalping
-
Shadow Trader
- Posts: 977
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Session mastery: common debates settled with data
The split between best hour per trade and best hour by total is the kind of thing that gets lost in these debates. People argue about London open when they're really comparing different measures.
On my own sample, smaller than yours at about four years, the 07:00 to 07:10 window was the worst ten minutes of the day after costs. Spread on GBPUSD in those first minutes was often double the 09:00 median, so the gross result looked great and the net didn't.
The Monday before 08:00 result is interesting. I'd guess some of that is the thin Asian Monday leaking into early London, with gaps from the weekend still being filled.
Your Friday answer made me laugh, but it's a real problem with this kind of data. Rules you stopped following stop generating data, so you can never check if they still make sense.
Have you ever run a small paper sample on Friday afternoons just to see if the old rule still holds?
On my own sample, smaller than yours at about four years, the 07:00 to 07:10 window was the worst ten minutes of the day after costs. Spread on GBPUSD in those first minutes was often double the 09:00 median, so the gross result looked great and the net didn't.
The Monday before 08:00 result is interesting. I'd guess some of that is the thin Asian Monday leaking into early London, with gaps from the weekend still being filled.
Your Friday answer made me laugh, but it's a real problem with this kind of data. Rules you stopped following stop generating data, so you can never check if they still make sense.
Have you ever run a small paper sample on Friday afternoons just to see if the old rule still holds?
-
Shadow Trader
- Posts: 977
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Session mastery: common debates settled with data
One debate I'd add to the list: whether the 13:30 US data window should be traded at all by morning scalpers. Your twice as many large losers in the overlap mostly around 13:30 and 15:00 suggests the answer is in the numbers already.
I ran mine with the data windows removed. Taking out 13:25 to 13:45 and 14:55 to 15:10 cut my overlap trade count by about a fifth, and the variance dropped to roughly the morning level.
The per trade result barely moved, so the large losers were just noise around the releases rather than part of the edge.
The catch is that removing windows after the fact is easy to overfit. I'd only trust it if the same cut works on a year of data you didn't look at while deciding.
When you split the overlap result, did the losers sit right after the release or a few minutes later on the second move?
I ran mine with the data windows removed. Taking out 13:25 to 13:45 and 14:55 to 15:10 cut my overlap trade count by about a fifth, and the variance dropped to roughly the morning level.
The per trade result barely moved, so the large losers were just noise around the releases rather than part of the edge.
The catch is that removing windows after the fact is easy to overfit. I'd only trust it if the same cut works on a year of data you didn't look at while deciding.
When you split the overlap result, did the losers sit right after the release or a few minutes later on the second move?