Thanks to Todd for telling me about the IRT homework study #3 which did the same thing, and that let me compare my results with theirs. I had an error in my Excel which made the time get out of sync and that compounded and ended up making the last period H/L more often. Doing these stats in Excel is a bit complicated, but worth the time & effort in my opinion.
I updated my previous post, but here is the updated chart:
This fits better with convential sayings that the first period is most often the H/L. Lately that may not seem like the case, we’ve had many mid-day reversals. But over several years, that is the correct pattern.
The last time period is the second most likely to be high low. And since by the last 90 minutes, we already know if the first few periods are likely to be high or low, we can guess the last period will be the opposite and use that for a directional bias if supported by context.
Another important implication of this is that trying to fade the intraday trend mid-session is not a good idea.