I've been tracking my roll in results over several thousand rollins because it seemed like I was constantly very unlucky.

The results are:

With one open - 22.5% success
With two open - 47.4% successs
With three open - 46.6% success
With four open - 63.9% success
With five open - 86.4% success

I'm not posting to bemoan my clear underperformance in rolling in, just trying to understand the results. If this was an issue of it being a relatively small sample size, wouldn't it be more random? All five rollins are significantly under what they should be long term.

Do I have the "normal" rollin percentages wrong? I thought they should be 31%, 56%, 75%, 89% and 97%.

This is depressing.

Could it perhaps have to do with how you're accounting for the cases where you have multiple checkers on the bar? If for some of your logged entries you had two checkers on the bar, you'd expect you success rates for re-entering both checkers to be somewhat lower, and if you don't track these separately you might get confused.

This numbers are so far off what would be expected that you must either be neglecting to count some instances where you entered from the bar or you're counting something incorrectly. The odds against a distribution like this after thousands of instances is literally tens of million to one.