Hi 2p2,

I've asked a few people on this, but have not come to anything concrete for or against. I'm hoping somebody with more expertise can help shed some light.

Example:
WR: 5bb/100
Std/Dev: 100
# of hands: 1
Observed winrate: -3kbb/100

The above should demonstrate a 5bb winner losing 30bb in a single hand. According to POKERDOPE, this should only happen 0.1328%, or 1 in 753 times.

Using the same method, the likelihood for losing 40bb in a single hand is 1/32258, and 50bb returns a 0% output.

How can this tool properly evaluate Poker variance, against the above outputs?

Thank you for taking the time.

For a single hand your std deviation would be waaaay higher than 100.

The common variance calculators are OK approximations.

Sent from my SM-G996B using Tapatalk

The calculations that these variance simulators/calculators use assume a normal distribution. By the Central Limit Theorem over a large number of hands the distribution of results will approach a normal distribution, but over a very small sample (like one hand here) the distribution of results is NOT normal and calculations can be waaaay off like this.

These tools should be accurate for larger samples. Not sure how large exactly but probably even like 10 hours live would allow for the calculator to make a reasonable model.

Note that these tools can tell you how likely you are to experience certain results given the winrate/SD you provide are accurate, but they don't work the other way around. You can't reliably determine how close your observed results should be to your true win-rate with these tools.

My take was similar to this, minus the depth of understanding you have. Thanks.

For online games, especially HU online games, short samples, particularly those of 30k hands or less, should be quite compromised. After all, these calculators do not factor in preflop setups...


P.s.

Thank you to the mod who moved the thread to this forum.

Seems accurate to me compared to the real results I have seen. Will be using that as a part of my therapy -- after I hit the 1 in 20 longer run, both ways, lol. Had a good time seeing it in these graphs, and looked further and there too liked what I saw. Looked all realistic to me.

1bb/100 with 8.5bb/100 variance is next to hopeless, like 55% to win in 20k hands, 1/3rd or so? in 200k hands (looks good here, but it isn't that easy). After 1M hands, the graph was fine but was more like all luck as far as I saw as it was at the same spot at 300k hands and 900k hands, and then luck boxed a huge up (after a similar down) the last 100k hands, making the path and the end look good. While 6bb/100 was pretty much all different and I very much agreed with that also.