wiki - Did you even look at the graph stuff?
Quote:
but I find the fact that the first player has an almost inverted set of stats compared to the other two a bit of a stretch.
you "finding the fact" unlikely has nothing to do with it. The type of variance you are looking at is not at all unlikely for the exact same player with the exact same win-rate and SD over that length of sample. Period.
Quote:
it happens that distinctly for three separate chunks of data over a period of six months and nearly a million hands.
Now I don't even know what you're saying. What is happening "that distinctly?"
Quote:
I wouldn't call those differences slight.
Except that the differences pretty much are. Again, you seem to be underestimating how much variance can be a factor for the exact same type of player.
If you go to the graph and just plug in for 3 or 4 samples (or 10 samples to get a slightly better idea) with win-rate of 0.5BB/100 and SD of 30 over 200k hands or something then there's a reasonable chance you will end up with variance EXACTLY as shown in the PTR winrates.
Quote:
Particularly if a robot is involved as that should be more consistent than a human player
You brought this up before and it's still not really relevant and you are overestimating the consistency of a robot over a human player. The EV-graph simulator stuff is even MORE consistent than the robot. Because the robot will be playing in a bunch of different games some of which he is at a greater advantage than others. The EV simulator is EXACTLY whatever you want it to be (0.5BB/100 with 30BB SD or whatever) as if it were playing on the exact same table against the exact same opponents ALL the time.
So if you want the ultimate consistency you should dismiss the bot at the live tables with different players making different adjustments to him and instead go with the EV graph simulator. And the results for that simulator are all over the place for every 100k or 200k hand sample you put in there.
Yes, they might or might not be bots. But it's clear the winrate does not get us any closer or further away from being able to draw such a conclusion either way. You "thinking" that a bot would have more consistent results over those smalish sample-sizes if it were the same bot is an opinion that doesn't really matter here. You are just guessing.
There are ways to actually scientifically measure whether an even more consistent player than a bot would yield the kind of consistent results you envision. And the science has proven your belief to be wrong.