Quote:
Originally Posted by poker2brain
@Mike
I have PT4 and I am able to run tests any way I want. Even though the distribution of cards seem to look fair, running a deeper analysis shows some odd things. I am a winning player when it comes to Allin situations, however some values just don't make any sense to me. Most players I believe are not fooled by their selective memory. They actually do loose their monster hands far more often then they should. However, we can just look at Allin situations to get valid data.
For my first test you might say "variance"! Well, maybe variance is a key factor when it comes to winning and loosing players online. Maybe variance will adjust based on the strength of a player. But why does variance target monster hands? How do you explain, that there are far more bad beats at tournaments compared to cash for 1 specific player name?
Test I.
Headsup Allin Any Street - Pair vs 2 undercards
Tracked: 1363
Won: 933
Split: 5
EquityExpected: 73,65%
EquityOccured: 68,64%
Allin Any Street - KK
Tracked: 645
Won: 359
Split: 22
EquityExpected: 59,64%
EquityOccured: 57,36%
Headsup Allin Any Street - KK vs AX or weaker Pair
Tracked: 429
Won: 303
Split: 7
EquityExpected: 72,73%
EquityOccured: 71,45%
The first data point you chose (that is an important point btw - you chose it because it is far off and does support your theory) seems pretty far off, I give you that. The two others are pretty close, or do I miss something? Certainly in no case something you could correctly recollect from memory, to address your point about riggies not having cognitive biases. Everyone has them.
Also, if you do an analysis, literally no data point will be exactly as expected. Some will be close, some below, some above expectation. A few will be pretty far off - that is exactly what someone analyzing a random sample would expect. To single out the data points that are far off in the desired direction makes no sense at all if you want a significant result.
To do a proper analysis, go ahead and formulate a work hypothesis like "My equity with big pairs all in preflop will be significantly smaller than expected the next 1000 instances", then play (collect data) and then analyse it. Looking at your data after collection, searching for the biggest deviations and singling them out is far from something a mathematician would ever call an analysis. Something you can find by definition in any random sample (single data points pretty far off) is nothing you can use to show it is not a random sample, you agree?
Quote:
Originally Posted by poker2brain
Test II.
Assuming that a Bad Beat is when Player A moves Allin and Player B calls Allin at an amount that he shouldn't, based on Equity*(Potsize+Bet). Player B makes a bad call and wins, Player A looses even though he was ahead of all the other players.
Bad Beats Cash Tables (5606 Allins):
12,18% received 5,22% send = 17,40%
Bad Beats Tournaments Tables (7508 Allins):
15,05% received 9,56% send = 24,61%
If I get that right you aren't even comparing expected with occurring equity here - you show a difference in the way players play between cash and tournament poker.
An example would be pushing a draw on the flop when you don't have the odds to just call and want to exploit the relatively huge fold equity compared to stacksize in a tournament, when you can just call in a cash game given the bigger implied odds. Getting it in with a draw will generally result in a way closer situation than getting it in with two made hands. That is exactly what one would expect, and exactly what you found.
Additionally, due to shallow stacks and comparibly small stack-to-pot ratio postflop-all-ins are generally closer equitywise in tourneys than in cash games, subsequently it's just logical that the worse hand wins more often.
You agree?
In conclusion, your "deeper analysis" is not deep at all, let alone in any way properly executed, plus you missapplied the concept of variance and compared apples to oranges.
Quote:
Originally Posted by poker2brain
Cash Allin Preflop - KK
Tracked: 238
Won: 123
Split: 8
EquityExpected: 59,66%
EquityOccured: 53,36%
Cash Allin Preflop (Being ahead of all other players)
Tracked: 1106
Won: 812
Split: 25
EquityExpected: 77,22%
EquityOccured: 74,59%
Cash + Tournament Allin Preflop (Being ahead of all other players)
Tracked: 5527
Won: 4113
Split: 203
EquityExpected: 77,27%
EquityOccured: 76,26%
The bigger your sample size gets, the closer your results are at expectation, which is exactly what one would expect given a random deal, and exactly what would not happen given a rig.
You wanted to show evidence that the deal is indeed fair, I assume? Well done, if so.
Last edited by franxic; 05-09-2014 at 06:12 PM.
Reason: 2 posts merged