Quote:
Originally Posted by tiltymcfish0
I'm talking about ALL sites - can anyone produce HEM all in luck graph from ANY site over at least a 150k sample where they run above average?
This would work fine if you knew all the folded hole cards and if HEM figured those dead cards into the luck graph. Then it would be a true test of whether a site is fair.
The trouble is we don't know the folded hole cards. That means we have to guess something about them. All the luck graph calculators, like HEM, assume that the folded cards are uniformly random. That makes the deck stub uniformly random too. This is the basis for the luck calculation.
But it's founded on a false assumption. The folded cards are NOT random. People fold aces less than deuces. So we shouldn't expect the luck graph to converge to the same thing it would with a random deck.
There is still the interesting pattern that winning 2+2 posters tend to underperform their EV computed from a random deck. That means the villains are overperforming their EV. It's not clear this is actually true, since 2+2 posters who underperform are more likely to mention it. But if it is true, then here is the hypothesis:
Hypothesis: Villains tend to play different cards preflop than heroes; specifically, they tend to hold cards that are over-represented in the deck stub -- like an ace -- when a hero and a villain choose to go all-in together. Thus, they tend to hit their outs more often than they would with a random deck stub.
-pyg