Quote:
Originally Posted by joka
The colors are uniformly distributed but there are simply less cards to chose from. Example: The chance of getting a monochrome board is higher if we play with 13 of each suit than if we play with 3 of each suit and much, much higher than if we play with 2 of each suit (=0 ldo). So, in the example you mention there are less of each suit remaining(!) which decreases the chance of a monochrome board.
(NOTE: too long and skip if don't like technical details)
Yes, but if there are two of each suit remaining, then the probability of a suited flop is zero... I think your example has gone extreme in an impractical direction, and is likely missing the point.
Here's another way to think about the situation. A bias in the flop texture will occur if the distribution of the remaining deck cards becomes polarized (its "entropy" is low). This is a conceptual issue, and is non-trivial. On the other hand, a bias caused by having fewer cards is essentially a technical issue, and is simply due to a loss of accuracy when attempting to approximate subsequent distributions of the deck as community cards are dealt by the deck's initial distribution (which by the way is the popular way to mentally calculate odds.)
In any case, suppose we forgo the above notes, and instead, delve further into the issue, and quantify it via examples/calculations.
Generally speaking, bias in the color of the flop caused by playing paired holdings happens on a significantly smaller order of magnitude than the one due to playing suited cards, which is why we can expect it will almost surely be absorbed by the latter. (of course, it's possible there are other more significant biases that will counter the latter, but this is not our concern here!)
Consider two examples.
Ex1. 6-handed, half the players are dealt hearts, the other half
dealt another spades. Then the probability of a suited flop is 6.49%. By contrast, if we assume no a priori knowledge about players' holdings, then the probability of a suited flop is 5.18%. A difference on this scale (6.49% vs. 5.18%) would be felt after running into about 6000 such hands.
Ex2. 6 handed, all players dealt high pairs. The probability of a
suited flop is 4.86%. As before, without a priori knowledge, the
probability of a suited flop is 5.18%. This difference would
be felt after running into about 100,000 such hands.
Note in Ex1., the probability of half the players being dealt a single
suit and the other half being dealt another is about 9*10^(-8) %, and
in Ex2. the probability that all players are dealt pairs is about
3*10^(-8)%. So, the scenario in Ex1. is about 3 times more likely,
which makes it easier to observe empirically (it requires going through about 50 times as many hands to observe the bias in Ex2.; on the other hand, good players may be more likely to see flops with pairs than with generic suited cards. But there are armies of bad players who like nothing better than play cards simply because they're suited.)
More generally, let P denote the probability of a suited flop given no a priori knowledge about the distribution of the remaining deck cards. Let N be the number of players. Suppose roughly half of the players are dealt one suit, while the other half is dealt another. Then the probability of a suited flop becomes about
0.5*P*(1-2N/52)^(-1)*(1-2N/51)^(-1)*(1-2N/50)^(-1)*(1+(1-N/13)*(1-N/12)*(1-N/11) ~~ P*(1-0.0079*N+0.0049*N^2+0.0004*N^3)
(e.g. in Ex1. above, N=6, P=5.18%, so according this calculation the new probability is ~ 6.47%, which is pretty close to 6.49% from before)
On the other hand, given N players, suppose each two players are dealt pairs of the same rank. Then the probability of a suited flop becomes about
P*(1-2N/52)^(-1)*(1-2N/51)^(-1)*(1-2N/50)^(-1)*(1-N/26)*(1-N/24)*(1-N/22) ~~ P*(1-0.0079*N-.0003*N^2-0.00001*N^3)
(e.g. in Ex2. above, N=6, P=5.18%, so according this calculation the new probability is ~ 4.87%, which is pretty close to 4.86% from before)
The thing to note are the coefficients of N^2 and N^3 in both equations, which are clearly much larger in the first equation (about 17 times larger for N^2 and 40 times for N^3). Remember, the first equation corresponds to the suit bias case). This suggests bias in the flop color due to playing paired holdings will be absorbed in the one due to suited holdings (if both occur and induce flops with relatively comparable frequencies, which are plausible assumptions as explained before).
Here is another perspective on the situation. Imagine a generalized deck with R ranks and S suits. So the total number of cards, call it M, satisfies M=R*S. Also assume R~S and both are large (so as to eliminate degenerative cases; e.g. R=1 S=10^10 is not interesting!). Let P denote the probability of a suited flop given no a priori knowledge about players' holdings. Then, given N players, if all of them are dealt the same suit to exhaustion, the chance of a suited flop is approximately
P*(1+1/R)
On the other hand, if all players are dealt the same rank to exhaustion, then the chance of a suited flop becomes approximately
P*(1-1/R^2)
So, the pair bias will decline like 1/R whereas the suit bias will decline like 1/R^2. The latter is a completely different (much faster) order of decline.
In any case, the basic point is this: bias in the color of the flop due to playing paired cards happens on a small enough scale that it'll be absorbed by the bias resulting from playing suited cards! In other words, deviations caused by the former will be difficult to measure empirically. This could've been anticipated from the start as pair bias arises for purely technical reasons.
p.s. damn, I didn't intend for this to get that long. I have other **** to do.