Quote:
Originally Posted by Klyka
I started calculating the answer to question 1, but, as BruceZ pointed out, I have some trouble with double counting, so I gave it up. I hope for him or someone else to solve that question.
However, question 1 can be further detailed into including or excluding the instances where two players are dealt the same pocket pair, in which cases the probability of all three players flopping sets equals 0, since all cards of that rank are out of the deck and therefore cannot flop. Therefore, I think we better exclude those instances.
Also, it can be further detailed into including or excluding the instances where more than three players are dealt pocket pairs, in which instances the probability that three players flop sets is bigger. For the sake of simplicity, I think we should exclude those cases too.
So, I'd like to specify question 1 into "what is the probability that exactly three players out of eight are dealt different-ranking pocket pairs", and I hope for someone else to show me how to calculate that probability without falling into the traps of double counting.
It's quite laborious to calculate this probability
exactly, i.e.,
"the probability that exactly three players out of eight are
dealt different-ranking pocket pairs", but as BruceZ would
point out, three or four decimal places of precision should
suffice.
One idea is to compute the probability that three players each
have different pocket pairs and estimate the probability that
the other five players don't have a pocket pair and this is the
idea that I'll be using. If the game were only three-handed,
the probability of these players each being dealt a unique pair
is then
[(13C(4,2))/C(52,2)][(12C(4,2))/C(50,2)][(11C(4,2))/C(48,2)]
= (1/17)(72/1 225)(11/188) = 198/978 775 ~
0.0002022936834.
This is the same probability that the first three players of an
eight-handed game will be dealt a pocket pair. Now, what is
the probability that the fourth player will not be dealt a pair?
[C(10,2)x4x4+10x4x6+C(3,2)x2x2]/C(46,2) = 972/ 1035 =
108/115 ~ 0.9391304348.
Now one could simply use this number and raise it to the fifth
power for approximating the probability that the other five
players don't have a pocket pair. Of course, it's only an
estimate because the probabilities won't be independent.
Another problem is that this estimate may be a little low since
if the fourth player doesn't have a pocket pair, the fifth
player is more likely (compared to the fourth player) not to
have a pocket pair.
It turns out that for the fifth player not to be dealt a pocket
pair given exactly three of the first four players have been
dealt a pair will be on average, slightly higher. There are
three cases for the nonpair of the fourth player and the
probability that the fifth player doesn't have a pair will vary
depending on the type of nonpair the fourth player has:
a) C(10,2)x4x4 = 720 combinations of a nonpair hand with
ranks different from the pocket pairs. For the fifth player not
to have a pocket pair, of the C(44,2)= 946 combinations he
could have, he has no pair on
C(8,2)x4x4+8x4x12+3x3+6x6+C(3,2)x2x2 = 889
of them or a probability of 889/946.
b) 10x4x6 = 240 combinations of a nonpair hand with one
rank different from the pocket pairs and one rank which is the
same as a rank of a pocket pair. There are now
C(9,2)x4x4+9x4x8+3x5+2x2+1x4 = 887
combinations for the fifth hand to be a nonpair.
c) C(3,2)x2x2 = 12 combinations of nonpairs with both ranks
of the hand being among the ranks of the pairs. There are now
C(10,2)x4x4+10x4x4+2x2+1x1 = 885
combinations for the fifth hand to be a nonpair.
On average, the probability that the fifth hand will not be a pair
is then
[(720/972)(889)+(240/972)(887)+(12/972)(885)]/946
~ 0.9391720826.
I'm confident that the probability of a sixth hand being not a
pair will be even higher (given that exactly three out of the
first five hands are unique pocket pairs). To "make an already
long story short", the probability of the first three players
being dealt unique pocket pairs and the other five players
being dealt "nonpairs" could be approximated by:
[198/978 775](0.9392)^5
and then this is multiplied by C(8,3) (since the three "pairs"
can be distributed in so many ways) to get a probability of
about
0.008279. [ Just to get close to four decimal places! ]
Anyone care to check this?