|
Originally Posted by JKDS
http://en.wikipedia.org/wiki/Binomial_theorem
???
you mean bayes theorem?
Yeah, a lot of folks use Bayes Theorem here and treat this as conditional probability, but we can make a simple assumption and just compute the probability directly. The Binomial Theorem works for repeatedly "coin flip" situations, whether the coin is fair or not.
Suppose we flip a coin with a 20% probability of heads. Let x = .2 (probability of success) and y = 1 - .2 = .8. Now, this may seem trivial at first, but note that x + y = 1. So (x+y)^5 = 1 for instance. This happy fact allows us to simply expand the binomial power and calculate the probabilities we want. If you remember Pascal's Triangle in high school algebra, it's the simplified version of the Binomial Theorem. And the process works for any probability simulation we can create with fixed chance of success.
So, imagine a TAGG opening in the CO with a known chance of success (open-raising) of x = .2. And imagine that we see him in this situation 5 times (after learning the iron-clad fact he open-raises 20% in the CO). What is the probability he open-raises 4 times or more out of 5 trials?
The probability of 5 successes is a straightforward (.2)^5.
The probability of 4 successes is a bit more interesting:
(5 choose 4)(0.2^4)(0.8)
The 4 successes each have probability 0.2, the failure has probability 0.8. But there are 5 ways to choose one slot out of 5 for the failure to occur (or 5 choose 4 ways to choose the 4 slots for the successes to occur):
F S S S S
S F S S S
S S F S S
S S S F S
S S S S F
Recalling Pascal's Triangle:
1
1 1
1 2 1
1 3 3 1
1 4 6 4 1
1 5 10 10 5 1
The bottom row allows us to expand the binomial (x+y)^5 by telling us the coefficients in the pattern where powers of x ascend from 0 to 5 and powers of y descend from 5 to 0:
(x + y)^5 = y^5 + 5 x y^4 + 10 x^2 y^3 + 10 x^3 y^2 + 5 x^4 y + x^5.
Again, recall that this value is just = 1, since x + y = 1. In the expansion, the pattern relates to the probabilities:
y^5 term - probability of 0 successes, 5 failures
x y^4 term - probability of 1 success, 4 failures
x^2 y^3 term - probability of 2 successes, 3 failures, and so on
Example: For a fixed chance of success x = .2, what is the probability of 5 successes in 8 trials?
P(X = 5) = (8 choose 5)(0.2^5)(0.8^3)
The "8 choose 5" simply calculates all the possible ways to permute the 5 successes into 8 slots.
Awright, given all that introduction, the setup for my post is simple.
Hypothesis: let's assume we have a TAGG reg who plays <25/15 over some reasonable sample of hands. We can think of his PFR = .15 as a chance of success, so x = .15. Then the probability of this event occurring 15 times OR MORE out of 50 trials is the sum:
P(X >= 15) = (50 choose 15) (.15^15)(.85^35) + (50 choose 16) (.15^16)(.85^34) + (50 choose 17) (.15^17)(.85^33) + ...
I did the sum in an Excel spreadsheet. There's an extension of the binomial theorem called the multinomial theorem which would allow us to take VPiP into account simultaneously with PFR, which I may get into later if you're still interested after all that.
Just to check the viability of this math, you can see the same calculations for other probabilities.
For x = .2, there is about a 6% of him open-raising 15 times (or more) out 50.
For x = .25, there is about a 25% of him open-raising 15 times (or more) out 50.
For x = .3, there is about a 55% of him open-raising 15 times (or more) out 50.
For x = .4, there is about a 94% of him open-raising 15 times (or more) out 50.
For x = .5, there is more than a 99% of him open-raising 15 times (or more) out 50.
Let me know what you think. If you have more questions, I can try to answer them.
|