Open Side Menu Go to the Top
Register
Getting AK vs AA n times in m hands Getting AK vs AA n times in m hands

10-20-2022 , 03:30 PM
I felt I was getting extremely unlucky getting AK in vs AA 5 times in the last 28 times I got dealt AK (suited and unsuited).
Wanted to check what the probability is this happens to see how unlucky this is.

I came to the following computations:

(n choose m) * P(getting dealt AK | 1 out of 5 vilains has AA)

Getting AK while a single other player has AA has a probability of 1/408.
We are at a 6 max table so this is 5/408.

Filling in the numbers we get

98280 * (5/408) = 0.002%

Is the correct number?
My math is a bit rusty so not entirely sure.
Getting AK vs AA n times in m hands Quote
10-20-2022 , 05:52 PM
The probability of getting AK v AA 5 times in the next 28 times you are dealt AK shouldn't be the question. Unless you are going to stop playing poker after the next 28 times you get AK, then it would be ok.
Getting AK vs AA n times in m hands Quote
10-20-2022 , 09:09 PM
Quote:
Originally Posted by 6betpot
AK in vs AA 5 times in the last 28 times I got dealt AK

(n choose m) * P(getting dealt AK | 1 out of 5 vilains has AA)
You meant to say P(someone has AA | you have AK) and that's what you calculated. But you can't just multiply by 28C5; you also need to raise the probability to the power of 5 and multiply by its complement to the 23rd. That would be the chance of it happening exactly 5 times, but really you need the chance of it happening at least 5 times.

P(one of 5 villains has AA | hero AK) = 5*3 / 50C2 = 3/245
P(>4 occurrences in 28 deals of AK) = 1 - binomialCDF(4; 28, 3/245) ≈ 1 in 46760, which coincidentally matches your answer of 0.002%.

It matched because you made a right out of two wrongs:
Quote:
98280 * (5/408) = 0.002%
98280*5/408 = 120441%
Getting AK vs AA n times in m hands Quote
10-21-2022 , 04:16 AM
Quote:
The probability of getting AK v AA 5 times in the next 28 times you are dealt AK shouldn't be the question. Unless you are going to stop playing poker after the next 28 times you get AK, then it would be ok.
I dont realy understand this? I'm just looking at how unlikely this was, as a measure of luck.

Quote:
98280*5/408 = 120441%
Yes.. I didnt wrote down well what I actually did in Excel.

I actually calculated

(((5/408)^5)*(1-(5/408))^23)*98280

Which i thought would be the probability that it exactly happens 5 times.

And indeed I was actually looking for this happening at least 5 times. Thanks!
Getting AK vs AA n times in m hands Quote
10-21-2022 , 09:26 AM
Quote:
Originally Posted by 6betpot
I dont realy understand this? I'm just looking at how unlikely this was, as a measure of luck.
When flipping a fair coin, the chance of getting heads on your next 5 flips is very low. But, if you flipped that same coin a million times it would be very strange indeed if you never got 5 heads in a row.

My comment was my way of stating that there is a better way to think about luck* than what most people ask about here. Yes, you were unlucky in your recent string of getting AK. But maybe you have been really lucky that this is the first time it has happened to you.




*Luck isn't something innate to someone. It is a description of past events.
Getting AK vs AA n times in m hands Quote
10-21-2022 , 12:02 PM
Yes that makes sense.

In the long run the more hands you play the more likely such a streak will happen and as the number of hands you play approaches infinity, the probability this happens in some 28 hand sample of getting dealt AK approaches 100%.
Getting AK vs AA n times in m hands Quote
10-21-2022 , 01:31 PM
Right. So the more interesting question, at least to me, is - If I play 100,000 hands a year, how likely is it that I will have at least one string like this?
Getting AK vs AA n times in m hands Quote
10-22-2022 , 03:48 PM
Yeah that's a tough one to calculate, but easy to simulate:
Code:
function aaak(p::Float64, hands::Int64, span::Int64, coolers::Int64, sims::Int64)
   successes = 0
   for s=1:sims
      result = [rand()<p for h=1:hands]
      for lasthand=span:hands
         if count(==(1), result[(1+lasthand-span):lasthand])>=coolers
            successes += 1
            break
         end
      end
   end
   return successes/sims
end
In 100 hands it's about 1 in 3700 to happen in one of the 28-hand contiguous subsets.
In 1000 hands it's about 0.34%
In 10k hands it's about 3.4%
In 100k hands it's about about 30%
Getting AK vs AA n times in m hands Quote
10-22-2022 , 04:59 PM
Is that 100 hands? 100 hands when you have AK? or 100 hands where you have AK and face AA?
Getting AK vs AA n times in m hands Quote
10-23-2022 , 12:20 PM
Oops sorry, those are hands where Hero has AK because I plugged in 3/245 for the p parameter. If we wanna talk about actual hands dealt, we can't simply use 1/6768 for p because the AK's don't need to happen in a span of 28 hands, it's just the AA's that need to happen in a span of 28 AK's. We can get an approximation by plugging in the average number of AK's. With 100k hands that's 1207, resulting in a probability of .004%. The more precise solution would be to take the sum of
P(AK=x)*P(success | AK=x) from like x=1060 to 1340 (given that the standard deviation is 34.5 AK's and the distribution is right-skewed).

I've greatly sped up the code, which is Julia btw:
Code:
function aaak(p::Float64, AK::Int64, span::Int64, coolers::Int64, sims::Int64)
   successes = 0
   for s=1:sims
      result = [rand()<p for h=1:span]
      c = count(==(1), result)
      if c >= coolers
         successes += 1
      else
         for h=span:(AK-1)
            firsthand = h%span + 1
            c = c - result[firsthand] + (result[firsthand] = (rand()<p))
            if c >= coolers
               successes += 1
               break
            end
         end
      end
   end
   return successes/sims
end
So for instance I typed aaak(3/245, 1207, 28, 5, 1000000)
Getting AK vs AA n times in m hands Quote
10-23-2022 , 01:14 PM
Quote:
Originally Posted by heehaww
.004%
Doh I mean 0.4%
Getting AK vs AA n times in m hands Quote

      
m