Specific hand results for AK -- how many hands do I need for accuracy?

[Shoe]

I just got poker tracker, and although I am a net winner, I am a small loser with the AKs and AKo, for the limited history I have (about 4,500 hands).

I have played AKs 18 times and lost an average of .21 BB's per hand.

I have played AKo 54 times and lost an average of .13 BB's per hand.

My question is, is this sample size big enough to determine that I am a net loser with AK and need to make changes? I will be reviewing my play as it looks like an apparent leak to me, but if not, how many times would I need to play a particular hand to feel comfortable with the accuracy of the results?

Also, how else do you guys identify leaks in your game and fix them?

Thanks.

[hotquietday]

That sample size is almost certainly not big enough to be statistically significant.

What you should do is calculate the standard deviation of the win rate for just that hand. Use that to estimate the standard error. Then you can get a confidence interval for your win rate with AK. I'm sure that, with a sample size this small, it will still contain positive values, which means that you may be winning with it but getting unlucky in this sample.

Of course, that is NOT to say that AK is not a leak for you. It may be. It is simply to say that you probably don't have enough evidence to know one way or the other.

HTH,
John