Statistically valid sample size in 'Holdem

In No Limit Holdem, to be a statistically valid sample size, how many rounds of pocket cards have to be dealt before every possible combination of starting
hand has had a decent chance of occurring??? The reason I was wondering, is because I thought that it might be relevant to observe a certain # of showdowns of online players that I play against in the same type of games (stakes/table size) that I play. It might give me a better look at what type of starting range to put them on, or at least, which starting hands they're comfortable enough to showdown under the right circumstances. I did something similar to this when I first started learning the ins and outs of 'Holdem. I believe I only observed about 300 showdowns before I stopped. That's likely too small a sample size, but I was never quite sure of my math. I'd appreciate any insight on this - Q7x

There are 2652 possible starting hands. So 350 is definitely too small. Really, you'd need to watch each player play each hand multiple times. But you probably don't want to do that - it'd mean spending days and days watching other people playing poker when you could be in on the action yourself!

deleted

Quote: Ibeatyouraces

1326 possible starting hands.

Why is this not 52*51?

deleted

Quote: rdw4potus

Why is this not 52*51?

You have it so order matters
2,7
7,2
does not matter. You have rags.
52*51/2*1

For the OP
1326 coupons in the coupon collector problem
I linked to BruceZ at 2+2 solution before in R
http://forumserver.twoplustwo.com/showpost.php?p=36694761&postcount=71

from the picture, looks like 18000 would be a very high confidence level
or a coin flip by 10000

Yeah, 18,000 looks like a fair amount of hands to observe to get a decent picture of what people are holding preflop and also showing down. It's been about a year since I did the last observations, but I'm starting to remember, roundabout, what I did in more detail. I think I started with an original number about the size of 18,000 and then thought that I could divide it by ten since I would observe full ring tables with 10 players. Then, the number 1,800 was still quite large so I think I divided it in half to dismiss all the lower/trash hands that would occassionally win, but I'd not want to play them anyway (72o, etc). Lucky for me, after around 300 observed showdowns on full ring tables, every additional observance would be redundant. This happened for about another 50 observances before I just settled on what I had observed as a roundabout, typical player range. I then PokerStoved the equity of 169 starting hands (me) vs the ranges (observed hands) of 1, 2, and 3 players. I stopped vs 3 players 'cause, on average, that's how many players showed down over a couple hundred rounds. I ranked the hands in order of %equity and grouped them into "fold", "call", "raise", and "all-in". It's a handy little chart that I still use occassionally, but I don't need to look at it much anymore. Once I started to remember the %equity figures, I rarely referred to it. Then, as I got a little better, I started varying my playing of those hands also on my position, players' actions before me, stack sizes, etc. Now that I've had time to think more on the problem, it would be way too time consuming for me to do something like that for Heads Up SnG's. Thanks again for everyones input on this subject. I think I'm going to just keep searching for some good beginner videos on the associated forums.