Sample Size with Expert Observation

[RocketManJames]

I know we've all talked about sample size and how 50K hands is just approaching where we can start making some judgements about win rates and so forth.

My question is... what sample size would be required if we had an expert observe the plays? Say Sklansky were to watch someone play 2500 hands. Would Sklansky be able to estimate how strong the player was, and therefore come up with a win rate that would be close to the true win rate as measured by a decent sample size?

Or am I being dumb here?

-RMJ

[Mason Malmuth]

Hi Rocketman:

Depending on the player, 25, not 2,500, hands might be enough.

Best wishes,
Mason

[Duke]

In the same vein as Mason, I think that if you only saw a few hands, and talked about them with the guy, you'd at least know if they're capable of winning.

~D

[Girchuck]

Does that mean, that if I detect my opponents make many mistakes, while I only make few mistakes, I can estimate my winrate on this table to be positive?

[dogmeat]

That depends on how many mistakes you actually do make, and whether there are some other good players at your table. If every player was worse than you, you could eventually make a very good judgement of what you should make if you played against the same lineup everyday for a year.

Dogmeat [img]/images/graemlins/spade.gif[/img]

[RocketManJames]

[ QUOTE ]
Depending on the player, 25, not 2,500, hands might be enough.

[/ QUOTE ]

So, if someone wanted to know their winrate with fairly good accuracy... wouldn't it make more sense to have an expert or near-expert evaluate his play for a few hours? Seems easy enough to do online. Sure beats having to play 50K hands to get an idea. And, maybe finding out after the 50K hands that you're a long-run loser.

-RMJ

[chris_a]

I think i read Sklanskdog himself write that for bad players you can assess it very quickly. You can tell how bad they are quickly, but it takes much longer to tell how good a good player is. This makes perfect sense since bad players are goign to be playing too many hands and you'll be able to tell quickly.

You can usually do it in a few hands if they are really bad but it might take as many as 30 or 40 or more if they are good to tell just how good. This makes sense since good and great players will be playing a similarly small number of hands.

[flair1239]

I think a good for instance, is if you are playing LHE and see the same player CC an early position raise twice within a orbit or so; even without seeing a showdown you are probably safe to assume he has many other basic deficiencies.

I agree for a tighter player it might be harder to evaluate, because IMO some of the most obvious signs of a bad player, take place before the flop (which is not to say that they are not making worse mistakes post-flop, it is just those are a little harder to spot).