Making reads based on minimal information

[Sherman]

The inspiration for this post comes from this thread link .

A common statement at the beginning of many hands posted in this very forum goes something like this, "Just got moved to this table an orbit ago, so no reads." In online tournaments, we often move tables so quickly that we have little time to get good reads on our opponents. The objective of this post is to demonstrate that it doesn't take many hands to get reads on opponents (especially in small buy-in MTTs) and that we can actually learn more than we think we can in less than 1 orbit.

In the thread linked above, another poster indicated that he didn't think that a player folding six hands in a row and then CC an EP raiser in MP doesn't mean much. By this, the poster implied that MPs hand range is still very wide including weak aces and all pocket pairs.

I disagree.

To get a better estimate of his calling range, let's use probability and what we already know. We already know that this particular player has folded 6 consecutive hands. Presumably, he was not sitting out and actively clicked fold on each of these hands.

Now let's use some probability to answer a few questions. In general, I tend to categorize my opponents as extremely tight (under 10.00 VPIP), tight (10-15 VPIP), middle (15-20 VPIP), somewhat loose (20-25 VPIP), loose (25-30 VPIP), very loose (30+ VPIP). These guidelines are not necessarily rigid, but I think they are good general categories.

In any case, given that this particular player has folded 6 hands in a row, what is the probability that this person is a tight player (10 VPIP). First, recognize that a 10 VPIP means that on average this person plays 10% of hands, meaning on average he folds 90% of hands. Thus, we simply take .90^6 and get .53. Thus, even though we have only seen 6 hands, we know that there is a .53 probability that this player has a True VPIP of 10 (that he plays 10% of hands). Now what is the probability that this player is somewhat loose (20 VPIP). Again take .80^6 (1-.20 = .80) and we get .26. Thus there is a 26% chance that this player (who has folded 6 hands in a row) has a true VPIP of 20. Taking it a step further, what is the probability that his true VPIP is 30 given that he has folded 6 hands?

.70^6 = .12

So, as you can see by doing a little math, we can realize that in just 6 hands, there are some pretty large differences in the probabilities that our opponent is loose or tight. In fact, according to my rough guidelines, this villain is 4.5 times more likely to be "tight" (vpip of 10) player than a "loose" (vpip of 30) player.

As a disclaimer, yes I realize that I don't know anything about position and if he adjust his play based on position etc. And yes these estimates are very very rough, however, I am very convinced that there is some truth to these numbers. The fact is, we can learn a lot from very little information from our opponents.

I have a meeting now, so this is somewhat incomplete, but hopefully you get the idea.

Sherman

[4CardStraight]

I generally agree with what you are saying, IE that VPIP and PFR quickly converge, however the math you are doing isn't correct.

When you take .90 ^ 6 , what you are saying is that a vpip10 player 53% of the time would not get a hand as good as top 10% across 6 hands. That is not the same as saying he is 53% likely to be a VPIP 10 player.

A different, but perhaps somewhat less bad approximation is to say that we can think conclusively that this must be about the best of seven hands he has seen. Does that mean it is a top 14% hand ? On average, yes, but given skew distribution of hand ranges through such a small sample size, this could well be as bad as top 25-30%...

I think the point of your post is valid tho, VPIP and PFR tend to converge the fastest of the statistics we gather on our opponents. I think most people agree vpip and pfr can converge even over 100 hands or so, although not perfectly of course... but are somewhat reliable.

I think to say that there is an "X%" chance that this player is Y tight tho, requires advanced statistical techniques that you are not using....

[mflip]

http://forumserver.twoplustwo.com/showfl...rt=all&vc=1

Great post from uNL on Bayes Theorem and this concept of small sample sizes.

[Sherman]

I'm not saying anything about the probability of him picking up playable hands in the top 10%.

I'm saying if he truly on average plays 10% of hands, what is the probability that he folds 6 hands in a row if his true VPIP is 10%? That is .90^6. It has nothing to do with what hands he gets dealt...I really don't care about that.

I really care about the reverse though, given that someone has folded X amount of hands, what is the probability that he is tight, loose, etc. Maybe I did do the math wrong, but it has nothing to do with being dealt particular cards. I don't care about that.

[alavet]

maybe man folded 6 times a row just in a ran of dead cards
who knows?
too low info to say about
ESPECIALLY in tournaments where people not going to call with ATo
i think average and _good_ vpip in the tournaments on the early\mid stages is about 17%

[Kelsey72]

[ QUOTE ]
http://forumserver.twoplustwo.com/showfl...rt=all&vc=1

Great post from uNL on Bayes Theorem and this concept of small sample sizes.

[/ QUOTE ]

Was going to link same. Great post [img]/images/graemlins/laugh.gif[/img]

[Jerro]

I like the post Sherman but I think that this is just a complicated way of stating what should be common sense...i.e. if he folds six hands in a row he is more likely to be tight than loose...
I don't think you can read into the exact figures i.e. 4.5 xs more likely etc without using regression and the sample size is just too small to get anything accurate...

Like everything else at the pokertable our decisions are battles of imperfect information....
It helps to know he is more likely to be tight but we must also be aware that he could easily be loose too...
Nice post...

[seke2]

Sherman, this is a very good post.

I was thinking about posting something just like this yesterday when someone posted asking how many hands you needed to get a read on someone. My first thought was "1 hand--and refine my read from there."

I think too many people are quick to ignore information if they can place a great deal of confidence in that information. The reality is, even the smallest amounts of information are helpful.

If you think of things always in terms of hand ranges, yeah, seeing this guy fold 6 times doesn't mean much...but maybe it means instead of your range for his preflop raise being any pair, any ace, any 2 broadway, any SC's, you can maybe say like, 55+, A5+, KT+, QJ, 76s+ ... not a huge difference, but now at least you've done a LITTLE bit of adjustment. It may not be right, but in a vacuum where you have imperfect information, it's the best you have.

[JFJB]

Excellent post!!!

I like this one a lot. This is a math demonstration of stuff that has been said before but here is some limited hint of proof.

Well done,