On theories and sample size

[wahooriver]

I have hung around this forum for the past 3-4 months. One common attitude is to discount any theory which one dislikes by claiming insufficient sample size.

If one takes this attitude to its extreme - we never have sufficient sample size to make any strong statements about poker statistics.

Nonetheless, we must each make multiple decisions each time we play poker. We must choose tables, choose starting hands, decide on raising, checking or folding multiple times per hour.

One need not have complete information to inform decision making. The science of decision making allows for making decisions under conditions of uncertainty. What we need are estimates of probabilities of various conditions preflop, postflop turn and river.

We have some data and can collect more data. We should make informed decisions on our own play based on results. Admittedly some results are confusing.

The big question though is how to interpet admittedly incomplete data. I would suggest these observations (like the LAG success at 15/30 thread) are hypothesis generating. We should become scientists and examine these hypotheses - collecting as much data as possible.

If you just listed Gus Hansen's starting hand requirements, would you consider him a World Class Player? One could argue that he is just lucky - but he sure has had a significant string of luck.

Being a LAG may help you when you do have a great hand. We are less likely to believe a LAGs bets - just because he plays crap and bluffs at times. If he can break even on the loose hands and win bigger on the tight hands - then overall he will be a bigger winner.

One can construct a rationale for LAG as well as a rationale for TAG. I am interested in better understanding how to adopt these ideas to the short handed game.

Perhaps I should try several thousand LAG hands playing 1/2 (where everyone else is loose but generally passive). I could learn a great deal about my own abilities to succeed postflop.

There are probably several successful strategies for SSHE (Harrington's book clearly states that for tournament play). I believe that we should respect that diversity and figure out which style works best for us.

End of rambling rant

[stigmata]

my rambling reply....

I have had similar thoughts. Looking at the SD of my play was really scary. I have 25k hands at 5t6m and I'm doing well, but in reality I could just be a break even player.

So where does that put me? Am I over-evaluating my own skills? It is all terribly self-doubting when you start looking at the true nature of the stats.

However, you have to have some confidence in your judgement. I know I make sound plays a good percentage of the time (not nearly enough), and it is this fact that convinces me I am a winning player - not the statistics.

Similarly, I think we have to exercise some judgement when evaluating the statistics. Hence everyone not believeing the LAG data.

On the other hand, I do think you need a certain amount of LAGiness to succed at the 6 max tables. I honestly believe that LAGs do significantly better than over-tight passives at the 6 max tables. But that is again based on judgement - e.g. you need lower starting hand requirements, you are more likely to pick up pots through raw aggression and get paid off when you actually have a hand. After all, using standard rating scores, all of the regulars on this forum are technically LAGs.

Anyhow.... not sure where my rant is going either /images/graemlins/laugh.gif

I think you need to exercise judgement about the nature of the game. After all, statistics can be used to prove anything if you interpret them badly.

[Grisgra]

Big difference between loose play in NLHE with deep stacks and loose play in limit HE. Nonetheless, I agree with your general point, at least at the higher limits where your opponents are tricky and trying to put you on hands and being LAGgy adds deception to your game that you just don't need at 1/2.

Maybe . . . I've just started the higher limits myself. I'll be keeping an open mind -- but I have a very hard time believing that being a TAG is going to cost me money.

[k_squared]

Another thing to consider about these statistical analysis of hands over a period of time is that both the games change and the players change such that their is no good way to quantify how successful and individual style of play is.

The fact is that our data includes a huge number of different situations, and even 'players.' I might have 40,000 hands in my database, but I have changed as a player, and so it is a mistake to think that the information shown from that database relates directly to the type of player I am today. I would even wager it would be nearly impossible to ever gather a set of data large enough to derive meaningful information, without the subject's play changing over the course of that time.

And that still doesn't account for the real life issue of the fact that the game changes, which means HOW TO PLAY IT CORRECTLY CHANGES, and so that skews how the data should be interpreted. If the game moves from very passive to very aggressive over a matter of months then your style of play should change.

But, suffice it to say the information can be valubaly analyzed inspite of these flaws. But, we must be honest with ourselves that the numbers we typically look at (our winrate, aggression factors...) cannot tell the whole story! They should be in large part a factor of the games we play in.

K-squared

[naphand]

[ QUOTE ]
One common attitude is to discount any theory which one dislikes by claiming insufficient sample size.

[/ QUOTE ]

This is how rational thinking progresses. For a theory to have respect it not only needs data to back it up, but it needs a rationale and provable logic. What made Einstein's Theory of Relativity accepted was not eons of accumulated data, but remarkable insight and an elegant and precise logic that could not be denied. It is possible to prove that time is different at close to the speed of light in about 2 minutes, on a piece of paper the size of a postcard, with only 1 assumption (the speed of light is constant) and high school grade math.

Most "theories" and "ideas" that pop up on this forum are not, in fact, based on logic or insight, but small data samples. The recent "LAG players make more than TAGs" post being a very typical example of this. And what underlies this? The ego of the individual who, rather than face learning a hugely complex game and face issues such as playing tight, big downswings etc. prefers to believe that they have "discovered" something new and unique that the other respected poker minds have missed. While this is possible, in 99% of cases, it is merely the fantasies of an overactive imagination and deluded sense of self-importance.

While ideas and hypotheses should be discussed, when they start with "over the hands I have seen..." or suchlike, it is pretty easy to skip the rest and move on to the next thread unless we are looking a very large sample sizes. A hypotheses is not an idea based on a few observations, it is a complete theoretical structure that necessarily includes prediction of results. If the hypothesis cannot predict results, it cannot be tested and is therefore of very little use. There are any number of "theories" that can be fitted to small samples of data; 20-20 hindsight etc. but just how useful those theories may be in predicting is another matter.

As an example (and I have talked of this before) there are a very few players on the forum who have successfully played a more LAG style of play at $5/$10 and above (Peter_Rus, strip and one or two other names) AND who have been able to present their case and use intelligent argument to quantify their play. There are plenty of players who play loose and have won, but whose posts clearly indicate their grasp of poker theory and game mechanics is weak. Both make the same claims, only a few actually have the ability/intelligence to understand its mechanics and play appropriately.

An issue that constantly crops up is win rate. People talk of not having enough data to know if they are even a winner yet. Well, mathematically you will need a substantial sample size to determine if you are a winning player, and a VERY substantial sample to determine your WR to any accuracy. This is the nature of statistics, and this is not even starting on the issues such as how much you improve as a player over that time, the nature of the games etc. Statistically it is a slow grind. However, it is certainly possible to determine if you are a winning player over as little as 20K hands. Just look at your opponents and how they play. With sufficient experience it is easy to classify players and understand if a table is weak or tough, and it is also not difficult to make comparisons between those players and ourselves. This is the basis of table selection, and also how we read a particular player and adjust our play accordingly. It does not require statistical analysis, just observational skills and some measure of our own abilities. It does not require complex theories or mathemtical modelling, it is a question of judgement.

If you have an idea that you would like to discuss then present it as such, just don't pretend it is a "hypothesis" or "theory". Questions and ideas are good, but will get nowhere without structure or data. It is not "disliking" but much more questioning and doubting, which is the way of all ideas.

[k_squared]

great post-naphand...

I think you are right that we can logically understand something to be 'correct' given certain assumptions are made. This is also true of poker. We could set-up a computer program to analze how certain types of players perform against one another given a set of environmental constraints. We could show that players with the same level of aggression who take into account position when choosing starting hands will do better than those who don't. But, the problem becomes then finding ways to apply this knowledge to the significantly more complicated scenario in which live players are involved. Live players are constantly changing styles of play based on how they are feeling, the table dynamics, how long they have been playing, how they are running... And while we might be able to quantify that it is correct to play hands X, Y, and Z given certain considerations, it is much more difficult to know when those situations are actually in place.

What I like about your post though is that you highlighted for me a clear distinction between two different ways of learning about, and understanding poker. We can do it experientially, i.e. we have seen so many hands and noticed A, B, C. Or logically, in so far as given certain premises, certain conclusions can be drawn that show we should play a certain way whenever those premises show themselves to be true. Without both of these analytic tools we will be missing something!

With only the experiential, it is very easy to fall into seeing patterns that will make you a losing player... "when I play 10-9 I win more than I lose... I should play 10-9 all the time." The problem being that our sample size will often cause us to find patterns that misrepresent the true strength or weakness of hands. It is frustrating but in a large sense the financial results of a given hands outcomes must be removed from our sense of whether or not we played the hand correctly. We often play the hand 'correctly' and lose money.

With only the logical we fall into the opposite trap of over simplifying play. Poker is very complicated. And many of the players don't truly act rationally in the framework of the game, which means our logical thinking will have a difficult (if not impossible) task of accounting for thier irrationality. The nuances involved in playing against other living beings leaves us constantly adjusting to how those other people play, and how their play changes.

Being able to combine your observation, and your knowledge of what was more likely to win leaves you in a much stronger place than simply using your experience, or simply playing from some hand charts.

-k_squared