Using Stats

One of the differences in online poker compared to live poker is that you can gather and use statistics about your own play and the play of your opponents. If you have these statistics and others don't, this is obviously to your advantage. But pretty much all of the regulars in online games have data-gathering, -interpreting, and -displaying programs such as Idle Miner and Poker Tracker/Ace Hud or Hold 'Em Manager. The advantage then falls to those who 1) gather more data and/or 2) are better at using the data they have.

While many statistics such as VP$P and PFR are straightforward to interpret, other commonly used statistics can have a variety of interpretations. The most important of these is Aggression Factor (AF). AF is defined, on each street, as ((bet + raise)/call). That is, it's the ratio of the times the player does something aggressive (betting or raising) to the times the player does something passive (call). Note that the number of times the player checks or folds has no impact.

For a given hand, a player's AF has little bearing on how strong they are when they raise. If a player has an AF of 5 but a VP$P/PFR of 9/6, they probably have a very strong hand, simply because their starting hands are so strong. But even a 45/30 player with an AF of 5 could theoretically have very strong hands every time he bets or raises if he is just check/folding all his weak to medium strength hands. To get a sense of this, you could look at his Went to Showdown %. If it is very high, then he's not folding very often, so the high AF and VP$P mean he must be bluffing a lot. Again, AF is a ratio. It gets high either by a player betting and raising a lot or calling very little.

AF, on its own, means pretty close to nothing. To interpret it, you really need to use VP$P, to understand the range of hands a player is heading into later streets with, and WTSD%, which gives you a rough idea of how often a player is either betting, raising, or calling.

Not only do you need a bunch of other stats to be able to interpret AF, it takes a long time to converge. That is, you need a lot of hands before the AF starts to really describe a player's behavior. There are a few reasons for this. First, most players don't see the flop that often. So if you're looking at flop AF and you have 1,000 hands on a player, there are probably only 200 hands or so that are relavent, because most of the time either the player folded before the flop, or everybody else did. Of those, only a fraction will impact AF. For instance, if a player calls a PFR and then check/folds, his AF is unaffected. Second, since it's a ratio, small changes in the denominator can make the number much different. For instance, if on the flop, a player has bet once but never called a bit, his AF is infinity. If the next hand he calls a bet, his AF is 1.0. It takes many hands for this kind of herky-jerkiness to calm down.

There are also just so many different situations post-flop that one statistic isn't very good at summarizing a player's tendencies. By contrast, preflop VP$P/PFR gives you a LOT of information because there are relatively few situations a player can face. Also, a player faces a VP$P/PFR decision in every hand, so it converges very quickly. These stats after 100 hands are only rarely different than they would be after 10k hands.

AF can be a useful statistic, but I think it's far overused, just because there are many different kinds of players that will arrive at the same AF through a different mixture of statistics. Really, using AF at all before you have at least 1,000 hands of data is pure guesswork. Once you get enough hands on a player for the AF to mean something, your own observations and notes will probably carry more meaningful and accurate information than AF anyway.

-BRUECHIPS