Optimizing HUD For Small Sample Size

With the benefit of the proverbial 20/20 hindsight, we can now safely say, that the invention of the poker tracking software turned out to be somewhat problematic. The popularization of HUD's and databases changed the poker landscape forever - in some ways for the better, in others... not so much. Poker Office, Tracker, and Manager allowed for the extremely rapid evolution of poker strategy.

Tracking software made collusion more difficult, with a big percentage of players at every limit keeping extensive ledgers containing all the hands played by them at the tables. Lastly, trackers helped confirm, once and for all, that if we look at poker from a long-run perspective, it is, in fact, a game of skill. Unfortunately, the ubiquity of tracking software also had its downsides. The rapid evolution of strategy made poker more interesting and attractive but also a lot harder to win at.

Poker Tracker and Holdem Manager established a crucial divide between players using HUD's and those who either wouldn't or couldn't ("HUD" and "HUD nots" if you will) causing many recreational players to feel targeted, cheated out of a fun experience and eventually quit the game. That last part had a non-trivial impact on the long-term health of poker ecosystem and in turn, poker rooms bottom line. Some poker operators now forbid the use of HUD's while others attempt to render them less effective by either providing players with access to anonymous tables, the ability to periodically change their screen name, or by recording hand history in such a manner that HUD users can only rely on the stats that they managed to gather during the course of a single session.

Given that this approach became fairly popular in recent months and is now favored by many major poker providers like partypoker, Microgaming Network etc. many poker players have to adjust to a new reality of extremely small sample sizes. In this article, we're going to discuss how to design the perfect session HUD and make the best out of this situation.

---

Keep It Lean, Keep It Mean

Playing on a site that makes the stats gathered outside of a particular poker session either useless (anonymous tables, the way partypoker records hand history on regular tables etc.) or unreliable (playing in the environment where players are allowed to periodically change their screen names) requires a new approach to the problem of HUD design.

This can be seen both as an issue and opportunity. It might, in fact, prove to be a blessing in disguise. Many poker players are guilty of simplistic "more is better" kind of reasoning when it comes to their game. We tend to play sessions that are too long, juggling too many tables at the same time that we can barely see behind gigantic HUD's containing way too many numbers. Optimizing our HUD for extremely low sample sizes is a great opportunity to get back to basics and rid ourselves of a bunch of unnecessary information weighing us down.

First of all, let's figure out the kind of sample size we can realistically rely on when playing at a site that only allows us to gather session stats on our opponents. The number of hands per hour per table will vary greatly depending on a number of different factors like the poker software, how loose your average player is etc. etc. but we can safely assume that it's somewhere in the neighborhood of 70-90 per hour. The optimal length of a cash game session is a topic so broad that it would merit a separate article, but let's assume - somewhat arbitrarily - that if you're a pro grinder with impeccable work ethic, you can maintain your concentration for three to four hours straight, sitting out for an orbit or two every hour or so. 300 to 350 hands would, therefore, constitute the absolute maximum amount of hands you can gather on a single player, at one table during the course of a single session.

However, this will only be true for poker players well on the right side of the normal distribution graph when it comes to talent, skill, discipline and work ethic. Most of us probably shouldn't ever play a session that's longer than one or two hours without taking at least a 5 to 10-minute break so the maximum of 80-150 hands is a far more realistic one. Furthermore, table lineups are fluid, players leave, others join, and tables break all the time. Lastly, given that the average playtime of a recreational player is far shorter than that of a dedicated regular, the players that we really want to gain information on will provide us with the lowest amount of it - which, depending on how long your sessions are, might be anywhere between 20-60 hands.

---

Core Stats

Now that we've established our target sample size of around 20-60 hands for recreational players and about twice that for the regulars, let's discuss the narrow range of the best 'bang for the buck' stats that will be the most useful in this specific context.

Hands - This obviously goes without saying, especially in light of what we discussed above.

VPIP and PFR - This shouldn't be surprising as both VPIP and PFR are the cornerstones of practically every HUD - at least as far as poker variants involving 4 or more players at the table are concerned. VPIP and PFR are the most useful stats in the low sample size environment because they allow us to identify the recreational players - especially those of loose-passive variety - fairly quickly.

While 30/25 player after 50 hands or 50/50 player after 10 hands can be anywhere on the recreational-regular player spectrum, a 40/5 player after 20 hands is most likely not a regular. Same goes for 80/60 player after 20 hands (when was the last time you got 16 out of 20 playable hands in your session?). VPIP and PFR are great at quickly exposing both extremely passive and extremely aggressive players which is crucial because those are the types of players we're going to win the most money playing against.

3bet - This is probably the most controversial and dangerous addition to the minimalistic HUD set-up. The reason for that is simple, we're used to looking at stats like 3bet as a representation of a certain range of hands. For example, a tight aggressive player with 6% 3bet most likely has a very honest and unbalanced value 3betting range that consist mostly of premiums and maybe the occasional blockers, while 11% 3bet range has to be more or less polarized simply because there's not enough value combinations to 3bet a linear range with that frequency. In a context of a session HUD we can never gather enough hands to treat 3bet as an accurate representation of certain player's range but we can use that stat to help us determine our opponent's player type.

The fact that a player made one 3bet in 10 attempts tells us very little about his 3betting strategy (this could've been a bluff 3bet with A5s or a simple case of value re-raise with rockets) but the fact that he's capable of 3betting at all makes it that much likely that he's not a passive recreational. Similarly, a villain 3betting in 5 out of his 10 opportunities is either tilting, a maniac or an extremely lucky regular on a heater.

Aggression Factor (AF) - Another stats that - similarly to 3bet - requires a considerable sample size to be truly representative but it's not without its use in a session HUD. If you think that you'll be able to rely on aggression factor after 40 hands to determine if a regular is bluff raising you on the turn, you'll pay for that assumption with big blinds from you win rate. However, AF can finish painting the picture about the recreational player's postflop game in a similar fashion to what VPIP and PFR can quickly tell us about his preflop tendencies.

The best thing we can hope for when using a session HUD, is that we're going to be able to quickly identify the targets at our tables. If you add the aggression factor of 1.0 to that 40/5 player mentioned above, or an aggression factor of 3.5 to their 80/60 counterpart, you get a much more complete picture of your villain's tendencies even when you only have 20 or 30 hands at your disposal.

---

Optional Stats

VPIP, PFR, 3bet, and AF might seem like an extremely limited set of stats and most of us are conditioned to think that we need more. However, when we only have access to a very small sample size, the more stats we use and the harder we rely on them, the more likely we are to actually sabotage our own winrate. That being said, if you're playing particularly long sessions or your aggressive style causes you to often engage in intense wars with other regulars (which is a suboptimal approach to making money at poker, especially at micros, but is often unavoidable at higher limits), here's a short list of optional stats you might consider adding to your session HUD:

Fold to 3bet - Combined with 3bet can help you navigate those preflop wars with other regulars.

Fold to Steal - Similarly to VPIP and PFR, it can be helpful when identifying very passive recreationals and very tight regulars, though it's not great and describing anything in between.

Flop Cbet, Fold to Flop Cbet, Raise Flop Cbet and/or Check Raise Flop - You can use some of those stats if you feel like you need some more general info about specific player's post-flop tendencies, as they do have some value in a low sample size situation. That said, unless you're playing unusually long sessions and/or a lot of tables, I wouldn't include those stats in your HUD, because more often than not, you won't really be able to use them.

---

Stats to Avoid

Lastly, let's talk a bit about the stats that you definitely shouldn't be including in your session HUD:

Turn and River Postflop Stats - As we've already mentioned, even something as simple as Flop Cbet is already pushing the envelope as far as it's usability in low sample size situations is concerned. Turn and River stats will, therefore, be virtually useless as you'll almost never have enough hands on your opponent to even merit a quick glance at them.

4bet/5bet - Given that we're already forced to redefine the way we use 3bet in our session HUD, 4bet and 5bet stats will only muddy the waters even further, as they require a lot more hands to provide us with any meaningful information.

Showdown Stats - Went to Showdown, Won Money at Showdown and Won When Saw Flop % are perhaps the most valuable stats a poker player can include in their HUD. They can tell us the most complete stories about our opponents. There are plenty of 22/18 regulars - some of them solid winners, some of them break even rakeback grinders and you can even find some big losers who look competent because they strictly follow some starting hand chart. However, when you see a player with Won When Saw Flop of 52% you know you're in for a wild ride. Unfortunately, showdowns are a very rare commodity in poker. This is especially true in the context of a session HUD and that makes showdown stats virtually unusable.

After reading this article, you should have a good idea how to construct your session HUD and optimize it for small sample sizes. Remember that it's not about trying to determine the exact ranges and frequencies. It's about quickly identifying recreational players, improving your table selection and getting a general idea about the type of villain you're playing against.