Long term online poker success with winning strategies – register for free!
The best strategies With the correct strategy, poker becomes an easy game. Our authors show you how to succeed, one step at a time.
The smartest thinkers Learn from and with internationally successful poker pros, in our live coaching sessions and in the forum.
Free poker money PokerStrategy.com is free of charge. Additionally there is free poker money waiting for you.
Evaluation and Exploitation of Stats
IntroductionIn this article
- Which dangers lurk when dealing with stats
- How to correctly interpret the stats
- Thich stats are useful and which are less so
Stats have become an important part of playing poker. By analysing a number of past hands, one is able to form a rough profile of his/her opponent, and use this to make better decisions.
When doing this, however, you need to consider various factors. Here you will get an overview of important stats that you should use when making decisions in the future.
In this article we will talk about the following points:
- Which stats play a decisive role
- Irrelevant stats
- Normal values
- Sample size
Further topics will be discussed at a later time.
A focus on stats harbours certain risks, this is something you should always be aware of.
The basic problem: Poker is a game with missing or at least incomplete information. This of course is one of the things that makes it so interesting, yet one would still like to break through this barrier and get as much information as possible without the need for a real check.
Ultimately, a mixture of misinterpretations, vague conclusions or insufficient sample size are the main reasons for a wrong adaptions of stats.
If these mistakes start occurring with regularity, you begin to rely on false assumptions and run the risk of putting less thought into the matter. The opponent is quickly identified as a fish, as extremely aggressive, or as a specific 3-better, even though the boundary conditions do not permit such a conclusion.
Still, many players are inclined to regard this information gap as closed and to refrain from any further assessments. This, however, should preferably not happen.
Even if you are able to avoid these kind of mistakes, always keep in mind that stats can change, that your opponent is more than capable of adapting, and that nothing can replace your personal impressions, your reads.
If the sample size is sufficient, you could theoretically assume that the stats are significant. But what does that mean exactly?
You are sitting at a table and in fact have a few thousand hands from your opponent. That should be a number with which to infer helpful information.
What you should however not forget:
- The stats come from different days, possibly even weeks.
- Villains can have good and bad days.
- Villains can change up their style.
- Villain may operate with table selection and play primarily against loose-passive opponents.
- Villain may play differently against you.
There are a lot of factors that need to be taken into account.
A simple example for this is the 3-bet-stat.
The 3-bet-stat is in itself an obscure stat. To really be able to get a rough range, one needs to look at the position stat. What does it mean, for instance, if your opponent 7% 3-bets? He 3-bets maybe 5% at CO, 10% at BU and very little in the blinds.
Or conversely: In position he tends to call more, but in the small and big blinds he 3-bets a lot. That changes so much that it makes it essential for you to look at the position stat.
Another factor also plays a role: Your opponent 3-bets a value Y in position X, yet this is related to all tables and all opponents. It does not mean that the player in a given situation [u]against you[/u] will 3-bet the same amount.
You also take your opponent into consideration when making moves (whom do you 3-bet and why?), your adversary will do the same.
If you take into account all these aspects, it will become clear that such stats are to be treated with caution, and that you cannot justify expensive moves purely with them in the long run.
What you should definitely take with you: Stats should always be critically scrutinised!
Under no circumstances should you trust them blindly, so keep you eyes open to look out for any possible misinformation.
Which Stats Play a Decisive Role?
While you can now take a look at some of the most important stats, there will also be mention of "normal" values, i.e. values that are within the normal range. We will also be taking a look at sample size.
"Normal" values are only very broadly defined, simply because such values vary dynamically with different variants (e.g. 6max => Fullring), different limits and different opponents. That is also why such a thing as good and bad values don't necessarily exist.
In the future we want to look at how to exploit these stats, and how you can use them to your advantage. That is why only stats that can be used in such a way will be relevant.
These two are the most common and most important stats. The VPIP (how often a player enters into a pot) and the PFR (how often he raises pre-flop) give you a first rough picture of your adversary.
Players are already categorised using only these stats, that you should know. You can find a short summary of the classic opponent types here.
The advantage of these stats: The sample size does not need to be especially large.
After just a few hands an impression can be formed that you can make use of. Whoever shows 80/0 stats after 20 hands is no TAG. The likelihood of this being the case is already extremely small. This does not mean that 20 hands is enough to decide this, but 100 hands are already a very good indicator.
At least the opponent type will fit in the most cases, i.e. the ratio of VPIP to PFR. This is not set in stone, but you will seldom get much more information with so few hands.
Standard values naturally do not exist, they have a high dispertion and vary from 8/5 Nit up to 90/80 Maniac. The information you can draw from that therefore provides an important impression on the overall picture of your opponent.
The 3-bet-stat is generally considered to be very important. The reason is simple: You open-raise a lot (depending on position), but there are not many hands you would be comfortable facing a 3-bet with.
Such an increase makes a large pot, takes away many of your options, and forces you to either fold, 4-bet, or to call and thereby play a 3-bet pot without initiative. And if you don't happen to be holding the nuts, these aren't the best situations you can imagine for yourself.
It is therefore important for you to be aware of how often your opponent 3-bets.
In the section "Problems" you have already read a lot on this subject. To recognise the rough range on a 3-bet-stat, it is important to look at the position stat and - as also previously mentioned - consider whether villain will 3-bet just as much (or as little) in this situation against you. That is anything but easy.
Just the fact of having to use primary position values makes it clear that the sample size needs to be fairly large. The most HUDs don't just show you the percentage, but also the number of observed cases. This is a good way of checking the significance of the values when interpreting stats.
Even after 1000 hands these values are based on what is actually a very small sample size. Of course, after 20-30 spots you will be able to recognise a rough tendency, from 50 onwards you have a value you can use, even though it can naturally still deviate from the theoretical ideal. Nevertheless, 10 from 50 holds a greater significance than 2 from 10.
You can consider these issues of sample size each time for yourself: With regards to VPIP, for example, each hand counts, as the villain always has the option of entering the pot. But how often will he be able to use a 3-bet in a cutoff? The observed spots will obviously be less (this effect will be greater post-flop).
What values can you expect? A general hard value range can be considered as QQ+/AK, which means a range of 2,56%.
It is therefore clear: The higher the 3-bet value, the higher the proportion of bluffs. Naturally, in many other situations, even weaker hands can be considered value-3-bets, however anyone who has 10% 3-bets over a long period of time will not have a higher value range against you. At some point it just does not exist anymore. He will therefore have to bluff a large proportion of his extra range, giving you the possibility to respond (in the form of calls or 4-bets).
With most TAGs, the 3-bet values lie between 6% and 9%, the position values of course deviate more strongly.
This value should be very interesting for you. In the previous section you will have realised that 3-bets against you are always problem spots, where one is desperate for information with which to be able to defend one's self.
As such, it makes sense that the most of your opponents will not like such spots. The ft3-b-value shows you how often your opponent folds to a 3-bet.
In this case it is not possible to look at position stats, since it would be theoretically important to know from where villain raised, and from where he has been 3-betted. Apart from the fact that such values are not offered, it should be clear by now that the sample size of such spots would be so small that the value would have no significance whatsoever.
Accordingly, the sample size for this does not need to be as high as a position-3-bet-stat for instance, since all spots in which villain received a 3-bet after a raise are taken into account.
After a few hundred hands you will already have enough spots together to be able to see an early tendency.
The values vary between 40% and 80%. If someone only folds 40%, they can be considered a loose player. He calls a lot and runs into a lot of difficult situations post-flop. A majority of players are not in command of such spots (of course there are exceptions).
Whoever folds 80% is making a big mistake. In the next section mathematical background knowledge will be explained. You can try considering how much villain may fold long-term for his bluff-3-bet not to become +EV in the long run. 80% is in any case too much.
There are no standards in No-Limit Hold em, but you will most often see a value of around 55-65%.
A large stumbling block is the following: Do you consider the 4-bet-stat or the 4-bet-range? What are the differences?
While values like VPIP, PFR or the 3-bet-stat additionally give you the range, in the case of a 4-bet the following conditions are required: Villain raises, someone 3-bets.
The resulting value therefore does not denote the actual range.
The correlation is as follows:
4-bet-range = 4-bet-stat * PFR
A player with a PFR of 20% and a 4-bet-stat of 10% therefore has a 4-bet-range of 2%.
Using the 4-bet-stat you can thereby directly work out what your opponent will roughly 4-bet. You know that QQ+/AK is in total 2,56%, a player with a 4-bet-range of 2% will therefore seldom 4-bet-bluff.
4-bet-spots are however also extremely rare, so you will need a few hundred hands in order to recognise early tendencies. Values between 1% and 4% will be most common, whereby a player with a 1-4% bet range can also write what he has in the chat.
This stat shows you how often a 4-bet of yours will achieve fold equity against his 3-bet-range. The values vary greatly, depending on the 3-bet behaviour of your opponent.
Here a small digression: Later articles will deal concretely with misinterpretations, and a value such as this is particularly susceptible to this danger.
To be able to judge this stat it is important to also look at your opponent's 3-bet-range. If the ft4-bet-stat is 0%, the player will look like a very loose player against whom you can easily push for value. But what if this opponent only 3-bets 2%? Whoever only 3-bets the nuts will not fold on 4-bets. This has nothing to do with going loose broke.
This just show you that it is always important to see these contexts or to view this stat critically.
This value should be looked at with caution. What does “steal” mean? Theoretically, every raise is a steal, as it is an attack on the blinds. One does also not differentiate (it can logically not be differentiated) between whether villain is really stealing or if he is raising for value.
This value can be misleading, generally it denotes a first-in raise from CO, BU or SB.
The values vary between 30-50%, however you should definitely make use of the position stats (how much does villain raise first-in from each position). The condensation of all 3 stats can be confusing and lead to false range estimations (since the BU and SB-first-ins, for example, can vary greatly).
These values denote folds in the respective position of the above defined steal.
This is now interesting for you, because you can see how often on average the blinds simply fold. The sample size does not need to be especially large, since the blinds need to be defended each orbit.
The values vary between 70-90%, however here you also need to check the boundary conditions each time.
If the BB folds to 80% of steals, he folds against raises from the cutoff upwards. That does not automatically mean that he folds that same amount when you steal from the small blind (and that villain could then play heads-up in position against you).
If you steal from the button, you naturally have to take into account that both players would have to fold, should you be making a pure steal attempt.
The continuation bet stat tells you how often the pre-flop aggressor is likely to make the bet that is “entitled to him” on his own initiative.
Here it should be considered that the value can hardly reach 100%. All spots are taken into consideration,even multi-way spots in which a pure bluff-c-bet does not make any sense. Should villain play in position and donk his opponent, in other words not recognising the initiative and betting himself, the villain does not even have the option of making a continuation bet.
The values usually lie between 55 and 70%, everything below that is relatively low, and as such, value-intensive. Everything over 70% should be analysed carefully, villain's bluff proportion will be much higher, he will have to compensate this somehow (for instance through high 2nd-barrel-values on the turn and further post-flop bluffs).
In order to accurately recognise such patterns, you should take these values equally into account. Important here: The continuation bet values on the turn and river refer only to the spots in which villain made his c-bet on the previous streets.
A c-bet-turn-value of 50% therefore means that villain bets a 2nd barrel in 50% of all spots in which he also bet on the flop. The c-bet-river-value shows how often villain barrels his way through, in other words bets what is the 3rd barrel on the river.
You can imagine that these spots do not, however, occur all that often. The c-bet-river value on its own therefore has no real significance, it must be combined with the amount of hands villain plays in general, and how often he bets through on the flop and on the turn.
A river-value of 100% is of course very strong if it comes from a 13/10er who doesn't often conti-bet and rarely uses a 2nd barrel.
These values show you, how your opponent responds to c-bets on average.
You can remember for future reference that a 2/3-potsize bet requires a fold equity of 40%. If your opponent folds much more, the bet will become in itself +EV (with regards to bluffs).
If you see, for instance, that your opponent folds 35% of c-bets on the flop but 50% on the turn, the whole move (continuation bet + 2nd barrel) becomes +EV. Moves like this should be your aim. Future articles will explore relationships like this more specifically.
There are likewise values which indicate how frequently villain reacts generally to flop/turn/river bets. Here you need to consider that all spots will be taken into account, also unraised pots or situations in which flop and turn have been checked. These values should therefore always be treated with caution, since they have quite a low informative value.
Definition Aggression Factor:
(% raise + % bet) / (% call)
The AF refers only to the ratio of bets (and/or raises) to calls. It does not, however, take into account how often villain folds (or checks). The calculation only tells you that someone betted or raised on eight, and called on two of 10 spots on the flop, resulting in an AF of 4.
The value has always been used and offers many insights. Generally, the AF should go down from street to street, and usually appears in the following ranges:
- Flop-AF: ~3,5 - 5
- Turn-AF: ~2,5 - 4
- River-AF: ~1,5 - 3
You should pay special attention to small deviations. The phenomenon of the River-AF being higher than the Turn-AF is of particular interest.
Reasons for this are primarily slow plays on the turn and/or a lack of value bets. In such cases,river bets are made more for value.
Alternatively, villain uses the river to bluff if he senses weakness in a villain.
This case is a good example for misinterpretation, because a high River-AF does not automatically mean that the villain bluffs too much. Here you need to pay attention to why the River-AF is higher,and with what hands villain is most likely to play – stats alone cannot help you in such a matter.
Definition Aggression Frequency:
(times bet + times raised) / (times bet + times raised + times checked + times called + times folded)
This value now also takes passive actions into account.
f one bets or raises (aggressive) five times and calls twice, checks once and folds twice in ten spots, the resulting aggression frequency would be 50%.
The “Went to Showdown” shows how often villain reached the showdown.
The value usually lies between ~22-28%. For every hundred observed hands, with a VPIP of approx. 20%, this would be five hands that go to showdown. This example shows you that even after a few hundred hands this value should still be treated with caution and can vary greatly.
One also has to establish a connection to the VPIP/PFR. Theoretically, a tighter player should have a higher WTS. He plays a smaller and therefore stronger post-flop range, that should want to reach showdown more frequently. Someone who plays very loose should, on the other hand, often try to win an unimproved pot, since his equity is not sufficient to reach (to win) many showdowns in the long run.
Therefore, a loose player with a high WTS would seem to indicate a bad player, i.e. someone who calls too much.
On the other hand, a tight player with a low WTS would indicate that you are able to eliminate many of the villain's made hands by the river, as long as they don't happen to be the nuts.
Along with the stats that have already been mentioned, there are also numerous others, however most of them should not be overvalued.
The crux of the matter is that many stats don't really have any significance. You have already gotten to know about the basic problems (e.g. sample size), but several values can't even be reasonably interpreted.
If you take a look at a stat like “check-raise river”, you will see that all problems come together. For one, this spot occurs so rarely that it can hardly be used. You also do not know if the villain has made this move for value or to bluff. Furthermore, a comparison to the alternatives (c/c river, c/f river, bet river) is not made.
The main problem, however, is that this stat cannot take into account what happened before. What was the pre-flop action? What happened on the flop or turn?
The c-bet-river-stat or the ft-c-bet-river-stat at least make clear what previously happened. But a check-raise-stat on its own is taken completely isolated and doesn't take into account initiative or anything else.
It is therefore important that you concentrate on the most important aspects, look at how to interpret that important stats, and use them to your advantage. For now you should just ignore the other stats.
You have now received a first insight into the problem of stats and should now know what problems lurk when interpreting them. The most important aspect here is that each value should be critically scrutinised at all times.
In the near future you will learn what exactly to do with these stats. When is a steal, a 3-bet, a 4-bet worthwhile? With what raise size can you achieve the required fold equity? How you can get a general picture of your opponent and combine important stats? Which adaptations can you bring into your own game?
Click for more information.