When it comes to fully analyzing the behavior of a poker player, there are different levels of analysis we can do. Some take seconds and others take hundreds of hours. I am going to take you through the process of conducting a full behavioral breakdown of a poker player. Please note that the analysis isn’t possible without the extremely labor intensive process of coding the behavior into a database. You can check out how we do that here: How We Code the Nonverbal Behaviors of Poker Players. 

One of the most impactful insights we had when studying behavior at the poker table is that the value of behavior is really only unlocked when you look at behavior on an individual basis. 

For example, in the Beyond Tells studies we found that certain behaviors were strongly correlated with marginal hand strength. The problem with this perception is that different players have different ideas of what a marginal portion of their range actually is. Behavior at the table is not only a function of hand strength, but also depends on a player’s experience, understanding, emotional control, etc. To account for this complexity inherent in human behavior, we perform analyses of both between-subject and within-subject variability, looking at behaviors across every player in the sample as well as doing individual analyses of variations in the behavior of a specific player over the session. Below I will discuss how we conduct a full breakdown of each player. 

Step 1: A Full Contextual Analysis of a Player 

A contextual analysis is when we look at all the actions a player takes at the table. When they check, call, raise, fold, etc. This type of analysis isn’t new and is quite popular in the online community with software such as Hold’em Manager or Poker Tracker. We look at all the standard metrics that online players have been looking at for ages (VPIP, PFR, 3BET, etc.) to get a general sense of how a player is approaching the game.

‍Understanding what hands someone is playing and how they handle certain spots can give us insights into their behaviors and emotional reactions at the table. For example, there is a big difference between a player who takes one hand to the river in a 300 hand sample and someone who is doing it a dozen times. This one observation can give you information on a player’s experience level and skill, range, perceived value of their hand, comfort with post flop play, and the likely response to an opponent’s post flop aggression. An understanding of the player’s actions and approach to the game is critical to correctly interpreting behavior at the table. It is therefore helpful to build our more involved behavioral analyses on this basic contextual analysis.

Step 2: Standard Quantitative Analysis 

I have defined certain behavioral measures, such as blink rate, card check frequency, card apex, and decision time, which are analyzed for every player in the study. These measures are quantified and correlated with both hand strength and physiological arousal. 

First, we look at the connection between these behaviors and a player’s preflop range. For example, a player may check their cards 2 or more times with only marginal hands and never do it with hands on the top of their range. 

Second, we can look at how these behaviors change throughout the hand in relation to a player’s relative hand strength, which is essentially how much equity they have in a given spot. For example, a player might be in a post flop spot with a very strong hand and we will look to see if there are changes in blink rate in that high equity spot. It’s important to note that experience and skill drive understanding of equity. A player can be in a spot where they have a ton of equity against their opponent’s range but they still perceive their hand to be “weak.” So a player’s behavior will be a function of their perceived strength or weakness, not necessarily their actual equity. This is why the contextual analysis discussed in step 1 is so important.

Third, we can look at the connection between these behaviors and a player’s current level of physiological arousal at the table, which we can use as an indicator of emotionality. We can measure physiological arousal using certain biofeedback devices such at the Empatica E4. (Note: This was not used on every single player in our studies. Only select players were measured. We also use the device when we are paid to do high level personal analysis.)

‍

‍

During a standard quantitative analysis you will reach conclusions like:

• A player has a 3x increase in blink rate when they perceive themselves to be bluffing. 
• A player who displays a delayed apex greater than 1.5 seconds is connected to marginal hands. 
• A player is significantly less likely to double check their cards when they have a pair. 
  

While these may be based on a relatively small number of hands, they can still be statistically significant, and are extremely useful in practical application. If you want a deeper discussion about that please check out Understanding Sample Sizes and Useful Information in Poker Tells Research. 

Step 3: Spot Check, Hypothesis Creation, & Re-Code 

Our third step involves us looking at the first 10 hands a player plays and 2-5 hands where a player gets into a big spot (i.e. river bluff, shove with nuts, etc.) and evaluating for more qualitative variations in behavior. During our spot check period, we will create a series of player-specific predictions so that we can re-code their behavior to account for individual variability in behavior (in the thousands of ways people move at the poker table). For example, the video below is a clip of a player placing his arm.

We might have multiple predictions about what it means. 

1. He never closes hands when in marginal spots.
2. He always executes this behavior but exerts more pressure in his hands when bluffing.
3. He forgets to perform this behavior at the top of his range. 
   

By re-coding a specific behavior we are able to look at how a player gives off information at an individual level. This allows us to find tells that are unique and player-specific. It also allows us to evaluate the practical value of player-specific tells, by measuring the accuracy of poker tell predictions that can be picked up by observing an individual over hours of play. Human behavior is quite complex so you want to adjust coding analysis standards so it can account for that level of complexity. This re-coding process ensures a highly detailed look at behavior and also makes our next level of analysis significantly more robust. An example of a re-code can be:

• Identifying and timing a player’s smiles at the table.
• Identifying and measuring the level of tension in a player’s arm in post flop spots. 
• Identifying, measuring, and timing the smile quality of a player. 
• And so on. 
  

Step 4: Execution String Analysis 

Once we have all the re-coded data we conduct a string analysis on a player’s behavior preflop and post flop. A string analysis is when we code behaviors on a linear timeline to look for differences in order and timing. I go over that extensively here: Poker Tells Analysis Method #1 - String Analysis. Below is an image of how the difference in a player’s string can be connected to hand strength.

The order of action is often connected to the order of thoughts at the table. This is how we use behaviors to reverse engineer how a player is thinking about their hands. The most useful strings are found preflop and include a player’s card check and the behaviors that immediately follow. However, you will also see strings post flop. An execution string analysis will account for every single time a player has to act, and will include exactly what they did, how long the behaviors took, and the exact order in which they took place. 

Step 5: Comparative Analysis and Understanding Concealment Strategy 

After our string analysis we do a comparative analysis in order to understand the method and level of concealment strategy and potential emotional leaks in a player. We might take a spot where a player is bluffing river and a spot where they have the nuts (the best possible hand) and watch both hands side by side. When we do this we are looking for several things.

First, we look to see if a player has what we call a concealment strategy. At the poker table you aren’t supposed to display open levels of emotionality. Concealment is a known element of the game and because of that players use specific strategies to conceal their behavior. However, they sometimes use different strategies or different magnitudes of that strategy when they are bluffing/strong/semi-bluffing, etc. They might always be still, but hyper-still in certain spots and only partially still in others. A comparative analysis makes this much easier to see. 

Second, we look for differences in emotionality and how that emotion manifests. Sometimes it’s easier to notice that there is a subtle change in smile timing when a player is bluffing vs when they are strong. Or a player’s neck might pulsate in one spot but not in another. A comparative analysis allows us to take a closer look at these subtle changes in behavior. 

It’s important to note that this level of analysis is the most useful when we have a decent number of moments to compare. There are players who are never bluffing in a 10-hour session and there are players who could be card dead and not be in many spots where they have a very strong hand. However, we can still provide players with a high level of insight into their behavior with just 3-5 instances of behavior. 

Step 6: High Level Facial Analysis

The face will move more than any other part of the body at the poker table, however, it’s a highly regulated part of the body, and players usually don’t flash open smiles when they have a strong hand or appear openly nervous when bluffing. That being said, a high level facial analysis can at times be useful if you want to quickly find how the face can give off information. A high level facial analysis is done with software such as Noldus Face Reader 5, which codes every single movement in the face using the Facial Action Coding System and produces a massive dataset that can be used to identify micro expressions, shifts in gaze direction, and a lot more.

‍

We don’t do this level of analysis on every player since for the most part identifying slight changes in the face isn’t practical at the table. This is something I discuss at length in the Beyond Tells 2.0 Training - Check it out for free here.

Practical vs Academic Overview

Once all the data is collected and analyzed, it’s often my goal to make the data practical. We will coach you through a systematic approach that you can use to read players at the table, and which can be customized based on your skill level and strengths. We also take large amounts of data and insights and distill it into a practical set of guidelines so that a player can conceal their behavior more effectively in the future. It’s important to note that this isn't the most straightforward process. Some players will have significant mental barriers, different understandings of how they deal with risk, and other factors that affect their ability to conceal their behavior. Our goal is to create the best possible method for a player to conceal their behavior at the table, but the implementation of this method will often take dedication and practice on the part of the player before the method becomes second nature. 

‍
When it comes to more academic explorations we are also open to sharing portions of our dataset. If you are doing a master’s or doctorate thesis and want to explore the possibility of studying behavior at the poker table, reach out here.