Understanding Data Sources in Poker Tells Research
At a macro level when looking at behavior at the table, you have 3 specific categories.
1. Player actions
2. Table events
Each of the instances in those categories occur at a specific time.
A player action is anything that a player does that is directly related to the game of poker. Raise, fold, call, etc.
Table events are any action executed by the dealer. They create the general flow of the game. For example, the moment a player is dealt their hole cards and what they have, when we see a flop, turn, or river, these would all be considered a table event.
Finally we have behaviors which are essentially all observable movement at the table and in some cases non-observable movement which is recorded via biofeedback devices.
All of this data is bound to a moment in time. We measure time by breaking 1 second down into 30 frames. We want a more granular level of time recording because a lot of actions are executed in less than a second and it’s valuable to have a standard when dealing with such a massive amount of raw video footage.
Table events and actions are very easy to deal with. It’s a labor intensive process but it’s very straight forward. You mark the time an event happened. Some of this is done manually and some of it is done with RFID technology. The more complex data source is physical behavior.
The poker table is one of the best places to measure behavior because you have repeated actions that are executed in a similar fashion. For example, a player has to check their cards no matter what. You can check your cards with one hand, two hands, one at a time, but the variations are relatively contained. The same with betting.
This aspect of contained variations is great because it allows for a large sample of individual moments that can be categorized in a way that makes comparison a lot easier. For example, in order to determine how a player bets we can use a coding methodology that looks at the speed of their bet and the style they used to execute it. We can then compare that speed and style to every single bet they made and look for bigger themes. You can read more about our coding methodologies by clicking here. Those methods create the behavioral data that is then analyzed with the event data and the table data.
Once we have our 3 data sources, we can start to look at the relationships.
For example, we can look at a smile and its relationship to certain dynamics in the game. A smile and hand strength, a smile and how much equity a player has, a smile as an indication that a player is going to play better, etc. However, doing this probably won't tell us much because smiles are complex and there are a lot of different ways a person can smile.
If we want to go into hyper detail, we can use software such as Face Reader by Noldus and break down every single facial action unit, create customized definitions of what constitutes a specific smile type, and do analysis that way. While this is interesting, it often becomes a bit of a rabbit hole because when it comes to facial movement while there is value, there is also a TON of noise.
A more practical and ultimately easier relationship to study is the difference between the perception of preflop hand strength and certain hand movements. The table action is when a player is dealt their hole cards, the action is a player’s first decision, and the behavior is the description of how they execute that decision. A player will have a perceived hand strength in every single hand they play. They receive cards several hundred times across a session, allowing us a very comprehensive understanding of how this particular moment can be used to understand the connections between behavior and a player’s hand strength.
While this can be technical, from a practical perspective you unlock so much more information when you build a narrative around a player’s behavior, something us humans are absolutely incredible at with training. If you are a poker player make sure you check out The Behavioral Edge, the first week of the Beyond Tells training and if you have any other questions about research please reach out.