Expert on analytical data Daniel Poston told about how statistical analysis helps to win in poker, business and life.
15 APR 2011 in the poker community called Black Friday — the day the U.S. government shut down the first three sites for online poker game. At that time professional poker online in the United States have played about 4 thousand people, and they began to look for other sites, particularly popular were Canada and Costa Rica. I’m from southern California, so Mexican Baja California is no stranger to me. There, in Rosarito, I decided to move to continue to deal poker.
When I was preparing to leave, I was often asked: “what are you going to do if it won’t shoot?”. To play online poker, you need to be able to handle the data, to know statistics and probability theory. Then I knew only one profession, which would be asked the same set of skills, so replied: “Go to wall street analyst”.
Then just released film “the Man who changed everything.” It is about baseball, and the action takes place in Oakland during the 2002 season. Then the team “Oakland athletics” using data Analytics strategy, similar to those used on wall street completely changed baseball with a limited budget, they won a record 20 games in a row, and at this point, the data analysis went to the people.
A year later, Thomas Davenport and Patil Dorji published an article “data Analysis: the hottest profession of the XXI century” (Data Scientist: The Sexiest Job of the 21st Century), and the site Glassdoor.com called the data processing best work in the US in 2016 and 2017.
What is common in poker and analyze the data?
I began to professionally engaged in the analysis of data in 2016, and I noticed for example that what I was doing, playing poker, similar to the customer segmentation.
When you select strategy is very important to determine where the enemy come from (geographical segmentation) as he thinks (psychographic segmentation) and how he plays (behavioral segmentation).
Over his career in poker I’ve learned is that these parameters can be reduced to a simple numbers — only two indicators I can tell me a good player or bad. To test this theory, I built a model to segment players according to the method of k-means — exactly the same company segmentyou their customers.
The data for this project I took from my own career. I was playing in cash no-limit Texas holdem with a buy-in amount (minimum fee) $25 (big blind us $0.25) to $200 (the big blind is $2). I usually play 15-20 tables simultaneously, each table has 8 or 9 players for a total of about 600 combinations per hour.
The most data I have buy-ins starting at $25, because it is the most popular game. I used data from 2013, then I won 387 373 donations $1913,13 — and it was a small part of played combinations.
Every time in online poker is played the next delivery, an entry is created in history, which shows that during the game did each of the participants. I used a software called Hold’em Manager (for other types of poker you can use Tableau) — it real-time downloads the history in a PostgreSQL database, allowing the player to analyze the game of your opponent. The result is displayed over the table:
I used statistical analysis to win in poker
In Texas hold’em each player is first dealt two cards — so you will get one of the 1326 card combinations. As you add cards you need to make assumptions about the range of combinations that may be your opponent, and it helps statistics.
For example, some players rarely raise preflop (that is, during the first round of betting), then they will have low rate Pre-Flop Raise (PFR). If the opponent has a PFR of 2%, I know that he can only be 26 of 1326 possible starting combinations. And since I know that they will only raise with good cards and combinations of aces and kings, this 28 options, I quite clearly imagine their cards at this point.
During the game I could mark any that interest me the change, then back to her tonight.
As I mentioned, there are two of the most important parameter of each player. These above mentioned the percentage of PFR and the percentage of “Voluntary put money in pot” (VP$IP). VP$IP is the frequency with which the participant plays when first offered the opportunity to bet or fold. The statistics for these two options it allowed me to understand the player will be a winner or a loser.
The Pareto principle, named after economist Vilfredo Pareto, States that in many cases, 20% of events cause 80% of the effects. This means that 80% of the company’s profits may come from 20% of customers and 80% of my profits from probably 20% of my opponents.
I allocated 20% of opponents, which I win (I call them “fish”), and 20% of those who lose most (they are, respectively, the “sharks”).
I built the model according to the method of k-means, smashing opponents into five clusters across the eight parameters after the segmentation has identified the segment with the highest concentration of fish, and another, with a high content of sharks.
For each segment I averaged VP$IP and PFR. The hypothesis I was: sharks VP$IP and PFR are very similar to mine, and fish high VP$IP and the maximum difference between these two parameters.
Among the sharks VP$IP opponents on average is 15.1%, and PFR of 11.7%. The top image illustrates the range of approximately VP$IP with the value of 15.1%, and the lower range when the PFR value of 11.7%. Yellow-shaded standard hands that cast the players in this category. As you can see, the images look like: they are mostly good starting hands. Sharks rely on two key principles:
- No reason to add money to the Bank when there are no good starting hands — it is better to fold.
- If you have a good starting hand, better to play aggressive and raise. An aggressive style of play gives you more chances to win than passive. The fact that the bets and raises open two ways to win — by a better hand or force the opponents to Pasu. Opponents can’t do a reset, if you do not bet.
At the poker table with such opponents are difficult and dangerous to compete: they can leave the less experienced player with nothing. And what if to draw an analogy with the world of business?
Suppose we run online retailer — for example, you sell widgets. Perhaps we can learn a lot about their potential clients, the number of pages on our web site they viewed and what pages interested them the most.
How each individual visitor interacts with the web site will show the statistics and will help to identify some trends. A category of visitors who browse a few pages, mostly ones that offer widgets with a low yield may indicate a model of the behavior of low-profit or unprofitable customers.
Evaluating this situation, we can avoid assigning resources to these potential customers.
Players-fish, on average, show a 43.8% for VP$IP (the top image) and 14% for the PFR (bottom image). Image vary greatly. These players voluntarily report the money to the Bank almost three times more often than sharks. This strategy indicates that they often start the game with a mediocre or weak starting hands — and even worse, stick to passive games.
Passive draw a bad hand at the poker table — a dubious pleasure, and honey, the money is guaranteed to go into other people’s pockets. My memory was not such that at the poker table sat at least two fish.
But back to our analogy of the online retailer for the sale of widgets. Might look like their most profitable segment? Customers from this segment, probably browsing a lot of pages on the site and stop at those that offer widgets with the highest return.
These buyers, bringing the business a lot of money, can get to the site via target links or may be interested in advertising a blog post. Sometimes the most revealing moment is the time that a buyer spends viewing the site.
Once we understand that the potential customer belongs to the high-yield category, our goal is to allocate resources to these customers, for example, make them part of a targeted marketing campaign or to settle with a consultant for the selection of optimal options.
I realized that poker game is life in miniature. What happens at the poker table, often occurs in many other aspects of life. I will share with you a couple of important revelations that I have learned over the course of his poker career:
- Everyone thinks they play poker, but most are badly mistaken. When you earn a living at the expense of people who are so wrong about yourself, at some point comes the understanding that we are all a bit lie to yourself. It takes a lot of effort to be honest with yourself and with others — but it’s worth it.
- To blame only you. So it sounds the perfect cure for lying to myself. To improve your playing skills in poker, you need to be self-critical. It is impossible to learn from mistakes, if you hang them by other players or circumstances. If you can’t sell your product after the presentation or does not take you on as a specialist on data analysis — take time to figure out where you made a mistake, and not repeat this error.
- Make decisions based on logic and reason. Ego and emotions are costly at the poker table.
- Gambling ultimately defeat those who most efficiently gathers and uses information. So don’t be lazy: take the time to learn about the company where you go for an interview and, if possible, the personnel officer, who will conduct it.
- Don’t be passive. The best strategy is usually selective aggression. Selective aggression at the poker table means making bets and raises with a strong hand from time to time to bluff, and not to dwell on the checks or Coly. In business it means actively conduct business: send offers and engage in further interaction, and not just hope that the deal somehow itself will draw.
If you’re wondering, for 387 373 played hands my their VP$IP was 15.6%, PFR — 12.2%, and the players-fish showed the highest percentage for VP$IP. Here is a link to codethat I prescribed to determine their opponents into one category or the other.