How I created a poker solver and made $500k at the age of 23
In 2013, I developed a program that calculates Nash Equilibrium, the optimal strategy for No-Limit Holdem, the most popular variation of poker.
This article details my journey from being a broke college dropout to solving the game of poker, gaining elite players’ trust, and earning half a million by 23.
How it started
I met a professional online poker player in 2008, and the game instantly sparked my interest. It felt like a perfect opportunity for me — I’m a math major and a video game fan, poker is based on probability theory, and most importantly, I love money.
In just a year and a half, I climbed from $2 to $1000 buy-in games, won $40k, and moved out of my parents’ house.
Thousands of professional players worldwide live off poker, with the top players earning millions annually.
One-on-one poker does have a mathematically optimal, unbeatable strategy, known as the Nash Equilibrium. This concept was developed by mathematicians John Nash and John von Neumann. It also performs extremely well in games with three or more players.
Proving the existence of the Nash Equilibrium in poker is relatively straightforward for a math student. However, this proof doesn’t provide a practical method for finding the equilibrium.
In high school, I stood out in programming competitions and could handle complex algorithms, but applying these skills to finding poker’s Nash Equilibrium proved elusive. This intriguing challenge occupied my mind for years, leading me to make sporadic attempts at developing a solution.
By 2012, there were no products on the market capable of calculating the Nash Equilibrium for poker, leaving me without any existing models to reference for my project.
Despite this, by January 2013, after six months of experimentation and coding, I had successfully developed a program for No-Limit Hold’em, the most popular poker variant.
My program calculated an approximation of the Nash Equilibrium, starting from the flop, with two active players and a limited set of allowed bet sizes.
The program started with a completely random strategy, which evolved as it played against itself, improving in each iteration by increasing the frequency of more profitable actions. It utilized the Monte-Carlo Counterfactual Regret Minimization algorithm to compute equilibria.
At that moment, I realized I had created something groundbreaking — a tool that could significantly improve poker players’ understanding of the game. This program would eventually be known among my customers as a “solver.”
I’m proud to be the developer of the first No-Limit Hold’em solver utilized by high-stakes professionals. To confirm its pioneering status, I conducted extensive interviews with top players and researchers, finding no evidence of prior instances of a similar tool used by professionals.
My work was fundamentally based on existing academic research, turning theoretical algorithms into a practical, usable tool for players looking to enhance their skills
Finding Customers
The next phase was to find a buyer.
22 years old, I had no experience in software sales, and setting the right price was a major challenge. I decided to sell my product at a premium to a select group of professionals playing at the highest stakes, instead of offering it at a lower price to a wider audience.
Without direct competitors in the market, figuring out the right price was tricky. The closest product, StoxEV (later known as CardrunnersEV), which only calculated Nash Equilibria for trivial scenarios, was priced at $50. Clearly, this price was too low for my more advanced program.
However, I did have one benchmark for reference. In the world of poker, securing a skilled coach is a rare find as the best players often prefer playing to teaching. For instance, the world’s 20th top player charges about $1,000 per hour for coaching. My software, in essence, functioned as a round-the-clock poker tutor, providing endless hours of practice.
After careful consideration, I decided to price my software between $50,000 and $100,000 per license. This was not just wishful thinking; there was genuine interest from a potential buyer, known in the poker world as:
Trueteller
Trueteller was an extraordinarily successful player who, in just a year and a half, rose to the highest stakes in No-Limit Hold’em (20bb stacks on PokerStars), winning over $2,000,000.
I obtained Trueteller’s Skype contact through a mutual acquaintance and was eager to speak with him. The prospect of discussing business with one of my toughest opponents, known for his secrecy and limited communication, was exciting.
In 2012, there was a unanimous consensus among poker players that after a check-call on the flop, one should always check the turn. This strategy is correct… most of the time.
However, the solver bet 11% of the time after a check-call on the flop, valuebetting trips of 9 and semibluffing flush draws — a highly unorthodox move in this context.
Trueteller suspected this was the correct strategy, but no one played like that. He also knew that I couldn’t have hardcoded it or overfitted the algorithm to produce this outcome.
*
My and Trueteller’s initial goal was to negotiate an exclusive deal for a six-figure sum. We were both aware of the product’s uniqueness and the substantial edge it could provide.
However, negotiations dragged on for a month and a half, largely because an exclusive deal demanded a level of mutual trust we hadn’t yet established.
Trueteller was deeply involved in a $40,000 buy-in game at the time, leaving him little time to talk. At some point he requested a week to consider our deal. Feeling shy, I didn’t remind him about it and instead waited two weeks. Frustration mounted during our next conversation:
I, somewhat irritated, suggested, “Given the length of these negotiations, and your current exclusive access, I think a fair compromise would be for you to pay $5,000 a week during this negotiation period.”
Trueteller, also showing irritation, countered, “Why don’t you pay me $2,000 an hour while we talk? That’s what I could be earning in the game right now.”
Concerned about damaging our relationship, I didn’t press the issue further. However, I realized I was losing valuable time and needed to start looking for other potential clients.
In an interview (in Russian) Trueteller spoke about his decision to purchase my solver.
Interviewer: “Tell us about the first solver you bought in 2013 for $100k. How did you know it would be useful? Were you already aware of an optimal strategy in poker at that time?”
Trueteller: “From my university days studying Mathematics, I understood that poker had a Nash Equilibrium strategy. Before solvers, I used to do rough Nash Equilibrium calculations on napkins. It’s not too hard to figure out for the river phase of the game. Then Oleg approached with his program that could calculate this equilibrium and we made a deal”
Searching for customers
In the beginning, my efforts to secure customers met with limited success. I reached out to several players I knew, suggesting a license fee in the five-figure range, but faced skepticism. They failed to see the value in the program and were not ready to invest more than a few hundred dollars.
Additionally, I approached a top-20 heads-up player through a poker forum to showcase the software. Despite my direct message and a follow-up through someone we both knew, he ignored my attempts to connect. I later discovered that his doubts about the calculation of the Nash Equilibrium in poker were so strong that he considered my offer a scam.
Alex Millar
Eventually I contacted Alex ‘Kanu7’ Millar, a top-10 world player, via direct message on the twoplustwo poker forum.
I showcased my software to him and his circle of high-stakes players. They showed interest, even considering an exclusive deal for using the program. The negotiations were protracted, and time was not on my side.
By then, I had devoted eight months to development and negotiations, without playing any poker. I was running out of money.
Determined to secure a fair price, I waited patiently for potential clients to deliberate, careful not to appear desperate. However, the exclusivity talks prevented me from approaching other potential buyers. I was aiming for around $200,000 and unwilling to settle for much less. With just $6,000 left and monthly living expenses of $1,500, the pressure was mounting.
The situation became almost absurd. I would sit in the living room, reading, with my computer in the bedroom and the speakers at maximum volume, listening for Skype notifications. Twice, I rushed to check a “Que-wee!” sound, only to find no message. I initially thought it was a glitch. Eventually, I realized it was the sound of an old refrigerator’s compressor in the kitchen. Under the strain, I had mistaken any similar noise for a Skype alert.
The Deal
By April 2013, my persistence finally paid off. I closed a deal with a consortium of six players led by Alex Millar, who was an absolute pleasure to work with. In addition, I secured a separate deal with Trueteller, with both parties aware of the other’s involvement.
I granted these players exclusive access to the program. The total compensation for the deal was $200,000, spread over a year and a half. Initial months were more expensive than later once, because we expected early usage to provide the most significant insights and value.
One of the initially more sceptical customers, Raul, became the biggest fan of my solver. At one point Raul flew off to an island for two weeks. During this time, he intentionally disconnected his WiFi to avoid distractions and dedicated his days to practicing with the program.
From Alex’s interview (at 30:24)
Alex: Even before Pio Solver came out, I already had access to a private solver…
Interviewer: How the hell did you get that one, brother? What do you mean “you had a private solver”? The Dream Machine?
Alex: Haha, yes, The Dream Machine came to life. I was messaged by a guy on twoplustwo. So nobody had heard of solvers at this point. At first I didn’t know what it was? He described in detail how the solver worked
[… continues at 32:44] Ended up talking to him a bit, and then I gathered a small team together, and we paid a certain six-figure sum for the exclusive rights to the program.
Raul also mentioned purchasing the solver in his AMA (in German) at https://de.pokerstrategy.com/forum/thread.php?postid=20235645#post20235645
Interviewer: Tell me about buying a solver in 2013. How did you come up with the idea in the first place?
Raul: The developer actually reached out to me through the community tool here on ps.de, but I ignored the message. Then, the contact was established through someone else. I was quite amazed when I discovered that NLHE could truly be solved
The Results
Inspired by my Holdem solver’s success, I developed an Omaha solver in the same year, another widely played poker variant. This new solver eventually sold for $300,000. Combined with the $200,000 from the Hold’em solver, my total earnings reached $500,000.
Since then, I’ve developed solvers for various other poker formats and am still open to inquiries from players in $5000+ buy-in games.
In 2015, two years after my initial launch, the first major competitor emerged, offering a No-Limit Holdem solver for $500 per license. Rather than entering a price competition, I shifted my focus to different projects. In hindsight, this decision is regrettable, considering the Holdem solver market now stands at around $10 million annually.
Successfully calculating the Nash Equilibrium in poker was a source of immense joy and pride. It was a challenge that had occupied me for four years, and solving it not only was a personal triumph but also helped top-tier players enhance their game worldwide.
This journey made me realize my true passion lies in developing complex programs rather than playing poker. I take great pride in having built this business independently, multiplying my earnings compared to the previous year.
Beyond solvers, I’ve since engaged in poker room development, trading tools, and investment ventures.
Currently, my interest has expanded to learning more about Large Language Models (LLMs), and I am actively seeking a residency position at an AI lab. I’m also happy to give a talk at your company.
In future articles, I’ll talk about the technical history of solver creation and other business adventures. To follow my next articles, subscribe to my Twitter
Demo
Want to put yourself into the shoes of the world’s best poker players? My solver is now public! Test out your game against the Nash Equilibrium and let me know what is your optimality score: holdem.olegsolvers.com
Acknowledgments
Special thanks to Ivan Bogatyy, without whom this story would not have been published .
For years, Ivan has inspired me with his epic publications detailing his daring and deeply technical exploits. He executed the world’s first blockchain frontrun (also known as an MEV sandwich), successfully hacked the cryptocurrency Grin, and engaged in some extraordinary NFT trades. For more of Ivan’s adventures, subscribe to his Twitter.
Thanks to Marc Lanctot for his generous and insightful responses to my queries, and for his Monte Carlo Sampling for Regret Minimization in Extensive Games paper which described the algorithm that became integral to my business.
Thanks to his co-authors Kevin Waugh, Martin Zinkevich, and Michael Bowling.
Thanks to Yoram Bachrach and Ian Gemp for their interest in my work and encouragement.
Thanks to spears from pokerai.org for advice.
Thanks to Wiktor Malinowski, Anton Makhlin, Mikhail Lysov for their help with the demo.
Thanks to Haseeb Qureshi, Ilya Laut, Andrash Gusti, Ashwin Ramachandran, Sergey Bartunov, Michael Johanson, Timofey Kuznetsov, Daria Feshchenko, Alex Elenskiy, and Andrey Shelomentsev for their help with the article.
