Hello everyone! This interview was perfromed by "forrhje" specially for his blog with two of Simple Poker founders Mikhail and Roman. This interview was translated from russian, enjoy reading!
Hey! Thank you for agreeing to chat with me! Can you tell us a little about your team first?
Hey! Our team consists of founding partners and company employees. The founders are four people, two developed this project through the first tool - Simple Nash. They are now focused on the development of solvers engines and solving algorithms/approaches. Roman is more focused on marketing strategies, relationship with partners, and leads user support. Mikhail is directly involved in product development. There are more than 10 people in the team at the moment.
How long has your brand been around?
The first solvers appeared a little earlier than the “Simple Poker” brand. Then the idea came that we can produce not one solver, but a whole line of solvers. The first solver was released in 2012. It was Simple Nash. Only 3 years later the “Simple Poker” brand was formed. When the first product was released, the team was much smaller. We all worked in different jobs and this project was not the main source of income.
How did it happen that in the end you all got together and decided to make poker software?
We studied together with Alexander, the person who implemented Simple Nash. Even then, we actively communicated and performed some tasks together. With Roman we worked together in an organization. We were friends and spent our free time together. I realized that Roman was very interested in the topic of poker, and we decided that it was possible to work in a team. Why a poker theme? One of the partners initially played poker actively. He was engaged in mathematics, was interested in research in areas of mathematics, including those related to poker. It was at that time that the first articles about poker solvers were published, and he decided to start working in this direction. But if you take the team as a whole, most people were not familiar with poker initially.
If people were not initially familiar with poker, then who is testing the software? You never know what your solver can count there ... Who checks?
Due to the fact that we have been developing our software for a long time, we already have a good understanding of the game, from the point of view of checking how it all works. We also have a set of benchmark calculations to check the results of the solver before it is released. You don't have to be a poker player to make sure the calculations are correct. Quite the opposite. Poker players may miss some bugs, but we do not. For example, one competing solver produced computational results that surprised us a lot. Later, we were able to find an error in his calculations which was not found by the players who used these calculations for over a year.
How did your brand grow?
The products released largely depended on scientific research in this area. First came Simple Nash, because there were materials on how to find equilibrium in a push-fold strategy. Then there were materials on how to find equilibrium in postflop strategies - Simple Postflop appeared. We tried to quickly implement the knowledge gained in the final product. But not fast enough. Simple Postflop's main competitor is PioSolver. He rolled out his solver on twoplustwo.com a week earlier than we did. In many ways, we associate its popularity with this. We bite our elbows, of course, that we were a little late with the release, although on GipsyTeam we released the beta a month earlier than Pio.
Do I understand correctly that PioSolver is mostly popular outside of Russia, and Simple Postflop has captured the Russian-speaking community?
Yes. Initially, our products were targeted at the Russian-speaking and English-speaking markets. Pio immediately focused only on the English-speaking audience. Plus, it was released a little earlier, so many users did not change the solver after purchasing Pio. Also, initially, Pio had a standalone version, which allowed an unlimited number of calculations on a computer, and we made the first launch of a cloud product that allowed making calculations remotely. We decided that many users do not have powerful enough computers to perform calculations on their machine. The poker players thought that it was much more convenient to do calculations on their computer, so they preferred Pio. We quickly released a version with local calculations, but already some users have chosen Pio. I also link our lag in the English-speaking segment with this. At that time, our product was faster and more functional, but due to a lack of time, it was quite difficult for us to fight in the English-speaking twoplustwo forum. We decided to concentrate on the russian segment.
How did the brand develop further?
Further development took place in two directions: user friendly applications for a wider audience to attract new people, for example, Simple GTO Trainer, GTOBase, GTO Sensei, and the implementation of new algorithms, for example Simple Preflop Holdem, plus the adaptation of current algorithms for other disciplines, for example Simple 3-way, Simple Omaha. Finding preflop equilibrium strategies for SPF and launching distributed cloud computing for complex preflop calculations were also added. Distributed cloud computing is a unique service. Now no one can compete with us in cloud computing. We can calculate both a difficult preflop and a large number of flops in a short time. As soon as a technology appears, we immediately strive to implement it in our projects e.g. as soon as the algorithm for 3-way spots appeared, we immediately began to implement it.
Where do all these calculation algorithms come from?
There are two main universities that are actively pursuing poker - the University of Alberta and Carnegie Mellon University. They publish articles with algorithms that allow us to figure out what works and how.
Can you explain in simple terms what a solver is and what is it for?
If we talk about a poker solver, then we describe some kind of game situation in which we can calculate the Nash equilibrium using the solver. Nash Equilibrium is a two-player strategy in which no player can deviate from their strategy in such a way as to increase their EV. That is, if one of the players deviates from equilibrium (strategy), he will lose his EV. And, since poker is a zero-sum game, if one player loses EV, the other automatically increases his EV. If it is even simpler, then this is some kind of strategy that cannot be exploited in any way. For poker, this means some balance for value bets and bluffs. It doesn't matter who is playing against us. If our opponent bets too much, then we will make more money with value hands. If opponent overfolds, then the expectation of our bluffs increases.
If we play strictly according to the solver, then at least we won't lose money to any player?
Yes. Right. This feature works for two players well. If we have more than two players in the hand, then if, for example, two out of three deviate, then they can execute a cooperative strategy, which can reduce the expectation for the third player, despite the fact that the first both deviate. Poker is a non-cooperative game, but theoretically it could be. In a 2-way spot it is impossible to beat a solver, in a 3-way spot and more, some theoretical chances appear if the strategies of several players cooperate, by chance or on purpose.
Poker is often compared to chess. Chess is also a game that two people play, but it is a game with open information. In poker, some of the information is hidden - the opponent's cards. At the moment, a person can no longer beat a computer at chess, even if the computer uses a smartphone engine. It is much easier to calculate a game with completely open information, that is why poker is considered a more difficult game. If a person thinks that he is able to beat the GTO strategy, then first let him try to beat the chess engine or he could try to do it in our Simple GTO Trainer.
What is the difference between the algorithms that solvers use?
The algorithm implemented in Simple Postflop is currently the most accurate of all the existing ones. When we talk about GTO in poker, this is not a perfect GTO that cannot be exploited. There is a deviation of some thousandths of the blinds. People deviate much more from equilibrium, and therefore cannot beat GTO strategy from a solver. The calculation can be made more accurate, but this will require hundreds of GB of RAM and a lot of time.
Algorithms are divided into vector and abstraction-simulation. The convergence of vector algorithms is mathematically proven. An example of a simulation algorithm is the Monte Carlo method, when a huge number of situations are simulated in order to find equilibrium. Comparison of simulation algorithms with vector algorithms showed that these algorithms can exist and produce good results.
Do solvers have the opportunity to be improved significantly?
In terms of solutions accuracy, no. In the speed of solving still yes. In 2015, Alberta University released their FL HU Holdem solution. They spent a month and a half on these calculations. Now all this can be calculated in less than 10 hours. Progress can still occur due to the development of computer capacities. Perhaps someday we will develop a solver that will not need to set sizing at all. It will calculate everything itself and give the best size.
Who can benefit from your product?
Solver can be useful at all stakes. I'll start with the high stakes. The game there is such that many strive to play in GTO poker. Solver produces, one might say, a basic balanced strategy. Knowing it, the player sees how his opponents deviate from it. Thus, he sees their leaks and can already start building his exploits. At the lower limits, there are probably simpler strategies that will yield higher win rates. But knowing GTO and why we can deviate from it gives a greater understanding of the game. It will not be superfluous if the player knows GTO. It is important to understand that GTO will not give the highest win rates because this is a basic strategy. Exploiting the leaks of opponents and the leaks of the players pool gives you more of a winrate, but it is important to understand what to start from. It should be noted that there are not only solvers, but also training programs based on solvers. If we talk about micro-stakes, comments “solver-bullshit” slip through. Recently there was a stream of Innerpsy and Trueteller. Trueteller played several hands in the Simple GTO Trainer and made mistakes in some of them. People started to write that the trainer was bullshit. In fact, people need to approach the solutions of the solver differently. If the decisions of people do not coincide with the GTO decisions, then this is a reason to think, "Why did this happen?" and try to figure it out, improve your strategy and understanding. Even a player from the highest limits may not play perfect GTO. Everyone decides for himself whether it is useful to him or not. But the important thing to note here is that there is a lot of feedback from high stakes players that a solver is what gives them an edge over other opponents. Many of the reviews come via personal message, as people don't want to publicly declare that a solver is the most useful thing there can be.
Is this software difficult to master?
If you just download the software from the website and start working with it, it could be very difficult at first. To help the players, we have articles on some programs, there is a YouTube channel with tutorials, and a support service. There are some people who coach to work with our software for money. I also saw training courses on how to work with solvers. Anyone who wants to can easily find the necessary material.
Low limits players are scared off not only by the complex interface, misunderstanding of how to work with all this, but also by the price of the product. How is pricing done? Are there any plans for cheaper products for the lowest stakes players?
I would not say that we have sky-high prices for our products. Most expensive - Simple Postflop local version for 2 computers costs $300. If a person plans to engage in some kind of sports professionally, then I think that equipment and inventory are much more expensive. And then a person has the opportunity to earn more money thanks to our software. So I don't think it's a high price to be honest.
The price of the product is set depending on the market situation, on the end users, on the cost of the product itself and its support. We are trying to reach as many people as possible, while making money ourselves.
People who do not want to spend a lot of money can afford to buy training packs in the Simple GTO Trainer / GTO Sensei. The price for such a pack is about $100 per year. It's less than $10 per month, which is very inexpensive. Prices are quite low for all products, including due to our knowledge of the Russian-speaking market.
Why did players spend tens of thousands of dollars on solvers before, but now it can be bought for several hundred dollars?
Just because it was a bespoke product. They paid for exclusivity. Today's solvers are in no way inferior to those $ 100k solvers. Rather, even better than them.
How do your products compare favorably with competitors' products?
We have the largest product line. Products have something in common. Having learned one product, learning another is not difficult. Our solver for 3-way postflop spots is the only one on the market that solves equilibrium without abstractions. Each product has features that competitors do not have. We were the first to implement the idea of a GTO trainer on the market.
How is contact with poker rooms when a new product is released?
Our software does not interact with the rooms directly. For our part, we are trying to make sure that our software cannot be used to gain an advantage directly during the game. We initially had a concept that our software is a tool for developing skills, and not just some kind of RTA (Real-time Assistance Software). When a new product comes out, we write to the support of the room. We ask what needs to be done so that our software appears in the list of allowed programs. They send their demands and we fulfill them. A normal dialogue can only be conducted with the support of PokerStars. The rest of the rooms most often do not even answer. Therefore, our software works in all rooms according to the rules set by Stars. There was a situation that the user was banned from Chico's network simply for our software installed on the computer. We tried to contact them several times, but did not receive a single answer. There is a very strange story in general.
But still there are players who try to use solvers to gain an unfair advantage by creating bots and prompts. How do you feel about this poker problem? How difficult is it to use your solver's calculations to write a prompt?
Our software is sufficiently protected from this. Solutions from our solvers are stored in a protected format. For example, competing software has an open data format and access to the engine is open. Probably, an RTA can be done using a solver, but it is definitely not convenient to do with our products. At one time, we made a decision that we will not have dumped reports that can be opened and used in other programs. For Pio, scripting is implemented in the full version for $1k. You can use the command line to access the solver and get a solution, access to the engine. It is not difficult for us to add this, but we do not. Many of the limitations of our software are related, among other things, to the fact that the solver cannot be used by the RTA to use its calculations. Of course, if someone takes a lot of screenshots of calculations and runs them during the game, we cannot do anything about it. But we do not provide the user with a convenient way to transfer data.
What you say only evokes a feeling of respect for you. You are great guys that, for your part, do not allow using your software for calculations used in RTA. How do you think the rooms should fight against bots or hints?
As far as I understand, dealing directly with RTA is quite difficult. Poker rooms take a different path. They add new game formats that have not yet been calculated. They remove old formats, for example HU games, which have already been calculated. Many poker rooms attract top users of solvers who act as consultants. The rooms themselves often ask the players to record videos. They have enough tools to figure out RTAs / bots. They use multivariate analysis: they track mouse movements, running processes, the frequency of players' actions, etc. It will be more difficult if the RTA does not work according to the GTO, but according to some unique exploit strategy. We can only hope for other security checks like recording games.
Advice from the developer. How to remember everything that the solver gives us?
You can round strategy frequencies in our calculations. They will be less accurate but easier to remember. You can come up with a human interpretation of the solver's strategy, try to derive some rules for yourself. Often practice strategy in training tools. The general advice is to work with software as often as possible.
What are your company's plans for the future?
It is difficult to make plans for the future at the moment. Our primary task is to finish what is unfinished. For what doesn't work very well, make it work very well.
Thank you very much for answering my questions! It was interesting!
Thank you! Bye!
If you enjoyed the interview, share it with your friends! Want to know more about our company and stay tuned with latest news, follow us at social networks: