How I created Draw Solver, made a million dollars, and helped to take down Phil Ivey
(This is part 2, part one)
In 2013, after we made a deal about Holdem solver, Trueteller mentioned that solving 2-7 Triple Draw (TDFL) could make me a million dollars. TDFL was an obscure poker variant regularly played with $2000 big blind on PokerStars – and there was no action at lower stakes to practice, which made solver even more valuable.
The first version of the solver relied on handcrafted strategies for preflop betting and the 1st draw, solving subtrees starting from the second betting round (mimicking the approach with solving flops in Holdem).
At some point we realized that tracking discarded blockers is crucial, and I managed to make solver remember a single discard, which later turned out extremely valuable.
I partnered with Trueteller and RaulGonzalez: I would have 40% of their result, gradually decreasing to 10% over two years (nowadays I charge more). At lower stakes, I would have a regular piece where I owe money if they lose. For the highest stakes, where bankroll swings were enormous and I could not take on losses, we devised an “Indirect Share” approach. This allowed me to earn a percentage of their EV with a freeroll arrangement. The percentage of freeroll was was calculated based on the volume of hands played.
At the $2000/$4000 tables, we encountered a rival group of professionals who also trained with their own private solver. However, theirs did not account for blockers, leaving their strategy vulnerable. By capitalizing on this weakness, Trueteller and RaulG won $700K over 40K hands against them. One of the key strategic differences were raises at turn after 1-1 discards.
Between 2013 and 2015, my bankroll grew from $20K to $1M. The swings were intense: RaulG’s best day saw a $704K win, while his worst day was a $346K loss. I did not handle such volatility well, as my previous poker career had much smaller stakes (I played $1K buy-in max and my biggest daily swing was $10K). To cope, I distracted myself by immersing in Dota 2, logging 2,000 hours in 2014 alone.
Here's RaulG's biggest pot. I never watched the games in real time, it would have been too stressful to see both of my partners fold to a $980 bet in $120K pot
In 2015, Trueteller used it to prepare for an 8-game match against Phil Ivey, studying with my solvers for TDFL, Razz and FL Holdem. Trueteller won $400K at the table over 40K hands, and even more with a cross-book. I made $200K from that match.
Later, during $3000/6000 Vegas live cash mix game, Trueteller’s dominance at Triple Draw was so overwhelming that opponents voted to exclude the game from the mix.
In 2017 I started working with Jungleman, who studied with TDFL, SDNL and Razz solvers, which contributed to his back-to-back wins of $50K WSOP Players Championship in 2021 & 2022.
Seeing my work help someone succeed on poker’s biggest stage was deeply satisfying.
Special thanks to Trueteller, RaulG, and Jungleman; also R Nikhil, Ian Chan, Joey Ingram, Ivan Bogatyy for their help with reviewing the draft of the story.
Read the full story here https://medium.com/@olegostroumov/how-i-...
For future stories and updates follow me on Twitter
–Oleg Ostroumov
Hello!
I've made a youtube video on history of poker solvers and I've included part of this story. Ejnoy!
Oleg,
Thanks for this thread. I'm curious if other players in the big mixed game work with their own solvers or have purchased your services. I'm also curious what games have the biggest gap in terms of solver play vs top players skills.
Ugh this all disgusts me. The OP disgusts me and the excitement shown by most of the posters makes it worse.
I just want to know:
A list of all the games he has solvers available for
And the price attached to each
I think I need this product!!! (and I didn't know I needed it until he made the thread/until I knew I needed it 😀 )
Oleg,
Thanks for this thread. I'm curious if other players in the big mixed game work with their own solvers or have purchased your services
Some people are purchasing my services.
Some have tools of their own, mostly rudimentary solvers or calculators as far as I know.
Some are just brilliant at raw learning and playing the game and exploiting opponents. There are few of them nowadays but they exist
I'm also curious what games have the biggest gap in terms of solver play vs top players skills.
Regarding the gap, I would say in NL Holdem the gap is the smallest. Top pros are damn good at NL Holdem. In all other games, there is gap.
Besides the loserate of a human against the machine, we also need to consider learnability. I would say:
Draw games are learnable
Razz and OFC are somewhat learnable
Omaha is hard to learn from pure solver training. There are just too many flops, ranges are hard to reason about, and patterns are hard to understand. It is also a UI issue, modern omaha training tools can't even show you a player's range on a single screen properly
Regarding Gus Hansen, he is a very strong and talented player. It's just he chose to compete against the absolute best, no surprise he lost
If Trueteller didn't tell Phil Ivey about his solver this is this is honestly kind of scummy
I wish the trainer was still working.
If Trueteller didn't tell Phil Ivey about his solver this is this is honestly kind of scummy
I used to play NL50, that's 50 cents, on PS 10 years ago. The entire reg zoom pool used PIO solver.
Isn't there some expectation when you're a world-famous pro flying around in your Falcon 2000 that opponent tool use might be a possibility?
I used to play NL50, that's 50 cents, on PS 10 years ago. The entire reg zoom pool used PIO solver.
Isn't there some expectation when you're a world-famous pro flying around in your Falcon 2000 that opponent tool use might be a possibility?
A couple of years could make a big difference as far as this goes.
In 2013 I could see someone not expecting solvers to be a thing.
In 2025 anyone who thinks online isn't filled with collision bots and rta is either naive or an idiot.
Draw solver demo is working at https://draw.olegsolvers.com – if it doesn't DM me a screenshot with a problem
I used to play NL50, that's 50 cents, on PS 10 years ago. The entire reg zoom pool used PIO solver.
Isn't there some expectation when you're a world-famous pro flying around in your Falcon 2000 that opponent tool use might be a possibility?
they absolutely did not, piosolver came out to the public near the end of 2015 lol
Well I had a copy around then and so did everyone else — or knew of and had access to it — in the HUNL forum, in which posted most zoom regs.
Edit: I should have clarified I am talking about the HUNL zoom pool. I never played 6 max then.
Draw solver demo is working at https://draw.olegsolvers.com – if it doesn't DM me a screenshot with a problem
Can you remove the need to login with google please. Seems completely unnecessary to demo something. thanks.
Very interesting read! Can't believe how much time has passed now since these events.
Were there any parts of the 2-7 solver's strategy that you found particularly interesting? The authors of a paper published in Science made a big deal out of the simplicity of the optimal HU-LHE preflop strategy (link). I'd like to hear if you found anything particularly striking, especially in the heads-up strategy.
Very interesting read! Can't believe how much time has passed now since these events.
Were there any parts of the 2-7 solver's strategy that you found particularly interesting? The authors of a paper published in Science made a big deal out of the simplicity of the optimal HU-LHE preflop strategy (link). I'd like to hear if you found anything particularly striking, especially in the heads-up strategy.
This is a good point.
I am v v interested in the story but due to flaws in my poker thinking I've neglected to consider the nuts and bolts of the solver.
(HU)NLHE is my game.
Therefore: outside of the incident with trueteller knowing the solver was legit when... the turn card paired the board and the solver recommended a lead.
Were there any parts of the NLHE solver strategy that was -interesting -nonintuitive -whatever?
How is someone just getting lucky to be born with a natural talent more impressive than someone putting insane amount of work to get good?
They weren't great players out of the womb. They figured out how to beat the game themselves rather than copying what a computer does. That is more impressive.
after rewatching the NOSEBLEED poker documentation on youtube I discovered the following conversation at min 53:
friend:
"did you get any action online (while beeing in london for a week for a break of vegas/wsop end of june 24)?"
seb86:
"... some 2/4 deuce against trueteller and san1ker (raulg) ..."
friend:
"how do they play by the way?"
seb86:
"they both decent now"
alexonmoon (from back of the car)
"its getting tougher"
another quote to this:
https://www.youtube.com/watch?v=Hq6675cy...
in this joeingram1 podcast from september 2016 he says at min 1:29 that his biggest downswing ever was a 2.5 Mio USD swing playing the germans in 2014
Another great story / thread Laegoose, thanks for sharing!
A while back, I started a thread here suggesting that Hold-em would be improved if you dealt two dead cards face-up at the start of the hand.
I struck me that, while this would have almost no effect on recreational poker play, it would make it an order of magnitude harder for solvers to find solutions, and several orders of magnitude harded to memorize them.
The fact that OP was only able to account for one dead card in his solver, but even doing this was able to significantly outplay solvers th
Was and still is a very good idea imo.
Very curious about the hand example posted, specially the 4 bet after first draw.
