Built by a player who
decoded the solver

The story

I play PLO. I built SolvePoker (it launched as SolvePLO) because studying it properly was harder than it should be.

What I wanted was simple to describe: the full solver answer for any board I'd just played, on screen in seconds, at a price that makes sense at the stakes most of us actually play. What existed was either priced for high-stakes pros, covered a fraction of the 1,755 flops, or meant running my own MonkerSolver sims, which takes serious hardware and days of compute per configuration.

So I built the thing I wanted. Complete board coverage, precomputed, in the browser, for less than a buy-in at PLO100.

The decode

SolvePoker exists because of one specific piece of work: I reverse-engineered MonkerSolver's binary output format.

MonkerSolver writes its solutions in an undocumented, compressed format designed to be read only by its own software. I spent months mapping it byte by byte: the encoding for strategy frequencies, the hand-indexing scheme, the action ordering, the tree structure, the compression layers. The result is a parser that reads Monker's own files and extracts the complete strategy, preflop through river, every hand, every action frequency, at the half-percent resolution the files store.

Then I precomputed solutions for all 1,755 strategically distinct flops and built a site to serve them. Pulling up a board is a lookup, not a solve; the data streams to your browser about a megabyte at a time.

In 2026 the same pipeline brought NLHE into the library, on the same subscription. Because the decoder reads solver files rather than approximating them, nothing on screen is interpolated, rebuilt from equity buckets, or smoothed by hand. If the solver mixed, you see the mix. The features page lists exactly what's inside, and what isn't.

How we decide what to claim

The rule on every page of this site is that a claim has to be checkable. The spec sheet says what's solved and what isn't, including the parts that cost us sales: no multiway postflop, no PLO5 solves, no HUD. The reference pages publish numbers straight out of the library, like c-bet frequencies aggregated across all 1,755 flops and every one of the 16,432 PLO starting-hand classes ranked, most with CSV downloads, so you can audit the data before you pay for the interface to it.

Pricing follows the same logic. Core is $60 a month and covers both games, because the players who most need this data are grinding stakes where study tools compete with buy-ins. The free tier is the trial: it never expires, and every preflop range in the library is in it. If the product is wrong for you, we'd rather you find out on the spec sheet than in month two of a subscription.

Support goes to [email protected]. A person reads it, and a report of a node that looks wrong gets answered faster than anything else, because that's the kind of bug we most want to know about.