HydraSolver

Two products

Which one do I want?

One of them solves poker. The other reads a solve that is already finished. Most people should start with the second, and it costs nothing.

Tournament equity

Chip EV is not money EV

A double-up is not worth twice as much when there is a pay jump underneath you. The solver knows the difference, and it is the same engine either way.

• ICM with future game Switch from chip EV to ICM and payouts, players left, stacks and future game all feed the algorithm, so a stack is valued in real money rather than in chips.
• Bubble factors and risk premiums, on the spot Visible where you are reading, so you can see how ICM reshapes each decision rather than only the headline EV.
• Progressive and fixed knockouts Fair, Total T$ and Disbursed P$ options for PKO, plus fixed knockout. The bounty and the payout structure are inputs to the solve, not a number added afterwards.
• Mystery bounty, valued honestly Priced at the mean of the envelopes still in the pool — a mean EV rather than a sampled draw — with envelopes left and chips-per-dollar kept in the model.
• Large fields, not only final tables Accuracy stays stable when the field is big, so a day-two spot is as solvable as a nine-handed final.

Solving

Solve on your terms, on the machines you have

Stop when you say stop. Pause, resume and re-solve at any point, and read what the accuracy actually is while it runs.

• Split one batch across several machines Point a second computer at the same shared folder and press Join batch — the boards are handed out from one manifest, and the results land together. A desktop and a laptop finish a texture set in a little over half the time.
• Stop on exploitability, drift, convergence or time Set per solve, so every board in a batch finishes on its own criteria instead of on the slowest one.
• Accuracy you can read while it runs Exploitability as the headline number, with convergence and volatility beside it, plus a certificate and a best-response probe. When the abstraction is coarse or the maps are legacy, a banner says so rather than letting a clean-looking number stand.
• The cores you choose, and a memory ceiling Pick the worker count, or run single-worker when two solves have to match exactly. Cap the solve at a share of system memory so the machine stays usable while it works.
• Reset regret and reset strategies, separately Two tools for two jobs: clear the learning without wiping the strategy, or start the strategy over. After a node lock, the second is much the faster way back.
• A headless runner for the long ones Run and resume a batch without the interface, so a big set can go overnight on a server that has no screen attached.

Node locks and best response

Lock a strategy, then attack it

Equilibrium is the start of the question, not the end of it. What beats the way your game actually plays is a different solve.

• Pin the frequencies you want and free the rest Lock how a seat plays at a node and the solver works around it. Locks are inventoried, and you can release them one at a time when you change your mind.
• Best response against a locked table Pick a seat and the solver attacks every other strategy at the table. You get that solve’s leaks ranked biggest first, each row clickable straight to the spot that leaks.

Saving and exporting

Take it with you

Nothing you solve is trapped in the application, and nothing about a save depends on us still being here.

• Resumable .hdr saves Reopen a solve and keep calculating from where it stopped, or open a browse snapshot scoped to the street you solved.
• Export to the browser Write a .hdv and open it in HydraViewer with nothing installed. Lite exports, EVs and cleaned paths are all options when you need them.
• Formats other tools read PPT .txt, columnar JSON, CSV, Pio format and range export and import. One street or the whole tree.
• A batch comes back as files A board batch saves as .hydrasim alongside a frequency-weighted aggregate.csv with texture rows included.
• Projects, recent trees and recovery A project manifest ties a tree to its assets with checksums, recent trees are one click away, and an interrupted session offers to restore itself.

HydraViewer

Free, in your browser, and nothing is uploaded

HydraViewer reads solved spots. There is nothing to install and nothing to buy before you can use it — open the site and drop a file on it.

• Nothing you open ever leaves the device Solves are read in the browser on your own computer. There is no server to send a strategy to, and no account to make.
• Ten-gigabyte files open without being loaded A real solve is far too big to hold in a browser tab, so it is read in byte ranges and never held whole.
• Works offline once it has loaded Installable, and the shell, your library and your local files all open with the connection cut. Only the shop needs the network.
• Your library survives the tab closing The shelf, your place in each bundle, your session and your notes all come back — held in your browser, never uploaded.
• A hundred solves sort themselves into games Each bundle states the game it was solved for, so an import lands as a drawer per game with each flop under its own preflop solve.
• Untrained means untrained, never 0% A bucket the solver never visited is hatched rather than shown as zero, because a confident zero would claim it chose not to act.
• Every board in a pack, folded into one report The mean across boards with the spread behind it — a 60% mean built from 20% and 100% is two strategies, and the mean alone cannot tell them apart.
• It prices the decision beside the solution What a bet needs to win as a bluff, what a call needs, and minimum defence against the frequency the solve actually folds.
• Check your own chart against the solve Import a range from another tool and see exactly where it disagrees, priced in EV.
• Publish a chart on your own page A read-only grid you can embed anywhere, carrying its data in the link itself — so your readers fetch nothing from us and we learn nothing about them.
• Readable without seeing it One tab stop, arrow keys through the grid, every cell announcing its full mix, and the same numbers as a plain table on request. A colourblind-safe palette that still composes with dark mode.

Free packs to start with, and paid packs from $19 to $89. Five packs take 10% off, ten take 20%, twenty take 30%.

Licence

One price, no tiers