Something to send the player who already has everything.

What we do with a hand database

Everything is computed before anything is made.

You send an export. What comes back is a finished piece — a video, a set of cards, a page of text — but almost none of the work behind it is the making.

1. Normalise the export. Every platform records a hand differently; the adapter maps yours onto one shape and proves it by reconciling every pot to the cent.
2. Compute everything that can be counted exactly — equity at every decision point, EV wherever money went in before the last card, rake and rakeback by his share of each pot. Where a spot needs equilibrium, TexasSolver and GTO Wizard are run for it, and the piece names which.
3. The opponent's cards are the one thing the file never shows. So his range is built from what he actually did, street by street, against the width the field opens — and its composition is printed class by class. Anything built on a range arrives as a spread — 30 to 55 per cent — not as a single confident 37.5.
4. Check it against what is being published about the game. New analysis is read as it appears; a conclusion that cuts against it has to survive a second, independent calculation before it reaches the page.
5. Select against the objective you set — not against the biggest pot — at the scope you asked for, and say why that hand rather than another.
6. Only then build the piece — video, cards, an infographic, text — in the voice your player will actually stay with. Every number in it was settled before any of it was made.

What a general AI cannot do

No room for mistakes.

It goes out under your brand to the one player who can check it — he was at the table, often with a tracker open. Nothing else your room sends him is about a subject he knows this well.

Solver licences are ours. The field figures and your corrections stay with your room and are never carried to another.

Not a guess. We ran one of our own sessions through a capable model, with code and the whole file in front of it. The sums came back right. So did these:

• MisreadThree sets and an overpair called hands he had crushed — sixteen combinations on the wrong side.
• Wrong setThe share of opponents who would pay, counted over hands with no pair at all.
• Wrong count“Four to eight outs” for a draw that had thirteen.
• HindsightThe verdict argued from how the hand ended — known only because the file held the ending.

You cannot catch any of these by reading — they are written exactly like the truth. The player who was at the table catches them anyway.

Next step

Start with a pilot.

One piece a week to twenty of your VIPs, for four weeks, with a comparable group alongside receiving nothing. We measure active days and volume played against that group — not views, not engagement. Fixed fee, credited against the first contract. Tell us which players and what you want moved, and we come back with the plan and the number. No data moves until there is an agreement in place.