How we solved your game.

Felt Reps doesn't borrow anyone else's charts. We wrote our own solver from scratch and ran it on the two live games we drill, $1/$3 and $1/$2.

What goes in

A solver only knows the game you describe to it. Before the flop, we described yours in dollars.

Every mix of stakes, buy-in, table size and straddle is a preflop solve of its own, 60 in all. Change a setting in the drill and the preflop answers change with it.

After the flop, we solved seven heads-up pots. At 100 big blinds: the button's raise called by the big blind, the cutoff's raise called by the big blind, the cutoff's raise called by the button, and the cutoff calling the button's 3-bet. At 150 big blinds, all of those but the button calling the cutoff's raise. Each one starts from the hands a nine-handed table with no straddle takes to that flop, and the same solve serves every table size and straddle setting.

The rake is the one we modelled: 10% of the pot up to $15, taken only when there's a flop. At $1/$2 with a $200 buy-in, the drill uses the $1/$3 flop solves scaled to the pot, so after the flop the cap it models there is $10. If your room takes a different cut, a few close answers could come out differently.

What comes out

At every decision, the solver says how often to make each play with each hand. Each of those numbers, how often to take one action with one hand at one decision, is a frequency. The flop solves and their follow-ups worked out 1.8 trillion of them, and the follow-ups go over some of the same ground.

Before the flop, there are 8,552 spots you can be dealt across the two games. The 60 solves behind them hold 897,190 answers, one for every hand a player can hold at each decision.

After the flop, there are 96,095 decision points you can be dealt, with every turn and river card counted on its own, and 5,107,567 answers, one for each exact pair of cards you can hold there.

How your answer is graded

Preflop, the play the solver makes 65% of the time or more is the answer. If no play gets there, the hand is close, and either of its top two plays counts. After you answer, you see how often the solver makes each play.

After the flop, the answer is the play that wins the most with your hand. A play that wins almost as much counts too, and the grade tells you, in your own money, what the other plays cost.

How we checked it

For every flop we solve, we ask one question. If a player knew our strategy exactly and played the best counter to it, how much could they win? We work it out for each of the two players, as a share of the pot when the flop comes down.

The bar is 0.3% of the pot for each player: 30 cents in a $100 pot. All 1,750 flop solves clear it.

Each decision later in the hand is measured too, against its own pot. If one goes over the bar but stays under 1.5%, the drill still deals it, tells you how close the solve is, and counts any play that close to the best as right. If it goes higher, the drill doesn't deal it.

• The player with more to gain, averaged over every flop: 0.29% of the pot.
• Both players, averaged over every flop: 0.20% of the pot.

Solver people call this number exploitability. It's the usual way to say how close a strategy is to GTO, game theory optimal, where zero would be exact.

Preflop is checked its own way, and the flop's 0.3% doesn't apply to it. Each preflop solve ran until no seat could gain more than a hundredth of a big blind a hand by changing how it plays, about 3 cents at $1/$3, and the worst of the 60 stopped at 0.0071 big blinds. That's measured inside the solver's own model of the game, so read it as how finished the solve is, not as a promise about your table.

How much work went in

There are 1,755 flops that play differently from one another. We picked 250 of them to stand in for the rest, and within each kind of flop the drill deals them as often as they come up at a real table.

We solved those flops in each of the 7 heads-up pots, 1,750 flop solves, plus 2,990 follow-up solves on top of them.

The last round of flop solving alone took 4,038 hours of processor time.

All told, that's 104,647 spots you can be dealt.

What's left out, and why

• Pots with three or more players after the flop. They aren't solved yet, so the drill doesn't deal them.
• At $1/$2 with a $300 buy-in, the button calling a raise from the cutoff, which we didn't solve that deep. Also at that depth: check-raises and river bets when the button raised and the big blind called, and flops of one suit when the cutoff raised and the big blind called. In those spots the big blind's calling hands are so few and so easy to read, or so much of the solve missed the check, that there's nothing fair to grade.
• Any decision after the flop that misses the check by a wide margin.

We don't grade what we haven't solved.

It keeps improving

We keep refining the solves. A better answer reaches the drill from our servers, so you never have to update anything.

GTO, and your table

This is the GTO baseline for your game. Exploits are how you beat your table, and you can't spot one until you know the baseline. The drill teaches the baseline; the fish at your table show you the exploits.

Try it free

The first 10 spots need no account and no card.