GTO, human: where the solver ends and exploitation begins
GTO is not «play like a robot.» It is a map of what happens when nobody leaks — and almost every table at $1/$2 through $5/$10 is a leak festival. Solvers teach you what *uncounterable* looks like. Exploitation starts the moment you can name which assumption in the model is false for the villain in seat four.
What solvers actually give you
A solver answer is frequency + sizing + range composition for a spot with fixed assumptions: stack depth, rake, bet sizes allowed, opponent strategy. It does not know your villain limps 40% from CO or triple-barrels air on paired boards. What it gives you is language: if you are supposed to defend 43% on the river and you fold 60%, you are donating — whether the opponent is balanced or not.
The three exit ramps from GTO
Preflop width
Pool opens too wide → 3-bet linear and squeeze wider than charts. Pool nits → steal and flat IP wider.
Sizing tells
Min-bet river = weak or blocking; pot on turn = polar. Solver assumes mixed sizes — humans broadcast.
Showdown tax
If villain never bluffs rivers, solver bluff catches become pure folds. If they never fold, value bet thinner and stop bluffing.
How I study without living in a sim
I run one solved tree per week — usually BTN vs BB single-raised pot or SB vs BB 3-bet pot — and export three frequencies: flop cbet, turn barrel, river bluff. That is the compliance target for anonymous pools. Everything else is note-driven: when a stat crosses a threshold (WTSD > 32%, fold to cbet < 38%), I pick one exploit and run it for 500 hands before adjusting again.
- Solver first on spots you spew most — for most regs that is 3-bet/fold and river bluff-catch.
- Exploit second only with a written hypothesis («folds too much to 1.2× pot»), not «felt like a fold.»
- Re-solve when you move stakes — pool texture changes faster than your ego admits.
I remember when I balanced against a calling station
I remember firing three barrels with A-high because «solver mixes here» against a guy who had WTSD 41% over 800 hands. He snap-called with third pair twice in one session. Solver line: +EV vs balanced defense. Real line: −4.2bb average on those runouts. The exploit was check back turn and value bet river thin — boring, printable, profitable.
Pick one solved spot this week, log 20 river decisions against your pool on NUTS POKER cash tables, and mark where villain ranges clearly are not solver nodes. The gap between chart and reality is where your edge lives — if you measure it instead of guessing.
