Node Locking Turns a Solver Into an Exploit Tool

Node locking means telling a solver that somebody will play badly, then asking what to do about it. You fix one player's strategy at a single decision — this opponent folds 70% of the time to a flop c-bet — and re-solve the rest of the tree. The solver drops the unexploitable line and hunts for the most profitable counter to that flaw. What comes back is an exploit you can take to the table.

What a locked node changes

A normal solve assumes two perfect opponents. If you have never seen the mechanics, how a poker solver works comes down to two strategies improving against each other until neither side gains by deviating.

Locking freezes one branch of that process. You hand the solver a fixed strategy for one player at one decision point and forbid it from changing. Everything else stays free to adapt.

The output is no longer GTO for either player. It is the maximally profitable response to a specific flaw, and it is wide open to counter-attack itself. That trade sits at the heart of all exploitative adjustments: you give up your own defence to attack somebody else's.

Lock the mistake, not the response. Fix only the frequency you have evidence for, and let the solver work out everything that follows.

What is actually worth locking

The lock has to come from something you observed. Most usable ones come straight from HUD stats like fold to c-bet, because those are the numbers where you already have a sample.

Three tests before a read becomes a lock:

• Sample. Flop stats need a few hundred hands, turn and river stats need thousands. A 100% fold to river bet over six hands is noise wearing a suit.
• Size. The gap from baseline has to be big enough to be worth chasing. A player folding 55% where equilibrium wants 50% barely moves your strategy. One folding 70% changes everything.
• Stability. Lock behaviour the player repeats without thinking. Habitual leaks survive; deliberate choices get corrected the moment you attack them.

A worked example: the opponent who folds too much

You are playing $0.50/$1. You open the BTN to $2.50, the BB calls, and the pot is $5.50 on a flop of K♦8♠3♥. Your HUD says this player folds around 65% against flop c-bets.

Lock the big blind to that 65% and re-solve. Two things happen immediately.

Your betting frequency goes near universal. At equilibrium a hand like 7♣4♣ has no business betting this board. Against a 65% folder it is free money: betting $1.80 into $5.50 needs to work about 25% of the time, and you are being paid at more than two and a half times that rate.

Your sizing shrinks. There is no reason to risk $4 to buy a fold you can buy for $1.80. Small and relentless beats large and occasional against anyone who folds on reflex.

The third change is the one people miss. Your strongest hands get worse, not better. K♥K♣ wants a calling station, and this opponent folds instead, so the solver starts checking some monsters back to keep the folding range in the pot.

Reading the output

Every locked solve moves the same three dials, and it is worth checking each one deliberately rather than eyeballing the colours.

Locked leak	HUD signal	What the solver does with it
Folds too much to flop c-bets	Fold to c-bet above 60%	Small bets with close to the entire range
Never folds on the river	High WTSD, low fold to river bet	Bluffs vanish, thin value bets grow
Calls opens far too wide	VPIP 40%+ with a low PFR	Wider value opens, bigger sizings, fewer speculative hands
Never 3-bets light	3-bet under 3%	Folds a lot more of your opens, stops paying off
C-bets every flop	C-bet above 80%	Check-raises widen sharply, floats increase

Notice that none of the counters are exotic. Node locking rarely invents a play you have never seen. What it does is tell you how far to push one you already know, which is the part intuition gets wrong.

Where node locking goes wrong

The failure mode is not bad maths. It is confident maths applied to a read that was never there.

Locking both players. If you fix strategies for the two of them, the solver has nothing left to optimise and you are simply watching two guesses collide.

Locking a node nobody reaches. A brilliant counter to a turn check-raise that occurs in 2% of hands earns you almost nothing. Lock the decisions that come up every orbit.

Forgetting they can adjust. Exploits decay. The moment a regular notices you are betting every flop small, your unbalanced range becomes the target.

Generalising the output. The strategy you just built is correct against one player, on one board texture, at one stack depth.

Locking preflop nodes

Postflop gets the attention, but preflop locks pay better, because preflop leaks repeat every single orbit and almost nobody at low stakes has fixed theirs.

Against a high VPIP opponent who calls raises with 40% of hands, lock their calling range that wide and re-solve your opens. Your CO and BTN ranges widen, your sizing goes up, and the hands that gain most are the ones that flop well against junk — suited broadways and pairs, not offsuit aces.

Run the mirror case too. Against a player who only 3-bets QQ+ and AK, lock that range and watch how many of your standard 3-bet calls turn into folds. That is one of the cheapest adjustments in poker and you will make it several times a session.

Start from the solver preflop range charts as your unlocked baseline, then lock one opponent tendency at a time. Comparing the exploit against the baseline is what tells you the size of the edge; the exploit on its own tells you nothing.

Fitting it into your study

You cannot node lock at the table. The work has to compress into something you can recall in three seconds under pressure.

1. Pick one opponent type per session. The over-folder, the station, the nit. Not a specific player, a type you face constantly.
2. Lock one node. Change a single number from the baseline so you can attribute the difference to it.
3. Write down the delta, not the strategy. "Against over-folders, c-bet everything for a third pot on dry boards" is portable. A grid of frequencies is not.
4. Check what it costs when you are wrong. Re-solve with the opponent playing well against your exploit. If the loss is small, the adjustment is safe to make on thin evidence.

That last step separates players who use solvers well from players who collect outputs. The best exploits are the ones that barely lose anything when the read fails, and node locking is the only tool that tells you which of yours qualify.