Node-Locking: Modeling Mistakes to Find Exploits
Teaches node-locking: fixing one player's strategy at a node to a realistic mistake and re-solving so the solver computes the best response, turning the equilibrium tool into an exploit engine.
Assumptions: 100bb 6-max online cash, no rake unless stated. Numbers are no-rake unless the lesson states a rake structure. Solver-facing statements are solver-informed approximations, not newly generated exact solver outputs.
Node-Locking: Modeling Mistakes to Find Exploits
Teaches node-locking: fixing one player's strategy at a node to a realistic mistake and re-solving so the solver computes the best response, turning the equilibrium tool into an exploit engine. Runs two locks on Ks7d2c BTN versus BB: locking BB to fold 60% versus a 33% c-bet (solver responds by betting its entire range including pure air, and EV jumps measurably) and locking BB to never fold middle pair through two streets (solver stops bluffing turn and river entirely and value-bets thinner). Explains how to choose lock values from population data rather than guesses, and that the locked solve's output is exploitative, not equilibrium, so it inherits exploitability. The quiz should test predicting the direction of the solver's strategic response to a described lock before seeing the output.
The job of this lesson is to make node-locking: modeling mistakes to find exploits executable at the table. Advanced players do not need another slogan about balance. They need a way to look at a hand, name the strategic pressure, choose a sizing or defense frequency, and then explain the decision after the session without rewriting history around the result. The focus words for this lesson are node, locking, for, exploits. If you can state what those words mean in range terms, the rest of the spot becomes much easier to audit.
Start with the complete strategy view. A poker strategy is not a single clever action; it is a plan for every hand class at every decision point that can occur. That includes hands that bet, hands that check, hands that call, hands that raise, and hands that fold. In equilibrium language, a strategy is good when the opponent cannot improve by changing only their own plan while yours stays fixed. In practical language, that means your baseline should not leave an obvious door open: too many folds to one size, too many bluffs in one river line, or a checking range with no strong hands.
When this lesson mentions solver numbers, treat them as examples of how outputs are read, not as fresh solves run for this file. The quantitative claims that are exact in this track come from the repo tools: odds, combos, and simple algebra. Solver frequencies are described as solver-informed patterns because exact frequencies depend on rake, preflop ranges, bet-size menu, stack depth, abstraction, and convergence tolerance.
The first principle is: The tree you build is the game being solved. In this lesson that means you should describe the range before you describe the hand. A single combo can be tempting or ugly, but the equilibrium question is always about the whole distribution. Which hands arrive here? Which hands are clearly strong enough to value bet or stack off? Which hands have showdown value but hate facing a raise? Which hands are too weak to check and win? Once those buckets are visible, the individual combo is no longer floating in space.
The second principle is: EV beats frequency worship. This matters because poker language often confuses two different claims. One claim is defensive: this line is hard to exploit because the opponent is not offered a clean counter. The other claim is offensive: this line wins the most against the actual person in front of you. Both can be true, but they are not the same statement. If a player folds wildly too much, the max-profit response is to bluff too much. That response is intentionally exploitable if they notice. The baseline remains valuable because it tells you where to return when the read weakens.
The third principle is: Aggregates teach rules, single combos teach exceptions. EV is the language that keeps this honest. Frequencies are not moral rules. A defender who folds below MDF against a value-heavy nit may be printing money, while a defender who proudly hits MDF with dominated bluff-catchers may be burning it. A bettor who matches a textbook bluff ratio with the wrong blockers can still lose EV. The question is always what the marginal action earns against the opponent's range, not whether the final percentage looks elegant in a chart.
The fourth principle is: Node locks create exploit outputs, not equilibrium outputs. The purpose of study is not to memorize every mixed combo. It is to compress the solution into stable rules without lying about the cost of compression. A rule like small bet range on very dry high-card boards works because the raiser has broad range advantage and the defender has too many weak continues. A rule like check heavily on low connected boards works because the defender owns more two-pair, straight, and set density. The reason travels better than the exact frequency.
Here is the at-table diagnostic. Before acting, ask four questions in order. First, who has range advantage, meaning whose whole range has more equity? Second, who has nut advantage, meaning who can represent the strongest hands at meaningful density? Third, what price does the sizing offer, expressed as alpha for the bettor or MDF for the defender? Fourth, what does the opponent actually do differently from a competent baseline? If you cannot answer the fourth question, play the baseline. If you can answer it with evidence, deviate in the direction the evidence supports.
A common mistake is to use GTO vocabulary as decoration after the fact. A player bets because they feel weakness, gets a fold, and calls it range pressure. Or they call because they do not want to be exploited, lose to a value-heavy line, and call it MDF. That is backwards. The theory has to make a prediction before the action: this size needs this fold share; this range can defend these hand classes; this blocker removes these calls; this node is under-bluffed by this pool. Without that prediction, the words are just expensive labels.
Another mistake is over-precision. If you did not run a solve with the exact preflop ranges, rake, stack depth, bet sizes, and convergence tolerance, do not claim exact frequencies. It is honest to say that a texture is commonly high-frequency small bet territory. It is not honest to say that a specific combo bets 73.2% in the game you played unless you have that exact output. The lessons in this track use exact arithmetic where the repo tools support it, and use approximations for solver patterns that are sensitive to setup.
Worked hand 1
In the first hand, the important habit is to decide what the bet or check accomplishes for the range. If Hero bets, the bet should either get called by worse often enough, fold out enough equity, or apply pressure to a range that cannot continue at the required frequency. If Hero checks, the check should protect a range that needs strong hands or realize equity with a hand that does not benefit from building the pot. The board and formation matter more than the hand name.
A useful review question after this hand is: which better hands fold, which worse hands call, and which hands raise? If the answer is that worse hands rarely call and better hands rarely fold, the bet is not value and not a bluff; it is usually a merge or a mistake. If the answer is that the opponent has many indifferent bluff-catchers and your range has credible value density, the pressure is more defensible. That distinction is the heart of Node-Locking: Modeling Mistakes to Find Exploits.
Worked hand 2
The second hand forces the exploit question. A baseline can recommend one action, but the opponent profile can make another action better. The disciplined process is trigger, direction, size, and reversion. Trigger means the evidence that justifies leaving baseline. Direction means which frequency changes: more bluffs, fewer bluffs, thinner value, tighter defense, wider defense, more raises, or fewer raises. Size means how far you move. Reversion means what evidence sends you back.
This is where many players become too brave. If the only evidence is that the last bluff worked, there is no stable trigger. If the database shows a meaningful overfold or the live opponent repeatedly shows capped folds in the same node, then a deviation is reasonable. The adjustment should still be built from the baseline, not from scratch. You widen the relevant bucket first: the hands with blockers, backdoor equity, or poor showdown value. You do not start firing every random combo just because a formula says a bluff can be profitable at some fold rate.
How to drill this lesson
Review five marked hands and write one line for each: formation, board texture, range advantage, nut advantage, price, and action. Then add a second line saying whether the action was baseline or exploit. If it was exploit, name the trigger. If you cannot name the trigger, score the hand as a baseline miss even if the result was good. This review format is intentionally dry. It removes the emotional story and leaves the strategic claim exposed.
For off-table study, compare similar boards instead of random hands. Put a dry ace-high board beside a low connected board. Put an in-position 3-bet pot beside an out-of-position 3-bet pot. Put a heads-up pot beside a three-way pot. Ask what changed in range advantage, nut advantage, SPR, and defense responsibility. The point is not to memorize fifty answers. The point is to see which variable moved the answer.
The final standard for this lesson is simple: you should be able to teach the concept without referring to the last hand's outcome. State the baseline, state the arithmetic if a price is involved, state the range reason, and state the exploit condition. If those four pieces are present, the decision can be improved. If they are missing, even a winning hand history does not prove the line was good.
One more practical note: keep the language humble. Equilibrium is a model of a specified game, not a certificate that one click is perfect in every poker room. Rake, pool tendencies, bet-size restrictions, and player mistakes all move the answer. Strong players use GTO study to build a hard-to-exploit default, then use observation to earn more than the default. Weak players either worship the output without context or reject the model because humans make mistakes. The profitable middle is to know the model well enough to know when you are leaving it.
When you are unsure in game, prefer the action that keeps your range coherent. That usually means value betting hands that can be called by worse, bluffing hands that have poor showdown value and useful blockers, checking enough medium strength to avoid a hollow check range, and folding hands that only satisfy a formula while losing badly to the actual range. Coherence is not passivity. It is the discipline of making every action belong to a range plan.
The lesson also has a psychological edge. Once you know the baseline, you stop treating every tough spot as a referendum on your talent. You are solving a repeated node. Did the node match a known family? Did your simplification apply? Did the opponent give a reliable reason to deviate? Did the price make sense? This turns poker from a stream of emergencies into a set of decisions you can train.
Use the quiz to check definitions, not memory theater. The correct answer should follow from the framework even if the exact board is new. If you miss a question, find which step failed: range shape, price, indifference, exploit trigger, or implementation. That diagnosis is more valuable than the answer itself because it tells you what to drill next.