Node-locking in Rail: the best response to a read

Lock one decision of an opponent in Rail's solver and see the best response against it: what it gains, and what it risks.

The exploit lab locks one tendency of the opponent on a river Rail solves on demand, keeps the rest of the opponent's strategy at the solve, and computes your best response. A lock is your read of a player, not a fact, and the lab works on the river only; the ranges rest on Rail's narrowing model.

Approximate: the ranges here are hand-written teaching ranges or rest on Rail's narrowing model, not on a solve of the whole hand. The direction is what to take away.

In this lesson

Lock an opponent's strategy at one decision in Rail's solver.
Read the best response against the locked strategy.
Compare what the exploit gains with what it loses if the read is wrong.
Choose locks that match what your own data shows.

Read first: Adjusting to the player pool · Advice, not a lock: every lesson opens.

Concepts used: GTO and exploitative play, EV, EV loss and how Rail grades

What a lock is

Rail's river solve is the baseline: both players playing as well as the solve can. A lock fixes one decision of the opponent at a number you choose — how much more or less he folds to your river bets, that he never raises, how much of his air he bets first to act — and keeps everything else he does at the solve. Rail then computes your best response: the strategy that makes the most against exactly that opponent.

Keeping the rest frozen is the honest model of a read: "I think he does this, and nothing else changes". It is also the lab's limit. A real player who folds too much may also bluff differently; the lab does not invent that for him.

Three numbers, on a game you can check by hand

Take Rail's smallest test game: a pot of 10, a bet of 10, a bettor holding the nuts or nothing half the time each, and a caller with one bluff-catcher. At the baseline the caller calls half the time and the bettor bluffs half his air, worth 7.5 to him. Lock the caller to fold 75%.

A bluff now earns 0.75 × 10 − 0.25 × 10 = 5, so the best response bluffs every air hand: 0.5 × 12.5 + 0.5 × 5 = 8.75.
The baseline against the same caller makes 6.25 + 0.25 × 5 = 7.5, so the read gains 8.75 − 7.5 = 1.25.
If the caller really plays the baseline, a bluff earns exactly nothing, and the response still makes 7.5: no cost.
If the caller sees it and calls everything: 0.5 × 20 + 0.5 × −10 = 5, a cost of 7.5 − 5 = 2.5 — twice the gain.

Rail's lab returns exactly these numbers on this game; the tests check it.

Reading the lab on a river

On a real river the lab shows the same three numbers in bb per river, plus the baseline's own cost against a counter, which is small. Below them, the decisions the response changes most, hand category by hand category: what the solve plays and what the response plays. Pick a read, set its number, and run it.

Rail's exploit lab: a river solved in your browser, one read locked, the best response.

The lab works on the river: a spot dealt from the charts, ranges narrowed on the flop and turn by Rail's heuristic model, bets of 33%, 75% and 150% of the pot and all-in.

Rules of thumb

A lock is a read: one decision fixed, everything else at the solve.
Read three numbers: the gain, the cost if wrong, the cost if countered.
The baseline's own cost against a counter is the floor to compare with.

When they break

A player rarely deviates in one place only: the lab shows one read at a time.
The lab covers the river; earlier streets are the baseline's.

Example hands

Examples come from two places only: your own analysed hands in this lesson's spots, and a hand Rail deals and grades itself. Never a hand from outside Rail.

A hand Rail deals

Rail scripts this hand from the lesson's own practice, always the same one, and grades it with the analysis. Decide first, then read Rail's verdict.

Practice

The lesson is mastered when every exercise below that Rail can grade is passed. Missed items come back later as review cards.

Exercise 1

Pass: 3 of 6 right

Six rivers, each with a read locked: play the hand Rail's best response plays against it.