River Decision Algorithm

Streets · updated 2026-03-08

This section can be turned into a practical river algorithm that keeps a GTO baseline while adding an exploit switch (the idea: know the theory, spot deviations, punish them).

Below is a version that can be applied at every river decision point.

1) Start with theory: what should happen in equilibrium

A. What type of node is this (this decides almost everything)

• You are OOP and choose the sizing: often a split strategy: block (10–25% pot) + big or overbet (75–150% pot).
• You are IP and choose the sizing: usually more weight on larger bets, but blocks and mixes appear on some textures.
• You are defending vs bet or vs jam: immediately compute alpha/MDF and your defense range.

B. Node math (fast)

• Against a bet B into pot P:
• MDF = P / (P + B)
• Fold percentage (alpha) = B / (P + B)
• For your river betting plan:
• Optimal bluff share ≈ alpha (same B / (P + B))
• Value share ≈ 1 − alpha

This is the standard risk versus reward logic.

C. Hand classes and sizing

• Block (10–25%): thin value plus some strong hands only if you are prepared to face a raise or 3-bet jam (block-jam) and the opponent actually raises.
• Big (75–150%) or jam: nuts and near-nuts plus the best bluffs (blockers to calls, unblockers to folds).

Key insight:

If the opponent under-raises blocks, the strongest hands should not hide in block bets. They perform better using big bets, overbets, or jams directly.

2) Exploit diagnosis: where exactly the opponent makes mistakes

You need opponent typing, otherwise exploits are unfocused. Basic archetypes with typical river leaks:

• Passive reg or nit: under-bluffs, under-value-bets, few river raises
• Overfolds vs large bets
• Under-raises vs blocks
• Recreational calling station: overcalls, under-folds vs value, random bluffs, often cannot raise thinly
• Remove bluffs, expand value
• Aggressive reg: capable of raising blocks and punishing checks, may over-bluff some lines
• Respect MDF and combo control, avoid auto-overfolding

3) Six river questions that cover most decisions

Q1) Is my range capped or uncapped?

• The capped player (cannot have nuts) folds more vs large bets.
• An uncapped range can polarize.

Q2) What are my two value classes and which sizing do they prefer?

• Thin value (top pair or second pair in the right nodes): block or medium.
• Nuts or near-nuts: big, overbet, or jam.

Q3) What must raise a block in theory, and does the opponent do it?

• If the opponent does not raise blocks, blocks become low-risk value:
• Bet thinner for value with blocks.
• Move the strongest hands into large sizings, because extracting value via raises is unrealistic.

Q4) Are there natural bluffs in this line?

• A natural bluff is a missed draw that logically arrives at the river.
• If natural bluffs are few or nonexistent, the population almost always under-bluffs.
• Bluff-catchers become clear folds, especially in live play.

Q5) Is my best bluff blocking calls or unblocking folds?

Priority order: 1. Unblockers to opponent folds, 2. Blockers to opponent calls (nuts and strong calling hands), 3. No showdown value.

Q6) How expensive is a mistake (ICM, stacks, stage)?

In MTTs, river exploits must account for ICM: - Under ICM, the calling threshold is higher, variance is reduced, and folds gain dollar EV. - Practical adjustment: near the bubble or final table, fewer hero calls, more thin value with small sizings, and more overfolding vs polarized jams from passive players.

4) Two exploit switches

Switch 1: Opponent overfolds

• Bluff aggressively with hands that qualify by blockers.
• Especially effective vs capped lines and passive branches.

Switch 2: Opponent overcalls

• Cut bluffs.
• Expand value and size up.
• Against calling stations: value betting dominates, bluffs approach zero.

5) Quick combo counting (for decisions vs jam)

Rule-of-thumb:

• Suited: 4 combos (minus blocked suits).
• Pairs: 6 combos (AA to 22), minus blockers.
• Offsuit: 12 combos, minus blockers.

Then apply: Value combos versus bluff combos, compare to pot odds via alpha or MDF, and decide.