The 7 Pillars of Poker Strategy
and what our solver says about each

Bet sizing is the least intuitive part of solver play. You can get range construction roughly right and still hemorrhage expected value by picking the wrong size — and the solver's sizing logic depends on conditions that most players don't check.

This pillar covers six sizing theories we tested against our model. Four confirmed cleanly: the wetness parabola (why small bets work on dry boards and wet boards but not the ones in between), the two-condition rule for large sizing, the rarity of flop overbets, and geometric multi-street sizing. One — the blocker-driven direction of sizing on flush-draw boards — confirmed with a refined direction (the original literature had the sign backwards, and the model corrected it). And one — river sizing splits by blockers — is partially verified because we can confirm the frequency pattern but not the underlying EV mechanism with current infrastructure.

Every table below measures CO vs BB in a single-raised pot at 100bb effective unless the table sweeps a different dimension. Position and stack depth matter for sizing; changing either changes the numbers.

Measurement conditions: 6-max NL, CO vs BB SRP, 100bb effective unless the table sweeps stack depth.

Why small bets on dry boards, big bets on wet, and small again on very wet

Sizing does not grow in a straight line as boards get wetter. It follows a curve — small on dry, larger on connected, then back to small (or check) on monotone. That non-monotonic shape is the most important sizing pattern in the solver's flop strategy.

Here is the solver's c-bet approach across 12 flop textures, all from the same spot:

CO's flop c-bet strategy by board texture. 100bb effective · 6-max NL · CO vs BB SRP

Source: cash-baselines.md lines 63–78 (Raw Data Flop C-bet, CO vs BB, SRP, 100bb)

The pattern: dry K-high boards get frequent small bets (83.6% at 2.4bb). Connected boards step up the average size but drop the frequency. Monotone boards drop both — K94ss bets just 32.2% of the time, and when it does bet, it sizes small (1.9bb).

Why does the gap grow?

The paired boards (KK5 at 1.8bb, 772 at 2.0bb) show the small-sizing extreme. CO bets frequently but tiny — BB's junk folds to any size, so the solver picks the size that maximizes call frequency from hands that are already behind.

You cannot bet big unless two things are true

Most players know to bet big when they have a strong range. Fewer realize that range strength is only half of what the solver looks at — and getting the second half wrong is the main reason sizing mistakes compound across streets.

The solver bets big when both of these are true:

1. You have more strong hands than your opponent (nut advantage)
2. Your opponent has hands that will actually fold (fold equity)

Miss either one and big bets stop working. Here is what that looks like across the board texture gradient:

On K72r, CO has total nut advantage — AA, KK, AK, every set, every AQ. BB's range is capped after just defending preflop. And BB's range is full of low-equity junk that will fold to any bet. Both conditions met. The solver bets 83.6% of the time.

On 654, the nut advantage flips. BB has more straights (75, 53) and more two-pair (65, 54) because those hands are in BB's defending range but rarely in CO's opening range. CO's c-bet drops to 56.4%.

On K94ss (monotone), CO might still have some nut advantage (AK, KK), but BB's range is loaded with flush draws and made flushes. Fold equity collapses. The solver bets just 32.2%.

The sizing data from the per-combo analysis confirms the pattern at the hand level. On 543, premium hands (AA/KK/AK) check 85.3% of the time — the solver does not force big bets into a board where CO lacks nut advantage. On AK6r, premium hands use the overbet sizing (bet8.2) at 18.75% — both conditions are present and the solver takes the large size.

Hand examples from cash-baselines.md D3 verdict and batches_cash_sizing_per_combo (K72r, 654, K94ss, 543, AK6r — all CO vs BB, 100bb SRP)

Flop overbets are vanishingly rare — and driven by bluffs, not value

If you think the solver bets bigger with its best hands and smaller with its bluffs, the data says the opposite. Across six boards, value hands and bluff hands converge on the same dominant bet size — and when they diverge, it is the bluffs that go bigger.

Here is the per-hand sizing breakdown from six tested boards:

Per-hand dominant bet size, CO flop c-bet. 100bb effective · 6-max NL · CO vs BB SRP

Source: cash-baselines.md D7 verdict worksheet (batches_cash_sizing_per_combo, 6 boards, CO cbet, 100bb)

On five of six boards, value and bluffs share the same dominant bet size. On zero boards does value use a larger size than bluffs. And on AK6r — the one board where the solver overbets — the overbets come from QJs (63%) and QTs (52%), not from AA (which uses bet1.8 at 96%).

This is the opposite of the common recreational pattern where players size up with their best hands "to get value." The solver treats sizing consistency as non-negotiable. Using different sizes for value versus bluffs creates a size tell that a competent opponent will exploit.

Geometric sizing: the pot grows roughly 2.5–3× per street

Across streets, the solver sizes each bet so the pot grows geometrically toward a target river pot. The progression is visible in the raw river data:

River bet behavior on three runouts, CO vs BB. 100bb effective · 6-max NL · CO vs BB SRP · river · specific boards per row

Source: cash-baselines.md Table 15 (batches_cash_river_strategies, 3 runouts, 2026-04-12)

The flop c-bet was preset at 3bb. The turn bet was 7bb. On the K72r runout, the dominant river size is bet19.1 (53% of the time) — a progression of roughly 3bb → 7bb → 19bb. Each street's bet grows the pot so the next street's bet can be a similar fraction of the new pot.

On the T98 runout, the solver reaches further. bet63.8 at 23% is an overbet — approximately 250% of pot — used on a dynamic board where the range is maximally polarized by the river. The geometric frame still applies: the solver is building the pot so that the final bet gets all the money in.

Shallow stacks compress sizing and widen frequency — but not in a straight line

At 20bb, there is no room for multi-street maneuvering. The solver responds by compressing sizing toward small bets and shoves — the mid-range bet sizes disappear. But what happens to frequency is less intuitive.

Here is the solver's c-bet rate on K72r across four stack depths:

CO flop c-bet frequency on K72r by stack depth. 6-max NL · CO vs BB SRP · K72r · stack depth sweeps rows

Source: cash-baselines.md Table 2 and D5 verdict worksheet (batches_cash_20bb_shortstack, batches_cash_deep_stack — 2026-04-12)

The relationship is not monotonic. Frequency peaks at 50bb (89.6%), not at 20bb. At 200bb, the solver bets less often — deeper stacks give the opponent more room to outplay you on later streets, so the solver checks more to protect against that possibility. And at 20bb, the frequency is slightly lower than 50bb because the SPR compression forces a different strategic logic: hands that were thin value bets at 50bb become shove-or-check decisions at 20bb, removing the middle of the betting range.

The A94r comparison sharpens the stack-depth effect. At 20bb, CO c-bets A94r at 98.4% — almost never checking. At 100bb, the same board gets 64.9%. The gap is 33.5 percentage points. When SPR is low enough that two streets of normal sizing get the money in, the solver shifts from selective betting to near-universal betting.

Hand examples from cash-baselines.md D5 verdict (A94r 20bb cbet 98.4% vs 100bb 64.9% — CO vs BB SRP)

River sizing splits by blockers — the frequency pattern is confirmed, the mechanism is under-isolated

On boards with flush-draw texture, holding a card that blocks your opponent's flush combos changes everything — not just whether you bet, but how big.

The data comes from per-combo suit filtering on T98ss (Ts9s8c), isolating combos with and without the spade blocker:

Per-combo c-bet behavior with and without flush blocker on T98ss. 100bb effective · 6-max NL · CO vs BB SRP

Source: cash-baselines.md D6 verdict worksheet (batches_cash_blocker_sizing, T98ss CO cbet, 100bb — 2026-04-14)

The direction is clear and consistent: holding the flush blocker shifts the solver toward smaller sizes. Without the blocker, the solver uses larger sizes. This was the opposite of the original literature framing, which suggested blockers "allow larger sizes." The model corrected the direction — holding a card that depletes your opponent's flush draws means even a small bet achieves the fold equity you need, because their calling range is already thinner.

The AA frequency effect is the most dramatic: 90.8% with the blocker versus 33.6% without. The solver almost never bets AA on this board unless it holds the spade.

This finding is board-dependent. On A94ss (As9s4c), the blocker effect was near-absent — KK, QQ, and JJ checked 87–99% regardless of suit. On that board, CO is range-disadvantaged (BB defends with many Ax combos), and the blocker's impact on the flush draw portfolio is small relative to the overall range dynamics.

This specific measurement is confirmed on the frequency and sizing layers, but the underlying EV mechanism is under-isolated — see the Research notes at the end of this pillar for why.