α Is Not Your Bluff %: Counting River Combos by Bet Size
June 25, 2026
Pot bet ≠ 50% bluffs. Walk the indifference math, rebuild the sizing ladder (33%–200%), and see where solvers quietly refuse the chalkboard.
- GTO
- value to bluff ratio
- bet sizing
- alpha
- MDF
- polarized river
- poker math
Scroll poker Twitter long enough and you’ll see the same wrong sentence twice a week: “Pot-sized bet → 50% bluffs.”
That number is α—the fold frequency a pure bluff needs to break even. It is not the share of bluffs inside your betting range. Confuse the two and you over-bluff every river you “balance.” DeucesCracked’s 2026 bluff-frequency writeup gets the split right: pot bet → about one bluff per two value. PokerSkill’s bluff-to-value glossary frames the same idea as matching the caller’s break-even equity. Plenty of other pages quietly swap α into the bluff column. Don’t copy them.
Build One Spot Before You Memorize a Chart
Pot 100. You fire 100 (pot). Villain is looking at a river with no draws left—classic polarized node: nuts-or-air from you, bluff-catchers from them.
Their call risks 100 to win 200, so they need 33.3% equity to break even:
[ \frac{B}{P + 2B} = \frac{100}{300} = \tfrac{1}{3} ]
Indifference for a bluff-catcher means: when you show up with a bluff that often, calling is zero EV. So one-third of your betting range should be bluffs, two-thirds value. Ratio: 2 : 1.
If you only have 12 clean value combos on this runout (sets, two pair, the nut straight—whatever the board actually deals you), you don’t invent air until the chart “feels full.” You attach:
[ 12 \times \alpha,\quad \alpha = \frac{B}{P+B} = \frac{100}{200} = 50% ;\Rightarrow; 6 \text{ bluff combos} ]
Twelve value, six bluffs. Bluff share = 6/18 = 33%. α was 50%, but α scaled the count of bluffs against value, it didn’t become the percentage on the pie chart. That identity—bluff combos ≈ α × value combos—is the only table shortcut worth tattooing.
Three Labels, Three Jobs
Keep these in separate mental drawers:
| Name | Formula (bet (B), pot (P); (s=B/P)) | What it answers |
|---|---|---|
| α | (B/(P+B) = s/(1+s)) | How often does a zero-equity bluff need folds? |
| MDF | (P/(P+B) = 1/(1+s)) | How often must the defender continue so you can’t print with pure air? |
| Bluff % of bet range | (B/(P+2B) = s/(1+2s)) | How many of your bets are bluffs if you want their catchers indifferent? |
Value share is whatever is left: ((1+s)/(1+2s)). Value:bluff simplifies to ((1+s)/s = 1 + 1/s).
RiverOdds’ sizing tables often land near the right bluff percentages for small and pot sizes, then their FAQ text redefines α as “bluffs in range”—so the same site can teach both the correct 33% and the broken 50% for a pot bet. BeyondGTO’s fundamentals series walks the indifference idea well, then prints Value:Bluff = Bet:Pot, which for a pot bet collapses to 1:1. Treat popular explainers as starting points, not oracles.
The Ladder You’re Actually Allowed to Use
Same formulas, seven common sizes. Numbers recomputed; no rounding theater.
| Size (s) | α (folds a pure bluff needs) | MDF | Bluff % of your bet range | Value:Bluff |
|---|---|---|---|---|
| 33% (⅓) | 25.0% | 75.0% | 20.0% | 4 : 1 |
| 50% (½) | 33.3% | 66.7% | 25.0% | 3 : 1 |
| 66% (⅔) | 40.0% | 60.0% | ≈28.6% | 2.5 : 1 |
| 75% (¾) | ≈42.9% | ≈57.1% | 30.0% | ≈2.33 : 1 |
| 100% (pot) | 50.0% | 50.0% | ≈33.3% | 2 : 1 |
| 150% (1.5×) | 60.0% | 40.0% | 37.5% | ≈1.67 : 1 |
| 200% (2×) | ≈66.7% | ≈33.3% | 40.0% | 1.5 : 1 |
Read the table left to right the way you actually decide at the table: pick a size because of nut advantage / SPR / how capped they are—then ask how many bluff combos that size buys you. Don’t reverse it (“I feel like bluffing a lot, so I’ll overbet”) unless the board and your nut share already support the polarization.
As (s \to \infty), bluff share approaches 50% from below. Even a grotesque overbet still wants at least as much value as air. If your river overbet range is majority bluffs by combo count, a competent caller just calls wider and you donate.
Half-pot is the one players under-bluff least often in theory discussions and still butcher live: 3 value per bluff, not 2. Two-thirds sits awkwardly at 2.5:1—fine to remember as “a hair under three-to-one.” Overbets are where under-bluffing is epidemic; 1.5× wants 5 value : 3 bluffs, not “I’ll only jam the nuts.”
When Solvers Refuse the Chalkboard
This math assumes a perfectly polarized river: value wins ~100% when called, bluffs win ~0%, one size, no future streets. Real solver outputs deviate for boring, important reasons:
- Thin “value” isn’t 100%. Second pair you’re betting “for value” loses often enough that the equilibrium tilts more value-heavy—or the solver checks some of those hands. Don’t force a 2:1 pie chart onto a merged mess.
- Blockers and removal. Combos aren’t equal. Solvers overweight bluffs that block strong calls and unblock folds; they underweight air that blocks the folds you need. Frequency can look “off” the formula by a few points and still be correct.
- Multiple sizes. Split a node into 33% and 125% and each bucket needs its own ratio. Mixing sizes without partitioning value/bluff is how mid-stakes regs invent free money for anyone with a HUD.
- Flop and turn. Semi-bluffs carry equity and future barrels. You can (and should) bet “more air” than the river formula allows, because a lot of that air is still a draw. The river formula is a river tool.
Acevedo-style modern theory and solver study culture both push the same caveat: indifference math is the scaffold; node-locked population tendencies and card removal do the finishing work.
Population Leaks (Use Sparingly)
MDF tells you what a balanced defender must do. Humans don’t.
- They fold more than α allows → your pure bluffs are +EV even above the chart. Add combos, or keep size and widen the bluff bucket. Don’t celebrate by sizing down into a merged range they suddenly love calling.
- They fold less (call light vs the price) → cut pure bluffs; thicken value; sometimes size up so the same sticky calling range pays more when you’re ahead.
Live read quality is usually garbage compared to a solver node-lock. Bias your default toward the chart, then nudge—don’t rebuild your entire river book around three showdowns.
If you want a mechanical check later, dump a polarized river into any trainer that shows combo counts and verify the bet bucket sits near (s/(1+2s)). PokerJudge’s scenario tools are enough for that sanity pass; you don’t need a sermon about it.
One last ugly truth: most losing players aren’t off by 2% from GTO. They’re potting rivers with near-pure value, or overbetting with a comedy club of missed draws. Fix the direction of the error before you sweat whether 28% or 30% is “more accurate.”