The Mathematics of Optimal Bluffing

🎯 Why Bluff Frequency Matters

If you never bluff, opponents always fold to your bets (knowing you only bet value). If you always bluff, opponents always call (knowing you rarely have value). The optimal strategy lies between these extremes.

A balanced range makes opponents indifferent between calling and folding — they cannot exploit you regardless of their strategy.

🛡️ Minimum Defense Frequency (MDF)

MDF tells us: "How often must I call (or raise) to prevent my opponent from profiting with a pure bluff?"

If we fold too often, opponent can bluff any two cards profitably. MDF prevents this.

$$\text{MDF} = \frac{\text{Pot Size}}{\text{Pot Size} + \text{Bet Size}} = 1 - \frac{\text{Bet}}{\text{Pot} + \text{Bet}}$$

Derivation

Opponent's bluff EV must be less than or equal to 0:

$$EV_{\text{bluff}} = P(\text{fold}) \times \text{Pot} - P(\text{call}) \times \text{Bet} \leq 0$$ $$P(\text{fold}) \times \text{Pot} \leq P(\text{call}) \times \text{Bet}$$ $$P(\text{call}) \geq \frac{\text{Pot}}{\text{Pot} + \text{Bet}}$$

MDF Table

⚖️ Optimal Bluff-to-Value Ratio

From the attacker's perspective: what proportion of your betting range should be bluffs?

Let \(b\) = bet size as a fraction of the pot. The optimal bluff percentage in your betting range is:

$$\text{Bluff\%} = \frac{b}{2b + 1}$$

Derivation

When we bet, opponent must be indifferent between calling and folding with their bluff-catchers.

Let \(V\) = number of value combos, \(B\) = number of bluff combos.

Opponent's EV of calling:

$$EV_{\text{call}} = \frac{V}{V+B} \times (-\text{Bet}) + \frac{B}{V+B} \times (\text{Pot} + \text{Bet})$$

Setting \(EV_{\text{call}} = 0\) (indifference):

$$\frac{B}{V+B} \times (\text{Pot} + \text{Bet}) = \frac{V}{V+B} \times \text{Bet}$$ $$B \times (\text{Pot} + \text{Bet}) = V \times \text{Bet}$$ $$\frac{B}{V} = \frac{\text{Bet}}{\text{Pot} + \text{Bet}}$$

With bet = \(b \times \text{Pot}\):

$$\frac{B}{V} = \frac{b}{1 + b}$$

Therefore:

$$\frac{B}{V + B} = \frac{b}{2b + 1}$$

Bluff Ratio Table

📐 Mathematical Proof in Detail

Full derivation from game theory first principles.

Start with the strategic form: Player A bets with value hands (\(V\)) and bluffs (\(B\)). Player B decides to call or fold.

Payoff Matrix

Using pot \(P\) and bet size \(S\):

• If A bets value, B calls: A wins +S, B loses -S
• If A bets value, B folds: A wins +P, B loses 0
• If A bets bluff, B calls: A loses -S, B wins +S
• If A bets bluff, B folds: A wins +P, B loses 0

For Nash equilibrium, both players must be indifferent:

B's Indifference (determines A's bluff frequency)

$$EV_{\text{call}} = EV_{\text{fold}}$$ $$\frac{V}{V+B}(-S) + \frac{B}{V+B}(P+S) = 0$$

Solving: \(B = \frac{S}{P+S} \cdot V\), giving \(\frac{B}{V+B} = \frac{S}{P+2S}\)

A's Indifference (determines B's calling frequency)

$$EV_{\text{bluff}} = EV_{\text{check}}$$ $$(1-c) \cdot P - c \cdot S = 0$$

Solving: \(c = \frac{P}{P+S}\), which is exactly MDF.

🔄 Multi-Street Applications

On the river, the math above applies directly. But across multiple streets, we need to think about how bluffs "compound."

Geometric Bet Sizing

If you want to get all-in over \(n\) streets with stack \(S\) and pot \(P\), the optimal bet each street is:

$$b = \left(\frac{S + P}{P}\right)^{1/n} - 1 \quad \text{(as a fraction of the pot each street)}$$

Example

Pot = $100, Stack = $400, 3 streets remaining:

$$b = \left(\frac{500}{100}\right)^{1/3} - 1 = 5^{0.333} - 1 \approx 0.71$$

So bet approximately 71% of the pot each street.

Cumulative Bluff Frequency Across Streets

If you bluff \(f\) fraction of your range on each street, after \(n\) streets your total bluff frequency is:

$$\text{Total Bluffs} = f^n$$

With \(f = 0.33\) (pot-sized bets), after 3 streets: \(0.33^3 = 3.6\%\) of your range are pure bluffs.

♠️ Practical Exploitation

GTO is the starting point. Exploit when opponents deviate:

When Opponent Folds > MDF

• Increase bluff frequency
• Use more aggressive bet sizing
• Add "thin bluffs" that you normally wouldn't bet

When Opponent Calls > MDF

• Eliminate bluffs, bet only value
• Use larger sizes for value extraction
• "Merge" your range (bet medium-strength hands for value)

$$EV_{\text{exploit}} = EV_{\text{GTO}} + \Delta_{\text{opponent error}}$$