Optimal Calling Strategy in a Simplified Poker Game
Players A and B each ante $\
The game proceeds as follows:
- Player A acts first and can either check or raise by betting an additional $\
Players A and B each ante $\
The game proceeds as follows:
Describe Player B's optimal strategy. Specifically, when A raises, for which values of B's number $b$ should B call vs. fold? Derive the equilibrium threshold.
How to Think About It: This is a toy poker game, and it teaches you the core mechanic of calling strategy: pot odds vs. the opponent's range. B doesn't know A's number, but B knows A is rational. A will raise with very strong numbers (value betting) and very weak numbers (bluffing), and check the middle. B needs to call just often enough to keep A honest -- call too rarely and A bluffs freely, call too often and A only raises strong hands and prints money. The equilibrium threshold is pinned down by pot odds.
Quick Estimate: When A raises, the pot is $\$6$ and B must risk $\ $ to call. That gives pot odds of $3:1$, meaning B needs to win at least
Approach: We solve for the Nash equilibrium of the full game, focusing on B's calling threshold.
Formal Solution:
Assume A plays a threshold strategy: raise with $a \in [0, x]$ (bluffs) and $a \in [y, 100]$ (value bets), check with $a \in (x, y)$. B plays a threshold strategy: call if $b > t$, fold if $b < t$.
Step 1 -- B's threshold from A's bluffing indifference.
A bluff is a number so low it always loses at showdown when called. So the EV of raising as a bluff does NOT depend on the exact value of $a$: A wins the pot only when B folds ($b < t$, net $+\
$\text{EV(raise as bluff)} = \frac{t}{100}(+2) + \frac{100 - t}{100}(-4) = \frac{6t - 400}{100},$
which is constant across the whole bluff interval. Checking, by contrast, is a losing hand going to showdown, with EV that rises in $a$:
$\text{EV(check, } a) = \frac{a}{100}(+2) + \frac{100 - a}{100}(-2) = \frac{4a - 200}{100}.$
Since EV(check) increases with $a$ while EV(raise) is flat, A's indifferent bluff is the largest one -- the boundary $a = x$ -- not $a \approx 0$. Setting them equal at $a = x$:
$6t - 400 = 4x - 200 \implies 2x = 3t - 100.$
Step 2 -- verify via pot odds.
At $b = t$, B's probability of winning against A's raising range must equal
$P(\text{B wins} \mid \text{A raised},\, b = t) = \frac{x}{x + (100 - y)} = \frac{1}{4}.$
This gives the constraint $3x = 100 - y$, i.e., $y = 100 - 3x$.
Step 3 -- A's value-bet indifference pins down $x$, $y$, and $t$.
A with $a = y$ is indifferent between raising and checking:
$\text{EV(check, } a = y) = \frac{4y - 200}{100},$ $\text{EV(raise, } a = y) = \frac{2t + 4(y - t) - 4(100 - y)}{100} = \frac{8y - 2t - 400}{100}.$
Setting them equal:
$8y - 2t - 400 = 4y - 200 \implies 4y = 2t + 200 \implies 2y = t + 100.$
We now have three equations:
$2 \cdot \frac{100 - t}{6} = 3t - 100 \implies \frac{100 - t}{3} = 3t - 100 \implies 100 - t = 9t - 300 \implies 10t = 400 \implies t = 40.$
Back-substituting: $x = (100 - 40)/6 = 10$ and $y = (40 + 100)/2 = 70$.
Step 4 -- the full equilibrium.
A's raising range has total length
Answer: B's optimal strategy is:
The threshold comes directly from pot odds: B risks $\
This is the fundamental mechanic behind every calling decision in poker and, more broadly, in any adversarial game with incomplete information. The defender (B) doesn't need to figure out whether the opponent is bluffing on any specific hand -- that's impossible with symmetric information structure. Instead, B commits to a calling frequency that makes the opponent's bluffs break even. The threshold is entirely determined by pot odds: if B must risk $\
The deeper lesson is that in equilibrium, the caller's strategy is pinned down by the bettor's indifference, and vice versa. This circular logic -- each player's strategy makes the other indifferent -- is the hallmark of mixed-strategy Nash equilibria. In quant interviews, this same structure appears in market-making problems (how much to shade your quote to deter informed traders), auction theory (how aggressively to bid), and adverse selection models (when to trade against a potentially informed counterparty).