Bet Tac Toe: The Mathematics of Bidding War

Apr 26, 2026
#game-theory #mathematics #tic-tac-toe #algorithms #richman-games

https://btt.onrender.com/

Tic-Tac-Toe is often dismissed as a solved game, but Bet Tac Toe (Bidding Tic-Tac-Toe) transforms it into a complex optimization problem. By introducing a finite resource (100 points) and an auction mechanic, the game shifts from simple pattern matching to a Zero-Sum Stochastic Game variant.

You can test these theories here: btt.onrender.com

The Rules of the Game

In Bet Tac Toe, the board is the familiar 3x3 grid. However, players don’t take turns in a fixed order. Instead:

  1. Starting Capital: Both players begin with 100 points.
  2. The Auction: For every turn, each player places a secret bet.
  3. The Move: The player with the highest bet takes their turn.
  4. The Cost (All-Pay): Both players’ bids are deducted from their respective balances, regardless of who wins the auction.
  5. Tie-Breaking:
    • If the bets are tied, the player with the lower total balance gets the turn.
    • If a tie persists in both bet amounts and total balances, a new round of betting is initiated.

The first person to get three-in-a-row wins. If the board fills up without a winner, it’s a draw.

The Gameplay Loop

graph TD
    Start((Start Game)) --> Init[Both players: 100 pts]
    Init --> Bid[Players submit secret bids]
    Bid --> Pay[Deduct Bid A from Bal A & Bid B from Bal B]
    Pay --> Compare{Compare Bids}
    Compare -- A > B --> AWins[Player A moves]
    Compare -- B > A --> BWins[Player B moves]
    Compare -- A == B --> Tie{Compare Balances}
    Tie -- Bal A < Bal B --> AWins
    Tie -- Bal B < Bal A --> BWins
    Tie -- Bal A == Bal B --> Bid
    AWins --> WinCheck{3 in a row?}
    BWins --> WinCheck
    WinCheck -- Yes --> End((Winner!))
    WinCheck -- No --> FullCheck{Board Full?}
    FullCheck -- Yes --> Draw((Draw))
    FullCheck -- No --> Bid

1. The Mathematical Foundation: All-Pay Richman Games

Bet Tac Toe, in this configuration, is an All-Pay Auction variant of Richman Games. This is significantly more “expensive” than standard bidding games where only the winner pays.

The “Burn” Rate & Deflationary Economy

Because both players lose their bids every turn, the total capital in the system ($C_{total} = Bal_A + Bal_B$) shrinks rapidly.

  • Turn 1: 10 points is 1/20th of the total economy (200 pts).
  • Turn 6: If 80 points have been “burned” by both players combined, 10 points now represents 1/12th of the remaining economy.

Points become more valuable as the game progresses, making mid-game efficiency paramount.

The “Loser’s Tax”

In a standard auction, the loser keeps their money. Here, the loser pays a “Tax” for failing to win the turn. This turns every auction into a high-stakes gamble. If you bid 20 and lose to a bid of 21, you are not just behind on the board—you are economically crippled.

2. The Algorithm: Bidding Minimax

Standard Tic-Tac-Toe uses Minimax to find the best move. In Bet Tac Toe, the algorithm must be modified. A “Bidding Minimax” doesn’t just return a score; it returns a Point Value for the square.

Step-by-Step Valuation

  1. Leaf Nodes: If a state is a Win, $R = 0$. If a loss, $R = 1$. If a draw, $R = 0.5$.
  2. Recursive Step: For each empty square, calculate the potential $R$ if you took it vs. if the opponent took it.
  3. Square Selection: You should bid on the square that has the maximum “Swing” ($R(s_{lose}) - R(s_{win})$).

The Twist: Because the points are paid to the bank (and not the opponent), the total money in the system decreases. This makes early-game points more “expensive” in terms of relative power than late-game points.

Finding the Optimum Play Strategy

To find the optimum play strategy, Bet Tac Toe is modeled as a Zero-Sum Stochastic Game using a variation of the Minimax algorithm combined with Richman Resource Theory. Here is the mathematical framework for solving the optimal loop:

  1. Define Terminal Values: First, map every possible board configuration ($s$). Assign a Value ($V$) to the terminal nodes:
    • Win: $V(s) = 1$
    • Loss: $V(s) = 0$
    • Draw: $V(s) = 0.5$
  2. Calculate Richman Values ($R$): For any non-terminal state, calculate $R(s)$, representing the proportion of total money needed to guarantee a win. This is found recursively: $$R(s) = \frac{R(s_{win}) + R(s_{lose})}{2}$$ Where $s_{win}$ is the state if you win the auction, and $s_{lose}$ if you lose.
  3. Identify the Swing: For each possible move, calculate the “Swing”: the difference between the $R$ of winning that square vs. losing it ($R(s_{lose}) - R(s_{win})$).
  4. Determine the Optimal Bid: The bid should be proportional to how much that specific square changes your probability of winning. In this All-Pay variant, you must calculate the bid that forces the opponent into a lose-lose situation by maximizing your expected value over the shrinking total capital.
  5. Account for the Tie-Breaker: The tie-breaker acts as an infinitesimal advantage ($+\epsilon$). If you have the lower balance, your effective power is slightly higher. This means the optimum strategy often involves bidding exactly the “Richman Swing” value, forcing the opponent to overpay by 1 point to win, shifting the wealth advantage to you.

First Move Analysis: R-Values of an Empty Board

By applying the recursive step starting from the empty board (where $R(empty) = 0.5$), we can calculate the exact Richman values for the first move. Here are the $R$-values and corresponding Swings if you or your opponent take each type of square:

  • The Center Square:
    • If you win it: $R \approx 0.336$ (You need ~33.6% of the capital to win)
    • If you lose it: $R \approx 0.664$ (You need ~66.4% of the capital to win)
    • Swing: $0.664 - 0.336 = \mathbf{0.328}$ (32.8% of total capital)
  • A Corner Square:
    • If you win it: $R \approx 0.381$
    • If you lose it: $R \approx 0.619$
    • Swing: $0.619 - 0.381 = \mathbf{0.238}$ (23.8% of total capital)
  • An Edge Square:
    • If you win it: $R \approx 0.379$
    • If you lose it: $R \approx 0.621$
    • Swing: $0.621 - 0.379 = \mathbf{0.242}$ (24.2% of total capital)

Notice that the Center has the highest “Swing”, meaning it drastically changes the balance of power. However, as noted in the “Center Bait” paradox, overpaying for this 32.8% swing can leave you bankrupt for the remainder of the game!

3. Deep Dive Scenarios

Strategic Decision Matrix

graph TD
    Decision{Identify State}
    Decision --> S1[Empty Board]
    Decision --> S2[Opponent at 2-in-a-row]
    Decision --> S3[You have a Fork]

    S1 --> P[0-Probe Strategy]
    P --> P1[Lose Tie-Breaker Adv]
    P --> P2[Gain Wealth Advantage]

    S2 --> B[Blockade Valuation]
    B --> B1[Bid R-Swing Value]
    B1 --> B2[Force Opponent to Overspend]

    S3 --> F[L-Fork Trap]
    F --> F1[Bid Balance - 1]
    F1 --> F2[Drain Opponent or Win Turn]

Scenario A: The “Center Bait” Paradox

On an empty board, the Center square is intuitively the strongest. However, its $R$-swing is not necessarily the highest.

  • The Math: If you spend 40 points (40% of your wealth) to take the center, your new wealth ratio is $60/160 = 0.375$.
  • The Consequence: You have lowered your $R$-capacity so much that your opponent can now win the next three edge/corner auctions by bidding just 15 points each, effectively surrounding your center.

Scenario B: The “L-Fork” (The Deadliest State)

In standard TTT, a fork is a winning move. In Bet Tac Toe, a fork is a Financial Trap. Suppose you have X’s at (1,1) and (1,3). The square (1,2) is a winning move.

  • Opponent’s Dilemma: They must win the auction for (1,2) to survive.
  • Your Strategy: You don’t actually need to win (1,2). You just need to make them pay for it. By bidding $CurrentBalance - 1$, you force them to spend almost everything. Even if they survive that turn, they will have 1 point left, allowing you to walk into your next victory for free.

Scenario C: The Tie-Breaker Squeeze

The rule: In a tie, the player with the lower balance wins.

  • Theorem: If you have 80 points and your opponent has 81, you effectively have 80.5 points.
  • Execution: If you identify a “Critical Bid” (a turn that decides the game), and the math suggests a bid of 20, you can bid exactly 20. Your opponent must bid 21 to beat you. This 1-point “Tie-Breaker Tax” compounded over several turns is often what separates grandmasters from casual players.

4. Optimal Betting Strategy: “The 0-Probe”

One of the most powerful algorithms for human play is the Asymmetric Opening.

  1. Turn 1: Bid 0.
  2. Outcome 1: Opponent bids 1-5. You lose the turn and the Tie-Breaker Advantage, but gain a significant Wealth Advantage. You now have 100 points vs their 95.
  3. Outcome 2: Opponent bids 0. You might win the turn (if balances are equal and the re-bet logic favors you) or simply stay even.

By bidding 0, you are gathering information about your opponent’s “Aggression Constant.” If they bid 20 on the first turn, you know they are playing a “Sprint” strategy, which you can defeat by playing a “Marathon” strategy (blocking only critical lines until they are broke).

Conclusion: The “Solving” of Bet Tac Toe

While a computer can solve Bet Tac Toe by mapping all $3^9$ states and calculating precise Richman values for each, for humans, the game is about Volatility Management.

Key Takeaway: Every square has a price. If your opponent is willing to pay more than the “Richman Swing” value for a square, let them have it. The goal isn’t to own the board; it’s to own the economy when the final, winning square is auctioned off.

*Which strategy do you prefer? The early-game aggression or the late-game squeeze? Play a round at btt.onrender.com