The Mathematics of Pinpoint

A deep dive into combinatorics, probability matrices, and game theory heuristics

Introduction to Grid Geometry

Pinpoint is played on a discrete 10×10 Cartesian coordinate space, yielding exactly 100 playable cells:

S = { (x, y) \in ℤ² | 1 \leq x \leq 10, 1 \leq y \leq 10 }

Unlike standard alignment games which treat the board as a uniform field, Pinpoint partitions the board into four distinct 5×5 quadrants. The division lines (represented visually as blue lines) act as scoring multipliers for specific structural configurations. In particular, the four quadrant centers:

C = { (3,3), (3,8), (8,3), (8,8) }

serve as focal points for double-value score triggers. Understanding the geometric distribution of these cells is fundamental to calculating placement utility values.

Combinatorics of Line Alignments

Points are scored by forming collinear sequences of same-colored pins. We can compute the total number of possible lines of length L on the 10×10 grid using combinatorial analysis:

Horizontal and Vertical Combinations

For a horizontal line of length L, each row has exactly 10 - L + 1 possible positions. Across all 10 rows, this yields:

N_horizontal(L) = 10 \times (11 - L)

Because the board is square, vertical combinations are symmetrical:

N_vertical(L) = 10 \times (11 - L)

For example, for a 5-in-a-row pattern, there are 10 × (11 - 5) = 60 horizontal and 60 vertical lines, totaling 120 orthogonal vectors.

Diagonal Combinations

Diagonal lines are divided into major diagonals (45° and 135°) and their parallel offsets. For a diagonal of length L, the number of combinations is calculated by summing the available segments across diagonals of varying lengths:

N_diagonal(L) = 2 × [ 2 × ∑_{k=L}^{9} (k - L + 1) + (10 - L + 1) ]

Let us compute the total possible 5-in-a-row diagonal lines:
For k = 5 to 9, the sum of (k - 5 + 1) is 1 + 2 + 3 + 4 + 5 = 15.
Substituting this back, we get 2 × [ 2 × 15 + 6 ] = 2 × 36 = 72 diagonal lines.

                Total Combinatorial Opportunity: For length L = 5, the board contains exactly 192 scoring vectors (60 Horizontal + 60 Vertical + 72 Diagonal). For length L = 6 (the Reversal trigger), the opportunity decreases to 150 total vectors (50 H + 50 V + 50 D).
            

Probability Matrix of Hand Distributions

A player's hand consists of 4 pins chosen at random from the four standard colors: Red (R), Green (G), Purple (P), and Yellow (Y). Because colors are replenished with equal probability, the hand distribution follows a multinomial probability distribution:

Pr(n_R, n_G, n_P, n_Y) = ( 4! / (n_R! n_G! n_P! n_Y!) ) \times (0.25)⁴

where n_R + n_G + n_P + n_Y = 4. We can classify all hands into 5 distinct partition classes:

Class [4-0-0-0] (Monocolor Hand): e.g., {R, R, R, R}.
Probability: 4 × (1/256) ≈ 1.56%.
Class [3-1-0-0] (Tri-dominant Hand): e.g., {R, R, R, G}.
Probability: 12 × (4/256) ≈ 18.75%.
Class [2-2-0-0] (Double-pair Hand): e.g., {R, R, G, G}.
Probability: 6 × (6/256) ≈ 14.06%.
Class [2-1-1-0] (Pair-hybrid Hand): e.g., {R, R, G, P}.
Probability: 12 × (12/256) ≈ 56.25%.
Class [1-1-1-1] (Rainbow Hand): e.g., {R, G, P, Y}.
Probability: 1 × (24/256) ≈ 9.38%.

The probability matrix reveals that over 75% of hands (Class 2-1-1-0 and Class 3-1-0-0) contain at least a pair or triplet of a single color. This makes planning color-aligned patterns highly viable, while the rare Monocolor Hand (1.56%) represents a significant tactical spike, allowing a player to lay down a major pattern sequence in a single round.

Game Theory and Strategic Utility

During each turn, a player evaluates the board to maximize their expected utility. We can model this decision using a heuristic valuation function:

U(x, y) = V_immediate(x, y) + w_d \times V_defensive(x, y) + w_s \times V_positional(x, y)

Where:

V_immediate is the direct points gained by completing a pattern (1 to 10 points).
V_defensive is the points blocked from opponents (preventing them from completing lines or squares).
V_positional represents the positional influence (e.g., proximity to quadrant centers or alignment vectors).
w_d and w_s are weight parameters adjusted dynamically based on game state.

Applying minimax tree search on this utility function is how Pinpoint's advanced AI bots compute their moves, calculating up to 3 turns ahead to identify and block double-threat placements (where an opponent can score on two different vectors in their next turn).

Play Pinpoint & Test the Math