Generated by Codex with GPT 5.6 Sol XHigh

A Fair Share of an Unfinished Game

Probability theory began with a practical argument: how should gamblers divide the stakes when chance interrupts their game? The puzzle, known as the problem of points, asks for more than a reward for whoever happens to be ahead. A fair answer must account for what could still have happened.

The article illustrates the problem with two players who each contribute \$50 and flip a coin. Heads scores for one player, tails for the other, and the first to 10 points wins the \$100 pot. If play stops with the score eight to six, returning both stakes ignores the leader’s advantage, but awarding the leader everything ignores the trailing player’s real chance of winning.

In 1494 Italian mathematician Luca Pacioli proposed dividing the pot in proportion to the points already scored. That gives the leader eight fourteenths, or about \$57.14. The rule looks reasonable until the interruption comes after a single toss: it then gives the entire pot to a player who has only the smallest possible lead in a long game.

Niccolò Fontana, known as Tartaglia, later tried to repair the idea by measuring the lead against the number of points needed to win. In the eight-to-six example, his rule returns the leader’s \$50 stake plus one fifth of the opponent’s stake, for \$60. But this method also breaks down near the finish. In a race to 200, a player leading 199 to 190 would receive only \$2.25 beyond the original stake even though the opponent could win only by scoring 10 points in a row.

Both approaches share the same flaw: they look backward at the score rather than forward at the remaining possibilities.

Pascal and Fermat Count the Futures

In the mid-17th century a French gambler brought the problem to Blaise Pascal, who enlisted Pierre de Fermat. Their correspondence produced two different methods that reached the same result. The agreement mattered because it shifted the question from intuition about a lead to a calculation over possible futures.

Fermat’s method lists every continuation that could decide the game. From an eight-to-six score, no more than five additional flips are needed, giving 32 possible five-flip sequences. A player may reach 10 before all five flips occur, but the remaining imaginary flips can still be included for consistent counting. The leader wins in 26 of the 32 sequences and therefore deserves 26 thirty-seconds of the pot: 81.25 percent, or \$81.25.

Pascal avoided exhaustive enumeration by reasoning backward. A game interrupted at a tie should be split evenly. If the score is nine to eight, the leader has a 50 percent chance of winning the entire \$100 on the next flip and a 50 percent chance of reaching a nine-to-nine tie worth \$50. The fair value of the lead is therefore the average, \$75. Applying the same calculation recursively yields \$93.75 at nine to six and \$68.75 at eight to seven. At eight to six, the average of those two next-step values is \$81.25, exactly matching Fermat’s count.

The central insight is that an uncertain position is worth the weighted average of its possible outcomes. The score matters only because it changes which futures remain possible and how likely each one is.

From Gambling Puzzle to Expected Value

Pascal and Fermat did not make uncertainty disappear. They made it measurable. Their methods supplied early foundations for expected value: multiply each possible outcome by its probability, then add the results.

That logic now reaches far beyond games. An actuary pricing insurance, an analyst evaluating an investment and a gambler considering a wager all compare uncertain futures in essentially the same way. Fermat’s enumeration and Pascal’s recursion are two routes to the same discipline: decisions under uncertainty should be priced by the full distribution of what might happen, not by the most visible fact about the present.

The unfinished coin game mattered because it forced mathematics to treat chance systematically. Once possible futures could be counted and weighted, risk became something people could analyze rather than merely guess at.