What Moneyball Actually Proved (And What Everyone Got Wrong)

The popular version of the Moneyball story goes something like this. A poor baseball team discovered that on-base percentage was undervalued, signed a bunch of players nobody else wanted, and beat the rich teams with maths.

That version is wrong in a way that matters, and getting it right explains almost everything about how sports analytics has developed in the twenty-odd years since.

Billy Beane’s insight was never that on-base percentage is the most important statistic in baseball. It was that the market for baseball players was mispricing something, and that a team with no money had to find whatever was mispriced and buy it before anyone else noticed. On-base percentage happened to be the mispricing available in 2002. Once every front office read the book, that particular edge evaporated, and Beane went looking for the next one.

Moneyball was a book about arbitrage. Arbitrage closes.

Where the idea came from

Bill James spent his nights as a security guard at a pork and beans cannery in Kansas, writing baseball analysis nobody had asked for. He self-published his first Baseball Abstract in 1977 and sold it through a classified ad. The field he effectively invented took its name from SABR, the Society for American Baseball Research, and for most of two decades it lived entirely outside professional baseball.

The resistance was not stupidity. Scouting was a craft with real knowledge behind it, and scouts could see things no box score captured. But the profession had anchored itself to statistics that measured the wrong things. Batting average treats a walk as a non-event. RBIs credit a hitter for the performance of whoever batted ahead of him. Both numbers were widely used to set salaries, which meant both numbers were creating price distortions large enough to drive a team through.

Beane, running the Oakland A’s on roughly a third of the Yankees payroll, drove through.

The part the story usually leaves out

The A’s never won a World Series during any of this. They made the playoffs and went out early, repeatedly.

That detail gets skipped because it spoils the arc, but it is the most instructive thing in the whole history. A 162-game regular season gives statistical advantages room to express themselves. A five-game series does not. Beane himself said his approach did not work in the playoffs, and he was right. Analytics buys you more expected wins across a long sample. It does not buy you a specific October.

Anyone selling you a model that promises a particular outcome on a particular night is selling something Moneyball never claimed to deliver.

What happened once everybody had the data

Every Major League Baseball team now runs an analytics department. Statcast, installed league-wide in 2015, tracks exit velocity, launch angle and spin rate on every pitch and batted ball. Wins Above Replacement is cited casually on broadcasts. The information asymmetry that made Oakland competitive no longer exists anywhere in the sport.

The same pattern played out across every league that followed.

Basketball absorbed it fastest and most visibly. Player Efficiency Rating gave analysts a single number for overall impact, but the durable change was simpler arithmetic: a three-point shot is worth fifty percent more than a two, so the long midrange jumper is a bad shot. Teams acted on that. League-wide three-point attempts climbed for fifteen straight years and the shape of the game changed in a way spectators can see without knowing a single statistic.

The NFL built Next Gen Stats to track speed, separation and coverage. Soccer adopted expected goals, which estimates the probability that a given shot becomes a goal, and xG now appears in matchday broadcasts. Hockey uses Corsi and Fenwick to count shot attempts as a proxy for possession. Tennis and golf run motion capture on swing mechanics and shot selection.

In every case the sequence was identical. A metric reveals a mispricing, early adopters profit, everyone copies it, the edge disappears, and the metric becomes table stakes. What began as an advantage ends as a prerequisite.

Where judgment still lives

Beane traded Jeremy Giambi in May 2002, in the middle of the season the book covers, for reasons that had nothing to do with his production. Giambi was getting on base. He was also creating problems in the clubhouse, and Beane decided the second thing outweighed the first.

That decision was not analytical. It was a judgment about people, made by someone who had spent his career around them, and it sits awkwardly inside a story usually told as a triumph of numbers over instinct.

The teams that do this well now treat the two as complementary rather than opposed. Data narrows the field and flags what intuition misses. People decide what the data cannot see: whether a player will handle a bigger role, whether a locker room can absorb a difficult personality, whether a coach can teach. Front offices that abandoned scouting entirely mostly regretted it.

The betting question

This is where the Moneyball narrative gets applied most loosely and most misleadingly.

The argument usually runs that since data transformed how teams evaluate players, data can transform how you evaluate bets. Odds comparison services like Betbrain exist because line shopping is real, and a bettor who takes the best available price across several books is genuinely better off than one who does not.

But the analytics revolution reached the sportsbooks first, and it reached them harder. Books employ quantitative traders, ingest the same tracking data the teams use, and move lines within seconds of sharp money appearing. The public information that once created edges is now priced in before a recreational bettor finishes reading about it.

The numbers are unambiguous. American sportsbooks retained 9.3% of $149.8 billion wagered in 2024, up from 7% in 2019, driven largely by high-margin parlays. At standard -110 pricing you need to win 52.38% of your bets to break even. Long-term profitability estimates cluster around 3 to 5% of bettors, and reporting on operator data has shown that a very small group of heavy losers supplies the majority of revenue.

So the honest translation of Moneyball into betting is not “use data and win.” It is “the market you are betting into has already done this, better than you will, with more resources.” That is the same lesson Oakland learned when the rest of baseball caught up. The difference is that the A’s could go find a new inefficiency. A retail bettor facing a 9.3% hold is not arbitraging a market. They are paying to enter one.

The thing worth taking from it

Moneyball is a better book than its reputation, because its real subject is how institutions cling to bad measurements long after better ones exist, and how briefly you get to profit from noticing.

Twenty-three years on, the analytics are universal, the edges are thin, and the advantage has moved to execution, player development and the parts of the game that still resist measurement. That is what happens to every good idea in a competitive market. It works, it spreads, and then it just becomes how things are done.