Effort Scores Beat Win Rates 2-to-1 in 700 Skill Matches

Why do we keep reaching for the scoreboard when the scoreboard keeps lying to us? In competitive settings—from chess clubs to sales floors to ranked online arenas—win rate is the default metric for judging skill, yet it is a notoriously noisy signal, shaped by matchmaking, luck, and opponent quality as much as by the player's own competence. A recent analysis of 700 recorded skill matches found something quietly subversive: effort-based scoring predicted future performance roughly twice as well as win rate did. The question worth asking is not whether effort matters—everyone says it does—but why a measure of how someone played outperforms a measure of whether they won.

The Statistical Problem With Winning

Win rate has an intuitive appeal that is hard to argue with: it is binary, it is final, and it requires no interpretation. That is precisely its weakness. A single match outcome compresses dozens of decisions, several of which may have been excellent, into one bit of information. When the sample is small—and in most real competitive settings, samples are small—that compression produces variance that swamps signal.

Daniel Kahneman and Amos Tversky's work on judgment under uncertainty documented how readily people substitute an easy question for a hard one. "Is this player good?" is hard. "Did this player win?" is easy, and the mind treats the answer to the second as the answer to the first. The substitution feels efficient and is often wrong.

There is a second distortion: outcome bias. Research by Jonathan Baron and John Hershey demonstrated that people rate the quality of a decision more harshly when it produces a bad result, even when the decision itself was sound given the information available. In competitive contexts, this means a player who takes a calculated risk and loses gets marked down twice—once for the loss, once for the audacity.

Effort scoring sidesteps both problems by measuring the process rather than the result. It asks: did this person prepare, adapt, sustain attention, and recover from setbacks within the match? Those variables are observable, they accumulate across many decision points inside a single contest, and they are far less sensitive to the randomness that decides close games.

What Effort Scoring Actually Measures

The term "effort" risks sounding soft, so it is worth being precise. In the 700-match dataset, effort was operationalized through a composite of behavioral indicators: time spent in deliberation before committing to a move, rate of strategy revision after opponent responses, persistence after a losing position, and consistency of engagement across the full match duration rather than only its opening.

None of these are moral qualities. They are proxies for cognitive investment, and they map onto what Anders Ericsson's research on deliberate practice identified as the engine of skill acquisition: focused, effortful engagement with material at the edge of one's current ability. A player who coasts through a match they are winning easily may post a strong win rate while learning almost nothing. A player who grinds through a loss, revising and re-engaging, may post a weak win rate while building the exact capacities that will show up six months later.

This is where the 2-to-1 ratio becomes intelligible. Win rate measures the past. Effort measures the trajectory.

The Reward-Loop Complication

There is a behavioral wrinkle here that deserves attention, because it explains why effort metrics are so rarely adopted despite their predictive power. Competitive environments tend to be engineered around variable-ratio reinforcement—the schedule B.F. Skinner found to produce the most persistent behavior, in which rewards arrive unpredictably after an unpredictable number of responses. Winning is a textbook variable-ratio reward: you cannot know in advance which match will deliver the payoff, so you keep playing.

The trouble is that variable-ratio schedules reinforce playing, not improving. They reward the behavior that produces the next win, which is often the behavior that avoids risk rather than the behavior that develops skill. A player optimizing for win rate will avoid stronger opponents, stick to familiar strategies, and quit matches early when the outcome looks decided. Each of those choices protects the metric and starves the trajectory.

Effort scoring inverts the incentive. If deliberation, revision, and persistence are what get measured, then the rational move is to seek out difficulty—precisely the behavior that builds competence over time.

Loss Aversion and the Cost of Protecting a Number

Loss aversion, one of the most robust findings in behavioral economics, holds that losses loom larger than equivalent gains. Applied to competitive metrics, it predicts something specific: once a win rate becomes a valued possession, players will defend it rather than risk it.

Consider a concrete case. In a study of competitive chess players, researchers found that players rated near a round-number threshold—say, a rating of 2000—played measurably more conservatively in games where a loss would drop them below the threshold, accepting draws in positions where they had winning chances. The rating, a proxy for win-based skill, had become a thing to protect rather than a thing to improve. The same dynamic appears in sales teams that sandbag forecasts, in students who choose easier courses to guard a GPA, and in athletes who manage statistics rather than compete.

An effort score is harder to defend because it is not a single number you can sit on. It is composed of behaviors you either exhibit in the current match or do not. You cannot coast on last season's deliberation.

Why the Ratio Matters More Than the Finding

That effort outperformed win rate is unsurprising to anyone who has coached. That it outperformed by roughly two to one across 700 matches is the part worth sitting with. It suggests the gap is not marginal—that in competitive environments of moderate complexity, we may be systematically misranking participants by relying on the most legible metric available.

If that holds, the implications extend beyond games. Hiring, promotion, and educational admissions all lean heavily on win-rate analogues: past titles, past sales figures, past grades. Each is an outcome measure subject to the same compression problem and the same outcome bias. Effort-based assessments—work samples, revision histories, evidence of how someone responded to a documented setback—are messier to collect and harder to compare, which is exactly why they get skipped.

Building Metrics That Point Forward

The practical question is not whether to abandon outcome measures. Win rate is not useless; it carries real information, especially at large sample sizes. The question is what to pair it with, and how to weight the pair.

Three design principles follow from the research. First, measure process variables that accumulate within a single contest, not just across contests—deliberation time, revision rate, persistence after adversity. Second, normalize for opponent quality, or the metric will simply reward easy schedules. Third, and most importantly, make the effort measure visible to the participant. A metric that only evaluators see cannot shape behavior; a metric the player sees can redirect the reward loop from protecting a number to building a capacity.

The forward-looking version of this is a scoreboard that shows two columns instead of one. The first column tells you what happened. The second tells you what is being built. For most decisions that matter—who to promote, who to coach harder, who to bet on for the next five years rather than the last five matches—the second column is the one worth reading.