FORMAX Research · Experiment 03
Experiment 03 · Distribution Analysis

The Variance Gap

Ten thousand simulations of each chain. The structured chain wins on the mean and on the variance simultaneously.

The common defense of high-entropy football is that direct play is the high-variance gamble that gives an outmatched team a chance to compete. This experiment tests that defense empirically. Each chain is run thousands of times, and the resulting distribution of expected goals per possession is plotted. The structured chain has both a higher mean and a lower variance than the direct chain. Low-entropy football is not the high-risk choice; it is the opposite. The statistical case for structure that the public discourse never makes is contained in the histogram below.

Ready
Structured Chain
Low-entropy, constructed
Mean xG / possession
Std deviation
Mean xG / match (80 poss)
Std dev / match
Match 95% interval
Direct Chain
High-entropy, kick & rush
Mean xG / possession
Std deviation
Mean xG / match (80 poss)
Std dev / match
Match 95% interval

Match-Level Outcome Distributions

Simulated possessions are aggregated into matches and the per-match expected goals are histogrammed. Bin width is automatic. The structured distribution sits to the right of the direct distribution (higher mean) and is narrower (lower variance). This is the central empirical finding of the experiment, and it inverts the common defense of high-entropy play.

Structured chain
Direct chain

The Three-Claim Verdict

The three claims the experiment is designed to test, with the empirical answer from the current simulation run.

Claim 1 · The Mean
Structured chain produces more expected goals per match than direct chain.
Claim 2 · The Variance
Structured chain produces less match-to-match variance than direct chain.
Claim 3 · The Floor
Structured chain has a higher worst-case (5th percentile) than direct chain.

What the experiment shows

The public defense of direct play rests on a variance argument: when you are outmatched, embrace chaos because chaos gives you a chance. The experiment shows that the variance argument is empirically wrong on its own terms. The high-entropy chain produces a wider distribution, but the increased dispersion does not translate into a higher ceiling for the underdog. It produces a lower floor without producing a higher peak. The team running the high-entropy chain is more likely to underperform its own mean than the team running the low-entropy chain.

The reason is mechanical. The structured chain's mean is higher because the chain reaches better shot origins more often. The structured chain's variance is lower because the sequence is more deterministic; the same starting state produces a more concentrated distribution of outcomes. Combining these properties, the structured chain dominates the direct chain on first-order and second-order statistics simultaneously. There is no statistical case for choosing the direct chain when the substrate supports the structured one.

The variance argument for direct play is empirically wrong. The high-entropy chain produces both a lower mean and wider dispersion that mostly extends downward. The structured chain is the lower-risk option, not the higher-risk one. Coaches who choose direct play because they believe it is the high-variance gamble against a better opponent are choosing a strategy that loses on both axes.