Why Do We Love Statistics That Aren’t Relevant?

We love irrelevant statistics because we are wired to find patterns, especially when the world feels unpredictable. Numbers give those patterns credibility. They turn coincidence into something that resembles evidence and make the past feel useful for predicting the future, even when the conditions that produced the numbers no longer exist.

In a 2008 study published in Science, Jennifer Whitson and Adam Galinsky found that people who experienced a loss of control were more likely to perceive patterns in unrelated or random information. Across six experiments, participants identified meaningful relationships where none existed, including false correlations in stock market data.

The research points to a basic human desire: when we cannot control events, finding a pattern helps us make sense of them. It does not establish why every irrelevant statistic appeals to us, but it helps explain why apparent patterns can be persuasive even without a meaningful connection to the outcome.

The critical question is not whether a statistic is accurate. It is whether the mechanism that produced it is still operating.

What does Christmas Day have to do with winning?

Sports broadcasting offers an almost unlimited supply of statistics that sound meaningful.

Imagine an NFL announcer explaining that a team has won 80% of its Christmas Day games over the past 40 years. That's impressive. It also might mean the team won four of five games.

The percentage is accurate, but the sample is tiny. More importantly, four decades span different players, coaches, opponents, rules, and strategies. Unless playing on Christmas creates some repeatable advantage, the holiday record tells us little about the current matchup.

Now consider another statistic. The team's offensive line has allowed unusually few pressures over its last six games, while today's opponent consistently generates pressure with four pass rushers.

Both statistics describe the past. The second identifies a mechanism that directly affects the game being played. Protecting the quarterback influences how an offense operates. A Christmas Day winning percentage offers no comparable explanation.

And 80% sounds far more convincing than four out of five. Precision strengthens that impression. An announcer reporting a 79.6% historical success rate sounds authoritative, but additional decimal places do not create relevance.

Why random events create convincing stories

Flip a fair coin 100 times. You will encounter streaks: several heads in a row, several tails, and sequences that appear surprisingly organized.

A run of six heads looks unusual. Yet with enough flips, runs are expected. The coin has not developed a preference for heads, and the mechanism governing the next flip has not changed.

Sports fans recognize a similar idea as the hot hand. A basketball player makes five consecutive shots, and the crowd expects the sixth to go in.

In a 1985 study, Thomas Gilovich, Robert Vallone, and Amos Tversky argued that people often perceive patterns in shooting streaks that resemble sequences produced by chance. Later research complicated that conclusion. Joshua Miller and Adam Sanjurjo identified a statistical bias in the earlier analysis and found evidence that genuine hot-hand effects can occur.

That distinction is important. Some streaks are random. Others reflect real changes in performance. Observing a streak alone cannot establish which explanation applies.

The task is to identify what caused the pattern, rather than assume the pattern explains itself.

How do we find so many extraordinary statistics?

Suppose a sports researcher examines every team's performance by day of the week, month, weather condition, stadium, opponent, kickoff time, and holiday.

Thousands of possible combinations emerge. Among them, some teams will have remarkable records under strangely specific conditions. Perhaps one team has won eight consecutive Thursday games in November when the temperature was below 40 degrees.

That's the problem of multiple comparisons. The more relationships you examine, the greater the opportunity to find exceptional-looking results through chance alone.

The statistic can be completely accurate. What the audience never sees are the thousands of combinations that produced nothing interesting.

This is closely related to data dredging, a problem discussed in the American Statistical Association's 2016 statement on statistical significance. Researchers warned that selectively reporting attractive results from many analyses can create misleading conclusions. Statistical significance itself is not a measure of practical importance.

Sports broadcasters want interesting facts, so unusual records get airtime. The selection process favors surprises, not necessarily relevance.

The same process can occur in business reporting, financial analysis, and scientific research. When people search enough combinations, something extraordinary eventually appears. Finding an unusual statistic is not the same as discovering an important relationship.

When history stops being useful

Consider two hypothetical companies, each with 25 consecutive years of revenue growth.

Company A sells equipment used to maintain electrical infrastructure. Its customers replace aging components, expand networks, and meet recurring maintenance requirements. The business has changed over time, but much of the demand supporting its historical growth remains.

Company B sells specialized equipment built around a technology that has been widely adopted for decades. Its growth came from an expanding market and limited competition. Now a cheaper replacement technology is gaining adoption, and its customers are changing how they operate.

Both companies have achieved the same historical record. Both can truthfully claim 25 years of uninterrupted growth.

Yet their histories mean different things.

Company A's record provides evidence about a business operating under conditions that remain recognizable. Company B's record describes success under conditions that are disappearing.

Neither streak alone predicts next year's revenue. Company A still needs to demonstrate continued demand, while Company B might adapt successfully to its new market. The difference is whether historical performance has an identifiable connection to current circumstances.

A statistic becomes less useful when the conditions responsible for the original result stop resembling the conditions we face today.

Test the mechanism, not the number

Historical data deserves attention when it helps explain something. The mistake is treating a record as an explanation rather than something that needs one.

Before relying on an impressive statistic, ask what produced it. Was there a causal relationship, a persistent competitive advantage, an environmental condition, or simply an unusual sequence of events? Then determine whether that mechanism still applies.

This also explains why accuracy and usefulness are separate standards. A 40-year-old record can be precisely calculated and historically important while contributing almost nothing to a present decision. A recent measurement can be less dramatic but more informative because it captures the conditions that matter now.

Statistics are powerful because they reduce complex events to something we can understand and remember. But that compression removes context. Without the context, we risk turning evidence into decoration.

The next time someone presents an extraordinary statistic, don't begin by asking whether the number is correct.

Ask what made it true, and whether that still matters.

📚 Bookmarked for You

Fooled by Randomness: The Hidden Role of Chance in Life and in the Markets by Nassim Nicholas - Explores how we mistake luck for skill, coincidence for causation, and random patterns for meaningful evidence. Particularly relevant to understanding why impressive historical statistics can tell us so little about what happens next

🧬 QuestionStrings to Practice: Finding the Mechanism

Use this sequence when a statistic is offered as evidence for a prediction or decision.

What exactly does this statistic measure?
↓
What caused the observed result?
↓
Which of those conditions still exist?
↓
What evidence would show that the relationship has changed?
↓
How much weight should the statistic receive in the decision?

A number becomes useful when we understand the relationship it describes, not simply because we can measure it.

---

A new question every day. Pause. Think. Write if you feel compelled.

Explore QuestionClass's Question-a-Day.

← Home