The Complex Forecast Illusion: More Math, Less Right

Here's a prediction.

"Given the convergence of geopolitical instability in Eastern Europe, persistent inflationary pressures driven by supply-chain realignment, and the lagging effects of quantitative tightening on housing credit markets, we project a 67.3% probability of a recessionary contraction in the next 14-18 months, with GDP declining 1.8-2.4% and unemployment peaking at 6.2% before recovery stabilises in Q3 of the following fiscal year."

Sounds smart, right? Specific. Detailed. The kind of thing you'd trust.

Here's another prediction: "The economy might slow down sometime in the next couple of years."

Same claim. Second version sounds vague and slightly useless. First version sounds like expertise.

Here's what the research shows: for complex systems — economies, geopolitics, markets, weather beyond a few days — the complicated version is usually not more accurate than the simple version. It just sounds like it should be.

What's Actually Happening

The Complex Forecast Illusion is the tendency to trust more complicated, precise-sounding predictions more than simpler ones — regardless of whether the complexity makes them more accurate.

And here's why this happens:

Precision signals effort. "67.3%" feels like someone calculated something. "Maybe 70%" feels like a guess. But in a system with genuine uncertainty, the difference between 67.3% and 70% is meaningless — both are approximations of something no one actually knows. The decimal doesn't add information. It adds the feeling of information.

Jargon signals expertise. Technical vocabulary sounds like someone has done their homework. Sometimes they have. But vocabulary can also be learned without understanding — see Chauffeur Knowledge.

Complexity implies considered analysis. A prediction with eight variables sounds like someone thought hard. A simple prediction sounds like someone didn't bother. But in genuinely complex systems, adding more variables often doesn't improve predictions — it adds more ways to be wrong while appearing to be precise.

The uncomfortable truth: in complex systems, the best predictors tend to be humble. They use ranges, not point estimates. They say "I don't know" a lot. They update when they're wrong. They don't project confidence they don't have.

The worst predictors sound the best.

Real-Life Examples

Economic forecasts. Banks and research institutes publish detailed economic predictions constantly — GDP growth to one decimal place, unemployment projections with confidence intervals. Studies tracking these predictions over decades show they are barely better than simple trend extrapolation, and sometimes worse. The economists often know this. The complexity is partially for communication purposes — it sounds like they're doing their job.

Stock market predictions. Analysts publish 12-month price targets with specific numbers. "Apple will reach $212.50 by Q3." Research consistently shows these predictions do not outperform random chance at that level of specificity. But they keep getting published because they feel authoritative. The specificity is a performance.

Weather forecasting. Here's where complexity and accuracy actually do track together — but only to a point. 24-hour forecasts are quite good. 3-day forecasts are reasonably good. 10-day forecasts are less reliable. Two-week forecasts are barely better than "what's normal for this time of year." The models are sophisticated; the accuracy doesn't scale with the sophistication. Atmospheric systems are genuinely chaotic beyond a certain horizon.

Social media "predictions." Someone posts a detailed thread with eight interconnected observations about a geopolitical situation, each point adding specificity to the last, ending with a confident conclusion. It feels well-reasoned. But complexity of argument is not accuracy of prediction. The connections might all be logical while resting on a flawed premise. You can build an elaborate wrong house.

Medical misinformation. Pseudoscientific health claims often use the vocabulary and structure of real science — mechanisms, pathways, statistics — without the actual basis. The complexity makes it harder to dismiss at first glance. Real scientific claims can also be complex. But complexity alone is not a quality signal.

How to Spot It

• Precision beyond what the system supports. When predictions are precise to decimal places in a system nobody understands well enough to justify that precision, the numbers are performing confidence rather than expressing it.

• No track record. The most important question for any forecaster: How have their past predictions held up? Not the impressive-sounding ones — all predictions. Forecasters who don't share their failure rate, or whose predictions are phrased so vaguely they can't really be checked, are not accountable to accuracy.

• Complexity that can't be simplified. Real understanding can be explained at multiple levels of depth. If a "complex" forecast can't be summarised in plain language, it might be that the complexity is load-bearing for the authority, not for the accuracy.

• No acknowledgement of uncertainty. Legitimate forecasters express uncertainty. They use ranges. They say "this assumes X, which could be wrong." Forecasters who present everything with equal confidence regardless of how far out they're predicting — suspect.

• The predictions always confirm what you'd expect. A complex forecast that happens to perfectly validate what the forecaster's client, employer, or ideology would want is worth extra scrutiny. Complexity can dress up motivated reasoning very effectively.

Your Challenge

Find a specific prediction someone made in the news in the last year — about the economy, a political outcome, a sports result, anything.

Research: What actually happened? Was the prediction right, wrong, or unprovable?

Then check: Did the complexity of the prediction correlate with its accuracy? Was the simpler version more or less right than the detailed version?

Track three predictions over the next month. Write down what was predicted and what actually happens. You'll quickly develop a feel for who is genuinely calibrated and who is performing confidence.

The more someone pretends to know about an uncertain future, the less you should trust the specifics. Accuracy lives in the range. Confidence lives in the decimal place.