DIGNALEGI

The reading room · Digna Legi

Does Forecasting Have Room At The Top?

A personal relevance score

80–100: high value. 70–79: worth the time. Below 70: below the usual publication threshold.

Evidence-reviewed score based on available publisher text. Evidence is sampled; the excerpt gives the main reasoning but not every calculation or counterargument.

Scores reflect one reader’s profile, not an objective quality rating. Best is a separate personal selection.

How scoring works →

This brief · about 3 min with detail

Original article ↗

Why read this

AI forecasting gains may look small in uncertainty-heavy domains yet still matter when measured against the right baseline.

AI brief · Checked against source text

The main idea

The author argues that small-looking gains in forecasting can still be large once measured against the right baseline and domain difficulty. Sports, designed to keep outcomes near uncertainty, make even meaningful improvements look tiny, while geopolitics may contain easier extremes and harder strategic questions. His central expectation is not miraculous foresight, but AI forecasters improving prediction markets by practically useful percentage points.

Some background helpful. Comfort with probability forecasts, statistical error scores, logarithmic scales, and comparisons against simple forecasting baselines.

Go a little deeper

Baselines change the meaning of progress

The same result can look trivial or impressive depending on what it is compared with. In the sports study, prediction markets only improved a measure of prediction error modestly, but the author argues that the improvement was large relative to what remained after simple information such as team strength was already included. The mechanism is compression: once a baseline captures obvious structure, the leftover signal is harder to extract.

Some domains are engineered to resist prediction

Sports are not merely noisy; they are commercially and institutionally shaped to keep outcomes uncertain. Salary caps, drafts, equal field positions, and balanced teams all push games toward coin-flip territory. That means a forecasting method can be genuinely better while producing only small absolute gains, because the domain has intentionally reduced exploitable differences.

Geopolitics may not inherit sports limits

The author rejects a naive transfer from sports to world events because the distribution of questions differs. A best-versus-worst sports matchup can still allow upsets, while some geopolitical questions may sit near certainty. He uses the gap between a simple baseline forecast and Metaculus professional forecasters to argue that real-world forecasting has already surpassed the uncertainty removal implied by the sports-based limit.

Small percentage-point gains can matter

The author sees practical value in moving a market from 50% to 54%, 56%, or 62%. Describing his own Polymarket use, he says he usually already knows that an event is neither extremely unlikely nor almost certain; he wants a more precise estimate within that middle range. He also expects better forecasters to help frame questions and identify neglected risks, extending their value beyond market accuracy.

A case from the article

What a percentage gain hides

The box-office example shows why percentage improvements can mislead. Markets improved over models that already knew how many screens a movie had and how much search interest it attracted, while receipts varied from tiny releases to blockbusters. Because the study measured performance on a scale that compresses huge dollar differences, a 6% gain may understate how much extra signal markets added beyond obvious predictors.

How the case is made

The case is made through metric reinterpretation, domain analogy, rough score conversions, and explicit comparisons among sports, movies, geopolitics, chess, and prediction markets.

Where the idea has limits

The author repeatedly marks the quantitative conversions as approximate: some reference points are guessed, the score scale is transformed, and the napkin math increasingly relied on AI assistance.

A question to take away · from Digna Legi

Where are you dismissing a useful forecast because its absolute improvement looks small in a deliberately hard domain?

What the original adds

The original contains the detailed conversion path behind the estimates, including error-score comparisons, transformed forecasting scores, and chess-handicap analogies that this brief compresses.

About this brief

AI-written, then separately checked for source support, useful detail and clarity. The author’s claims and our editorial question are kept separate. The original remains the author’s work. How we select and summarise →

Digna legi. Worth reading.