How the score works
Every prediction source speaks a different language. Clarivate names people or it doesn't. Tyler Cowen publishes a ranked list. A pool gives vote shares. A betting market gives a probability. The KmikeyM Power Ranking turns each of those into one number, the composite, so the board can be sorted, argued about, and replayed.
Step 1: each observation becomes a heat from 0 to 100
| Kind | What the source said | Heat |
|---|---|---|
named | The source lists this person | 100 |
rank | Position r in a list of n | 100 × (n − r + 1) ÷ n. First of five is 100, fifth of five is 20. |
share | A vote share from 0 to 1 | share × 100 |
prob | A market or bookmaker probability from 0 to 1 | probability × 100 |
If a source publishes again about the same person, the newer observation replaces the older one. The old one stays in the record.
Step 2: sources carry weights
The weights are our editorial call. They are published here, from the same file the build reads, so what you see is what the score uses.
| Source | Kind | Weight | Why |
|---|---|---|---|
| Clarivate Citation Laureates | named | 3 | The strongest quantitative predictor. Each September Clarivate names a handful of economists whose citation records look Nobel-class. Dozens of past picks have gone on to win, often years later. |
| Clarivate alumni (prior years) | named | 1 | Everyone Clarivate named in an earlier year who has not won yet. Still a real signal, since most Citation Laureates who win do so years after being named, but a weaker one than this year's list. |
| Polymarket | prob | 3 | A prediction market with real money on it. Its 2026 Economics Nobel market opened October 6, six days before the announcement, with seventeen names and more volume on day one than Kalshi's whole season. We record the price Polymarket displays, the midpoint of its book; best bid, ask and volume ride in the note. Kept separate from Kalshi so the two can disagree in public. |
| Kalshi | prob | 3 | A regulated US prediction market. Its 2026 Economics Nobel market opened July 1 with nine names and added Emmanuel Saez, Thomas Piketty and Sendhil Mullainathan on October 7. We record the price Kalshi displays, the last trade, and fall back to the midpoint of the best yes bid and ask for contracts that have never traded. Bid and ask ride in the note, since the book is thin. Kept separate from Polymarket so the two can disagree in public. |
| Ladbrokes | prob | 2 | Bookmaker odds, converted to an implied probability. Bookmakers price to balance their book, not to be right, so they carry less weight than a market. No UK book had posted a 2026 economics market as of October 6; they price Literature and Peace and skip this one. |
| Marginal Revolution | rank | 1.5 | Tyler Cowen's annual picks, in his order. One well-read economist's judgment, published before the announcement every year. |
| HSE prediction contest | share | 1 | HSE University's annual open prediction contest. Aggregates informed opinion, reported as shares of entries. |
| KmikeyM desk | named | 0.5 | Our own calls. Deliberately light, so the desk can flag a sleeper without bending the board. |
| Prophet Arena AI consensus | prob | 1.5 | Ten frontier language models (Claude, GPT, Gemini, DeepSeek, GLM, Kimi) each forecast the nine Kalshi names from a shared research briefing, and the site averages them. We record the consensus. It is the market's own question put to machines instead of money, so it earns half a market's weight. Closes October 13. |
| Octagon AI model | prob | 0.5 | A single AI model's probabilities over the Kalshi names, with an undisclosed method and a last-updated date of July 28. Stale and sparse, so it carries little weight, but it is a second machine opinion and it disagrees with the first. |
| Richard Tol's model | rank | 1.5 | A statistical reconstruction of every past prize, published as a paper, that scores 161 living candidates on citations, coauthorship with earlier laureates and the historical rotation between fields, then lists the top ten. Published each September since 2025. The only source on the board that is a model of the committee rather than a guess about the economists. |
| Scott Cunningham | rank | 1.5 | An annual ranked list from the author of Causal Inference, the Mixtape, published every October since 2021 behind a paywall. We record the order as reported publicly, since we cannot read the post ourselves, and say so in the note. |
| Nicholas Decker | named | 1 | A long-form economics writer who posts one confident prediction a year with an essay behind it. In 2025 he named Berry, Hausman and Pakes, then wrote that the committee would instead go to Aghion and Howitt, which it did. Second year on record. |
| Maia Mindel | named | 0.5 | An economics writer's picks posted as a thread on X. We have the names secondhand, from a roundup, and could not retrieve the thread itself; the weight reflects that. One year on record. |
| Tomas Bata University contest | share | 0.5 | An annual guessing game run by the economics department at Tomas Bata University in Zlin, parallel to HSE's. Entries by email through October 11, results announced with the prize. Enters the board when the tally is published. |
Step 3: the composite
A candidate's composite is the sum of weight × heat across sources, divided by the total weight of every source that has published anything so far. A source that has not published yet is ignored entirely, so a Clarivate name in September sits high until the markets open. When a new source comes online, everyone it ignores drops and everyone it names rises. That reshuffle is the season, and the replay shows it.
Tiers
Lock is 70 and up. Contender is 40 to 69. Darkhorse is 15 to 39. Field is everyone else we track. A Sleeper badge marks a KmikeyM desk call that the composite has not caught up to yet.
Worked example: Ariel Pakes
As of Oct 21, 2026, these sources have published something (only they count):
| Source | Weight | Heat for Ariel Pakes | Weight × heat |
|---|---|---|---|
| Clarivate Citation Laureates | 3 | 0 (not mentioned) | 0 |
| Clarivate alumni (prior years) | 1 | 100 | 100 |
| Polymarket | 3 | 18 | 54 |
| Kalshi | 3 | 21 | 63 |
| Prophet Arena AI consensus | 1.5 | 21 | 31 |
| Octagon AI model | 0.5 | 8 | 4 |
| Richard Tol's model | 1.5 | 0 (not mentioned) | 0 |
| Scott Cunningham | 1.5 | 50 | 75 |
| Nicholas Decker | 1 | 100 | 100 |
| Maia Mindel | 0.5 | 100 | 50 |
Total weight of active sources: 16.5. Sum of weight × heat: 477. Composite: 477 ÷ 16.5 = 28.9.
Arguing with it
Every weight is a line in a public file and every change is a dated update, so if we change our minds the replay shows exactly when. If you think a source deserves more or less, that argument is the content. Write in.