Math & Stats @ Purdue (May 2027) Quant research, sports analytics, prediction markets
West Lafayette, IN
Joined May 2022
- Tweets366
- Following212
- Followers57
- Likes455
Pinned Tweet
Before a ball was kicked at this World Cup, I published 11 falsifiable predictions about how these markets would behave, with the thresholds set in advance. Today I graded all of them in public. 6 passed, 2 failed, 3 inconclusive. 🧵
The two failures both reproduce what the 2026 microstructure papers found. The documented goal-overreaction edge did not survive: PnL negative, arbitrage away. And the law of one price broke my own threshold, a persistent 3.9pp cross-venue gap where I predicted under 1.
3 of the 11 were inconclusive for data reasons I could not fix, all documented in an addendum. One tournament, one sport. Six weeks of tick capture, eleven calls made before kickoff, graded in the open, hits and misses. Everything is in the repo: thesavagecoder7784.github.io…
I took de-vigged winner prices for all 48 teams from a retail sportsbook (William Hill), a betting exchange (Betfair), and Polymarket. De-vigged, so this is pure pricing with the margin stripped out, not the bookmaker's cut.
William Hill priced the longshots roughly 20 to 25% above the sharp consensus, and its favourites a few percent cheap. Textbook favourite-longshot bias. Rank correlation between price and retail markup is -0.5 to -0.8, p under 0.005, stable across dates.
The tell is who does NOT do it. Betfair and Polymarket agree with each other, no bias, their gap is pure noise (p over 0.2). So a brand-new prediction market prices longshots exactly as lean as a sharp exchange. The padding is a retail thing. thesavagecoder7784.github.io…
The favourite-longshot bias, longshots overpriced and favourites too cheap, is one of the oldest findings in betting markets. Before the World Cup kicked off, I lined up the winner odds from three places to see who still had it. Only the retail book does. 🧵
Spain 1-0 France, first half of the semi. My model went in calling this a coin flip. Spain breaking the deadlock swings them from 50/50 to about 80% to reach the final. One goal, and the tightest game of the tournament is 4-in-5 Spain.
The semis are now set, and my four most likely semi-finalists, Spain, France, England, and Argentina, are the exact final four. But the stranger pattern is underneath. The chart is below. 🧵
It is a verdict on my own model too, not a victory lap. I nailed the final four, but I under-rated the survivors, Switzerland at 24 percent is in the last eight, and over-rated the fallen.
Every probability frozen in an append-only ledger before a ball was kicked, graded in public on July 19, hits and misses. The full board, the commit, and the code: thesavagecoder7784.github.io…
Prabhat M retweeted
Before a ball was kicked, my model's four most likely semifinalists were Spain, France, England, and Argentina. Spain and France are already through. If England and Argentina win tonight, my pre-tournament top four becomes the exact final four of this World Cup.
