Hi HN. Few days ago I posted about the NBA Edge Index. Today I'm looking at the NFL. NBA markets are almost perfectly calibrated. NFL markets are not. Heavy favorites underperform by 14 points.
The NFL Edge Index tracks how teams perform against Polymarket odds, compounding into a rating across seasons.
HOW IT WORKS:
Edge function pulls Polymarket's API for NFL game markets. Then we capture last valid price before kickoff (filtered 5-95% to exclude illiquid noise). After games finalize, it compares outcomes to pre-game odds. Each team starts at a baseline rating, and if they beat expectations as an underdog, rating climbs. Lose as a favorite, it drops. The rating compounds into a trajectory across seasons.
THE FINDING - NFL FAVORITES ARE OVERPRICED:
80%+ implied: 75 games, market expected 85.1%, actual 70.7%
70-79% implied: 161 games, market expected 74.7%, actual 67.1%
60-69% implied: 201 games, market expected 64.5%, actual 53.7%
Heavy favorites underperform by 14 percentage points. "Any given Sunday" is measurable.
For comparison, I ran the same analysis on 5,600 NBA games. Basketball markets are almost perfectly calibrated - 80%+ favorites win at ~82%. The NFL gap suggests something structural: parity, injury variance, weather, or thinner markets.
TEAM-LEVEL FINDINGS:
Best underdog performers - Kansas City: 76.5% win rate as underdogs (13-4). Market priced them at 31% average.
Detroit Lions: 65% win rate as underdogs (13-7). Market consistently undervalued them.
Worst as favorites - Carolina Panthers: Lost 87.5% of games when favored. 14 losses in 16 favored games. New York Giants: 76.9% loss rate as favorites.
Biggest upset - Dallas beat NYG at 7.5% implied odds (Nov 2023). Market gave NYG 92.5%.
EDGE ACCURACY:
70.9% accuracy when model had 70%+ confidence (161/227 games).
I'm sorry, it's my bad that I didn't post any details into comments.
The converter actually works pretty well even with large and complex designs. You should really try it out. There are still some features that we're working on.
Very interesting post! Our team also applied to YC. We didn't even get an interview :/ although we have a user base, paying customers and a product that got 20k sign-ups the first 48 hours. Read more here: https://medium.com/avocode-stories/ec1a87294204
"Right now Avocode is focused on the web but we would like to offer the same experience for mobile developers too" - Question 12 http://avocode.com/faq.html
The NFL Edge Index tracks how teams perform against Polymarket odds, compounding into a rating across seasons.
HOW IT WORKS: Edge function pulls Polymarket's API for NFL game markets. Then we capture last valid price before kickoff (filtered 5-95% to exclude illiquid noise). After games finalize, it compares outcomes to pre-game odds. Each team starts at a baseline rating, and if they beat expectations as an underdog, rating climbs. Lose as a favorite, it drops. The rating compounds into a trajectory across seasons.
THE FINDING - NFL FAVORITES ARE OVERPRICED: 80%+ implied: 75 games, market expected 85.1%, actual 70.7% 70-79% implied: 161 games, market expected 74.7%, actual 67.1% 60-69% implied: 201 games, market expected 64.5%, actual 53.7%
Heavy favorites underperform by 14 percentage points. "Any given Sunday" is measurable.
For comparison, I ran the same analysis on 5,600 NBA games. Basketball markets are almost perfectly calibrated - 80%+ favorites win at ~82%. The NFL gap suggests something structural: parity, injury variance, weather, or thinner markets.
TEAM-LEVEL FINDINGS: Best underdog performers - Kansas City: 76.5% win rate as underdogs (13-4). Market priced them at 31% average.
Detroit Lions: 65% win rate as underdogs (13-7). Market consistently undervalued them.
Worst as favorites - Carolina Panthers: Lost 87.5% of games when favored. 14 losses in 16 favored games. New York Giants: 76.9% loss rate as favorites.
Biggest upset - Dallas beat NYG at 7.5% implied odds (Nov 2023). Market gave NYG 92.5%.
EDGE ACCURACY: 70.9% accuracy when model had 70%+ confidence (161/227 games).
Live dashboard: https://attena.xyz/nfl
If you're trading on Polymarket, I'd be curious if you find this data valuable.