Max Nudelman

US hospital charges · 2,906 hospitals, 540 procedures

Every American hospital publishes what it charges, and that number decides what an uninsured patient owes. It explains about 6% of what is actually paid, and 85% of its variation is explained by which hospital you walked into rather than what was wrong with you. The median hospital bills 4.4 times what it collects.

Two ways of counting the same thing · 118 countries, 1990 to 2023

Countries report the emissions released inside their borders, which is not the same as the emissions their consumption causes. Of the 40 countries that reported a cut, 34 saw the emissions driven by what they consume fall by less, and for 11 of them it rose. Globally the two measures are the same number, so the accounting does not hide emissions, it just decides who holds them.

Home advantage, decomposed · a natural experiment across 11 divisions

Home teams win more often, but the usual explanation is a list rather than an answer. COVID emptied every stadium in Europe at once, which let me test those explanations against each other. The home team's edge in shots halved. Its edge in fouls and cards vanished completely, and the same referees reversed their own bias.

Volleyball Nations League 2025 · finding the errors in my own work

I wrote a volleyball statistics analysis for a class, published it, and later found that one of its central metrics was mathematically impossible and one of its main conclusions was circular. I traced the first to a mislabelled column in the source data and the second to scoring each stat against a total that already contained it. This is the rebuild, and what the data actually says once both are fixed.

Elo and Poisson forecasting · interactive dashboard

I built a forecasting model for the 48 team 2026 World Cup from scratch, starting with importance weighted Elo ratings fit over the full history of international results and feeding those into a ridge regularised Poisson attack and defence model that simulates every group and produces advancement odds. It runs as a self contained interactive dashboard.

Odds movement and bookmaker margin · Premier League, 7 seasons

I took seven seasons of English Premier League prices from 16 bookmakers, stripped out the margin that every price has built into it, and tracked each one from the opening line to the close. Selections whose price shortened before kickoff beat their opening implied probability by 3.4 points, while the ones that drifted missed theirs by 5.0. The same result shows up independently at Bet365, Pinnacle and the best available price, which is the reason I think it's real.

Bettor segmentation and account risk policy · 7 European leagues

This is the call I was making by hand at Fanatics, done properly with data. Closing line value identifies an adverse account in five settled bets, while realized profit never does. The more useful finding is the second one: restricting too many people costs the book four times more than the sharps ever take.