The Oscars Model

A logistic regression trained on precursor award results from the past 30 years, run in three stages that pick up signal as the season fires. Every number here comes from testing it on years it never trained on. ← The Academy Awards

The record

Leave-one-year-out cross-validation. No ceremony influences its own prediction.

Winner hit rate
77%
178 picks · 30 yrs · 1997+ · Core 6
AUC-ROC
0.900
discrimination
Brier score
0.085
calibration
What the model produces
This season's forecast
Every Big-8 category with current odds, the precursors driving each call, and how the shape of the nominations moves the favorite. The stage switches automatically as the season progresses.
See the current forecast →
Year by year
Thirty years of ceremonies, category by category, with every nominee, the pick the model would have made, and what actually won. The misses are in here too.
Browse 30 years →
How it works

Oscar forecasting is a small-N problem with one very strong signal. Candidates turn over completely every year and only one nominee can win a category, so there is not much to learn from volume. What does carry is precursor awards: the guild results, plus BAFTA, the Golden Globes and Critics' Choice.

The model runs in three stages. Before the season it works from career pedigree, genre and festival premieres. Through the fall it forecasts nominations. Once nominations are announced it switches to win probabilities among the confirmed nominees, which is where the guild results do most of the work.

The headline hit rate covers the Core 6 categories from 1997 on. Both screenplay categories are reported separately rather than pooled in, because they land around 46 to 47 percent against 54 to 79 percent for the other six, and averaging them together would hide where the model actually has signal.

Read the full methodology →

What it cannot see

The model only knows what has already been announced. Early in the season, before the guilds vote, it is working from pedigree and release timing, which are weak signals next to a PGA result. The odds move a lot in January for a reason.

Both screenplay categories are close to a coin flip. WGA results are in the data, but they diverge from the Academy often enough that those two calls should carry much less weight than the rest. Only the Big-8 are modeled at all, so the craft categories are not forecast here.

Campaign spending, branch politics, and the second and third choices that decide a preferential Best Picture ballot are all invisible to the model.

When two nominees sit close together, treat the gap as noise. A 77 percent hit rate means roughly one in four of these calls is wrong, and the close ones are where that lands.