Veneto Classic 2026

The fagIAno · frozen October 17, 22:30, before the start

Who wins Veneto Classic, according to our models

SHA-256 fingerprint b299a0213aec… What this means →

What we do

• ✓done: The 5 editions since 2015, 3 of them finishing in Bassano del Grappa
• ✓done: Route, climbs and timings: 99 locations
• ✓done: The entry list: 146 riders from 21 teams
• ✓done: Hour-by-hour weather at the key points
• ✓done: Our model: how the race can end
• ✓done: Riders' model: passed the test on races it had not seen
• ○in progress: Starters confirmed on the eve of the race

Our predictions

1. How it ends: the winner arrives alone, at 50% (between 18.9% and 81%).
2. The margin: without a dominant rider, between 14" and 17" over second place.
3. Who: according to the rider model, the most likely winners are Brandon McNulty (28%), Christian Scaroni (18.3%), Sakarias Koller Løland (5%).

Two scenarios

Where it is decided

The favourites will probably attack on the La Tisa: 0.3 km at 15.2%, peaks of 20%, summit 19.1 km from the finish. It is the hardest climb of the finale.

This is a reading of the route: length and gradient of the climbs in the finale.

How it can end

How big a gap? In the years without a dominant rider (2023, 2024) the solo winner had between 14" and 17" on second place. Every year between 53% and 66% of the starters finish.

The favourites

The test: on the 159 races it had not seen, the model has a log loss of 2.88, against 3.58 for the UCI ranking (lower is better).

The weather

At the expected passing times between 20 and 22 degrees, wind up to 8 km/h with gusts up to 17, rain at most 15% likely. The weather should not change the race.

The models we use

How it can end la fagIAnata’s model

From the 3 editions finishing in Bassano del Grappa, the probability of each outcome with its 90% interval: with few editions the interval is wide, as it should be.

Who is strongest The riders model

Each starter’s chance of winning, frozen before the start. We publish it because the model passed the test: on the 2023 races it hadn’t seen, it predicted better than the UCI ranking.

The benchmark UCI one-day race ranking

The riders model has to predict the races it has not seen better than the UCI ranking does: we measure it with the log loss. If it fails, we do not publish its probabilities.

How big a margin, how many finishers Statistics on the times

Gaps and finishers in past editions, from the UCI times: the minimum and the maximum, no model involved.

Where it is decided Climb difficulty index

The climbs of the last 30 km ranked by hardness, from length and gradient. A simple rule, not a statistical model.

The weather Numerical forecast

Hour-by-hour temperature, wind and rain from the weather models, at the altitude of each point on the route and at the time the bunch passes.

Our method → The record of every race →