Every prediction here was published before the race and settled against the result afterwards. The exceptions are the pages marked post-hoc: boards rebuilt after the race, which are labelled as such and left out of the record. Where market probabilities appear, they are the benchmark we score ourselves against — a measure of how much real information the model captures.
Live now
Nightly model run and results record for Tour of Guangxi.
1/6 stages settled · 13 Oct 2026 – 18 Oct 2026
Next up: Stage 2 (Wed 14 Oct) · prediction →
Recent predictions
Coming up
Retrospectives
How the three-week campaign scored — stage by stage, against results and against the market benchmark.
How the three-week campaign scored — stage by stage, against results and against the market benchmark.
The board and the result, side by side. Reconstructed after the race and labelled as such.
8 settled stages with predictions frozen before each start, plus the GC tracker as it evolved.
6 settled stages with predictions frozen before each start, plus the GC tracker as it evolved.
The board and the result, side by side. Reconstructed after the race and labelled as such.
The board and the result, side by side. Reconstructed after the race and labelled as such.
3 settled stages with predictions frozen before each start, plus the GC tracker as it evolved.
The board and the result, side by side. Reconstructed after the race and labelled as such.
5 settled stages with predictions frozen before each start, plus the GC tracker as it evolved.
5 settled stages with predictions frozen before each start, plus the GC tracker as it evolved. Reconstructed after the race and labelled as such.
4 settled stages with predictions frozen before each start, plus the GC tracker as it evolved. Reconstructed after the race and labelled as such.
The board and the result, side by side. Reconstructed after the race and labelled as such.
The board and the result, side by side. Reconstructed after the race and labelled as such.
The board and the result, side by side. Reconstructed after the race and labelled as such.
The board and the result, side by side. Reconstructed after the race and labelled as such.
4 settled stages with predictions frozen before each start, plus the GC tracker as it evolved.
The board and the result, side by side. Reconstructed after the race and labelled as such.
The board and the result, side by side. Reconstructed after the race and labelled as such.
2 settled stages with predictions frozen before each start, plus the GC tracker as it evolved.
4 settled stages with predictions frozen before each start, plus the GC tracker as it evolved.
4 settled stages with predictions frozen before each start, plus the GC tracker as it evolved.
20 settled stages with predictions frozen before each start, plus the GC tracker as it evolved.
21 settled stages with predictions frozen before each start, plus the GC tracker as it evolved.
The full write-up: seven pricing heads, one pricing layer, the day-type retrieval channel, and how the outputs are compared with the market.