Cycling race predictor · Vuelta a España 2026

Stage 16 — 2026-09-08

The same production stack priced under 4 course readings, against the bookmaker's pre-stage line. Each scenario is one opinion about what kind of day this is; the results section settles the argument.

4 scenarios 150 riders priced models retrained d911de60f7 book snapshot markets_2054.json

← System overview, season boards & the full race views

The fork

4 readings of the same road

Course profile is an input the model can't see from a map — it's set from reference races (or overridden by hand). Different readings redistribute the same probability mass across rider types.

Bunch sprint

The gallop — the hills are all in the first half and Huelva is flat

sprint1.00

mountain0.00

hilly0.00

rolling0.00

break0.00

TT0.00

keyed: bunch sprint 0.6 / survivor gallop 0.3 / fast-finisher power day 0.1 (board order by rider-type retrieval; prices unchanged)

The base case and the book's card. The organiser files the day as hilly, but every metre of the climbing is early: the race spends its first forty kilometres going up and down through the Sierra de Aracena, the last rise (about 4.5 km at 4.5%, uncategorised) tops out near km 90, and the remaining ninety kilometres run downhill and then flat across the plains of Huelva to a sea-level finish by the La Rábida monastery. No categorised climb anywhere on the stage. References are the campaign's pure-speed anchor set — flat GT and WT sprints at race_sprint_score 1.00 with winner groups of 127 to 181 — the same board that had Brennan #1 before Xeraco and Lorca. Inherits the config-root day key.

Top 5 (blend): Brennan Matthew 50.9%; Meeus Jordi 9.0%; van Aert Wout 6.4%; De Schuyteneer Steffen 2.7%; Coquard Bryan 2.0%

Wind thins it

The coast thins it — a reduced group after the estuary headwind

sprint0.52

mountain0.01

hilly0.46

rolling0.01

break0.00

TT0.00

keyed: survivor gallop 0.5 / bunch sprint 0.3 / split sprint 0.1 (board order by rider-type retrieval; prices unchanged)

The middle case. Two thousand metres of constant up-and-down in the first half is real work on the day after a rest day, the sprinters' trains are the thinnest they have been all race (Pedersen, Groves and Laporte gone, Red Bull riding for a podium), and the run to the coast is reported to carry a headwind over the final twelve kilometres — the condition under which a bunch arrives strung out and smaller than it looks on paper, or splits outright on the exposed estuary roads. References are the lumpy-finale reduced-sprint archetype (sprint ~0.51, hilly ~0.46): Philipsen TdF 2023 stage 7, Merlier Giro 2024 stage 3, Coquard Suisse 2024 stage 2, Milan Renewi 2024 stage 1, Merlier UAE 2025 stage 5. A headwind finish argues against a long-range split holding, which is why this board is the hedge rather than the read.

Top 5 (blend): Brennan Matthew 49.3%; van Aert Wout 13.7%; Meeus Jordi 12.2%; Coquard Bryan 2.7%; De Schuyteneer Steffen 2.3%

Break survives

The escape survives — one sprint team, ninety flat kilometres, nobody else with a reason

sprint0.00

mountain0.38

hilly0.00

rolling0.62

break0.00

TT0.00

keyed: hard puncheur day 0.5 / mountain escape 0.3 / rouleur power group 0.2 (board order by rider-type retrieval; prices unchanged)

The branch this campaign has learned not to dismiss — it is what actually happened on stage 15, and the failure there was keying a sprint family when one team owned both the sprinter and the escape candidate. The same configuration is live again: Visma have Brennan and van Aert, and van Aert has won two of the last three stages from moves. Against that, the road is far friendlier to a chase here than at Córdoba: ninety kilometres of flat after km 90 and a headwind run-in are the two things that most reliably bring a break back. References are hilly and medium-mountain GT stages won from an escape solo or two-up with the peloton minutes back: van Aert Vuelta 2024 stage 10, Izagirre TdF 2023 stage 12, Cort TdF 2022 stage 10, Vendrame Giro 2024 stage 19, Bettiol Giro 2026 stage 13. The GC top ten is cut on standings so the whole book tail is free to be ranked.

Top 5 if the break wins (top 10 on GC removed): Bisiaux Léo 5.9%; Johannessen Tobias Halland 5.0%; Buitrago Santiago 4.3%; Debruyne Ramses 4.1%; Tejada Harold 3.6%

Hindsight (post-hoc)

Hindsight (post-hoc) — the day as raced: 102 riders together and Brennan wins the sprint

sprint0.49

mountain0.02

hilly0.32

rolling0.06

break0.12

TT0.00

keyed: bunch sprint (board order by rider-type retrieval; prices unchanged)

Post-race reconstruction, NOT a prediction — run with --no-log and never write it to prediction_log. The road produced the clean mass sprint the official key led with: 102 riders finished on Brennan's time, Coquard and Braet completed the podium, and the GC remained unchanged. References are five disjoint 2026 mass sprints with groups of 29 to 117: Brussels Cycling Classic, Heistse Pijl, Boucles de la Mayenne stage 3, Giro stage 21 and Giro stage 18. The test is whether exact knowledge of a very large bunch changes anything material from the prospective pure-speed board that already ranked Brennan first.

Top 5 (blend): Brennan Matthew 56.0%; van Aert Wout 10.0%; Meeus Jordi 9.9%; Coquard Bryan 1.7%; De Schuyteneer Steffen 1.3%

The board

Win probabilities, scenario by scenario

Blend win probability per scenario (cascade × LambdaRank, DNF-adjusted at the pricing layer), next to the devigged Kambi (Unibet) win line captured the morning before the stage. "Max edge" compares the most favourable scenario against the book — treat it as scenario-conditional, not a free bet.

Read "Max edge" as directional, not as a price. The blend is a set of per-rider marginals and does not sum to one book — across this field each reading totals Bunch sprint 85%, Wind thins it 98%, Hindsight (post-hoc) 92% — while the devigged book column sums to exactly 100%. Putting the two side by side flatters the model, and Max edge is inflated by about that factor: divide a rider's probability by his reading's total before calling anything an edge. Of our heads only the LambdaRank one is natively book-preserving.

Rider	Team	Bunch sprint	Wind thins it	Break survives	Hindsight (post-hoc)	Book odds	Book devig	Max edge
Brennan Matthew	Team Visma \| Lease a Bike	50.9%	49.3%	0.7%	56.0%	1.85	36.2%	+19.8pp
van Aert Wout	Team Visma \| Lease a Bike	6.4%	13.7%	3.4%	10.0%	8.00	8.4%	+5.3pp
Meeus Jordi	Red Bull-BORA-hansgrohe	9.0%	12.2%	0.1%	9.9%	7.00	9.6%	+2.6pp
De Schuyteneer Steffen	Lotto-Intermarché	2.7%	2.3%	0.1%	1.3%	41.00	1.6%	+1.1pp
Coquard Bryan	Cofidis	2.0%	2.7%	0.1%	1.7%	9.00	7.4%	-4.8pp
Romele Alessandro	XDS Astana Team	1.1%	1.7%	0.2%	1.1%	26.00	2.6%	-0.8pp
Dainese Alberto	Soudal Quick-Step	1.7%	0.8%	0.0%	0.5%	41.00	1.6%	+0.1pp
Hofstetter Hugo	NSN Cycling Team	1.0%	1.2%	0.1%	0.9%	41.00	1.6%	-0.4pp
Cort Magnus	Uno-X Mobility	0.5%	1.1%	0.3%	0.7%	19.00	3.5%	-2.4pp
Laurance Axel	Netcompany INEOS	0.8%	1.1%	1.8%	0.9%	67.00	1.0%	+0.1pp
Brustenga Marc	Equipo Kern Pharma	1.1%	0.3%	0.0%	0.3%	126.00	0.5%	+0.5pp
Aular Orluis	Movistar Team	0.9%	1.1%	0.3%	0.5%	81.00	0.8%	+0.2pp
Albanese Vincenzo	EF Education-EasyPost	0.4%	1.0%	0.3%	0.6%	101.00	0.7%	+0.4pp

Break survives is shown conditional on the break winning — the top 10 riders on actual GC are removed and the remaining probabilities renormalize within the break pool. That column isn't comparable with the book line or the other readings, so it's left out of Max edge; a "–" means a GC captain we assume won't be up the road.

The gallop — the hills are all in the first half and Huelva is flat — top 30, all heads

#	Rider	Blend	Cascade	LambdaRank	Podium	Top 15	Book odds
1	Brennan Matthew	50.9%	39.9%	62.8%	67.2%	86.8%	1.85
2	Meeus Jordi	9.0%	10.9%	10.1%	35.1%	78.3%	7.00
3	van Aert Wout	6.4%	4.2%	8.1%	21.9%	62.0%	8.00
4	De Schuyteneer Steffen	2.7%	2.4%	3.3%	9.1%	63.2%	41.00
5	Coquard Bryan	2.0%	2.2%	2.3%	7.7%	80.6%	9.00
6	Dainese Alberto	1.7%	1.5%	2.1%	8.9%	50.8%	41.00
7	Romele Alessandro	1.1%	3.8%	0.7%	11.2%	65.1%	26.00
8	Brustenga Marc	1.1%	1.0%	1.3%	7.0%	50.9%	126.00
9	Hofstetter Hugo	1.0%	1.2%	1.1%	8.0%	78.9%	41.00
10	Aular Orluis	0.9%	0.6%	1.2%	6.1%	67.5%	81.00
11	Laurance Axel	0.8%	2.3%	0.6%	7.8%	47.6%	67.00
12	Nys Thibau	0.7%	2.3%	0.5%	7.3%	45.3%	41.00
13	Tronchon Bastien	0.7%	1.6%	0.6%	8.2%	63.2%	41.00
14	Govekar Matevz	0.7%	1.5%	0.6%	8.6%	53.2%	101.00
15	Braet Vito	0.6%	1.1%	0.6%	6.1%	61.7%	101.00
16	Miquel Delgado Pau	0.5%	1.2%	0.4%	4.8%	52.2%	101.00
17	Macias Cesar	0.5%	0.7%	0.5%	5.5%	57.9%	201.00
18	Cort Magnus	0.5%	1.3%	0.4%	6.7%	55.2%	19.00
19	Renard Alexis	0.4%	1.1%	0.3%	4.6%	48.6%	151.00
20	Albanese Vincenzo	0.4%	0.2%	0.5%	2.9%	57.2%	101.00
21	Canal Carlos	0.2%	0.2%	0.3%	2.0%	19.0%	401.00
22	Oliveira Ivo	0.2%	0.2%	0.2%	3.0%	26.0%	81.00
23	Camprubí Marcel	0.1%	0.1%	0.1%	2.3%	35.7%	201.00
24	Mulubrhan Henok	0.1%	0.1%	0.1%	1.5%	16.7%	151.00
25	Garofoli Gianmarco	0.1%	0.1%	0.1%	0.8%	9.3%	401.00
26	Debruyne Ramses	0.1%	0.1%	0.1%	0.8%	12.2%	201.00
27	Brenner Marco	0.1%	0.2%	0.1%	1.4%	8.4%	n/a
28	Vader Milan	0.1%	0.1%	0.1%	0.8%	17.8%	601.00
29	Fedorov Yevgeniy	0.1%	0.2%	0.0%	0.6%	17.6%	301.00
30	Hayter Ethan	0.1%	0.3%	0.0%	1.7%	15.9%	151.00

The coast thins it — a reduced group after the estuary headwind — top 30, all heads

#	Rider	Blend	Cascade	LambdaRank	Podium	Top 15	Book odds
1	Brennan Matthew	49.3%	41.9%	52.4%	67.9%	87.4%	1.85
2	van Aert Wout	13.7%	6.1%	15.9%	37.3%	69.4%	8.00
3	Meeus Jordi	12.2%	12.7%	12.3%	35.9%	74.1%	7.00
4	Coquard Bryan	2.7%	3.5%	2.5%	11.3%	78.0%	9.00
5	De Schuyteneer Steffen	2.3%	2.1%	2.4%	7.6%	59.8%	41.00
6	Romele Alessandro	1.7%	5.7%	0.8%	14.3%	59.1%	26.00
7	Hofstetter Hugo	1.2%	1.8%	1.1%	8.7%	73.4%	41.00
8	Cort Magnus	1.1%	2.0%	0.9%	14.0%	56.6%	19.00
9	Laurance Axel	1.1%	2.4%	0.8%	9.2%	49.2%	67.00
10	Aular Orluis	1.1%	0.7%	1.2%	7.1%	61.5%	81.00
11	Albanese Vincenzo	1.0%	1.5%	0.9%	10.2%	65.1%	101.00
12	Nys Thibau	1.0%	2.7%	0.6%	9.2%	44.4%	41.00
13	Braet Vito	0.9%	1.5%	0.8%	7.9%	57.8%	101.00
14	Dainese Alberto	0.8%	1.6%	0.7%	7.9%	47.3%	41.00
15	Tronchon Bastien	0.8%	1.8%	0.6%	8.8%	54.4%	41.00
16	Govekar Matevz	0.8%	1.7%	0.6%	8.9%	45.4%	101.00
17	Miquel Delgado Pau	0.6%	1.2%	0.5%	5.1%	52.5%	101.00
18	Canal Carlos	0.4%	0.2%	0.5%	3.2%	20.6%	401.00
19	Debruyne Ramses	0.4%	0.3%	0.5%	3.7%	19.5%	201.00
20	Renard Alexis	0.3%	0.1%	0.4%	2.3%	39.9%	151.00
21	Brustenga Marc	0.3%	0.3%	0.3%	3.3%	59.3%	126.00
22	Mulubrhan Henok	0.2%	0.2%	0.3%	2.2%	19.8%	151.00
23	Hayter Ethan	0.2%	1.0%	0.0%	2.4%	11.9%	151.00
24	Fisher-Black Finn	0.2%	0.2%	0.2%	2.6%	13.8%	151.00
25	Camprubí Marcel	0.2%	0.2%	0.2%	2.8%	27.9%	201.00
26	Brenner Marco	0.2%	0.3%	0.2%	2.4%	13.3%	n/a
27	Oliveira Ivo	0.2%	0.2%	0.2%	2.8%	21.9%	81.00
28	Macias Cesar	0.2%	0.1%	0.2%	1.6%	43.9%	201.00
29	Cobo Ivan	0.1%	0.1%	0.1%	1.4%	12.7%	301.00
30	Garofoli Gianmarco	0.1%	0.1%	0.1%	1.0%	15.2%	401.00

The escape survives — one sprint team, ninety flat kilometres, nobody else with a reason — top 30, all heads · top 10 on GC removed

#	Rider	Blend	Cascade	LambdaRank	Podium	Top 15	Book odds
1	Bisiaux Léo	3.1%	2.5%	4.7%	7.9%	39.2%	501.00
2	Johannessen Tobias Halland	2.6%	3.2%	3.7%	10.4%	37.8%	201.00
3	Buitrago Santiago	2.3%	1.9%	3.4%	6.3%	21.6%	801.00
4	Debruyne Ramses	2.1%	5.8%	2.2%	12.6%	38.0%	201.00
5	Tejada Harold	1.9%	2.7%	2.6%	7.3%	36.6%	801.00
6	Widar Jarno	1.9%	1.5%	2.8%	6.4%	37.5%	501.00
7	van Aert Wout	1.8%	3.1%	2.3%	12.2%	23.3%	8.00
8	Vermaerke Kevin	1.7%	1.7%	2.5%	6.4%	26.5%	126.00
9	Riccitello Matthew	1.5%	1.2%	2.3%	3.1%	20.0%	n/a
10	Zana Filippo	1.4%	1.6%	2.1%	5.6%	25.8%	601.00
11	Romeo Ivan	1.4%	1.8%	1.9%	2.8%	7.9%	101.00
12	Brenner Marco	1.3%	2.1%	1.7%	6.8%	18.1%	n/a
13	Paret-Peintre Valentin	1.2%	1.9%	1.6%	6.2%	23.4%	501.00
14	Tuckwell Luke	1.1%	0.4%	1.8%	3.4%	12.3%	1001.00
15	Fisher-Black Finn	1.0%	1.9%	1.3%	5.3%	14.6%	151.00
16	Vansevenant Mauri	1.0%	1.2%	1.5%	4.7%	19.0%	801.00
17	Almeida Joao	1.0%	1.4%	1.4%	4.0%	18.1%	801.00
18	Laurance Axel	1.0%	1.9%	1.1%	5.5%	16.6%	67.00
19	Leknessund Andreas	0.9%	0.4%	1.5%	6.6%	15.2%	101.00
20	Nys Thibau	0.9%	2.5%	1.0%	6.4%	18.0%	41.00
21	Berrade Urko	0.9%	1.4%	1.2%	6.4%	24.9%	401.00
22	Rodriguez Carlos	0.9%	0.6%	1.4%	2.3%	10.2%	n/a
23	Svestad-Bårdseng Embret	0.8%	0.5%	1.2%	2.0%	11.8%	n/a
24	Nordhagen Jørgen	0.8%	1.2%	1.1%	6.8%	19.1%	801.00
25	Kron Andreas	0.8%	0.4%	1.2%	2.0%	14.1%	101.00
26	Castrillo Pablo	0.8%	0.4%	1.2%	0.9%	4.1%	401.00
27	Martin Guillaume	0.8%	0.2%	1.2%	1.6%	14.0%	801.00
28	Mulubrhan Henok	0.7%	0.4%	1.1%	3.0%	14.7%	151.00
29	Bilbao Pello	0.7%	1.3%	0.8%	3.7%	17.8%	501.00
30	Dunbar Edward	0.7%	1.5%	0.8%	5.4%	24.0%	n/a

Hindsight (post-hoc) — the day as raced: 102 riders together and Brennan wins the sprint — top 30, all heads

#	Rider	Blend	Cascade	LambdaRank	Podium	Top 15	Book odds
1	Brennan Matthew	56.0%	42.6%	64.0%	69.3%	85.8%	1.85
2	van Aert Wout	10.0%	4.9%	12.1%	30.1%	62.9%	8.00
3	Meeus Jordi	9.9%	11.5%	10.3%	36.8%	72.7%	7.00
4	Coquard Bryan	1.7%	2.7%	1.6%	9.3%	71.3%	9.00
5	De Schuyteneer Steffen	1.3%	1.5%	1.4%	6.2%	52.4%	41.00
6	Romele Alessandro	1.1%	3.8%	0.5%	8.9%	52.7%	26.00
7	Hofstetter Hugo	0.9%	1.4%	0.8%	9.2%	69.5%	41.00
8	Laurance Axel	0.9%	2.2%	0.6%	8.1%	47.5%	67.00
9	Nys Thibau	0.8%	2.4%	0.5%	8.0%	42.1%	41.00
10	Cort Magnus	0.7%	1.6%	0.5%	8.3%	50.1%	19.00
11	Govekar Matevz	0.6%	1.7%	0.4%	8.0%	42.8%	101.00
12	Albanese Vincenzo	0.6%	1.1%	0.6%	6.6%	59.2%	101.00
13	Braet Vito	0.6%	1.1%	0.5%	6.2%	52.2%	101.00
14	Tronchon Bastien	0.6%	1.6%	0.4%	6.8%	51.6%	41.00
15	Aular Orluis	0.5%	0.5%	0.6%	8.9%	63.5%	81.00
16	Dainese Alberto	0.5%	1.1%	0.4%	6.6%	44.1%	41.00
17	Miquel Delgado Pau	0.5%	1.1%	0.4%	4.3%	50.2%	101.00
18	Renard Alexis	0.4%	1.1%	0.3%	4.2%	49.5%	151.00
19	Canal Carlos	0.4%	0.4%	0.4%	3.0%	21.6%	401.00
20	Brustenga Marc	0.3%	0.5%	0.3%	6.9%	55.5%	126.00
21	Hayter Ethan	0.3%	1.5%	0.0%	3.6%	14.7%	151.00
22	Debruyne Ramses	0.3%	0.3%	0.3%	3.8%	20.4%	201.00
23	Mulubrhan Henok	0.2%	0.3%	0.3%	3.2%	20.6%	151.00
24	Macias Cesar	0.2%	0.2%	0.2%	2.7%	46.5%	201.00
25	Camprubí Marcel	0.2%	0.2%	0.2%	3.2%	31.8%	201.00
26	Oliveira Ivo	0.2%	0.2%	0.1%	3.5%	25.6%	81.00
27	Brenner Marco	0.1%	0.3%	0.1%	3.5%	15.9%	n/a
28	Covi Alessandro	0.1%	0.1%	0.1%	1.5%	12.3%	201.00
29	Fisher-Black Finn	0.1%	0.2%	0.1%	2.7%	14.8%	151.00
30	Küng Stefan	0.1%	0.2%	0.1%	1.0%	10.8%	101.00

After the stage

Results — won by Brennan Matthew

Pos	Rider	Team	Bunch sprint win p	Wind thins it win p	Break survives win p	Hindsight (post-hoc) win p
1	Brennan Matthew	Team Visma \| Lease a Bike	50.9%	49.3%	0.4%	56.0%
2	Coquard Bryan	Cofidis	2.0%	2.7%	0.0%	1.7%
3	Braet Vito	Lotto-Intermarché	0.6%	0.9%	0.0%	0.6%
4	Meeus Jordi	Red Bull-BORA-hansgrohe	9.0%	12.2%	0.0%	9.9%
5	Cort Magnus	Uno-X Mobility	0.5%	1.1%	0.2%	0.7%
6	van Aert Wout	Team Visma \| Lease a Bike	6.4%	13.7%	1.8%	10.0%
7	Hofstetter Hugo	NSN Cycling Team	1.0%	1.2%	0.0%	0.9%
8	Albanese Vincenzo	EF Education-EasyPost	0.4%	1.0%	0.2%	0.6%
9	Dainese Alberto	Soudal Quick-Step	1.7%	0.8%	0.0%	0.5%
10	Macias Cesar	Burgos Burpellet BH	0.5%	0.2%	0.0%	0.2%

Scenario scoring — probability given to the actual winner

Scenario	Blend win p	Book devig
Bunch sprint	50.89%	36.20%
Wind thins it	49.34%	36.20%
Break survives	0.36%	36.20%
Hindsight (post-hoc)	56.03%	36.20%

Reflection

What we make of it

The road chose the official branch cleanly. Matthew Brennan won from Bryan Coquard and Vito Braet, with 102 riders — 68% of the surviving field — credited on the winner's time. The oracle therefore labelled the day bunch. Our official key of bunch 0.55 / res_3 0.30 / res_1 0.15 was exact and covered with 0.55 on the truth; the independent agent key was also exact, with 0.60 on bunch. This was not a boundary call like stage 2: the group was five times the menu's 20% bunch threshold.

The main board also got the winner right. Brennan was first on the published bunch board at 62.8% on the book-preserving head, versus 36.2% after removing the market margin; the wind-thins-it board still had him first at 52.4%. Meeus was second on both boards and finished fourth, while Coquard rose from seventh on the bunch board to fourth on the wind board and finished second. The model's strongest disagreement with the market — that Brennan was even more dominant than 1.85 implied — went the right way.

What did not happen matters for the carry-forward. The early hills, depleted trains and estuary wind neither reduced the winning group nor opened the door to the escape. Stage 15's lesson remains specific: Visma could choose van Aert when a late climb supplied a launchpad. Here ninety flat kilometres after the last rise left the peloton time and road to organise, and Visma chose Brennan. The chase-incentive test works when paired with a real launch point; team optionality alone is not enough to move a flat day out of the sprint family.

The classification did not move: Mas still leads Gall by 2:06 and Roglic by 2:10, with Carapaz at 4:24. Timo Roosen was the only new abandon, leaving 149 riders. Stage 17 is flatter again — Dos Hermanas to Sevilla, 185 km, no meaningful climbs — so bunch is the mechanical macro read. The live risk is weather rather than terrain: extreme heat could shorten the course and strong winds could make the approach nervous, but neither is a reason to hedge the official day key away from bunch before there is an announced route change.