Cycling race predictor · Tour de France 2026

Stage 3 — 2026-07-06

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 184 riders priced models retrained 2026-06-28 book snapshot markets_1212.json

← System overview, season boards & the full Tour 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.

Healy/Carapaz days

Breakaway-climber days — stages Healy & Carapaz won

sprint0.00

mountain0.45

hilly0.00

rolling0.54

break0.01

TT0.00

Real mountain-breakaway wins: Carapaz TdF'24 s17 & Giro'25 s11 & Vuelta'22 s20, Healy TdF'25 s6 & Giro'23 s8. They average rolling ~0.55 / mountain ~0.44 — the model-space signature of a break day in the hills.

Top 5 if the break wins (top 20 GC removed): Riccitello Matthew 6.0%; Uijtdebroeks Cian 5.5%; Van Wilder Ilan 4.6%; van der Poel Mathieu 4.0%; Kuss Sepp 3.6%

Mountain break mix

Synthetic mountain-break mix (mountain 40 / rolling 35 / break 25)

sprint0.00

mountain0.40

hilly0.00

rolling0.35

break0.25

TT0.00

Terrain scores overridden on the Healy/Carapaz reference average — heavier mountain weight, explicit breakaway share.

Top 5 (blend): Pogacar Tadej 32.1%; Seixas Paul 10.9%; Evenepoel Remco 6.7%; Vingegaard Jonas 4.7%; del Toro Isaac 4.6%

GC mountain day

TdF summit-finish archetype — GC contest counterfactual

sprint0.00

mountain0.53

hilly0.00

rolling0.47

break0.01

TT0.00

Real TdF summit finishes 2021-25 (the artifact-run mountain archetype). The read where the GC teams race for the stage and the break is caught.

Top 5 (blend): Pogacar Tadej 42.0%; Evenepoel Remco 12.8%; Vingegaard Jonas 11.0%; Seixas Paul 8.3%; Lipowitz Florian 5.5%

Hindsight (post-hoc)

Hindsight (post-hoc) — stages that finished like this one

sprint0.00

mountain0.37

hilly0.00

rolling0.61

break0.02

TT0.00

Post-race reconstruction, NOT a prediction (never logged). The five DB stages whose finish structure most resembled stage 3's shattered GC group (1/10/19/27 riders within 0/5/30/60s): Evenepoel Vuelta'23 s3, Roglic Tirreno'23 s6, O'Connor UAE'24 s3, Pogacar UAE'22 s4, Ayuso Giro'25 s7 — all uphill GC-selection finishes. Answers: if we had known it would be a hard summit selection, what would the model have said?

Top 5 (blend): Pogacar Tadej 38.2%; Evenepoel Remco 10.4%; Vingegaard Jonas 9.3%; Seixas Paul 6.8%; Lipowitz Florian 6.8%

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.

Rider	Team	Healy/Carapaz days	Mountain break mix	GC mountain day	Hindsight (post-hoc)	Book odds	Book devig	Max edge
Pogacar Tadej	UAE Team Emirates-XRG	–	32.1%	42.0%	38.2%	2.75	23.9%	+18.1pp
Evenepoel Remco	Red Bull-BORA-hansgrohe	–	6.7%	12.8%	10.4%	26.00	2.5%	+10.3pp
Vingegaard Jonas	Team Visma \| Lease a Bike	–	4.7%	11.0%	9.3%	26.00	2.5%	+8.5pp
Seixas Paul	Decathlon CMA CGM Team	–	10.9%	8.3%	6.8%	34.00	1.9%	+8.9pp
Lipowitz Florian	Red Bull-BORA-hansgrohe	–	2.6%	5.5%	6.8%	201.00	0.3%	+6.4pp
del Toro Isaac	UAE Team Emirates-XRG	–	4.6%	5.2%	4.6%	19.00	3.5%	+1.7pp
Ayuso Juan	Lidl-Trek	–	2.4%	2.1%	2.3%	201.00	0.3%	+2.1pp
Martinez Lenny	Bahrain Victorious	–	1.8%	1.9%	1.7%	51.00	1.3%	+0.7pp
Yates Adam	UAE Team Emirates-XRG	–	0.8%	1.0%	1.6%	301.00	0.2%	+1.4pp
Johannessen Tobias Halland	Uno-X Mobility	–	1.4%	1.4%	1.5%	41.00	1.6%	-0.1pp
Carapaz Richard	EF Education-EasyPost	–	0.9%	1.5%	1.4%	67.00	1.0%	+0.5pp
Hindley Jai	Red Bull-BORA-hansgrohe	–	0.9%	1.5%	1.1%	601.00	0.1%	+1.4pp
Pidcock Tom	Pinarello-Q36.5 Pro Cycling Team	–	1.5%	1.2%	1.1%	34.00	1.9%	-0.5pp
Gee Derek	Lidl-Trek	–	1.0%	1.4%	1.0%	126.00	0.5%	+0.9pp
Arensman Thymen	Netcompany INEOS	–	1.1%	1.2%	1.2%	151.00	0.4%	+0.8pp
Skjelmose Mattias	Lidl-Trek	–	1.0%	1.0%	1.2%	41.00	1.6%	-0.4pp
Jorgenson Matteo	Team Visma \| Lease a Bike	–	1.0%	1.2%	1.0%	201.00	0.3%	+0.9pp
Riccitello Matthew	Decathlon CMA CGM Team	6.0%	0.9%	1.1%	0.9%	501.00	0.1%	+1.0pp
Vauquelin Kévin	Netcompany INEOS	–	1.0%	0.7%	0.6%	29.00	2.3%	-1.3pp
Grégoire Romain	Groupama-FDJ United	–	0.9%	0.8%	0.9%	19.00	3.5%	-2.6pp
van der Poel Mathieu	Alpecin-Premier Tech	4.0%	0.9%	0.6%	0.6%	13.00	5.0%	-4.2pp
Healy Ben	EF Education-EasyPost	2.0%	0.3%	0.4%	0.3%	23.00	2.9%	-2.4pp
Vacek Mathias	Lidl-Trek	1.6%	0.3%	0.1%	0.1%	15.00	4.4%	-4.1pp
Simmons Quinn	Lidl-Trek	0.6%	0.1%	0.0%	0.0%	26.00	2.5%	-2.5pp

Healy/Carapaz days is shown conditional on the break winning — the top 20 GC contenders 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.

Breakaway-climber days — stages Healy & Carapaz won — top 30, all heads · top 20 GC removed

#	Rider	Blend	Cascade	LambdaRank	Podium	Top 15	Book odds
1	Riccitello Matthew	1.1%	0.8%	1.1%	4.6%	41.9%	501.00
2	Uijtdebroeks Cian	1.0%	0.9%	1.0%	4.6%	43.8%	501.00
3	Van Wilder Ilan	0.8%	1.0%	0.8%	3.3%	48.9%	251.00
4	van der Poel Mathieu	0.7%	1.3%	0.6%	6.5%	17.6%	13.00
5	Kuss Sepp	0.6%	0.5%	0.7%	3.4%	15.7%	n/a
6	Bernal Egan	0.6%	0.4%	0.7%	2.9%	28.3%	201.00
7	Foss Tobias	0.6%	1.4%	0.4%	3.2%	27.5%	151.00
8	McNulty Brandon	0.6%	0.7%	0.5%	2.1%	22.5%	51.00
9	Higuita Sergio	0.5%	0.5%	0.5%	2.5%	30.6%	301.00
10	Izagirre Ion	0.5%	0.4%	0.5%	1.7%	20.2%	201.00
11	Storer Michael	0.4%	1.1%	0.3%	4.2%	30.0%	67.00
12	Aranburu Alex	0.4%	0.4%	0.4%	1.0%	8.9%	81.00
13	Schmid Mauro	0.4%	0.6%	0.3%	2.0%	10.9%	51.00
14	Jegat Jordan	0.4%	0.3%	0.4%	2.4%	32.9%	151.00
15	Healy Ben	0.4%	0.5%	0.3%	1.2%	8.1%	23.00
16	Van Gils Maxim	0.3%	0.6%	0.3%	1.4%	5.6%	34.00
17	Van Eetvelt Lennert	0.3%	0.4%	0.3%	2.1%	25.7%	67.00
18	Baudin Alex	0.3%	0.2%	0.4%	1.5%	36.1%	41.00
19	Berckmoes Jenno	0.3%	0.5%	0.3%	1.6%	12.2%	151.00
20	Castrillo Pablo	0.3%	0.2%	0.3%	1.4%	8.2%	201.00
21	Tejada Harold	0.3%	0.6%	0.2%	1.5%	14.7%	101.00
22	Harper Chris	0.3%	0.2%	0.3%	0.9%	9.8%	401.00
23	O'Connor Ben	0.3%	0.3%	0.3%	1.7%	12.9%	n/a
24	Vacek Mathias	0.3%	0.3%	0.3%	2.1%	10.0%	15.00
25	Plapp Lucas	0.3%	0.2%	0.3%	2.5%	12.3%	n/a
26	Pedersen Mads	0.3%	0.3%	0.3%	1.0%	2.3%	126.00
27	Paret-Peintre Valentin	0.2%	0.2%	0.3%	1.3%	7.8%	34.00
28	Benoot Tiesj	0.2%	0.2%	0.2%	0.8%	8.2%	401.00
29	Martin Guillaume	0.2%	0.1%	0.2%	0.8%	11.3%	301.00
30	Steinhauser Georg	0.2%	0.4%	0.1%	1.1%	8.0%	151.00

Synthetic mountain-break mix (mountain 40 / rolling 35 / break 25) — top 30, all heads

#	Rider	Blend	Cascade	LambdaRank	Podium	Top 15	Book odds
1	Pogacar Tadej	32.1%	38.9%	32.4%	63.4%	81.3%	2.75
2	Seixas Paul	10.9%	9.2%	12.0%	17.1%	55.3%	34.00
3	Evenepoel Remco	6.7%	3.6%	7.9%	13.0%	44.7%	26.00
4	Vingegaard Jonas	4.7%	3.4%	5.3%	17.3%	58.4%	26.00
5	del Toro Isaac	4.6%	0.9%	5.8%	6.9%	33.2%	19.00
6	Lipowitz Florian	2.6%	1.1%	3.2%	4.0%	37.4%	201.00
7	Ayuso Juan	2.4%	0.9%	3.0%	4.4%	32.6%	201.00
8	Martinez Lenny	1.8%	1.2%	2.1%	4.1%	33.4%	51.00
9	Pidcock Tom	1.5%	1.0%	1.7%	4.2%	39.2%	34.00
10	Johannessen Tobias Halland	1.4%	1.1%	1.5%	6.7%	59.1%	41.00
11	Arensman Thymen	1.1%	1.2%	1.1%	3.8%	33.7%	151.00
12	Skjelmose Mattias	1.0%	0.8%	1.2%	3.7%	31.0%	41.00
13	Jorgenson Matteo	1.0%	2.2%	0.8%	4.2%	20.4%	201.00
14	Gee Derek	1.0%	0.8%	1.1%	5.0%	26.9%	126.00
15	Vauquelin Kévin	1.0%	1.1%	1.0%	2.7%	17.2%	29.00
16	Riccitello Matthew	0.9%	0.7%	1.0%	4.3%	37.5%	501.00
17	Grégoire Romain	0.9%	1.1%	0.9%	3.8%	35.2%	19.00
18	Hindley Jai	0.9%	0.6%	1.0%	4.4%	28.7%	601.00
19	Carapaz Richard	0.9%	0.7%	1.0%	3.2%	39.2%	67.00
20	van der Poel Mathieu	0.9%	1.4%	0.8%	7.4%	20.5%	13.00
21	Yates Adam	0.8%	0.3%	1.0%	1.6%	27.2%	301.00
22	Van Wilder Ilan	0.7%	1.2%	0.7%	3.5%	47.1%	251.00
23	Piganzoli Davide	0.7%	0.6%	0.8%	3.6%	41.2%	501.00
24	Uijtdebroeks Cian	0.7%	0.6%	0.7%	3.0%	35.1%	501.00
25	Bernal Egan	0.6%	0.6%	0.6%	2.9%	29.9%	201.00
26	Kuss Sepp	0.5%	0.8%	0.5%	2.8%	15.6%	n/a
27	Foss Tobias	0.5%	1.7%	0.2%	3.1%	23.3%	151.00
28	Higuita Sergio	0.5%	0.6%	0.5%	2.7%	30.7%	301.00
29	McNulty Brandon	0.4%	0.8%	0.4%	1.8%	18.0%	51.00
30	Storer Michael	0.4%	1.0%	0.3%	3.5%	24.7%	67.00

TdF summit-finish archetype — GC contest counterfactual — top 30, all heads

#	Rider	Blend	Cascade	LambdaRank	Podium	Top 15	Book odds
1	Pogacar Tadej	42.0%	45.0%	35.2%	74.3%	88.5%	2.75
2	Evenepoel Remco	12.8%	11.4%	11.4%	36.3%	75.7%	26.00
3	Vingegaard Jonas	11.0%	8.5%	10.1%	43.3%	83.6%	26.00
4	Seixas Paul	8.3%	8.0%	7.2%	14.8%	62.9%	34.00
5	Lipowitz Florian	5.5%	3.2%	5.3%	14.3%	70.6%	201.00
6	del Toro Isaac	5.2%	2.0%	5.2%	10.4%	57.4%	19.00
7	Ayuso Juan	2.1%	1.7%	1.9%	7.1%	49.6%	201.00
8	Martinez Lenny	1.9%	2.2%	1.6%	6.5%	49.3%	51.00
9	Carapaz Richard	1.5%	1.4%	1.3%	5.3%	52.7%	67.00
10	Hindley Jai	1.5%	1.4%	1.3%	5.1%	46.4%	601.00
11	Johannessen Tobias Halland	1.4%	1.0%	1.3%	8.6%	64.0%	41.00
12	Gee Derek	1.4%	1.3%	1.2%	9.0%	35.2%	126.00
13	Pidcock Tom	1.2%	1.3%	1.0%	5.2%	49.6%	34.00
14	Arensman Thymen	1.2%	1.0%	1.1%	4.2%	43.3%	151.00
15	Jorgenson Matteo	1.2%	2.1%	0.8%	5.8%	21.5%	201.00
16	Riccitello Matthew	1.1%	1.2%	1.0%	5.8%	44.4%	501.00
17	Yates Adam	1.0%	0.9%	0.9%	4.4%	34.5%	301.00
18	Skjelmose Mattias	1.0%	1.2%	0.8%	4.8%	40.3%	41.00
19	Bernal Egan	0.8%	0.6%	0.8%	4.2%	33.4%	201.00
20	Uijtdebroeks Cian	0.8%	0.7%	0.7%	3.5%	51.5%	501.00
21	Grégoire Romain	0.8%	1.4%	0.5%	7.5%	58.5%	19.00
22	Kuss Sepp	0.7%	0.7%	0.6%	4.4%	22.7%	n/a
23	Piganzoli Davide	0.7%	0.4%	0.6%	5.2%	52.2%	501.00
24	Vauquelin Kévin	0.7%	0.8%	0.5%	3.7%	24.4%	29.00
25	Van Wilder Ilan	0.6%	0.9%	0.5%	3.5%	54.1%	251.00
26	van der Poel Mathieu	0.6%	2.0%	0.1%	8.6%	19.6%	13.00
27	Higuita Sergio	0.5%	0.6%	0.4%	2.8%	31.3%	301.00
28	Foss Tobias	0.4%	0.8%	0.3%	3.1%	29.2%	151.00
29	Healy Ben	0.4%	1.1%	0.2%	2.6%	9.7%	23.00
30	McNulty Brandon	0.4%	0.5%	0.3%	2.0%	18.2%	51.00

Hindsight (post-hoc) — stages that finished like this one — top 30, all heads

#	Rider	Blend	Cascade	LambdaRank	Podium	Top 15	Book odds
1	Pogacar Tadej	38.2%	44.4%	34.0%	75.3%	88.5%	2.75
2	Evenepoel Remco	10.4%	9.7%	9.9%	32.9%	70.5%	26.00
3	Vingegaard Jonas	9.3%	7.9%	9.1%	32.7%	74.2%	26.00
4	Seixas Paul	6.8%	6.4%	6.4%	14.4%	60.3%	34.00
5	Lipowitz Florian	6.8%	3.1%	7.2%	13.3%	65.1%	201.00
6	del Toro Isaac	4.6%	2.2%	4.9%	12.4%	57.1%	19.00
7	Ayuso Juan	2.3%	1.7%	2.3%	7.6%	47.3%	201.00
8	Martinez Lenny	1.7%	2.0%	1.5%	6.0%	42.7%	51.00
9	Yates Adam	1.6%	1.1%	1.6%	5.7%	30.6%	301.00
10	Johannessen Tobias Halland	1.5%	0.9%	1.6%	3.7%	44.0%	41.00
11	Carapaz Richard	1.4%	1.3%	1.4%	5.5%	50.2%	67.00
12	Skjelmose Mattias	1.2%	1.0%	1.2%	4.3%	33.2%	41.00
13	Arensman Thymen	1.2%	0.8%	1.2%	3.8%	39.0%	151.00
14	Pidcock Tom	1.1%	0.9%	1.1%	4.1%	40.7%	34.00
15	Hindley Jai	1.1%	0.8%	1.1%	4.7%	28.1%	601.00
16	Jorgenson Matteo	1.0%	1.7%	0.8%	4.6%	19.7%	201.00
17	Gee Derek	1.0%	0.7%	1.0%	5.5%	25.7%	126.00
18	Riccitello Matthew	0.9%	0.6%	0.9%	3.5%	38.5%	501.00
19	Grégoire Romain	0.9%	1.6%	0.6%	8.6%	53.4%	19.00
20	Uijtdebroeks Cian	0.8%	0.7%	0.8%	3.4%	46.3%	501.00
21	Van Wilder Ilan	0.8%	1.0%	0.7%	3.5%	50.5%	251.00
22	Piganzoli Davide	0.8%	0.5%	0.8%	5.1%	44.5%	501.00
23	Bernal Egan	0.7%	0.9%	0.7%	4.3%	32.2%	201.00
24	Vauquelin Kévin	0.6%	0.8%	0.6%	2.7%	24.7%	29.00
25	van der Poel Mathieu	0.6%	1.7%	0.3%	8.4%	18.2%	13.00
26	Kuss Sepp	0.5%	0.4%	0.5%	3.6%	17.6%	n/a
27	Higuita Sergio	0.5%	0.6%	0.4%	2.8%	29.6%	301.00
28	McNulty Brandon	0.4%	0.4%	0.4%	1.9%	20.7%	51.00
29	Foss Tobias	0.4%	0.8%	0.3%	2.2%	28.2%	151.00
30	Izagirre Ion	0.4%	0.2%	0.4%	1.6%	19.5%	201.00

Angle

If the break goes — competitive riders outside the GC pool

Top 20 riders by the latest GC-model blend (tdf_gc.json) removed — captains don't get in the break. Remaining riders under the 'Healy/Carapaz days' read; the conditional column renormalizes their win probability over this pool (P(win | a non-GC rider wins), pool holds 18% of the unconditional book). The model can't see who actually makes the break — read this as a fitness-for-the-day ranking, not break-formation odds.

Rider	Team	Healy/Carapaz days win p	P(win \| break wins)	Book odds	Book devig
Riccitello Matthew	Decathlon CMA CGM Team	1.1%	6.0%	501	0.1%
Uijtdebroeks Cian	Movistar Team	1.0%	5.5%	501	0.1%
Van Wilder Ilan	Soudal Quick-Step	0.8%	4.6%	251	0.3%
van der Poel Mathieu	Alpecin-Premier Tech	0.7%	4.0%	13	5.0%
Kuss Sepp	Team Visma \| Lease a Bike	0.6%	3.6%	n/a	–
Bernal Egan	Netcompany INEOS	0.6%	3.4%	201	0.3%
Foss Tobias	Netcompany INEOS	0.6%	3.3%	151	0.4%
McNulty Brandon	UAE Team Emirates-XRG	0.6%	3.1%	51	1.3%
Higuita Sergio	XDS Astana Team	0.5%	2.6%	301	0.2%
Izagirre Ion	Cofidis	0.5%	2.5%	201	0.3%
Storer Michael	Tudor Pro Cycling Team	0.4%	2.4%	67	1.0%
Aranburu Alex	Cofidis	0.4%	2.4%	81	0.8%
Schmid Mauro	Team Jayco-AlUla	0.4%	2.2%	51	1.3%
Jegat Jordan	Team TotalEnergies	0.4%	2.0%	151	0.4%

Angle

The book's card — its favourites under our readings

Shortest book prices for the stage, with our blend win probability (and rank) under each scenario. 'free' = outside our top-20 GC pool, i.e. the type the peloton might let up the road.

Rider	Book odds	Devig	Healy/Carapaz days	Mountain break mix	GC mountain day	Hindsight (post-hoc)	GC status
Pogacar Tadej	2.75	23.9%	29.7% (#1)	32.1% (#1)	42.0% (#1)	38.2% (#1)	GC #1
van der Poel Mathieu	13.00	5.0%	0.7% (#24)	0.9% (#20)	0.6% (#26)	0.6% (#25)	free
Vacek Mathias	15.00	4.4%	0.3% (#44)	0.3% (#37)	0.1% (#59)	0.1% (#51)	free
del Toro Isaac	19.00	3.5%	4.9% (#5)	4.6% (#5)	5.2% (#6)	4.6% (#6)	GC #6
Grégoire Romain	19.00	3.5%	0.9% (#21)	0.9% (#17)	0.8% (#21)	0.9% (#19)	GC #18
Healy Ben	23.00	2.9%	0.4% (#35)	0.3% (#35)	0.4% (#29)	0.3% (#36)	free
Vingegaard Jonas	26.00	2.5%	5.3% (#4)	4.7% (#4)	11.0% (#3)	9.3% (#3)	GC #3
Evenepoel Remco	26.00	2.5%	7.2% (#3)	6.7% (#3)	12.8% (#2)	10.4% (#2)	GC #2
Simmons Quinn	26.00	2.5%	0.1% (#63)	0.1% (#75)	0.0% (#75)	0.0% (#73)	free
Vauquelin Kévin	29.00	2.3%	1.0% (#17)	1.0% (#15)	0.7% (#24)	0.6% (#24)	GC #10

After the stage

Results — won by Pogacar Tadej

Pos	Rider	Team	Healy/Carapaz days win p	Mountain break mix win p	GC mountain day win p	Hindsight (post-hoc) win p
1	Pogacar Tadej	UAE Team Emirates-XRG	29.7%	32.1%	42.0%	38.2%
2	Vingegaard Jonas	Team Visma \| Lease a Bike	5.3%	4.7%	11.0%	9.3%
3	Carapaz Richard	EF Education-EasyPost	1.0%	0.9%	1.5%	1.4%
4	Seixas Paul	Decathlon CMA CGM Team	10.0%	10.9%	8.3%	6.8%
5	Johannessen Tobias Halland	Uno-X Mobility	1.6%	1.4%	1.4%	1.5%
6	Van Eetvelt Lennert	Lotto-Intermarché	0.3%	0.3%	0.4%	0.3%
7	Lipowitz Florian	Red Bull-BORA-hansgrohe	4.5%	2.6%	5.5%	6.8%
8	Evenepoel Remco	Red Bull-BORA-hansgrohe	7.2%	6.7%	12.8%	10.4%
9	del Toro Isaac	UAE Team Emirates-XRG	4.9%	4.6%	5.2%	4.6%
10	Ayuso Juan	Lidl-Trek	1.9%	2.4%	2.1%	2.3%

Scenario scoring — probability given to the actual winner

Scenario	Blend win p	Book devig
Healy/Carapaz days	29.73%	23.87%
Mountain break mix	32.06%	23.87%
GC mountain day	41.96%	23.87%
Hindsight (post-hoc)	38.18%	23.87%

Reflection

What we make of it

The road picked the GC-summit branch, but tighter at the front than any of our reads assumed. Pogacar won by 2 seconds from a ~10-rider lead group -- Vingegaard, Carapaz and Seixas all within 2s, del Toro/Ayuso/Evenepoel/Lipowitz at +4 -- with the field shattered behind (only 10 riders within 5s, 27 within a minute). A hard summit selection settled by an uphill punch, not a long-range solo.

All three pre-race reads out-priced the book on the winner (devig 23.9%): Healy/Carapaz days 29.7%, mountain-break mix 32.1%, and the GC-summit archetype 42.0%. The tdf_mountain read was both the right paradigm and the most confident -- and it topped even the post-hoc hindsight reconstruction (38.2%). On this stage the model was simply more right than the market about the strongest rider, and choosing the summit paradigm was enough to capture it.

Course knowledge was not the binding constraint here -- the winning paradigm was already in the mix and beat hindsight, so there was nothing structural the model failed to see. The one miss was magnitude, not identity: tdf_mountain projected Pogacar ~41s clear where he actually won by 2, so the time-gap head over-estimates how much a week-1 summit fragments an otherwise intact GC group. Carry forward: on these finishes trust the winner-probability read but expect tight front gaps rather than solos. Quiet positive -- the model had Seixas (4th, +2s) correctly in the front group, so its read on the breakout U23 is tracking.