As with the TDF I decided to test whether the stats predicted the outcome. Off the bat I would expect it to be less true here because the pool of riders in contention at the ToA is effectively much deeper since as a C1 race it is not selecting any riders from the very top of the curve.

And indeed that is the case. I ended up looking at 4 ways of ranking riders.

The stats below are ranks based on

OVL
TDF - matching what I used in that post - MO + TT + RS+ RC
TDF Mod - drop TT add HI based on observing the outcomes
Custom - my blend where I sought to explain these results, it is MT/HI/RC, with MT weighted heaviest
Ave - Average of the four

I looked at the top 20 finishers plus Kuss (for personal reasons) and minus Girdlestone. I think this is a pretty good representation of the top contenders. There are a few others like Koch and Pantano Gomez who's stats put them in this frame but they were both clearly working for top contenders and would be expected to underperform. I also excluded a couple of all-rounders who made the preview, like Roe and Novak, because I think TT just didn't prove to be that important in this parcours and without that stat those two are not competitive.

	Current GC	OVL Rank	TDF	TDF Mod	Custom	Ave Rank	Diff From Ave
Nerz	1	6	5	12	6	7.3	6.3
Warchol	2	2	8	3	5	4.5	2.5
Carapaz	3	5	1	3	3	3.0	0.0
Potocki	4	19	12	17	15	15.8	11.8
Aru	5	12	12	5	10	9.8	4.8
Novak	6	1	6	1	1	2.3	-3.8
Min	7	4	14	6	8	8.0	1.0
Denifl	8	10	19	21	2	13.0	5.0
Chamorro	9	7	16	11	14	12.0	3.0
Ranaweera	10	3	2	6	4	3.8	-6.3
Lunke	11	13	2	6	9	7.5	-3.5
Garby	12	14	6	9	11	10.0	-2.0
Eg	13	17	4	15	17	13.3	0.3
Gilanipoor	14	20	16	18	21	18.8	4.8
Grmay	15	15	15	9	13	13.0	-2.0
Elissonde	16	11	8	2	7	7.0	-9.0
Azman	17	8	21	12	20	15.3	-1.8
Diaz	18	18	11	18	19	16.5	-1.5
Warbasse	19	9	8	12	12	10.3	-8.8
Hoem	20	21	19	18	18	19.0	-1.0
Kuss	22	16	16	16	16	16.0	-6.0
	Correl	0.554	0.393	0.416	0.643	0.605

First none of the correlations are that great. Not surprisingly the custom model is the best although the average does OK. So there is clearly some randomization in the results (or some stat influence I have missed). However, I think that has to be taken in context of how close the riders are. If you look at the stats there are roughly two clustered groups, and most of the top cluster finished in the top and vice versa. So in that sense riders probably finished in the set of riders they should have although their positions within that set were hard to explain.

Two of the biggest underperformers, Warbasse and Elissonde were part of leadership duo's, in the latter case maybe Elissonde's support help explain Potocki's over performance. In the former both the AI's choice of leader and the value of using the other in support seem questionable.

Ranaweera and Kuss were underperformers (Kuss apparently should have finished 16th). Nerz and Aru clear over performers.

Denifil and Novak both have one outlier stat (low RS and TT, respectively). Neither of those values are in the custom blend because overall they don't seem to have much explanatory value. But I think in the case of outlier values, they matter, so these two riders have significant variation from their predictions which all either under or over weight these two stats.

Most of the rest are within a few places of their average prediction.

One thought I have is that the fairly large daily form may interact with recovery and explain why certain riders do consistently well or poorly. A rider who has a bad form uses more energy to stay in the race and then starts worse the next day. And if you are poor recovery that would be worse, so some early bad days snowball for a poor recovery rider like Kuss while the opposites happens for a better recovery rider like Potocki.

In response to BBl's comment: I think the limited influence of TT was not surprising. There weren't that many TT KMs, the 2nd one came very late when recovery would have been a major factor and in contrast there was a lot of selective climbing. I think the ToA has always been a climbers race, I wanted it to have a little more TT this year. In retrospect I wish I had put the second TT earlier and made it a little longer. But this was never going to be a TT dominant race.

I can't explain who chases. Why was McCormick chasing, we were even less suited than Carlsberg. Generally I think the chasing is weird. It seems to me the leader's team should be the default chaser but instead they only chase when the lead is directly threatened and then some random team gets stuck chasing the rest of the time.