Likely Winner of the Tour, 2016

[SeriousSam]

This thread is for discussing how the day's stage affects who we think is going to win the Tour.

Some people are analysing this carefully, putting considerable effort into sifting through all the available data, looking at historical records, estimating riders' power curves, building predictive models. Others have very sharp intuitions and instincts. Both types are likely to want to profit from their superior predictive ability, putting money where their mouth is. The forces of competition cause the market odds to reflect their beliefs, giving us a continuously updated estimate of what's going to happen, with a good empirical track record. A natural starting point for discussing who's going to win as the Tour unfolds.

So here are the odds over the past few days, taken from bet365 and betfair.

Here's the picture thus far, with the blue line indicating the start of he Tour. Not much has happened yet. A Froome v Quintana duel looked to be the most likely outcome before the grand depart, and that remains the case. The most significant GC development so far was that Contador crashed twice, which has already caused him to lose sufficient time to take him out of contention even if he were to recover completely. Though unfortunate, the damage to the intrigue of the GC competition is limited because, as the graph shows, he was unlikely to be a key protagonist to begin with.

Porte lost a lot of time as well and Pinot was gaped. It is thus Aru, freed today from the specter of Nibali challenging his leadership, who has replaced Contador as the distant third favourite.

 

[jmdirt]

This thread is for discussing how the day's stage affects who we think is going to win the Tour.

Some people are analysing this carefully, putting considerable effort into sifting through all the available data, looking at historical records, estimating riders' power curves, building predictive models. Others have very sharp intuitions and instincts. Both types are likely to want to profit from their superior predictive ability, putting money where their mouth is. The forces of competition cause the market odds to reflect their beliefs, giving us a continuously updated estimate of what's going to happen, with a good empirical track record. A natural starting point for discussing who's going to win as the Tour unfolds.

So here are the odds over the past few days, taken from bet365 and betfair.

Here's the picture thus far, with the blue line indicating the start of he Tour. Not much has happened yet. A Froome v Quintana duel looked to be the most likely outcome before the grand depart, and that remains the case. The most significant GC development so far was that Contador crashed twice, which has already caused him to lose sufficient time to take him out of contention even if he were to recover completely. Though unfortunate, the damage to the intrigue of the GC competition is limited because, as the graph shows, he was unlikely to be a key protagonist to begin with.

Porte lost a lot of time as well and Pinot was gaped. It is thus Aru, freed today from the specter of Nibali challenging his leadership, who has replaced Contador as the distant third favourite.

Sam, aren't there enough "who's going to win" threads for the TdF already (plus the discussion in each stage thread about what the days events means to Paris)?

 

[Netserk]

Given that Brexit had already happened (Everyone was done voting, but they weren't counted yet) when the quote in my signature was posted, I think it's fair to have some skepticism when it comes to the bias of the market.

I think it's quite obvious that before the Tour started, Contador had more than three times as big a chance to win as Porte. Duh.

Edit: By the way, SeriousSam, isn't the literature quite limited regarding the betting market's predicative power?

 

[burning]

purito with odd 34.00 to win pais vasco

Here, you can make your money:

Sergio Henao 2
Nairo Quintana 2
Simon Spilak 7/2
Michal Kwiatkowski 4
Ilnur Zakarin 14
Tejay Van Garderen 25
Simon Yates 25
Ion Izagirre 25
Joaquim Rodriguez 33
Thibaut Pinot 66

Purito:

“La contrarreloj no es lo mío, pero voy a salir a hacerla a tope para tratar de estar por lo menos en el podio. Esta crono es muy complicada, de hecho a estas alturas todavía no hemos decidido si saldremos con la cabra o con la bici normal porque hay un tramo de bajada importante y luego dos subidas muy explosivas. El recorrido es muy exigente y ya llevamos encima una buena paliza". “Las diferencias no son nada”

That was before final TT in Pais Vasco 2015

 

[SeriousSam]

jmdirt:
Uncertainty quantification using the odds has proved polarising in this forum, with some reacting with almost allergic hostility. In the interest of forum harmony, I thought it's best to post the graph in a thread that can be avoided, though the mods are free to move things into one of those threads if they think it's preferable.

Netserk:
A single data point doesn't provide much evidence regarding the predictive ability of some mechanism. But for those unaware of it, I suppose the Brexit failure, or the failure to predict Contador's sequence of crashes at this Tour, is a useful reminder that infallible predictions of the future are not currently available.

Regarding the literature, depends on what you mean. Cycling specific? I'm not aware of any study. In general? The efficient market hypothesis has spawned an absolutely enormous amount of theoretical and empirical work.

burning:
A single data point doesn't provide much evidence regarding the predictive ability of some mechanism.

 

[Netserk]

I'm talking about empirical studies specifically regarding the *betting* markets' predicative power (in general, not just sport). From what I understand, there isn't enough evidence to support that the betting market is more accurate than polls for example (when it comes to elections). The market is great in general, of course, but it's pretty obvious that the betting market is biased (otherwise it would be impossible for *anybody* to make a profit from betting in the long run), it's not like there's a clear correlation between market share and insight for the actors.

 

[Cookster15]

Save us Nairo Quintana you are our only hope.

 

[PremierAndrew]

The betting market is influenced heavily by the number of bets received by a bookmaker, and doesn't necessarily reflect the likelihood of the outcome being true.
For example, some Swiss bookmakers had Federer as favourite once Djokovic crashed out of this year's Wimbledon, even though universally, Murray was the favourite

 

[kwikki]

Barring accident Froome or Quintana. Quintana will need to make a move very soon

 

[Cookster15]

Re:

I'm talking about empirical studies specifically regarding the *betting* markets' predicative power (in general, not just sport). From what I understand, there isn't enough evidence to support that the betting market is more accurate than polls for example (when it comes to elections). The market is great in general, of course, but it's pretty obvious that the betting market is biased (otherwise it would be impossible for *anybody* to make a profit from betting in the long run), it's not like there's a clear correlation between market share and insight for the actors.

You are right. We just had an election in Australia where poll after poll predicted a too close to call result and yet betting markets consistently had the favourite at 6:1 on - suggesting a comfortable win. The election was last Saturday and we still haven't declared the winner 8 days later. In this case polls were a better indicator than the betting markets of the likely outcome.

 

[SeriousSam]

Deleted a few irrelevant riders so they don't clutter up the graph.

 

[Warhawk]

Re:

Deleted a few irrelevant riders so they don't clutter up the graph.

Have to admit it seems a tad bizarre that you can use "irrelevant rider" in reference to Alberto Contador and I can't really come up with an argument to the contrary.

 

[DFA123]

Re:

Deleted a few irrelevant riders so they don't clutter up the graph.

I think you could delete two more to be honest.

 

[Samson777]

Re: Re:

Deleted a few irrelevant riders so they don't clutter up the graph.

I think you could delete two more to be honest.

This

 

[SeriousSam]

Keeping them for now due to Aru's Lazarus-like powers of recovery and Porte's mercurial talents.

I confess to being surprised that the market has not reacted in Quintana's favour, when it was seemingly expected that Froome would gain time which he didn't.

Could this be one of those legendarily rare cases when the market is not completely accurate?

 

[Valv.Piti]

I've been surprised by the fact that Froome stands much stronger post Pyrenees than pre Pyrenees. So somewhat yes, to your last question, but that should maybe also be contributed to Aru, Contador abandon and no real contender apart from Quintana emerging in the Pyrenees.

But you wonder if he is too low right now. 1.65 at bet365 is the opposite of value IMO.

 

[SeriousSam]

Re:

I'm talking about empirical studies specifically regarding the *betting* markets' predicative power (in general, not just sport). From what I understand, there isn't enough evidence to support that the betting market is more accurate than polls for example (when it comes to elections).

Polls estimate a population proportion, namely who people would vote for if the election were today. Even if we knew that proportion (ie no sampling variability or bias in polls), we wouldn't know the outcome of an election because people don't always tell the truth, change their minds, or won't vote. However, we could use historical data on how that popuation proportion maps into election outcomes at various times-to-election and construct a model that spits out a probability that one party wins. Academics, and Nate Silver, have done that (they also take into account that polls are estimates, subject to variance and bias), and it is those numbers that should be compared to prediction market probabilities, not polls directly.

However, if you insist on comparing estimated vote share with chance to win, this has been done, and markets
obviously perform much better, even for polls immediately before the election, but especially way before the election (when neither bias nor variance, but the discrepancy between what people say they'd do and what they will do is the dominant source of error), as found in Berg et al (2008), Prediction market accuracy in the long run, International Journal of Forecasting.

Not a particularly surprising or interesting result in my view, but there you go.

The market is great in general, of course, but it's pretty obvious that the betting market is biased (otherwise it would be impossible for *anybody* to make a profit from betting in the long run), it's not like there's a clear correlation between market share and insight for the actors.

Sure, it's possible to outperform the market in the long run. Dave Brailsford, for instance, could certainly outperform the Tour market, thanks to his information advantage over the public. That doesn't require the market to be biased (here's an example: someoone throws a die. my information is whether it's in [1,2,3] or [4,5,6]. Dave's information on it is whether the number is [1,2], [3,4] or [5,6]. Even if I deal perfectly with the information I have, with no bias (ie I'm correct on average), Dave will outperform me thanks to his superior information about the outcome)

Can the market be outperformed with public information only? That's much more difficult but I'm sure there are people out there that can do it. This does imply a bias.

 

[Red Rick]

Re:

I've been surprised by the fact that Froome stands much stronger post Pyrenees than pre Pyrenees. So somewhat yes, to your last question, but that should maybe also be contributed to Aru, Contador abandon and no real contender apart from Quintana emerging in the Pyrenees.

But you wonder if he is too low right now. 1.65 at bet365 is the opposite of value IMO.

I'd lay that, and its not thin imo

 

[Inquitus]

Re: Re:

I'm talking about empirical studies specifically regarding the *betting* markets' predicative power (in general, not just sport). From what I understand, there isn't enough evidence to support that the betting market is more accurate than polls for example (when it comes to elections).

Polls estimate a population proportion, namely who people would vote for if the election were today. Even if we knew that proportion (ie no sampling variability or bias in polls), we wouldn't know the outcome of an election because people don't always tell the truth, change their minds, or won't vote. However, we could use historical data on how that popuation proportion maps into election outcomes at various times-to-election and construct a model that spits out a probability that one party wins. Academics, and Nate Silver, have done that (they also take into account that polls are estimates, subject to variance and bias), and it is those numbers that should be compared to prediction market probabilities, not polls directly.

However, if you insist on comparing estimated vote share with chance to win, this has been done, and markets
obviously perform much better, even for polls immediately before the election, but especially way before the election (when neither bias nor variance, but the discrepancy between what people say they'd do and what they will do is the dominant source of error), as found in Berg et al (2008), Prediction market accuracy in the long run, International Journal of Forecasting.

Not a particularly surprising or interesting result in my view, but there you go.

The market is great in general, of course, but it's pretty obvious that the betting market is biased (otherwise it would be impossible for *anybody* to make a profit from betting in the long run), it's not like there's a clear correlation between market share and insight for the actors.

Sure, it's possible to outperform the market in the long run. Dave Brailsford, for instance, could certainly outperform the Tour market, thanks to his information advantage over the public. That doesn't require the market to be biased (here's an example: someoone throws a die. my information is whether it's in [1,2,3] or [4,5,6]. Dave's information on it is whether the number is [1,2], [3,4] or [5,6]. Even if I deal perfectly with the information I have, with no bias (ie I'm correct on average), Dave will outperform me thanks to his superior information about the outcome)

Can the market be outperformed with public information only? That's much more difficult but I'm sure there are people out there that can do it. This does imply a bias.

The market is based on total bets placed, a mix of uninformed, informed and lets say super informed. You can certainly outperform the market as you only need to be somewhere inbetween informed and super informed to come out on the right side of the line. If we take today for instance, uninformed people might still have been willing to bet on Contador, but the super informed would have known about his fever before either the uninformed or the informed (who found out on twitter or eurosport or whatever)

 

[Red Rick]

^bingo

Meanwhile, I'm waiting for the odds guru to come in and share his wisdom

 

[Valv.Piti]

Re:

^bingo

Meanwhile, I'm waiting for the odds guru to come in and share his wisdom

I think you can do it yourself if you so choose to, its available out here. But it shouldn't come as a surprise Froome's chances has risen once again (they did so after Peyresourde, Arcalis and Montpellier, this time the most) and he now is favoured more than he ever has been in this Tour. And rightly so, IMO.

 

[SeriousSam]

 

[Oliwright]

Quintana has shown no Spine in this tour, no attacks, didn't want to work to follow froome on descent and then bitches about the stage after he looses time.

I can't see him winning right now, even though i thought he would at the start of the race.

 

[Valv.Piti]

Quintana has shown no Spine in this tour, no attacks, didn't want to work to follow froome on descent and then bitches about the stage after he looses time.

I can't see him winning right now, even though i thought he would at the start of the race.

In an assessment of who is the most likely winner of the Tour, your dislike for the diminutive Colombian has barely any relevance for the outcome

 

[Gigs_98]

I really don't get those odds. They were basically even, before the mountains and only because of a descent and crosswinds where Froome gained combined a little bit more than 30 seconds he is now suddenly way more likely to win? If anything I think that Quintana's chances got bigger since he didn't loose any time in the Pyrenees like in previous years. Of course the time loss is aggravating for Quintana but I still doubt that the tour will be decided by 30 seconds.