Thesis Final Draft - Breakaways in the Tour de France and Giro d'Italia

[cnighbor1]

''Thesis Final Draft - Breakaways in the Tour de France and Giro d'Italia




Hello pedalheads,


Just in case you're having trouble falling asleep at night, here is my nephew Jeremy's senior thesis (Econ-Harvard) on breakaways in the Tour and Giro. Jeremy is the most ardent cycling fan, especially regarding the Tour de France, of any non-cyclist I've ever met. Peg (my cycle-fan wife) and I began talking with Jeremy about breakaways about 2 years ago and we discussed the various breakaway strategies and hypothesized about the optimal kind of breakaways and number of riders necessary for success. That, plus Jeremy's innate curiosity and math-econ background, launched him into what I'm sure is the most scientific analysis of how and when breakaways succeed in the grand tours. Peg and I were actually quite surprised at the number of successful breakaways when the breakaway group is larger than 6. We had both guessed that breakaways with 7 or more riders might be successful 15-20% of the time. It turned out to be much greater than that! On the other hand breakaways with fewer than 3 riders almost never succeed. Anyway for your geek-like cycling information here is Jeremy's thesis. Peg and I will probably be the only ones to read the whole thing!




Yours in cycling minutiae and all the best - Moss


>
Sent: Thu, Mar 8, 2012 9:31 am
Subject: Thesis Final Draft


Hi Everyone!



I thought you all might like to see a final copy of my thesis. It's long and boring, so don't feel at all obligated to read it! In fact, here's a one-sentence summary of the 49 pages: More breakaway riders is good for the breakaway; more sprinters is bad for the peloton; we can kind of model this.




Glad to have it done! Also, if you see any typos or minor mistakes, don't tell me, since I've already turned it in and can't really do anything about it now.




Hope everyone's doing well!

Jeremy



1 Attached file| 1.1MB








Patashnik.Thesis''
I add thesis in a minute or two

[cnighbor1]

here is thesis
charles

[esldude]

Very interesting. Thanks for sharing.

Have read only a few pages so far. It does strike me there is some statistical noise in an issue he addresses somewhat. That being that race leaders are riding against overall contenders more so than wanting to win any given stage. For future topics he could analyze one day races. There is no prize other than winning that one day. It might give a purer look at what makes for an effective breakaway. I doubt the results would differ too much, but then there are always surprises to be found.

[MattTuck]

I'm going to read this tomorrow, but I'm saving it tonight.

Is your nephew going to be cycling's answer to Baseball's Bill James and Peter Brand?

I can see it now, JVV calls up your nephew... "there's a break up the road, should we chase?"

He should shop it around to see if a studio will option the film rights. ;)

[cachagua]

It is widely held that Islam is the youngest of the world's major religions, but in fact that distinction belongs to Economics.

But -- entertaining!

[tiretrax]

Very interesting. Thanks for sharing.

Have read only a few pages so far. It does strike me there is some statistical noise in an issue he addresses somewhat. That being that race leaders are riding against overall contenders more so than wanting to win any given stage. For future topics he could analyze one day races. There is no prize other than winning that one day. It might give a purer look at what makes for an effective breakaway. I doubt the results would differ too much, but then there are always surprises to be found.

There are lots of successful single breakaways in one day races, such as Tom Boonen, Johan Vansummeran, and Fabian Cancellara the past three years at Paris-Roubaix.

For his Masters, I'd like him to figure out why the Red Sox suck.

[dumbod]

There are lots of successful single breakaways in one day races, such as Tom Boonen, Johan Vansummeran, and Fabian Cancellara the past three years at Paris-Roubaix.

For his Masters, I'd like him to figure out why the Red Sox suck.

I don't think that a Masters is going to do the trick. If ever there were a topic that required multiple PhDs, it would be the suckitude of the Sox

[jeduardo]

I thank you for sharing this. I'm quite interested in learning just how he went about defending his thesis. I hope to read through it this evening. :)
:)

[Joel]

Very interesting work, and I too will sit and read the whole thing!

[laupsi]

wow, thanks for sharing. looks like the perfect night time, look foward to reading, full of interesting tid bits read I've had in a long while. :)

[tannhauser]

I skimmed the thesis; it's pure math and has little recognition of cycling as a sport. Too much statistical analysis, very little (if any) focus on riders' form, terrain, weather conditions, where the race is on the calendar, whether the break riders are out to win or soften the peloton for team mates.

The list goes on forever.

This kind of mathematical super-analysis is coming very much to the fore in other sports, yet universally ignore human factors. At best they are moderately predictive in a vacuum but have zero applicability to how a game or race will actually play out.

[esldude]

I skimmed the thesis; it's pure math and has little recognition of cycling as a sport. Too much statistical analysis, very little (if any) focus on riders' form, terrain, weather conditions, where the race is on the calendar, whether the break riders are out to win or soften the peloton for team mates.

The list goes on forever.

This kind of mathematical super-analysis is coming very much to the fore in other sports, yet universally ignore human factors. At best they are moderately predictive in a vacuum but have zero applicability to how a game or race will actually play out.

This sounds like the kind of criticism Billy Beane got at Oakland. Right up until he was able to show that such analysis was useful. Maybe more useful than the reliance on feel good human factors.

Also reminds me of analysis of shooting streaks in basketball. Not hard to come up with several famous examples. Players talk about knowing they couldn't miss, feeling it, just wanting to be fed the ball. And they were right, at least up until they were wrong and missed. Statistics showed that over hundreds of shots, a given player's shooting percentage pretty accurately predicted how many streaks and how big the streaks would likely be. Despite how the player felt, despite the human factors, a given player didn't seem to break out of any statistically predictable ranges. Does that mean the stat guy knew which day they would start one of those streaks? Nope. Then again, you couldn't have asked the player and found out either. They only 'know' while in the middle of a streak or when it is over.

[Louis]

I skimmed the thesis; it's pure math and has little recognition of cycling as a sport. Too much statistical analysis, very little (if any) focus on riders' form, terrain, weather conditions, where the race is on the calendar, whether the break riders are out to win or soften the peloton for team mates.

Perhaps, but lots of those variables are nearly impossible to quantify (and tend to even themselves out over time) so he had to leave them out. If, despite this omission, he can still get a somewhat predictive model then he's still doing pretty well with the variables he has included.

[MattTuck]

I skimmed the thesis; it's pure math and has little recognition of cycling as a sport. Too much statistical analysis, very little (if any) focus on riders' form, terrain, weather conditions, where the race is on the calendar, whether the break riders are out to win or soften the peloton for team mates.

The list goes on forever.

This kind of mathematical super-analysis is coming very much to the fore in other sports, yet universally ignore human factors. At best they are moderately predictive in a vacuum but have zero applicability to how a game or race will actually play out.

This was essentially the argument against the use of sabermetrics in baseball for years... I'm not saying that those factors you mentioned aren't important, but before you can take all those things into account, you need to start off with a more simple model.

I'm not sure how you'd characterize rider's form ex ante. I suppose you could use recent race results, but even that might prove problematic.

I think the general model of a breakaway rider deciding whether to work for the break or defect is an interesting game theory problem.

Overall, I was impressed with the work, not because it was elucidating to the sport of cycling, but because he actually went through the data of past races and tried to create something original and novel.

To your point about weather, I don't think it would be hard to add to his analysis, an independent variable of temperature and/or precipitation, and see if it explains any of the variation in winning. With regard to terrain, he was only looking at flat stages. So I'm not sure what you mean by him not accounting for terrain.

This is a senior thesis, not a Ph.D. thesis. I was actually pretty impressed with the amount of work that went into coding the various stages. It must have been 8 parts tedium and 2 parts amusing to go back over the live blogs of 100+ TdF and GdI stages.

Edit: Was typing this at the same time as the two above posts, I think we are in agreement.

[tannhauser]

Yes it is interesting as game theory, but it is very different from racing. But game theory is just that.

What I'm talking about is strengths and weaknesses of riders, whether they're good sustained mountain climbers, better at short hard efforts, whether they're good in the cold rain or whether they suffer inordinately, etc.

These are somewhat quantifiable and may give probable outcomes, but who saw Iglinsky winning LBL? The way a race plays out often hinges on missing one Power Bar.

Also look at team goals and individual goals within the season. In team sports the goals are clear: playoffs or win it all. In cycling, though also a team sport, individual goals play against team goals all the time, rendering statistical analysis somewhat moot.

Anyway, an impressive amount of math for a small result, I'd say.

[tannhauser]

PS I'm all for analysis, but it has to be balanced and applicable to the task. If all kinds of real-world variables are thrown out, it becomes a paper worthy, or not, of a senior thesis. I.e. somewhat nice brain food but that's it.

[tannhauser]

I'm not sure how you'd characterize rider's form ex ante. I suppose you could use recent race results, but even that might prove problematic.


This specifically - that's exactly my point. It is guess work at best.

[slowgoing]

I would have rather seen a study that takes into account WHO is in the breakaway, like hemocrat average of breakaway v. peleton. IMO, who is in the breakaway is the most important factor on whether it is successful.

[tannhauser]

I would have rather seen a study that takes into account WHO is in the breakaway, like hemocrat average of breakaway v. peleton. IMO, who is in the breakaway is the most important factor on whether it is successful.

Exactly. I was going to edit my response to reflect this. You put Voeckler and Vino in the same break the peloton starts to get its skivvies in a wad.

[Grant McLean]

IMO, who is in the breakaway is the most important factor on whether it is successful.

Another interesting angle, not raised in the paper as far as I can tell,
is that "success" of the breakaway need not be measured as a win.
A significant number of riders in Grand Tours have no hope of a stage win,
but go on breaks to get their sponsor on TV. This angle should have been
included in the paper, as it is an essential component of the 'game' the
riders play in their tactics. In that sense, all breakaways are successful.

-g