As Stage 9 brought in another handful of dramatic tumbles and sent out another handful of top names, the most compelling storyline at this year’s Tour de France continues to be the crashes. Everything from the weather, to “muppets” to too many bikes has been blamed, but I can’t help but wonder if this year has actually been any more dangerous than the others.
After all, so much attention has gone to crashes this year because so many GC riders have been taken out. But is there really an increase in overall riders down? It’s rare that more than a passing nod is given to a tumble that takes out a few domestiques, but as far as overall safety is concerned, I think one rider’s abandon is as good (or bad) as any other’s.
So I’ve compiled some data for all the Tours de France since 1997 (cyclingnews.com doesn’t go back any further), looking at the percentage of riders who’ve gone home after nine stages. Obviously, it’s not a comprehensive study—early climbs and drug scandals have also played a role in thinning the pack, and not all crashes result in abandons—but I think it’s a decent ballpark metric.
Avg Attrition: 8.13%*
Avg Attrition w/198: 11.11%*
Avg Attrition w/180: 5.67%
Std Dv: 3.10%*
+1 Std Dv: 11.23%* (higher rates in red)
-1 Std Dv: 5.04%* (lower rates in green)
The numbers say some interesting things. The first is larger fields definitely increase the number of abandons—the rate of attrition by the 9th stage in a 198-rider field is almost a full standard deviation above the average since 1997, while 180-rider fields fall almost a full standard deviation below it.
As much as I’d like to see as many teams as possible contesting the sport’s biggest prize, it might just make for a better race if a few more people stayed home. Perhaps the 8-rider-teams solution floated by Craig Lewis might be a good way to get as many sponsors involved in the Tour while maximizing rider safety.
The second big takeaway is that this years race hasn’t been as brutal as you might expect in terms of sending riders home. Through nine stages, 2011 is just a touch above the 15-year-average, and well below what you’d expect for such a full field. Certainly the GC contenders have been overrepresented in the early departures, but that higher visibility doesn’t necessarily reflect a more destructive event.
The third thing that stands out to me is that—and I readily confess to falling back on the TREND(); function here—there is a slight trend toward lower attrition rates over the past 15 years (through nine stages, not correcting for field size*):
While I wouldn’t say that the ASO has ever been a tremendous advocate for rider safety, I believe this trend reflects the increasing level of sanity they’ve applied to routing each year’s Grand Boucle. Continuing in a direction that began with the end of split stages, organizers have promoted shorter routes as a way to stave off doping, while extending the “safe” zones at the end of flat stages to preserve the campaigns of GC riders caught behind crashes.
The end result of these changes has been—statistically speaking, anyway—a less destructive race, and 2011, for all the carnage we’ve seen out on the roads thus far, has been yet another step in that trend. It’s certainly felt like a more dangerous race, and viewer reactions (mine included) have helped foster that sense. But looking at the numbers, it’s pretty clear that reaction is not reflective of a greater number of crashes, but more a result of a greater public awareness of and affection for the athletes involved.
This post initially misreported the number of starters at the 2007 Tour as 198. Dossards 1-9 were omitted that year, making for only 189 starters, despite dossard 219 being the highest awarded. Numbers and figures marked (*) have been corrected from their initially reported values; the conclusions of the post remain largely unchanged.
thoughts on “Has The 2011 Tour de France Really Been More Dangerous?”
Seems to me that the attrition numbers for 1998 may be a bit inflated due to the Festina affair.
In addition, I think some (much) of the reaction is not necessarily based on just the number of crashes, but also on the number of crashes involving high profile riders.
Oh, man, you’re totally baiting me, aren’t you Cosmo? 😉
The new intermediate sprints arn’t helping, alot of crashes there as teams jockey for position just as they do at the end
It seems useless to use general abandons to talk about abandons due to serious crashes. If you used abandons-due-to-crashes, this might tell a different story. Might not.
Cool high quality post. thanks!
Your figures are wrong though.
2007 there were only 190 starters.
And 1998 was because of the Festina scandal.
2003 was the centenary tour. The race was a completely different route and they went to the mountains first which saw 7 riders eliminated on and another 7 on stage 8 to Alpe d’Huez
@ Cosmo: this is “obviously it is not a comprehensive study” you write. Really, it’s not a study at all — or it it is, it’s more about the pitfalls of trying to react instantly to the news out there…
Why include in your statistic riders who had to give up due to doping or anything else than a crash for that matter? Doing so completely defeats any “conclusions” you make.
Is the urge to say something “clever and counter-intuitive” so strong that you’d bother to write a post that may (or may not) be utterly meaningless?
What’s wrong with waiting up till you find out the numbers for riders who left the Tour because of crashes before drawing conclusions?
Interesting way of putting things in perspective. As you and others have said, perhpas total ‘abandons’ are not the best metric of how dangerous a race is but it’s still nice to see this year’s numbers compared to year’s previous. Of course it remains to be seen whether Flecha and Hoogerland will start on Tuesday.
Whether or not this year has seen more carnage than others, here’s wishing for an exciting and safe last two weeks.
Oliver wrote: Why include in your statistic riders who had to give up due to doping or anything else than a crash for that matter? Doing so completely defeats any “conclusions” you make.
Wait – you just drew a possibly erroneous conclusion without citing evidence. If you want to show that what Cosmo did “completely defeats” his conclusions, then YOU’RE going to have to compile the additional, complicated information to prove that.
Or, you could just point out that you don’t think it’s a “decent ballpark metric” due to the reasons you cite. That would be okay, too.
The evidence has already been cited matty.
the 2007 start figures are wrong because cyclo has included the astana riders that did not take the start.,
The 1998 figures are wrong because an entire team was booted out after the festina affair and several other riders also left due to doping allegations.
Thats perfectlyt good evidence.
And as already said, 2004 was a one off, with the alps in the first week.
Great article, and I think you did a very appropriate depth and detail of work here!
“…the average since 1997, while 180-rider fields fall almost a full standard deviation below it…”
Can we get a T-value?
I’m sure I’m not alone in thinking this exact same thing….and it’s so nice that you put a post together to clarify all of this. Well done…this is fantastic stuff. I was just thinking earlier today how this year has got to be one of the worst for crashes and if it was because of a more dangerous route or something. Thanks for setting this straight.
Crashing in front of the cameras are enavitably a lot worse than in a text dialogue, and the perception weighs more heavily because of the more intesnce media coverage from every angle available. I must admitt though; the crashes have been “spectaculaire!!” I’m more concered about the lack of judgement exercised by the director sportives. Especially Sargent putting a diasabled Van de Broken back on his bike and Burneel allowing severly concust Chris Horner to ride 25km in a dissoriented confused state. Then the follow day to say that Horner didn’t start because he has a severe haematoma on his leg; this heamatoma was so severe in fact that it gave him concussion! And so intense that it even affected Johan Burneels rational! What a load of BS, derriere covering! JB was totally irresponsible in allowing Horner to finish the stage, after addmiting to the media that he knew what state Horner was in at the crash site! This shows a lack of care of duty to his riders and then openly lying to retreve his position.
1998 start number is wrong, should be 189. Festina booted. Two riders excluded because of time. So your thirty abandons become ten. Only three of those ten are listed as crash or injury related.
I agree Rod re 1998. This year I think we’re all a bit focused on the danger of the sport. Esp after Wouter’s death. Also, I feel that although this year’s edition of Le Tour may be statiscally no more dangerour than other years- it has claimed the hopes of a number of potential tour contenders- which adds to the attention.. Wiggins, van den Broek, Vino, Horner, etc etc..
Interesting. But what about the seriousness of the injury’s? It look’s like this year the injury’s are more severe then other years? Can’t you try to find out that?
International fame for this site at long last! This “study” is getting linked by cronies of A.S.O. (read journos from l’Equipe and other French media). Why? Because the (dubious) conclusions of this post fit right in with the talking points of A.S.O. and l’Equipe: this Tour is not unusually unsafe… Move along, there’s nothing to see.
And I thought Cosmo was a muckraker…
From 1993 to 1999 you also have to deduct one for Mario Cippollini who used to quit at the end of the first week once the sprinting was out of the way 😀
First off this is the first year in recent memory that they have not had a Prologue, you need to adjust accordingly and look at up to stage 8 or your numbers are meaningless.
Okay beyond that lets look at the read years.
Starting with 2003 specifically
prologue (remember stage 8 this year is more like stage 9 in most years)
198 starters/198 finishers
198 starters/198 finishers
196 starters/196 finishers
196 s/ 196 f
stage 4 (team TT)
196 s/ 196 f
196 s/ 196 f
195 s/ 194 f
(vicioso DNS and Baldato DNF)
MOUNTAINS – woops this throws it all off because so far the 2011 tour hasn’t hit the big mountains yet
anyway big attrition here, but due primarily to the heat and the climbing combined.
194 s/ 187 f
Oh hey look another GRUELING mountain stage (as they say). Seriously this stage was a big pain in the ass.
184 s/ 180 f (minus 1 for Time Cut so 179 f)
That’s it. That’s quite a decrease in numbers.
But of all those how many big names were lost? sure Tyler was still riding after stage 8 with a busted collar bone.
Back to your numbers.
Bottom line is sure looking at 2003 the number is higher but after 9 days of the tour they only have an attrition rate of 10% (to compare to this year’s number).
there were two huge mountain stages on day 8 and day 9.
We’ve had no significant mountain stages yet.
I would say your analysis is wrong AND misleading. Esp considering how wide of an audience you have.
Re-adjust your calculations and your statistics and it would make more sense. Rookie mistake forgetting about the prologue not being counted as a stage (but it still counts as a day in the tour and should be factored in to be accurate).
If you want to talk about the relative danger of bicycle racing at some point you need to differentiate between factors that can be reduced with proper planning and factors that are somewhat less controllable. Rain and crosswinds must be dealt with, but narrow roads with roundabouts and tight corners leading to field sprints (intermediate or final) are errors of planning, as are level train crossings with active traffic (Paris-Roubaix 2006), roads crossing tidal estuaries (Passage du Gois 1999) and roads with murderously dangerous conditions (Tourmalet 1910).
Somewhere in between those extremes lie the numerous errors in human judgement of those on the road themselves. Riders can never be completely insulated from making mistakes, but race organizers have a responsibility to protect them from the errors of others. Restraining fences must become more frequently used in sections where overwhelming fan presence threatens the safety of the riders and/or outcome of the tour. Upcoming narrow sections should be clearly labeled ahead of time on route maps and GPS and the movement of motorcycles and cars in proximity to riders should be more carefully orchestrated and controlled in those sections. Motorcyclists, drivers or errant fans who cause injury should not just be forcibly removed from the tour premises, they should be subject to criminal prosecution.
This is really bad information. Horrible information in fact. No one should be looking at it even for novel purposes. It doesn’t include people that dropped without injury, it doesn’t include people still riding but hurt from crashes, and it doesn’t include the amount of GC contenders hurt by the crashes.. These charts are misleading. They are something fox news would post up.
You should be ashamed of yourself for having the audacity to put up so bad data.
Seems that the main issues of concern
-number of crashes (including those that do not cause abandons, such as Hoogerland)
-number of crashes that have affected GC contenders
-number of crashes that have occurred in the otherwise “non-dangerous” parts of a race (pack moving in a straight line on a dry road as opposed to, say, bunch sprints, hairy high speed descents, etc.)
I don’t think the data you have provided would allow you to draw conclusions on any of these three points. As others have noted, you really need to know the reason for the abandons, who they affected, and at what part in the course they occurred.
Sorry, Cosmo, although your posts are usually quality material, this one is more than flawed. I’m really tempted to say, that this kind of unreflected math and assumptions is the one that lead to all those economics crises of the recent years. Although, in your case I think it’s just that you simplified it a bit too much for time reasons, resulting in an incorrect conclusion, while in the economy case i believe it was mischievous deception of many for the profit of a few.
The main problems of your numbers are as said,
how many abandons are results to crashes and
how many crashes were there in overall.
Without these details the figures are worthless even if you take out the 1998 and 2003 Tour as special cases, because there is no correlation between abandons and crashes as you assumed for simplicity, although knowing that it isn’t completely correct. But that is just a simplification you cannot make. So all you can take from these numbers, with three exceptions, of which two have a valid reason, everything is within a 5% margin, so everything is as usual and one cannot really draw any conclusions about the Tour 2011 being more dangerous or not.
But from a subjective point of view, I think, this year’s course is much safer than last years, because there are no cobbles, cobbles are always a bit of (bad) luck, and the carnage day of last year in the Ardennes. That course was really stupid. Those roads are part of my training rides, and I can tell you, the way that stage was planned was already stupid if it were dry, and the rain really brought out the worst. For example riding Le Stockeu the way like in LBL is perfectly fine, but riding it backwards is like asking for crashes.
Also, at least the media wants to make you believe there are more GC contenders than last year, that is why in my opinion we have more crashes by those riders. While it last year was Schleck brothers, soon reduced to AS, vs AC with Evans with outsider chances, this year partly because of AC’s fall during stage 1 and maybe being at least looking a bit tired from the Giro, everybody who is capable of a top 10 placing is already a candidate for the win/podium. So we also have an inflation of “crashing GC contenders”.
But seriously what is it with the convoy hitting riders? I’m sure this has happened before though I never witnessed nor heard of it, but already twice this year, I think ASO should do something about that. Stricter rules of behaviour in the convoy or similar and some fines additionally to the ejection of the Tour for compensation of the teams would be some approaches I’d think about.
Thank you for taking the time to look into something I have been wondering about, numbers of riders correlating to numbers of crashes. I think based on the other comments that there is obviously a lot of interest in this topic. Perhaps one of the dissenters will do their own study which will include all variables and be perfect.
Per cyclingnews 2010 had 197 starters
Here you go.
1998 – 188 started stage 1
Withdrawals (discounting Festina)
St2 – 3 abandoned
St4 – 2 dnf
St6 – 2 dns
St8 – 2 dnf
St9 – 2 dnf
Total withdrawals 11
5.86% attrition rate.
2003 – 198 riders started stage 1
For the purposes of comparison I have counted stages 1-6 and 10-12
Discounting stages 7-9 which were in the Alps and not comparable.
St2 – 2 dns
St6 – 1 dnf, 1 dns
St10 – 1 dns
St11 – 2 dnf 2 hc
4.55% attrition rate
2007 – 189 riders started stage 1
For the purposes of comparison Stage 8 has been replaced with stage 10. Stage 8 was a mountain stage with 2 HC climbs and 1 Cat 1 Climbs and saw 8 elimitations, mostly for being outside the time limit.
St1 – 1 dnf
St2 – 1 dns
St3 – 1 dnf
St4 – 1 dnf
St5 – 1 dnf, 1 dns
St6 – 1 dns
St7 – 3 dnf
St9 – 1 dns
5.8% attrition rate
On that basis this year is high, but no higher than several other years.
I think the difference this year is that its team leaders being wiped out rather than domestiques
If we’re looking at how dangerous these first 9 stages are compared to other years, you really have to look at more than just withdrawals and look at the reason riders withdrew in a given year. Whether or not it’s a mountain stage or a flat stage shouldn’t matter. A crash that causes injuries that result in a rider withdrawing from the Tour is more dangerous than a crash that results in road rash but allows the rider to continue.
The only way to compare the danger of the past tours is to look at how the crashes have contributed to the riders withdrawing from the race. If you look at those numbers — and omit the riders that withdrew because of illness, doping, saddle sores, etc. — the fact that this tour has been the most dangerous in the past decade becomes clear.
We can argue about the validity of the data forever; however I think we are missing the point. In the past 15 years we have seen plenty of close calls with cars and crashes with spectators, but I don’t remember anybody being run over by a car, or dragged of their bike by a motorcycle .
Cosmo – I really enjoyed this post. You’ve taken available data, pointed out flaws in methodology and drawn some conclusions that are reasonable based on the data presented. To ask for more would simply be churlish. So far as I can tell, most of those who criticise your method appear to have decided to do so because your results don’t support their preferred conclusion that this is “the most dangerous tour for decades”. So thanks.
You got a slightly confused mention in The Guardian today.
I just heard it suggest on Dutch radio that the number of crashes increased ever since riders started to use earpieces. Sounds reasonable enough to me.
Enjoyed the post, Cosmo, even if it the numbers might need a little tweaking. It’s an interesting and fun item to consider. (I usually go to bbc.co.uk when I’m feeling like a grownup). Keep up the good work. Good to see some folks are taking time away from dropping the hammer on the local beginner group rides and talking endlessly about their power data at the local bike shop to take your article way too seriously. But then maybe I should be taking all things cycling more seriously. Anyway, gotta go run the bike through the car wash.
I actually thought the the historical abandon rate for the first 9 would be higher due to the sprinters quitting. So in over all numbers, this year in in line.
But, really, riders in the breakaway getting hit by a car?!???? Sorry – that makes this Tour more dangerous.
The referred to blog
He has recalculated the attrition rate as follows.
2002: 189 starters, 5 withdrawals due to crashes, 2.65% attrition
2003: 198 starters, 6 withdrawals due to crashes, 3.03% attrition
2004: 188 starters, 10 withdrawals due to crashes, 5.32% attrition
2005: 189 starters, 8 withdrawals due to crashes, 4.23% attrition
2006: 176 starters, 4 withdrawals due to crashes, 2.27% attrition
2007: 189 starters, 9 withdrawals due to crashes, 4.76% attrition
2008: 180 starters, 4 withdrawals due to crashes, 2.22% attrition
2009: 180 starters, 5 withdrawals due to crashes, 2.78% attrition
2010: 198 197 starters, 9 withdrawals due to crashes, 4.55% 4.57% attrition
2011: 198 starters, 16 withdrawals due to crashes, 8.08% attrition
Thank you dim. The manner in which the article’s data is analyzed leaves much to be additionally dissected to be able to draw any conclusions from. Your data is much more digestible and shows what many of us feel, a greater than usual departure from the tour due to crashes. The tour organization needs to understand that when the fan favorites are lost too overcrowding (riders, support vehicles, press, narrow roads); it simply is not as interesting to watch. Answer maybe wider roads initially and through natural attrition, narrower roads later in the tour or the correction of the obvious.
@MTBCraig: Lemond was hit from behind by a car when he was riding for Z. He nearly got run over. I’m sure it has happened a few more times in the past.
The other day, there was some mention of the crashes on Dutch TV. Lubberding noticed that some riders would stop on the left hand side of the road, with the team car following suit. That is against common lore, and against common sense. His point was not so much that it happened, but rather that no-one objected. It would be wiser, he said, to pay more attention to correcting these things than imposing more rules.
quite a few riders have mentioned that all the team directors are saying to their riders to get up the front because of the dangerous conditions, if everyone is trying to gain that spot up front at the same time as ordered crashes will happen. Ban the radios ?
To ask for more would simply be churlish.