Giving up: Abandon Rates since ’98

When Katusha’s Michael Mørkøv climbed off his bicycle on the road to Luchon, ending his involvement in the eighth stage of the 2016 Tour de France, he was writing his name into (admittedly very cycling specific) Trivial Pursuit answer books for years to come.

morkov-abandon

The reason? Mørkøv was the first man to abandon the entire race, which meant that the rolling cavalcade circus that comprises the world’s greatest bike race had gone seven days and 1,401.5km without losing a rider. This was cited as a record, and one that would likely never be beaten. Obviously, this was an invitation to test that assertion…

Cyclingnews, everyone’s favourite cycling website has the stage results data for all three Grand Tours going back to 1998 (it actually goes back to 1996 for the Tour, but I went for 1998, when they started doing all three Grand Tours), which meant I could have a look at how many riders left the race for whatever reason. Then, this data could be mined for a look back at the last 19 odd years of cycling, plus, of course, to see whether Mørkøv reall had made history.

Firstly, let’s look at some of the reasons actually abandon or leave a race:

Injury/illness

pb-120706-tour-01.photoblog900
The infamous “Metz Massacre” of 2012 saw the majority of Garmin Sharp wiped out, and ensured a steady trickle of riders would abandon as they succumbed to its after effects

The most common reason. Whether it be roadrash, a broken collarbone, or some respiratory ailment, riders will most likely leave the race simply because they are unable to turn the pedals. Still, Robbie McEwen, who finished the Tour with broken vertabrae, and Geraint Thomas, who had a fractured pelvis, have shown that it usually has to be pretty bad before giving up – they aren’t footballers after all.

Fatigue

2009_vuelta_a_espana_stage1_time_trial_alexandre_vinokourov_astana
Alexander Vinokourov retired from his comeback race, the 2009 Vuelta, complaining of fatigue. Insert your own punchline here…

Usually combined with injury or illness, three weeks of hard racing can take its toll on some, especially young riders. Burning 5,000+ calories a day for three weeks is obviously hard, although during the period under investigation, the number of rest days have varied from zero to three.

Pre-planned

vueltaespana0007
Mario Cipollini’s participation at the 2003 Vuelta was a necessary requirement of his team’s invitation, so he turned up, and abandoned after the opening team time trial. Not quite Mark Cavendish leaving the Tour to prepare for the Olympics…

Of course, some riders actually want to leave Grand Tours early, having been using them as training for future races. This is most common with sprinters at the Giro trying to avoid the infamous final mountain week, or riders at the Vuelta using the event as a tune up or the Worlds. Later, we’ll have a look at the Tour in Olympic years to see if the same effect can be seen.

Extreme Weather

stelvio
Grim conditions on the Stelvio in 2014 forced a number of abandons

Grand Tours are keen on visiting the high mountains, and in the giddy heights of the Alpes and Dolomites, the weather gods cackle at the attempts of man to conquer them. Whether it be snow, hail, wind or even just extreme temperature changes from the previous day, the weather has a way of grinding down the resolve of a rider, and providing the proverbial straw to break the camel’s back. Who wants to be freezing for days on end, after all?

Outside the time limit

tdf-s19-01
Mark Cavendish and a bunch of other sprinters were outside the time limit on stages of the 2011 Tour, but were penalised with points deductions rather than disqualification

A lot of fuss was kicked up when 91 riders finished outside of the time limit at the end of stage 15 of the 2016 Vuelta, with many complaining the riders should have been ejected from the race. Rather than listen to the Twittersphere, the organisation wisely allowed the riders to remain in the race rather than risk the farce of having a peloton of just 73 limp into Madrid. Whilst it is rare for riders to actually be ejected en masse (it did happen a couple of times in the Giro, with over 40 riders culled from the pack in 1998 and 2003 for being outside the limit), it is more likely for

Whilst it is rare for riders to actually be ejected en masse (it did happen a couple of times in the Giro, with over 40 riders culled from the pack in 1998 and 2003 for being outside the limit), it is more likely for individuals, with a particularly cruel example being when Robert Hunter was “Hors Difficile” on the penultimate TT day of the 2006 Tour, with a saddle sore preventing him from sitting in the saddle. He was thus denied the privilege of riding into Paris as part of Floyd Landis’ yellow jersey winning team (ooh er.)

These days, big groups are more likely to get deductions in the points competition, and now that the precedent has been set not to remove huge groups, it’s unlikely we’ll see any massive eliminations for some time.

Disqualification

nibalicar
Vincenzo Nibali takes a not so subtle sticky bottle

Being disqualified can be achieved in numerable ways, and is not discriminatory to rider status. Both Chris Froome and Vincenzo Nibali have been expelled from Grand Tours, although Froome insists his disqualification, for hanging onto a race motorbike, was moot as he had already abandoned the race. Punching, headbutting, throwing water bottles, taking a lift in a car and overzealous shoulder barging have all been reasons for rider’s forced removal from races in the last few years.

Failing a drugs test

162252340
Riccardo Ricco is one of those who holds the dishonour of being booted off the race for failing a drugs test, having been caught using CERA at the 2008 Tour

Thankfully not so common these days, riders used to get booted out of races ofor failing tests with alarming regularity, particularly during the UCI’s “health check” period pre-EPO test, where the 50% limit on haematocrit caught out numerous riders. Sort of related is being pulled by your team due to not being entirely truthful about your location due to trying to avoid drug tests, as a certain Mister Rasmussen found out in 2007.

Tragedy

remembering_ww

Sadly, on one occasion, a rider has passed away during the race. Wouter Weylandt died instantly after crashing in the 2011 Giro, and the official results  record Weylandt as “DNF” – the only rider to do so on that stage. The #108 campaign and subsequent poignant procession by the peloton the next day thankfully balanced the personality lost by the cold accuracy of that results sheet.

Data

So how exactly does the data look? Well, there are various different visualisations, but first, let’s see the main data.

Some explanation – each year has five rows, which show the following things:

  1. The distance, in kilometers, of each stage. The accuracy of this data is occasionally questionable – different sources cite different lengths, while the Giro used to have a trend for getting their stage distances wrong by significant factors. In the end, I went for what was on the Italian Wikipedia pages.
  2. The number of riders remaining after each stage. As different races have different numbers of entrants, this varies – the number at the start is shown under the year in the left-hand-most column. If a rider withdrew before the race began, they are not counted in this number.
  3. The percentage of the original number of starters left in the race. As field sizes are different, the proportional amount of leavers is a more useful measure. This row shows how that progresses.
  4. The percentage of the remaining field lost each particular day. This shows the relative chunks of the field lost each day as a percentage, rather than as a number. For instance, if there were 100 riders in the race, and 1 left, it would show 1% whilst if there were 2 riders remaining and one left, it would show 50%.
  5. The raw number of riders who left the race on each stage.

The red lines represent when rest days fell – obviously, they are flexible, and vary in number from zero through to three. There is then a small table at the bottom which shows averages and sums for the number of leavers for each stage, as well as an average stage length calculator. Conditional formatting is used to highlight the gradient differences.

Tour de France

tourabandons

The formatting for this is somewhat lessened in impact by the 1998 race – on one day, 33 riders exited stage right as the race degraded into a battle between the anti-doping authorities and riders. Given the next two largest days in terms of rider loss are a -17 (also 1998, as Festina left) and -11 (in 2002), this is seemingly a once in an era, anomalous event.

However, in terms of abandons we can see that as expected, the majority happen in the final week of the race, when it traditionally hits the mountains. It is of note, though, that the shift from a traditionally flat first week of the race to a more varied, “lumpier” model, in seeking to appease Christophe Prudhomme’s desire to have a “Tour that can be won on any day.” has not seen increased losses of riders in the first week as might be expected from a move to more exacting terrain. Sure, years like 2000, 2002 and 2005 took 4 or 5 days to lose a rider, but there is little real pattern.

Giro d’Italia

giroabandons

A couple of mass time cut cullings colour the data in the same manner as the Tour for the Giro, but otherwise it is pretty much the same story. It does hold the record for the highest losses in a single day though – 45, who were removed for being outside the time limit, which seems crazy until you consider people were genuinely annoyed the Vuelta didn’t cut the field by 90 odd riders in 2016…

It’s arguable the Giro shows a bit more of a spread of abandons then the Tour throughout the three weeks, which saw its abandons backloaded, whilst people seem happy to pull out of the Giro in the first week.

Vuelta a Espana

vueltaabandon

The Vuelta doesn’t have the extremes of the Tour or Giro, with a high of “just” 19 riders abandoning being its high point. Riders do leave earlier though – whilst the largest number left the Tour and the Giro on stage 17, for the Vuelta, stage 11 is the stage on which the most riders have abandoned. This looks to be for two reasons – the historical position of the Vuelta before the Worlds has led to it being a preparation race for many, with riders leaving after a week and a half to prepare more specifically for their event.

All Grand Tours, 1998-2016

abandonsall

Above is a representation of all three Grand Tours on one page, in descending order of “age.” It uses the percentage of the field remaining figure, plus some formatting, to show the gradual fall off in numbers as the races progress. It should be immediately obvious that younger races have lower abandon rates, as shown by the large splodge of reds and dark oranges in the top right corner.

The right hand columns show various ranks and standard deviations for all the tours in comparison to one another, as well as showing two key stats:

  • The average percentage of the field that makes it hope from a grand tour is 75.55%
  • 46 riders, on average, do not complete a grand tour.

We can look at individual grand tours in situ, of course. The statistics for all are as follows:

stats-abandons

Tour de France

tourabgraph

The Tour has an average completion rate of 78.8%, the highest of the three Grand Tours. You often here injured riders quipping “you don’t abandon the Tour”, which combined with its well known prestige, seems why it see four percentage points more worth of its field finish than the Giro or Vuelta.

Giving the completion percentages a standard deviation also suggests that all bar two Tours fall within it, meaning it is very consistent in the number that finish, although the most recent and least recent Tours investigated, 2016 and 1998, are both at extremes of the spectrum and thus significantly alter this statistic.

The % finishers graph is also distorted by theses, seemingly showing a steady upward progression that would look fairly consistently flat if the 1998 and 2016 Tours were removed.

tourrates

On the 2016 Tour, it is certainly true that it gets the furthest of any Grand Tour without losing a rider – 1401.5km, nearly double the previous best of 644, which in itself was pretty impressive. But the highest rate of finishers? No. That goes to the Vuelta…

Vuelta a Espana

vueltaabgraph

The Vuelta has, until recently, suffered from a lower quality field, and being seen as a “World’s preparation race” The charts here suggest that the race is keeping riders deeper into the race than previously, although it still falls short of the other races, losing on average 26.8% of its field in every race.

vueltarates

Whilt the Tour managed 1,401km before an abandon after 8 stages, the Vuelta has gotten 7 stages before losing anyone in 2014 – going only 896 kilometres though, mainly due to the Vuelta almost always starting with a time trial of some kind and thus limiting the length. The race also has a habit of losing riders on the second stage, meaning that on average, it travels only 198km before losing a rider, compared to the 310 the Tour manages.

More interestingly though, the 2016 Tour, whilst managing to get 8 stages in without losing anyone, ended up with 87.9% of the field finishing, which, whilst a record for the Tour, isn’t a record for Grand Tours (in this period anyway). The 2012 Vuelta managed a 88.4% finish rate – the record for any Grand Tour covered in this period.

 

2012_vuelta_a_espana_stage8_alejandro_valverde_wins1
The 2012 “Three Amigos” Vuelta – the highest finishing rate of any Grand Tour between 1998 and 2016. Perhaps because it was more about short super steep climbs then the long, packed mountain stages of tradition?

 

Oddly then, despite its usual guaranteed good weather, the Vuelta loses the most riders, the highest percentage of riders, and loses them quicker than the other Tours. Sure, it does have a slightly larger than average field size, but the difference (4) is negligible.

Giro d’Italia

giroabgraph

The standard deviation figures for the Giro suggest that it has the lowest variance between its abandon rates, which is impressive given it has two massive drop offs (the red bars in the chart) caused by the organisation cutting those who finished outside the time limit. Otherwise, the race has been fairly consistent.

girorates

Like the Vuelta, the Giro has a habit of losing riders after the opening time trial, leading to it really dropping off in terms of how far it manages to go without losing anyone, managing 30km less than the already poor Vuelta at just 162 km – 4.6% of the average total.

Do larger fields cause more abandons?

 

gettyimages_576768350_670
“Skybots” caused a rethink in team sizes, as did a belief that big fields were causing more crashes.

 

This is a bit of a loaded question, in relation to the decision by ASO, RCS and Flanders racing to reduce Grand Tour team sizes from 9 to 8. Their logic was that this would “improve the safety conditions for the riders with a smaller peloton on roads equipped with more and more street furniture,”  as well as to “make it more difficult to dominate a race as well as enhance conditions for events to offer better racing for cycling fans.”

Clearly, there are more variables at play – the course, the weather, the context etc, but we can still have a quick look at whether the size of the field affects the abandon rate, if at all.

vers

In a word, no, it doesnt. There doesn’t seem to be any direct connection between field size and the number of finishers – if there was a negative relationship, as inferred by organisers, the graph would form a line from the top left to the bottom right, while if anything, it does the opposite – the few lower field sizes end up with lower rates of finishers. But there isn’t enough data for a firm conclusion, and the variance in a field of 198 between 88% and 55% shows that it is other factors that decide the “danger” of races, rather than the size of the field.

Do the Olympics have an effect?

 

jan_ullrich_winning_road_race_sydney_2000
Jan Ullrich is the only Olympic champion to enter the Vuelta the same year as winning the Gold Medal. He did not finish in 2000.

 

The rapidly changing dynamics of cycling over the years mean direct comparison is a bit moot, but we can have a look at whether the following annedotal propositions about the Grand Tours are correct:

  • The Giro will have a higher completion rate as riders, particularly for the track (up until 2012 anyway) use it as training to get a “base” for the Olympics.
  • The Tour will have a lower completion rate as riders abandon early to go and prepare specifically for the Olympics
  • The Vuelta benefits from those who peaked for the Olympics with an higher quality field, who are more likely to use this fitness to finish the race.

So, does any of this get borne out in the statistics.

olympic-effect

Yes, sort of, especially if you remove the 2016 Tour’s seemingly anomalous results. If you do that, then the results get borne out almost exactly as predicted by anecdote, although realistically, we are only looking at a data set with only 15 figures in, so it’s not a very reliable predictive model.

Is the abandon rate reducing?

gt

This graph suggests it is, although it is heavily skewed by an appalling 1998 that saw a maximum (maximum!) of 58% of a field finish a grand tour. The trend is generally upwards though, although this is in a fairly limited range. This suggests the notion that races are more dangerous (again, this depends on many more variables than just the size of the field) because their field sizes have been increasing is, well, flawed.

Are rest days more likely to see an abandon?

 

restday6x4
Rest days are more traditionally known for making riders have their picture taken whilst reading papers and drinking coffee.

 

The figures collected only show the numbers who finish a stage, so if a rider abandoned on the rest day, they would only show up when they didn’t finish (as they didn’t start) the next stage. Nevertheless, we can look at whether the numbers who abandon after a rest day are higher than any other day. We won’t include any “early transfer” rest days, such as the Giro’s in recent excursions to Ireland, the Netherlands and Denmark, in order to avoid warping  the figures.

restdays.JPG

This basic analysis produces a chart that shows the average number of riders lost on non-rest days, and the average number lost either side of the rest days, for each individual Grand Tour. It then takes the average who left on rest days as an average of the number who didn’t – so if the resulting percentage is over 100%, more riders abandoned either side of the rest days than didn’t (I realise this is a pretty simplistic way of adjudicating this.) Red bars are for Grand Tours that only had one rest day counted, as they either only had one, or one was a very early rest day taken to ensure a transfer fom foreign climes.

The results show that 37 of 56 Grand Tours (the 1998 Giro had no rest days), some 66%, have featured rest days where more players left than on non rest days. It would be good to do a two tailed test to find the statistical significance of this at some point, as this could just be a statistical fluke, but first impression do indeed suggest people are keener to leave on a rest day, oddly enough.

Other Charts

averageriderslostperstage

 

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s