Racing at Increasing Speed
by John R Petrocik
I am so lucky. Mr. Petrocik, the author of the following piece, sent me a fascinating note summerizing his observations about the increases in the average speeds of the Tour de France over the last century. He kindly agreed to share his thoughts in the following essay:
Every cyclist who pays attention to the Tour de France knows that speeds have increased over the 112-year history of the race. The following figure highlights interesting irregularities in that trend.
All of the lines in the figure above are calculated ordinary least squares estimates (“regression slopes”). Throughout the essay, I refer to them as slopes, lines, rates of increase, and trends interchangeably.
The first dozen races (to 1914) produced no trend in the winning speed. The earliest were completed at about 25 kph. The next few were faster, with winning speeds of 28 to 29 kph, while the last four held before WW1 finished with measurably lower average speeds – though faster than the earliest races.
Tour speeds were slower but consistent when racing resumed in 1919, averaging a bit more than 24 kph for the next ten years. The winning speed of the 1927 Tour spiked up almost 10 percent, and continually increased for the next twelve years. The trend (represented by the line through the data points) for the decade before WW2 marks it as the period during which Tour speed increased at a faster rate than at any time in the history of the event, as the steepness of the slope demon-strates.
Jean Alavoine, third-place in the 1909 Tour de France with a bike of the era
The 1947 Tour was won at almost the same speed as the 1939 race and the pre-WW2 trend of faster winning speeds continued. The rate of increase from 1947 through 1963 was less than the post-WW1 trend, the speeds were more variable, and there were many instances of lower speeds than a preceding race. Overall though, the Tour be-come faster.
That reversed with the 1964 Tour when, for the second time, the win-ning speed significantly declined. The drop-off in 1964 compared to 1963 was so large and the trend so much more flat that it took eight years to make up the decline. It was not until 1971 that the winning speed matched the value posted for the 1963 Tour. So while the trend between 1964 and 1989 was positive, the rate of change was more shallow than it was in preceding periods (compare the magnitude of the slopes). Also, the results were more scattered than they had been in the past. Some years posted losses, with surges on other occasions, and the race outcomes clustered around the trend less closely than was the case previously.
Felice Gimondi racing to win the 1965 Tour de France.
Speeds bumped up again in 1990, and noticeably dropped after 2006. Everyone reading this essay can probably explain this surge and de-cline. Performance-enhancing drug use did not begin in 1990 (am-phetamines were used in much earlier years, steroids and related drugs were apparently familiar prior to the 90s) but the availability of even more powerful PEDs become a centerpiece of professional bike racing in the 1990s.
The abrupt bump in 1990, followed by the noticeable drop after 2006, provides another facet for this statistical portrait. The rate of increase in each successive race’s speed during the 1990-2006 period is quite similar to the rate of increase during the 1964-1989 period. However, the 1990-2006 years have an average speed that is between 1.7 and 3 kph greater (depending on whether an adjustment is made for some arguably deviant values) during the 1964-1989 period. Put differently, if the 1964-1989 trend had continued through 2006, the average winning race speed might have been as much as 3 kph slower in the 1990-2006 races.
There are not enough races since 2006 to estimate a reliable slope for these years. That the average speeds declined is obvious in the figure. The average of the observed winning speeds for the 2003 through 2006 races is 41.98 kph. The observed winning speed for 2008 through 2014 averaged 40.2 kph.
Depending on your opinion about whether Tour racers currently use PEDs this small decline has a couple of plausible interpretations. It is worth noting that the recent performances are averaging above what is predicted by the 1964 through 1989 trend. The short black line on the far right of the figure is the expected speed of the Tour winners since 2006, given the trend of the 1964-1989 era. A comparison of this line with the orange points on the right is the comparison underpin-ning my previous statement that current race winners are outper-forming the speeds predicted from this trend of incremental increases in race speed prior to the 1990s.
The troubled 1998 Tour de France final podium, from left: 2nd place Jan Ullrich, winner Marco Pantani and third-place Bobby Julich
Some Concluding Thoughts
Increases in winning Tour de France speeds have several causes. Con-tinuously improved bicycles and equipment, more serious training regimens, changes in the Tour’s courses around France, improved roads, and racing programs that kept rider fitness high are certainly major influences. Drugs also contributed. Professional cycling, per-haps the European peloton especially, has been associated with per-formance enhancing drugs for so long that they cast a confounding shadow on any effort to understand the increases in Tour speeds over the century and what might be said about the several discontinuous trends.
The data in the figure do not permit a decomposition of the influences. However, such analyses might be attempted by adding data about plausible contributors to race speed (equipment, roads, distances, etc.) to results for different types of stages (“flats,” mountains, HC mountains, time trials, prologues). Time and speed information for particular pieces of a stage (the final climb up Alpe d’Huez, for exam-ple) could also yield useful information about relative contributions to TdF speeds over the century. It might even be possible to back-out an estimate of the net effect of PEDs on race speeds.
The bump in speeds after 1990 given the similarity of the slopes for 1964-1989 and 1990-2006 indicates how these estimates could be ob-tained. A standout feature of the two periods is that the slope for the speeds from 1964 through 1989 is almost identical to the slope for the drug era that began to get a grip after 1990. That is, to me, a surprise since everyone insists that the drug era saw a huge -- and increasing -- increase in speeds (the “hockey stick” metaphor). The data do not show that at all. The intercept changed: the average speed bumped up a small amount. Drug use made the peloton faster but the annual increase in overall Tour speed parallels the annual increase in speed from the mid-1960s through the 1980s. After that (drug-induced?) bump in Tour speed, the annual average speed increased at the same rate that speeds increased during the preceding 25 years. It looks like the drug era added a constant increment to TdF speeds, but nothing like what one might have expected from the breathless rhetoric that is often used to recount what PED use did to rider performance.
A plausible interpretation is that all the drugs apparently taken by so many in an arms-race pattern during the 1990s made people a bit faster overall, maybe much faster in some stages, and fast enough to defeat those who did not use PEDs. But it seems to have contributed a constant to a trend of increased speed that reflected better training, equipment, and so forth.
The preceding comment is rumination, not a conclusion. Overall race speed may not provide the best information about drug use. HC climbs might be better observatories. However, this essay addresses the overall race and plausible net effects. It is a “quick and dirty cut” at analyzing TdF speeds over the history of the race. The data are only a historical profile. They establish nothing about whether drugs are currently used, or how much they are used. Since riders are still be-ing caught, drug use has not ended. However, a more detailed analy-sis of rider speeds across different stages over the years might give an insight into the magnitude of the net contribution of PEDs to cycling performance. That is worth knowing. Save your ire.