Did Chris Froome dope to win the 2013 TdF? My opinion is that he did. Here is why.
- Break down the key question into its parts and summarize the result in a set of ICE hypotheses,
- Conduct a preliminary screening of the hypotheses to decide if it’s reasonable to discard some,
- Gather evidence,
- For each hypothesis, examine the evidence to see what additional requirements are necessary, and
- Use the parsimony principle (Ockham’s razor) to decide which hypothesis makes more sense.
So let’s go through those in detail. The first figure shows our diagnostic map (also called ‘diagnostic issue tree‘ or a why issue tree). Mapping out the stem question, we eventually converge on three hypotheses.
Next, we do a preliminary screening of these; is there any hypothesis that I can reasonably discard? In this case, we think that H3—Froome cheated in a way other than doping—is rather improbable, so we decide not to test it in depth. The important part here is to remember that this is an educated guess and that H3 exists so that if, after analyzing H1 and H2, we find that neither provide a reasonable explanation for all the evidence, we go back and look at H3.
We’re now ready to examine H1 and H2 in their context. To do so, let’s start by gathering relevant pieces of data, which, when connected with the hypotheses, will become evidence. The figure below summarizes this data.
(I’m attaching this information in a PDF file with working hyperlinks so that you can access the cited sources if you’d like: Context surrounding Froome, Sky, and the 2013 TdF 2013-08-01.)
The key in this evidence-gathering step is to be as impartial as possible while gathering this information and present it in as neutral a position as possible. To help me do so, I submitted an earlier version of the file to a couple of online forums and asked for feedback.
There is evidence compatible with the ‘Froome won it clean’ hypothesis
To test hypotheses, various approaches are available, including the analysis of competing hypotheses (ACH), recommended by some CIA analysts (see for instance Morgan Jones’ Thinker’s Toolkit), and hypothesis mapping.
In both, the analyst looks at each piece of evidence and decides if it supports the hypothesis under consideration, opposes it, or has no effect on it.
Here, we propose an alternative approach: let’s take each hypothesis for granted in turn, consider each piece of evidence, and see if that piece of evidence requires another requirement for the hypothesis to hold true.
So, let’s start by assuming that Froome didn’t dope to win the Tour. I’m taking this as a fact, a given.
The first piece of evidence is that he has been performing at the top of the sport for several months. The supporting evidence gives his race results since the beginning of the year. I suppose it is safe to assume that anyone who knows a little bit about cycling will agree that such a palmarès is indeed remarkable so, for Froome to place so well in these races in a single season requires that he has an ability to perform at top form for extended periods. No disagreement? Great, let’s put that in a gray box and move to the next piece of evidence.
The second is a statement that he didn’t just beat his rivals, he dominated them. Various items support this claim, which requires that Froome be much better than his rivals, who are the best cyclists in the world.
Also relevant is the recent analysis of some of his power files by an expert (for those of you who are not familiar with cycling, pros usually ride with a power meter that measures how much power they produce). The expert concluded that Froome has an excellent VO2max and recovery abilities. Since we assume that Froome is clean, such physiological characteristics imply that he has natural abilities near human limits.
Summarizing so far, we capture that “Froome has top natural abilities that make him one of the best in his generation.” Next, let’s compare his development to those of champions that didn’t benefit from ‘game-changing’ drugs.
Speak with anyone that follows pro cycling, and there is a clear understanding that doping has been in the sport from its very early days. However, in the late 80s or early 90s, a new drug—EPO—arrived. It was so effective that it could transform ‘middle of the pack’ or ‘even back of the pack’ riders into champions. Whereas the pre EPO-era doping improved a riders’ performance by some measure, EPO metamorphosed riders. The point is that there’s no much point trying to find a ‘clean’ base line in recent champions; instead, one should look at elder examples to find what resembles a ‘clean’ champion.
When we do so, looking at the likes of LeMond, Fignon, Merckx, or Anquetil, we find that, unlike Froome, they all displayed huge potential from an early age. Froome, Sky, and others have provided several explanations for this: he grew up away from cycling countries and joined the pro peloton late, for instance, might explain why it took some time for him to develop in his early pro days. But further reasons are ventured: he was misused by his various teams previous to joining Sky and he has an exotic disease that explains his erratic results. As before, we’re capturing these conditions that must be true for the ‘clean’ hypothesis to hold true.
The next point is critical: not only did Froome’s development not resemble that of past champions, it actually is quite similar to known dopers’. But here we’re assuming that Froome is ‘clean’, therefore this must be a coincidence. That’s another condition to explain his performance. We capture it, and move on.
We continue to review the evidence in the same fashion, assuming that Froome rode ‘clean’ (I’ve attached the full analysis in a PDF file at the bottom of this page). Some pieces yield another condition that must be met, some don’t yield anything. Either way, we capture the implications. When we’ve gone over all the evidence, we summarize the additional conditions in a single box.
That is, in the light of the information we know about Froome, his development, his current performance, his team, his rivals, etc., to hold true, the ‘clean’ hypothesis requires a number of additional conditions to be met. Once we’re done here, we can move to the second hypothesis.
There is also evidence compatible with the ‘Froome doped’ hypothesis
Before we start, I should say that I’m considering a wide definition of ‘drugs’ for this hypothesis, to include also ones that might not be specifically on the WADA prohibited list. That is because sometimes WADA lags behind dopers and may not prohibit drugs because they might not even know about a drug yet. Therefore, I’m lumping in this branch the use of any drug that the use of is unethical or against the spirit of the rules.
So let’s go through the same exercise, but this time, assuming that it is a fact that Mr. Froome doped.
Under this condition, several cases arise, so we first make three assumptions: that Froome’s team and his team director know about his doping; that when journalist David Walsh calls the team ‘clean’ he does so in good faith; and that Froome, his team, and Brailsford do not want to be caught doping. (We could avoid making these assumptions and study various alternatives under the ‘doped’ assumption, but at this stage, in the interest of simplicity, we focus on this particular sub case.)
So we take for a fact that Mr. Froome doped. What do we get in terms of additional conditions?
First, in terms of his performance and his physiological data, we get that he doesn’t have to be one of the top riders in his generation. In fact, many agree that performance-enhancing drugs are so powerful that, given an average pro rider that responds very well to the drugs, and given the right doping regimen, you could get a champion (support). So we capture that as a condition: to be compatible with the evidence, the ‘doped’ hypothesis requires that Froome have reasonable natural abilities (for a pro cyclist), respond well to doping, and have very good drugs.
The analysis then speeds up: since we assume that Froome doped, the fact that he didn’t develop like the non-doped champions (which I use as a shorthand for “probably-doped-but-without-game-changing-drugs champions”) isn’t surprising; it doesn’t add a new condition.
Similarly, the fact that Froome’s development resembles that of doped winners doesn’t add any requirement.
Under this hypothesis, the added requirements come later. For instance, Brailsford originally promised full transparency. Since we assume that Froome dopes, this requires that his team was confident that they could cover up the doping or back out of that promise.
In the end, the complete set of requirements for this hypothesis to match all the data is:
We have compared in turn two hypotheses that cannot be both true. So we need to decide which is more likely.
Mr. Froome, meet Mr. Ockham
A fundamental component of the scientific method is the principle of parsimony (also know as the principle of simplicity, the principle of economy, and Ockham’s razor). This principle recommends that, “among theories fitting the data equally well, scientists choose the simplest” (source: Gauch, p. 269).
This is a principle that we all use on a daily basis without taking notice. It is because of Ockham’s razor that, when you have a headache, you assume that you have just a headache, not brain cancer. Or that if the fuel indicator of your car lights up, it is because you’re running low on fuel, not because your friendly neighbor decided not to like you anymore, sneaked into your garage at night, and tampered with the gauge. (Note that, as powerful as it is, the parsimony principle is fallible: sometimes, the complicated theory is the right one.)
Now that we have introduced the razor, let’s go back to the situation. Our dichotomy isn’t whether Mr. Froome rode ‘clean’ or not. These two hypotheses are incomplete because, in such succinct form, they do not fit all the data of the case. Rather, we get to chose between two theories:
Theory 1: Froome is ‘clean’ and:
1. Froome is one of the best cyclists in his generation:
– he has top natural talent,
– he has a top training program,
– he does the necessary work to follow the program,
2. Furthermore, Froome’s display of dopers-only characteristics isn’t because of doping:
– doping isn’t why he experienced a sudden breakthrough in performance,
– doping isn’t why he is the only current rider able to climb as fast as convinced/confessed dopers
– doping isn’t why he is able to be ‘head and shoulders’ above his rivals in a Grand Tour,
– doping isn’t why he has been able to develop as one of the best at both climbing and TTing,
– his lack of improving as a young pro rider isn’t because of his lack of talent,
3. Furthermore, any opaqueness on Froome’s (or his team’s) behalf isn’t motivated by hiding doping:
– hiding doping isn’t why he stopped providing data without an explanation when it would be instrumental to help establish his credibility,
– hiding doping isn’t why Sky didn’t make good on its original transparency pledge,
– hiding doping isn’t why Sky isn’t submitting data to trusted, qualified third parties,
– hiding doping isn’t why Sky mysteriously rescinded their earlier all-access invitation to Kimmage to the entire 2012 Tour,
4. Furthermore, developing a doping program isn’t why Sky hired Leinders (a doctor linked to doping programs in other teams).
Theory 2: Froome is ‘doped’ and:
1. Froome can benefit greatly from doping:
– he has reasonable talent,
– he responds very well to doping,
– he has very good drugs,
2. Froome is willing to dope,
3. Furthermore, Froome/Sky have hidden the doping well:
– they have hidden the doping from David Walsh, a recognized anti-doping crusader whom they invited to the 2013 TdF,
– they have hidden the doping from Fred Grappe, to whom they provided some power files,
– they have hidden the doping from the bio-passport program,
– they have hidden the doping from other anti-doping programs,
4. Furthermore, Froome/Sky are confident they can continue to hide the doping well:
– they can hide the doping from WADA despite inviting them to “have everything we’ve got”,
5. Furthermore, Froome/Sky have silenced any leaks:
– they have ensured that no former or current Sky people have accused them,
– they have ensured that no other people associated with the doping (doctors, providers) have accused them.
Putting it all together
I will let you decide which theory is the simpler for yourself. For my part, based on this analysis, I find Theory 1 vastly more complex than Theory 2.
Theory 1: not only must he be the best in the world—that, by itself is fine, someone necessarily is—but his performance breakthrough must be a coincidence, as must be Sky’s unexplained lack of transparency, and the hiring of Dr. Leinders must not motivated by developing a doping program…
Theory 2, on the other hand, with only a hint of cynicism, amounts to little more than ‘business as usual’ in the pro cycling field of the past twenty years.
As such, it is my opinion that Chris Froome doped his way to the top stair of the 2013 Tour de France podium.
Like any intelligence analysis, this one isn’t bulletproof, and it is possible that I made some mistakes. Here are the ones I can think about:
- I am biased and this bias clouds my judgement. For instance, there are a lot more requirements for the ‘doped’ theory to be true but I failed to see them. In all fairness, I am probably biased, at least to some extent; being truly unbiased is extremely hard (see (Nickerson, 1998)).
- The parsimony principle does not apply in this case.
Sky’s Brailsford asked a few days ago in an exasperated tone what he could do to help establish that Mr. Froome raced ‘cleanly’. Unfortunately for them, he won’t ever be able to prove him clean (just as, in theory, you can’t accept a hypothesis to be true, you can only fail to reject it), but if I were him, I’d take the necessary conditions of the ‘clean’ hypothesis above and provide a valid substitute for each so that that hypothesis becomes simpler than the ‘doped’ one. That starts with providing the transparency that he has promised repeatedly: if, indeed, Mr. Froome and the rest of Sky are so ahead of their competition legally, share all the data with knowledgeable, trustworthy third parties.
Case – TdF Diagnostic hypotheses 2013-08-14-1234–full analysis-version of 2013-08-14.
Gauch, Hugh G. Scientific Method in Practice: Cambridge University Press, 2003.
Nickerson, Raymond S. “Confirmation Bias: A Ubiquitous Phenomenon in Many Guises.” Review of General Psychology 2, no. 2 (1998): 175.