A six is not a six: the Landis case Handwriting argument of Landis's lawyer shot down Luc Devroye, McGill University, Montreal, Canada October 13, 2006
	How many 6's can you see on this sheet? I say one, and I will prove it to you.
.....	¶ The Tour de France 2006 was decided in the 17th stage to Morzine, when Floyd Landis obliterated the competition. He told everyone that he would escape the pack early, and on the first serious climb of the day, he did. No one could follow him. Landis rode alone in 33 degree heat for five hours. Even the combined efforts of all other riders were not enough to catch him. Cycling had not seen a performance like this since the days of Eddy Merckx. ¶ A few days after the Tour had ended, word came from the UCI, the International Cycling Federation, that Floyd Landis had tested positive for testosterone during that 17th stage. One possible reading, the ratio of epitestosterone to testosterone is eleven (!!!) times what it is in regular humans. He risks being stripped of the title. ¶ Floyd hired some lawyers and went on the war path. Objective number one, quite predictable in such situations: try and discredit the lab report of the UCI. On September 11, 2006, Howard Jacobs, attorney for 2006 Tour de France Champion Floyd Landis, submitted a Motion for dismissal to the United States Anti-Doping Agency (USADA) Independent Anti-Doping Review Board. On Floyd Landis's web site, one can download that motion, as well as some documents from WADA, the World Anti-Doping Agency. That site (log in as PublicAccess, password PublicAccess) also has a PowerPoint presentation created by Arnie Baker, M.D. ¶ In that Power Point presentation, a big deal is made about inconsistencies in the numbering of various samples. The only thing we wish to point out is that the arguments in that presentation are easily dismissed by carefully examining some handwriting. So, at least that part of the Landis case is weak, very weak. ¶ The lawyers have a lot of chemo-analytical questions regarding the tests. There may or may not be any merit in those scientific arguments. I cannot predict how this case will turn out, and I am not taking sides at this point. A blog about the Landis case with many links.
The doping test	¶ Test samples of some riders were collected in Morzine. Three were given sample numbers 994178, 994179, 995474. ¶ Nine flacons (test tubes) were collected, with the numbers handwritten on the page below. There is mention of seven urine samples and two blood samples. At least, on the left, it is easy to locate the three numbers again, 994178, 994179, 995474, together with six other numbers. These are the samples sent on for further testing in a lab. ¶ The next page shows the tear-off test tube barcodes and their numbers, the same numbers again. The Floyd Landis sample is 995474. The lawyer agrees that indeed, that is his number. ¶ Blown up, with Landis's barcode on the right:
Landis: not me!	¶ If you have read this far, and you have not seen anything fishy, then you have passed the handwriting test. The Power Point presentation published by the Floyd Landis camp on October 12, 2006, and the lawyer's motion both argue that the doping lab cannot be trusted because thay cannot even get their numbers straight on the labels. The lawyer says possibly, we may not even have Landis's sample on this sheet, because Landis has number 995474, not 995476. ¶ Here is that Power Point page by Arnie Baker. Note that the numbers in the handwritten list should correspond to those on the barcodes, and none of those even ends in a 6.
Ductus	¶ Let us look more carefully at the way in which the author writes 4 and 6. We know that 994178 and 994179 should be on the sheet, because they were found in the barcodes list shown earlier. So we know what a 4 looks like. In fact, it is quite likely that 994 178, 994 179 and 994 180 are all barcodes, as they are consecutive numbers, and maybe came in a batch. We will also accept that the first number ends in 69, as there can be no mistake with such clearly printed numbers. ¶ The ductus, the way of stroking the letter, of the 4 and the 6 is similar in our sample. The 6 starts out curved, and remains curved, not quite performing a 360 degree turn. It ends well above the baseline. In case of the 4, the stroke path starts out straighter, and ends below the baseline, doing at least a 360 degree turn. So, the trailing digit in Landis's alleged non-number is a 4, not a 6. Moreover, the simple study of the stroke path shows that the sheet has only one six, the one in 994069. There are seven 4's, including the last digit of 995474. The lawyer is wrong. Arnie Baker is wrong. Landis is wrong.
	Luc Devroye School of Computer Science McGill University Montreal, Canada H3A 2K6 luc@cs.mcgill.ca http://cg.scs.carleton.ca/~luc http://cg.scs.carleton.ca/~luc/fonts.html