“I love it when plans go well,” said John. Hannibal Smith and perhaps Tadej Pogačar thought so smiling on the podium after his unchallengeable victory in the Giro. But our carefully laid plans are often twisted and ruined by a reality that reminds us that we live surrounded by disruptive events. This was expressed regretfully by Matteo Jorgenson in his statements after the massive crash that occurred in the fifth stage of the last race. Dauphiné Liberé. A Jorgenson who days later threatened Primož Roglič’s triumphant plans in the same race, in a memorable final stage brilliantly executed by Ineos.
This is not the first time that we have emphasized the hazardous circumstances that make up cycling events, the importance of team adaptation and leadership structures to respond to these circumstances. We have extended these reflections to our meetings and lunches, concluding in the curiosity to empirically verify the role of leadership structures in the victory of a cycling team in its highest competition: the Tour de France.
When Movistar showed up to the Tour with the legendary trident Landa-Nairo-Valverde, the most circumspect fans raised an eyebrow of skepticism at three leaders, while the most sanguine ones mocked what they called eusebiades (by Eusebio Unzue, the team’s magician). They were a generation of followers who had been fed the stories of the Tour of the great leaders of the past, its legends. Fausto Coppi was asked by his gregarious every morning: what are you in charge of today, captain? And one of them, Sandrino Carrea, the most faithful of his soldiers, cried on the podium of the 1952 Tour dressed in yellow thanks to a distant breakaway and looked for Coppi with his eyes as if to apologize and tell the very championthis jersey does not belong to me. Merckx was a tyrant with the whole peloton and with his team, like Armstrong or Chris Froome in his own way, even Indurain or Marco Pantani. The team was for them, and if they failed or fell, there was no more. There was no second in the team and no plan B. The change that the three-headed Movistar anticipated, the pandemic made it the norm and Jumbo refined it with its dominance in the last Vuelta, turning the victory into a team game between Roglic, Vingegaard and Kuss. The biggest teams – Sky, Ineos, Jumbo, UAE – began to take an intermediate step from centralised to decentralised leadership when they decided that the leader’s deputy or even the third in the team should not give up once their task was finished and should let themselves go, saving their strength for the next day, but should continue to work as if they were the leader, thus having an additional tactical weapon in case the leader falls or stumbles.
The specialized literature differentiates between centralized and decentralized leadership structures. Centralized structures are organized around a single leader, and the rest of the team members subordinate their actions to the requirements of the leader, enabling and reinforcing said leadership over time. Alternatively, decentralized structures are organized around several leaders, in which the leadership role alternates over time between several team members who move flexibly between follower and leader roles, being leaders or team members depending on the circumstances or changing task requirements. If you ask yourself which is better? the answer is that it depends on the team’s experience with the task it performs and its context. When the team understands its task as simple, with few changes and the context in which it is performed is predictable, centralized structures usually allow objectives to be achieved very efficiently. But when the team understands its task as complex, notices many changes and its context is unpredictable, decentralized leadership structures offer more flexibility and adaptability for the team to achieve its objectives.
Thus, to predict team performance in the Tour we selected several variables: team quality (total UCI points of team members), initial leadership structure (number of leaders at the start of the race), leadership structure during (number of leaders during the race), qualitative disruptive events (whether the leader(s) had suffered a crash, injury, crash or retirement) and quantitative disruptive events (number of retirements that occurred in the team, not including the leader(s). The initial and during-race leadership structure were inferred from our analysis of news stories, while the rest of the variables were obtained from the platform. ProCyclingStats. In particular, we obtained the performance measure from ProCyclingStatsScore (PCSS) of the 22 member teams in the selected editions (23 in the 2021 edition). The PCSS is a synthetic index that combines stage victories, days in the general classification, days with the regularity jersey, days with the mountain jersey, days with the white jersey, the number of stages in those in which its runners finish between the top three and ten, in addition to the days in the race of the best overall runner. For our analyzes we select the period between 2018 to 2023 (including both). In 2017 the calculation of UCI points changed and the number of members in the Tour teams was reduced from nine to eight and both modifications could affect the results.
We use decision trees, an algorithm machine learning A supervised model that generates a set of rules for making decisions. Decision trees work in two steps: first, they generate a model based on the data with which it is trained and second, they offer decision rules with a result that allows predictions to be made with new data in the future. We use this technique because it is easy to visualize (as an inverted tree), understand and interpret, making it widely used in decision-making processes.
With this analytical logic we train two models represented as tree 1 and 2 respectively. Tree 1 summarizes a first model that uses all the chosen predictor variables to estimate their effect on Tour performance, and clearly shows us the role that team quality has on Tour performance. According to the regression analyzes associated with the decision tree, 35% of the performance in the test is explained by the quality of the equipment (UCI points). No surprises up to this point, a team with a lot of quality is going to perform at a high level.
Considering the role of team quality in Tour performance, we trained a second model controlling for team quality, that is, making it equivalent between them (something that could make some sense among the top teams in the standings), to see how the other variables behaved when predicting performance. The generated model explains a smaller percentage of variability in performance when the variable with the most weight is subtracted, but its simplicity offers interesting complementary observations, as summarized in tree 2.
The analyses carried out using archival and historiographical data allow us to draw three conclusions:
First, we need a team made up of excellent riders to win the Tour. Having more than 7,900 UCI points is only within the reach of the best teams in the peloton, such as UAE, Visma, Red Bull, Ineos, Lidl-Trek, Alpecin, Soudal or Bahrain. But quality is not enough, because very good teams (with more than 7963 UCI points) obtain better results if they face more than one disruptive event during the test. This tells us about the concept of antifragility, which captures how teams improve by adapting to adversity and suggests that winning the Tour has more to do with adapting to disruptive events than with being fortunate enough that they do not occur. It’s winning, versus being in the top five.
Second, starting the Tour with decentralised leadership structures improves the performance of teams in the race. This effect is seen both in good teams (more than 6900 UCI points) and in more modest teams (less than 3680 UCI points). Therefore, a team like Ineos, starting the Tour with a decentralised leadership structure would finish in the top five, even with a chance of a podium, instead of finishing around eighth place. Similarly, more modest teams like Total Energies or Astana, starting the Tour with decentralised leadership structures could finish in the middle of the table, instead of at the bottom. It is a noticeable difference.
Finally, teams with decentralized leadership structures increase their tolerance threshold for disruptive events. It is surprising to see how the performance of decentralized structures with up to three abandonments is greater than that of centralized ones without suffering disruptive events. We are talking about finishing the overall in fifth or ninth position. In theory the performances should be equivalent, but the Tour is very tough, full of unexpected events, and rewards decentralized leadership structures that facilitate adaptation. We see this best by seeing how teams with centralized structures that suffer disruptive events fall to twelfth position overall. Centralized structures limit team adaptation and result in suboptimal performance.
These results are a first step towards understanding the role of leadership structures and disruptive events on performance in cycling teams. Our data do not allow us to observe the adaptive processes that occurred in the teams analysed, the events they experienced or their actual leadership structure. Perhaps this approach will spark interest in the peloton and future research will allow us to learn more about it.
We are not dedicated to betting, nor do we intend to win the game at the corner bar. But considering the results obtained with our trees, here is our prediction: The Tour will be won by a team that has more than 7900 UCI points, more likely if it starts with a decentralized leadership structure, even suffering a couple of disruptive events. The winning team must have a compact lineup, without major differences between its riders, shown in the median of UCI points with which the 22 teams start the race in Florence. The low median would exclude from the general classification Alpecin of Philipsen and Van der Poel, Lotto of Van Gils and De Lie, and Soudal of Evenepoel and Landa. The UAE of Pogačar, Adam Yates, Almeida and Ayuso presents an extraordinary 22,910 UCI points, almost triple the threshold of excellence; and, of them, 9,513 belong to the Slovenian, who alone owns more points than 18 entire teams. The quality is yours, but our model, however, confirms the intuition of the fans and reinforces the Visma of Vingegaard, Jorgenson and Van Aert, which is clear about the logic of decentralized leadership and has already begun its adaptation when suffering a disruptive event ( Kuss’s covid) before even starting the Tour. The third team should emerge from the duel between the Ineos of the four good guys with Rodríguez, Thomas, Pidcock and Bernal, and the Red Bull of Roglic, Vlasov and Hindley.
Let the road say the rest… Live the Tour!
Ramon Rico Munoz He is a professor of Business Organization at the Carlos III University, Madrid.
Ramon Rico Cuevas He is a PhD student in AI and Data Science, Utrecht University (The Netherlands)
You can follow Morning Express Sports on Facebook and Xor sign up here to receive our weekly newsletter.
.
.
_
