What does determine success in a fantasy soccer team (at least, in the very specific fantasy game I’ve chosen for this analysis)? Each player in a team is given an individual score, based on a combination of
– direct outcomes of his behaviour (goals, own goals, yellow and red cards,…)
– indirect evaluations (based on the overall outcome of the match, such as goals conceded by the team as a whole)
– decisions taken by the trainer, such as substitutions during the game
Wrapping it uyp, we’re building a numeric model for the performance of each single player during a match. Is it a good model? It depends on what our objectives are, and it is indeed a reasonably good one as long as we’re concerned with a fantasy league: there is a good correspondence between knowledge and results (choosing the strongest players, instead of 11 random ones, gives on average a better result). At the same time, a certain degree of randomness is present, allowing league players with different levels of knowledge to compete on a comparable level.
What if we tried to use the same model to evaluate the performances on the field or, even worse, to make predictions about the outcome of future matches? I believe this would be completely useless, as there are too few numbers and too few matches for us to figure out any trends.
Could a different model provide a better basis for evaluations and predictions? Yes, of course, but we’d probably need;
– a larger range of variables and parameters to be recorded, in order to account for a higher degree of complexity;
– a larger sample (a national league, with dozens of matches, would surely provide a more robust basis than a World Cup);
– some attempts at a “formula” to summarize the observations into a small set of numbers (a vote, a prediction for the number of goals scored by each team, or something along these lines);
– an evaluation of the results of our formula against real-world data.
That would be fascinating…