If you were attempting to forecast the outcome of a general knowledge quiz competition with predictors such as age, education levels, IQ etc there seems to be several obstacles to using say conditional logistic regression, or other classifications into Y=1 (100% winner) or Y=0 100% loser, even if you came second in the quiz.
The winner of a particular quiz depends on the questions asked. The winner depends not only on own knowledge but on the other competitors knowing fewer correct answers The quiz round may be very low scoring, the winner just gets a few more questions correct so has little winner ability. The winner may not win any subsequent rounds against the same competitors (is the winner truly a logistic Y=1 winner, in that case?) If the local winner goes on to compete against quizzers from other States then he/she may outclass them or be outclassed. Who knows until after the competition?
I have been advised that conditional logistic regression can deal with this. I am puzzled as to how. Ranking the competitors in each round would not seem to work as they could be any position, winner to last, in subsequent rounds. You could add up the ranking each round and the lowest number would be the overall winner and so on and regress on the final outcome data but that might not tell you anything forecasting the overall winner of an inter state competition.