3 thoughts on “Runner Rank Explained ft. Matt Abbott the Data Guy”
Timothy Hanniford
Hmm, perhaps the word “rank” implies ordinal, i.e., percentile. Shortfall of the current system is that it’s commonplace to see runners with 75% rank that appear in the top 10% of all their races. But every race they do has a 3-sigma deviation first place result which skews everyone’s “rank” to the right. Why doesn’t Ultra Signup use percentile? (Also include DNFs in said percentile, because if someone finishes 50/50 on a 100-entrant race with 50 DNFs, their percentile should be 50% not 0%.)
have you ever thought of weighting rankings by various factors? for example (a) more recent races carrying more weight than older races , (b) bigger (and presumably therefore more competitive) races carrying more weight than smaller races or (c) throwing out “outliers” just like you throw out DNFs (eg if 90% of a runner’s scores are between 50% and 70% but they have, for example a 100% or a 20% ranking you would either throw out that score or discount it’s weight)
Hmm, perhaps the word “rank” implies ordinal, i.e., percentile. Shortfall of the current system is that it’s commonplace to see runners with 75% rank that appear in the top 10% of all their races. But every race they do has a 3-sigma deviation first place result which skews everyone’s “rank” to the right. Why doesn’t Ultra Signup use percentile? (Also include DNFs in said percentile, because if someone finishes 50/50 on a 100-entrant race with 50 DNFs, their percentile should be 50% not 0%.)
Great breakdown! Are there any videos or articles on how “Target” is calculated after registering for a race?
have you ever thought of weighting rankings by various factors? for example (a) more recent races carrying more weight than older races , (b) bigger (and presumably therefore more competitive) races carrying more weight than smaller races or (c) throwing out “outliers” just like you throw out DNFs (eg if 90% of a runner’s scores are between 50% and 70% but they have, for example a 100% or a 20% ranking you would either throw out that score or discount it’s weight)