I'm trying to reconcile two things in my mind. The recently referenced "skill/luck" analysis of Tom Tango versus Elo.
Given the luck/skill analysis that after approximately 34 games, the future crystallizes around skill (if interpreting that correctly) and presumably your year...weren't the Rangers essentially done before Christmas at 16-17-1? I might be misinterpreting/overs simplifying the meaning of these numbers, but even today as we approach the midpoint, 40% chance still strikes me as too high for the Rangers.
Put another way, how does Elo and the Tango skill/luck analysis comport with one another? No rush and thanks again.
I think they arrive at the same place via different directions. It's tough to say how much regression is "padded" into Elo because it is based on small incremental, Bayesian-ish changes to a team's estimated talent. But here's a comparison of the two methods from baseball: https://tht.fangraphs.com/elo-vs-regression-to-the-mean-a-theoretical-comparison/
The skill-vs-luck point basically means a team's record tends to be more skill than luck after that 34-game point. But luck still plays some sizable role immediately after, even if it's not the majority. Also, it's important to remember that rule of thumb is a generalization across all teams -- particular teams who have injured goalies and seem to be quitting on their coach (speaking hypothetically) may not live up to projections based on a previous sample ;)
Thank you...extremely helpful. I think I misconstrued "additional games" under the previous Tango chart as superfluous...when that isn't the case. Appreciate the outstanding follow-up explanation as always.
I'm trying to reconcile two things in my mind. The recently referenced "skill/luck" analysis of Tom Tango versus Elo.
Given the luck/skill analysis that after approximately 34 games, the future crystallizes around skill (if interpreting that correctly) and presumably your year...weren't the Rangers essentially done before Christmas at 16-17-1? I might be misinterpreting/overs simplifying the meaning of these numbers, but even today as we approach the midpoint, 40% chance still strikes me as too high for the Rangers.
Put another way, how does Elo and the Tango skill/luck analysis comport with one another? No rush and thanks again.
I think they arrive at the same place via different directions. It's tough to say how much regression is "padded" into Elo because it is based on small incremental, Bayesian-ish changes to a team's estimated talent. But here's a comparison of the two methods from baseball: https://tht.fangraphs.com/elo-vs-regression-to-the-mean-a-theoretical-comparison/
The skill-vs-luck point basically means a team's record tends to be more skill than luck after that 34-game point. But luck still plays some sizable role immediately after, even if it's not the majority. Also, it's important to remember that rule of thumb is a generalization across all teams -- particular teams who have injured goalies and seem to be quitting on their coach (speaking hypothetically) may not live up to projections based on a previous sample ;)
Thank you...extremely helpful. I think I misconstrued "additional games" under the previous Tango chart as superfluous...when that isn't the case. Appreciate the outstanding follow-up explanation as always.
I fed into that some with the chart headline, apologies for that. They're definitely not unnecessary, just less necessary. I remember some story that said if baseball wanted 95% certainty that the best team won a series, you'd need a best of a thousand format or something, lol. https://princetonsportsanalytics.wordpress.com/2014/10/10/the-mlb-division-series-should-be-1101-games-long/
So you can always play MORE games if you want luck to be less of a factor.
oh no