Since 2007-2008, there have been 165 playoff series. The winner in 5v5 regular season Corsi has won 94 of those series and the winner in regular season points has won 94 of those series. Both have a 56.97% chance of predicting the winner of a series; 6.97% more useful than just flipping a coin.
So, I was wrong in saying Corsi had twice the predictive power; that was from data I remembered reading in a study from roughly 2 years ago. Over the past 2 seasons, 13 of the 30 series played have been won by the Corsi winning team, while 20 of the 30 series played have been won by the regular season winning team.
Prior to the past two years, the Corsi winner had won 60% of the playoff series, and the regular season winner had won 54.81% of the playoff series. So, prior to 2016-2017, Corsi was 10% more effective than flipping a coin, while regular season points were 4.81% more effective than flipping a coin. At that point, CF% could have been considered twice as effective as regular season points, but it appears that regular season points percentage has caught up.
I just gathered these numbers this morning so feel free to double check them:
Year | CF Winners | Home Winners |
2018 | 7 | 11 |
2017 | 6 | 9 |
2016 | 10 | 9 |
2015 | 8 | 9 |
2014 | 12 | 6 |
2013 | 10 | 9 |
2012 | 8 | 6 |
2011 | 7 | 10 |
2010 | 9 | 8 |
2009 | 9 | 9 |
2008 | 8 | 8 |
Total | 94 | 94 |
Win % | 0.56969697 | 0.56969697 |
[TBODY]
[/TBODY]
At any rate, with teams with higher regular season 5v5 CF% winning 56.97% of the playoff series played since 2007-2008 when CF% was tracked, maybe a guy whose CF% this season is currently 39.5, and whose CF% relative to his teammates on a poor Ottawa team is -3.3%, is not actually all that good.