Do Young Teams Cover the Spread Less Often in Their Playoff Debut?
Posted by Neil Paine on April 7, 2010
Listening to Bill Simmons talk to Chad Millman on the B.S. Report, they mentioned how a good play was to bet for a young team (like this year's Thunder) to come out and lay an egg in their first playoff game because they feel pressure and don't know how to handle it yet. They pointed to last year's Blazers-Rockets Game 1 as a situation where a young team came out at home, favored, with a crowd full of energy, and they fell flat, failing to cover the spread. So my first thought was, is this a real trend or just one of those "seems like it happens a lot" theories that don't hold up under close scrutiny?
Luckily, we can test this. We have playoff box scores dating back to 1991, so I used that data and determined every team's minute-weighted age and experience (measured by previous career playoff games and previous career playoff minutes) in their first game of the playoffs. I also calculated "point spread" by taking the difference in regular-season SRS and adding 3.4 points (the margin by which home teams won on average during the regular season from 1991-2010) to the home team or subtracting 3.4 points from the road team. A team was considered to have "covered" the spread if their actual point margin exceeded the predicted spread via SRS +/- HCA.
Then I plugged the whole thing into a logistic model and ran a regression to predict the likelihood of covering based on team age & experience. Guess what happened? Nothing was found to be significant at anything close to a 5% level. Not age, not previous playoff games, and not previous playoff experience... In other words, any time a young team comes out and tanks their first game of the playoffs, it's just a coincidence that they were young and that they tanked -- an old team could just as easily tank, and a young team could just as easily play well. In short, any trends you think you see with age/experience and opening playoff games are merely due to random chance.
So you can go ahead and write this theory off as one of those that "seems" true, but is a case of our memories playing tricks on us.
April 7th, 2010 at 3:26 pm
Neil Payne - Mythbuster!
April 7th, 2010 at 4:53 pm
Unfortunately, though, unlike Grant Imahara I don't get to blow anything up, and I don't get to work closely with Kari Byron.
April 7th, 2010 at 9:40 pm
That's likely a pretty poor estimator of point spreads. Why didn't you use real ones? Covers.com or goldsheet.com have them going back well into the 90's.
April 7th, 2010 at 10:54 pm
Why did you predict the likelihood rather than graphing the actual residuals? Or are they the same?
April 8th, 2010 at 1:24 pm
That's a good question -- why didn't I try to predict the magnitude of the error in predicted point spread instead of simply using the binary 1 if covered, 0 if not covered? So I set up another regression where I used Age, Previous Playoff G, and Previous Playoff MP as the independent variables and "Diff" (defined as Actual Margin - Predicted Spread) as the dependent variable. This would detect if maybe the younger teams were "not covering" by a wider margin than the older teams, all else being held equal... And again, none of the variables were significant in predicting the magnitude of the error. This one looks really busted.
April 8th, 2010 at 3:28 pm
Gary,
I believe it's actually a pretty good predictor of the spread for a metric that is automated and doesn't account for injuries, etc. When sharps talk about having "power ratings", SRS is essentially what they're using -- some kind of points-based rating that is then used to determine what kind of margin you should give to the underdog. Obviously the spread is determined by having equal action on both sides, but I think a large # of bettors are going to be using some kind of formula (or thought process) like this at least as a baseline to determine relative team strengths, as are the books when they set the opening lines.