mrjames
Professor
Posts: 6062
Loc: Montclair, NJ
Reg: 11-21-04
|
01-03-16 10:46 AM - Post#198441
Since we're into January now, here's a pre-Ivy glance at the title odds with league play getting underway this weekend.
I'll share three different views of the race. The way to read each is that there is a row for every team. In that row, you'll first see the odds of a shared title, then the odds of a solo title. From there you get the average Ivy wins per sim and the 95% confidence interval (upper and lower). Finally, I tacked on the playoff odds (not the odds of that team getting there, but the odds for the league to see a playoff).
The first is KenPom's view. KenPom's ratings are "preseason biased" which essentially means that there is some weighting of the preseason expectations in the ratings still, thus dampening the effect of teams that have moved significantly off those initial ratings. KenPom argues that since the preseason ratings provide predictive power even after we've seen a ton of current season games that this weighting is important. Here are the implied odds:
share solo avgwins Upper95 Lower95 playoff
brown 0.005 0.002 5.065 8.31 1.82 0.23
columbia 0.176 0.077 8.328 11.54 5.12 0.23
cornell 0.002 0.000 4.692 7.90 1.49 0.23
dartmouth 0.009 0.001 5.207 8.48 1.93 0.23
harvard 0.328 0.200 9.131 12.41 5.85 0.23
penn 0.003 0.001 4.979 8.13 1.83 0.23
princeton 0.231 0.120 8.705 12.00 5.41 0.23
yale 0.531 0.369 9.893 13.08 6.70 0.23
Next are the College Basketball Reference sims. These simulations use the ratings that are based on this season's performance only. So, any differences here are basically a sign of how much the preseason anchor is affecting the KenPom rating. Now, again, there's evidence that the preseason ratings are still predictive (which is why KenPom keeps them), so there's a debate about which view is more accurate:
share solo avgwins Upper95 Lower95 playoff
brown 0.002 0.002 4.538 7.61 1.47 0.259
columbia 0.120 0.046 8.575 11.70 5.45 0.259
cornell 0.000 0.000 3.469 6.40 0.53 0.259
dartmouth 0.000 0.000 4.642 7.65 1.64 0.259
harvard 0.454 0.271 10.339 13.13 7.55 0.259
penn 0.000 0.000 4.525 7.71 1.34 0.259
princeton 0.276 0.135 9.549 12.49 6.61 0.259
yale 0.466 0.287 10.363 13.18 7.54 0.259
Finally, I did some preliminary adjustments based on KenPom's summer work where he showed that 64% of PPP was based on a team's offense and 36% was based on an opponent's defense. So, in this final view, I adjust the offensive and defensive ratings used in the projections of every game to account for the offense having 64% of the control and defense 36%. Which is good news for Columbia fans:
share solo avgwins Upper95 Lower95 playoff
brown 0.047 0.018 6.076 9.59 2.56 0.274
columbia 0.312 0.170 8.277 11.73 4.83 0.274
cornell 0.035 0.011 5.527 9.04 2.02 0.274
dartmouth 0.046 0.017 5.813 9.38 2.24 0.274
harvard 0.257 0.143 7.957 11.54 4.37 0.274
penn 0.039 0.010 5.915 9.49 2.34 0.274
princeton 0.297 0.160 8.154 11.67 4.64 0.274
yale 0.325 0.197 8.281 11.79 4.77 0.274
Regardless of which system you look at, there are four teams with a shot at the title, and four teams playing to sneak into the upper division. The dividing line is bizarrely clear. It's almost impossible to get two groupings of four where each member of the group is this close to the other members.
|
Columbia Alum
Junior
Posts: 247
Age: 38
Reg: 11-15-11
|
01-03-16 11:44 AM - Post#198443
In response to mrjames
Very cool, appreciate the analysis, I do think the real odds are more uncertain like your last table, since these are Gaussian distributions based on performance so far but factors such as injured players, new injuries and improvements / deteriorations make basketball more uncertain. If Columbia returns a couple of its injured starters for most of league play, the top 4 will truly be wide open. This is shaping up to be as exciting and uncertain a season as I have ever seen it.
|
penn nation
Professor
Posts: 21193
Reg: 12-02-04
|
Re: Ivy Title Odds
01-03-16 12:00 PM - Post#198445
In response to mrjames
penn 0.003 0.001
So you're telling me there's a chance.... 
|
mrjames
Professor
Posts: 6062
Loc: Montclair, NJ
Reg: 11-21-04
|
Ivy Title Odds
01-04-16 10:32 AM - Post#198513
In response to Columbia Alum
Very cool, appreciate the analysis, I do think the real odds are more uncertain like your last table, since these are Gaussian distributions based on performance so far but factors such as injured players, new injuries and improvements / deteriorations make basketball more uncertain. If Columbia returns a couple of its injured starters for most of league play, the top 4 will truly be wide open. This is shaping up to be as exciting and uncertain a season as I have ever seen it.
That's an entirely fair point. We just saw what injuries did to Harvard. That being said, injuries are often relatively ignored because once a team can adjust, the impact is relatively minimal (just look at the Vegas odds), unless the team is completely exposed at the position (e.g., Harvard at PG). In most cases, relatively minimal changes don't make a difference. In the case of a four-team, incredibly tight race, though, a 2- or 3-point change to the spreads for every game could send a team from the top to clear fourth in the odds. So, this is the interesting case where injury uncertainty (or even an illness knocking a player out for a key weekend) could play an outsized role in picking the champ.
I do want to clarify one thing. The final table is not meant to be more even, but only ends up that way on accident. The KP Adjusted Odds table gives more credit to offense (which is considered more in control of PPP for the game) than defense. Six of the eight Ivy teams are in the bottom half offensively nationally, and the two contenders from among that group are downweighted because their defensive prowess isn't valued as much.
If the top Ivy contenders were way better offensively than the rest of the league (and the league's teams were closer on the defensive end), the KP Adjusted Odds table would actually be more unbalanced than the other tables.
|
JadwinGeorge
Senior
Posts: 357
Age: 75
Reg: 12-04-15
|
01-04-16 11:12 AM - Post#198515
In response to mrjames
Ivy League returns to #15 in most recent kenpom ratings among all D1 conferences. Tigers and Crimson essentially trade places with Princeton moving into top 100 and Harvard dropping out. This should be a wide open race, indeed.
|
whitakk
Masters Student
Posts: 523
Age: 32
Reg: 11-11-14
|
01-04-16 11:21 AM - Post#198519
In response to JadwinGeorge
And the league's absolute KenPom rating is now on pace to be the highest ever. Remarkable when considering how much turnover there's been this year -- only five of 14 All-Ivy-something players returned.
http://www.nycbuckets.com/2016/01/ivy-league-we ekl...
|
jadwin
Sophomore
Posts: 191
Age: 74
Reg: 01-14-15
|
01-04-16 11:46 AM - Post#198522
In response to whitakk
Based on pre-season play, the nyc buckets article does a good job of ranking teams in logical order based on their play in the pre-season. Harvard's play has been a pleasant surprise and Columbia's play been somewhat a surprise in the other direction although everything can change during Ivy League play.
Yale has been the most consistent team other than the Albany game, Sears illness plus jet lag. Princeton has been reasonably consistent as well. Harvard and Columbia appear to be more of the wild cards but time will tell how it all plays out.
As to the other four teams, can anyone surprise?
|
SRP
Postdoc
Posts: 4910
Reg: 02-04-06
|
01-04-16 07:03 PM - Post#198557
In response to jadwin
In honor of the return of Han Solo in the new Star Wars movie, let me be the first to say "Never tell me the odds!"
With that out of my system, those numbers seem close to intuition. Interesting that the Tigers have the highest chance of winning or sharing the title in the simulation where they have the lowest expected win total. Of course, that simulation also has the smallest expected-win gap with Yale and favorably changes the sign of the gap with Harvard.
|
