Saturday, March 20, 2010

Prediction System Bracket Challenge: Rd 1 Update

As mentioned previously, I set up a group on ESPN’s Tournament Challenge to track how the bracket predictions from about a dozen different ratings/prediction system are doing.  Here’s a quick review of where everything stands after Friday.

image

As you can see, everybody picked Kansas except for Yet Another Basketball Blog’s cupcakes-removed efficiency ratings.  Some of you might be saying “Wait!  I thought Duke was #1 in Pomeroy!”  Well, you’re right.  But they weren’t on Wednesday evening, when I filled out the brackets - Kansas was in front by about 0.001.  I’m pretty sure that’s the only day in the last month where that was true, but this is a good example of the fact that rankings and ratings are considerably different.  Pomeroy’s ratings say that Kansas and Duke are essentially equal, and both are considerably ahead of the pack.  If you only look at the rankings, all you see (today) is that the top three are Duke, Kansas, and Wisconsin.  Quite a difference.

At any rate, the biggest surprises in the above table have to be that the RPI is tied for first, while the bracket based on Vegas Watch’s attempt to mimic the sportsbook markets is tied for last.  However, look at the PPR (Possible Points Remaining) – Vegas Watch is actually tied for first.  That’s kind of the whole point of his approach – a lot of these first round games are tossups, and it’s just luck whether you pick them right, which is why you need to keep in mind the overall odds of a team advancing to a certain round, instead of merely looking at their most likely matchup at that point.

The fact that Team Rankings.com and two of Sagarin’s ratings (and specifically NOT the Elo rating) are tied at the top, is definitely not a surprise, as these guys plus Pomeroy are what I’d consider the Big Three of college basketball ratings.

I’m happy to report that my Similarity Scores system is tied with Pomeroy (which is the foundation for the similarity analysis) - though we’ve gotten there via differing picks, and Pomeroy has one more Sweet 16 team alive.

My comrade over at UFR isn’t going to be quite as pleased, as the Martin Manley rankings are only 1 game from last, and have the fewest potential future points.

In the grand scheme of things, the first round isn’t too important, so I’ll be more curious to see how things stand after the whole weekend is over.  I anticipate a comeback by Vegas Watch – and of course I’m rooting for my Similarity Scores to pass Pomeroy.

Similarity Predictions: 2nd Round (Sunday)

I'll start with the prediction chart for all of Sunday's games, then get to the previews for the Big 12 games, which are also cross posted over at Upon Further Review.  If you are unfamiliar with my similarity predictions, read about the methods here.  Also, sorry about the ugly tables - I tried a new method of posting them, and it didn't work so well.  You live, you learn.


SIMILARITY vs. POMEROY
GAME
POMEROY
SIMILARITY
Syracuse
-
Gonzaga
Syr +8.5 (78%)
Syr +11 (84%)
Ohio St
-
Georgia Tech
OSU +4.5 (68%)
OSU +7.5 (77%)
Maryland
-
Michigan St
MD +3.5 (64%)
MD +9 (81%)
West Virginia
-
Missouri
WVU +2.5 (60%)
MU +3 (62%)
Wisconsin
-
Cornell
Wisc +8.5 (82%)
Corn +2.5 (61%)
Pittsburgh
-
Xavier
Xav +1.5 (56%)
Xav +2 (57%)
Purdue
-
Texas A&M
Pur +0.5 (51%)
Pur +4.5 (69%)
Purdue w/o Hummel
-
Texas A&M
Pur +0.5 (51%)
A&M +3.5 (79%)
Duke
-
California
Duke +8 (78%)
Duke +9.5 (82%)

Friday, March 19, 2010

Similarity Predictions: 2nd Round (Saturday)

Just a quick note on which games my similarity system differs from the straight Pomeroy predictions.  Let’s start with the summary chart of all 8 games, and then I’ll include notes on a couple of them.

SIMILARITY vs. POMEROY (diff. of at least 10%)

GAME

POMEROY

SIMILARITY

Villanova
-
St. Mary's
Nova +4 (64%)
Nova +2 (57%)
Butler
-
Murray St
But +3.5 (65%)
But +5.5 (71%)
Tennessee
-
Ohio
Tenn +8 (79%)
Tenn +5 (70%)
Kansas
-
Norther Iowa
KU +8.5 (84%)
KU +7 (78%)
Baylor
-
Old Dominion
Bay +3.5 (66%)
Bay +4 (67%)
New Mexico
-
Washington
Wash +2.5 (59%)
NM +0.5 (52%)
Kansas St
-
BYU
BYU +0.5 (52%)
KSU +7.5 (74%)
Kentucky
-
Wake Forest
UK +7.5 (78%)
UK +6 (72%)

Most of these are pretty close to Pomeroy, with the general trend being that the games should be a little closer than expected.  I covered Kansas State vs. BYU in detail over at UFR, so you can head over there for analysis.  The New Mexico-Washington game has a different winner with my system, and that prediction is mostly due to the fact that Washington has underperformed on offense when facing teams like New Mexico, while the reverse is true for the Lobos.

Similarity Predictions: Saturday Big 12 Games

I’ve got a new post up over at UFR that breaks down Saturday’s Big 12 games from a similarity scores perspective.  Check it out!

Thursday, March 18, 2010

Final Four Characteristics

Just a short post to share something I noticed when digging through data last night…

In the Pomeroy Era (2004-09), there have been 24 Final Four teams.  All of them except 2006 George Mason and 2006 LSU fit the following critera:

  • 30th or better in Adjusted Offensive Efficiency
  • 30th or better in Adjusted Defensive Efficiency
  • 15th or better in Pythagorean Rating
  • 150th or better in defensive eFG%
  • 150th or better in offensive 2PFG%
  • 150th or better in defensive 2PFG%

Here are the teams in each region that fit these criteria:

  • EAST: Kentucky, West Virginia, Wisconsin
  • WEST: Syracuse, Kansas State, BYU
  • MIDWEST: Kansas, Ohio State
  • SOUTH:

Yes, I left the South blank on purpose.  Duke is disqualified due to their terrible shooting percentage within the arc (46.9%, 205th nationally).  However, they make up for this by being a fantastic offensive rebounding team (40%, 10th nationally), so given the lack of alternatives, I’d slot them in as the best fit in the South by these criteria.

Wednesday, March 17, 2010

Prediction System Bracket Challenge

As a fun (and non-statistically significant) way to pit my similarity-based predictions against other prediction/rating systems, I created an ESPN Bracket Challenge group – System Smackdown.  I filled out brackets based on the ratings listed below – for each game, I simply picked the higher rated team, unless otherwise noted.  If you know of any more ratings or predictions that should be included, drop me a line.

The Competitors

Hopefully, I’ll have time to post a score update after every round.  If not, I should at least be able to manage every weekend.

I should note that I didn’t realize – until after I took the time to fill out all the brackets – that I can’t change the scoring system, so I’m stuck with the default 1-2-4-8-16-32.  That overvalues the later rounds, in my opinion, so I’ll probably end up showing what the standings would have been with various other point systems.

Good luck to all the (unbeknownst to them) competitors, and may the best system win!

Similarity Previews: 1st Round

[NOTE: The Big 12 games from below are cross posted over at Upon Further Review, as well as some extra analysis on two other Big 12 games that fall below the threshold for inclusion here - Texas-Wake and Baylor-Sam Houston St.]

If you saw the similarity previews for the KC-area teams, you might have noticed that the predictions created by my similarity-score-drive system were quite similar to the straight Pomeroy predictions.  Naturally, games where these systems disagree are the more interesting ones – the testing ground for the similarity prediction system.  Luckily, there are about a dozen 1st round games where the difference between the two is at least 10%.  I’ll take a look closer look at those here.

#3 New Mexico Vs. #14 Montana

image

There’s a lot of green up there on New Mexico’s chart – they’ve not only been good, they’ve been consistently good against teams like Montana, on both offense and defense – about 3 points per possession (PPP) better in each. On offense, it’s due in large part to limiting turnovers.  Montana (and their comps) force TO’s at a below-average rate, while New Mexico has the 15th-lowest TO% in the nation.  On defense, meanwhile, New Mexico is ranked 5th in Reb%, while Montana and friends struggle on the boards.  Put that all together, and you get an easier-than-expected win for New Mexico:

image

Vegas is about halfway in between these predictions, at New Mexico –9 as I type this.  Here’s one case where the difference between standard Pomeroy and similarity-based predictions becomes a financial matter.

Monday, March 15, 2010

Similarity Previews: KC-Area Teams

[NOTE: Some of this is cross posted over at Upon Further Review, a great KC-area sports blog.]

With the success of the similarity predictions so far (WARNING: extremely small sample size):

image

I figured I’d continue down this path, at least until I start to see some bad results.  I’ll be doing a comprehensive preview of the first round games that highlights the instances where the similarity prediction dovetails from the standard Pomeroy prediction, but first I’m going to tackle the match ups involving teams that are of interest to UFR readers.  Any system that pegs the KU and KSU games as anything other than huge mismatches probably involves throwing things off the Empire State Building, so I won’t spend much time on those games.  The more interesting game is Missouri-Clemson, so that’s the one I’ll start with, and focus on.


Upset Warning: Extra Possessions

I mentioned a couple times during my Tickets Punched series that rebounding and turnovers are key stats to look at for first round underdogs.  I wasn’t just pulling this assertion out of thin air, and I’ll lay out what evidence I have here.

Let’s start with the theory.  Teams in the first round of the NCAA tournament are often nervous, and I would expect that underdogs who usually win by outshooting their opponent might perform poorly due to nerves.  But teams that can overcome poor shooting by getting themselves extra possessions (via turnovers and rebounding) should do well.  And as a bonus, the other team’s nerves might assist them in forcing turnovers.

Last year, to test this idea, I looked at the performance of seeds 11 through 13 in the 2004-2009 tournaments.  Splitting the teams up based on their seasonal TO% margin ([def TO%] - [off TO%]) and Reb% margin ([Off Reb%] - [opp Off Reb%]), I found that teams who were positive in both were far more likely to pull upsets.  Picking teams who fit this description did an excellent job of predicting upsets last year.  Unfortunately, I lost that research in a hard drive crash, so I did something similar tonight, just so I have some kind of evidence to post.

This time, I used raw rebounding margin and turnover margin, because that was already in a database (from Bracket Science) that included NCAA seed and win information for the last 4 tournaments (2006-09).  Once again, I looked at seeds 11 to 13.  Overall, they won 17 of 48 games (35%).  There were 28 teams that had positive rebounding and turnover margins, and they won 14 games (50%), leaving 3 wins for the other 20 teams (15%).  This time I noticed one more criteria that seemed to split the teams, and that was RPI SOS.  Narrowing it down to only teams with a SOS of 150th or better, the upset tally improves to 12 of 17 (71%), which means other teams are 5 of 31 (16%).  Here’s a summary chart, for those who like numbers more than words:

image

And here is a list of all the teams from the past 4 years that met the criteria, as well as how they did in the tourney:

image

I’m not suggesting you pick all 6 of the 2010 teams to make it out of the first round… but I am suggesting, you shouldn’t be surprised when one of them does.

When your bracket falls apart after taking my advice, feel free to email me with angry screeds!

Sunday, March 14, 2010

Tickets Punched: #27 - #31

This last post in the Tickets Punched series gives you the offensive and defensive comps for today’s four championship games, plus one that I somehow overlooked till now.

OHIO


image

The only one of these past comps to make the tournament was 2009 Mississippi State, and their offense was shut down in the first round by Washington.  The conventional wisdom is that teams who are weak inside (below average OReb% and 2P%) tend to do poorly when push comes to shove.  If true, that wouldn’t bode well for Ohio.


Similarity Predictions: 3/14

These predictions were 3 for 3 yesterday, nailing Washington over Cal.  Let’s keep going with today’s Big 6 games.

Tickets Punched: #18 - #26

This series is turning out to be more writing than I have time for, so most of today’s teams will be posted without comments.  The charts are all still here, though, and those are really the important part.