skip to main |
skip to sidebar
An exciting weekend, but not for this writer.
Weekends like this one are exciting to watch. Underdogs come up with big upsets, powerhouses are shocked, and endless buzz is created. It's a lot of fun. It's a little less fun, however, when you've trumpeted up your foolproof prediction method and get to watch it all fall to pieces. Let's review:
West Virginia 73, Duke 67. Duke was rated a full two and a half points higher in the SPI - high enough to earn them a spot in the finals in my bracket. West Virginia ruined my life again this year, knocking Duke out in the second round. They knocked off UConn a few years back in the second round as a 7 seed. I'm not too keen on those Mountaineers.
Stanford 82, Marquette 81. One point the other way and we aren't having this discussion. I went out of my way to talk about how overrated Stanford is and how underrated Marquette is. So much for that.
Realistically, those are the only two big upsets I had. I have 12 of the 16 remaining teams correct so far (only wrong on Washington State, Villanova, Stanford, and West Virginia), but the teams that I lost had a big impact. It's rough. It just goes to show you that there isn't a foolproof way to predict something as inherently chaotic as the NCAA basketball tournament. I'm still doing better than the wife (35-13 to her 33-15), but I'd be surprised if she doesn't end up outperforming me. Next year, when you see my article come up about my failsafe predictions for March Madness, cut straight to the chase and ask who my wife has winning it all. (She picked Tennessee, incidentally.)
AP Photo - Chuck Burton
Lots of upsets, and big ones, too.
This is the downside of going public with a bracket you think is really good - when it takes a nose dive, everyone finds out. There were four big upsets today, and my metric only got two of them right. And even that's debatable. Let's look at the numbers.
Western Kentucky 101, Drake 99. The SPI had #14 Western Kentucky (112.4489) narrowly beating #16 Drake (112.0300). It wasn't an easy choice. Still, it's hard to argue with a scoring differential of 14.8 over your last ten games, even if they come against Sun Belt competition. I had complete faith in the SPI for this one. (That means I spent the entire overtime glued to my computer, watching the score update every few seconds and screaming when Drake pulled ahead with seconds to play.)
San Diego 70, Connecticut 69. This one floored me. San Diego is one of the worst teams to make the field (104.4046, #59 SPI) and UConn is, while not top-tier, pretty solid (110.1566, #24 SPI). I thought this one was a no-brainer. And somehow, San Diego kept it close the whole time and pulled out a squeaker in overtime. Just goes to show you that no statistics-driven metric is perfect.
Siena 83, Vanderbilt 62. I liked Siena in this matchup and was so confident I didn't expect anything other than a Siena blowout. I got it, but a quick look at the SPI chart surprised me. Vanderbilt was ranked considerably higher than Siena in pretty much every category. Look at the numbers:
SPI Vanderbilt 108.0898 Siena 106.6667
SOS Vanderbilt .5644 Siena .5143
Scoring differential Vanderbilt 7.2 Siena 5.9
Not really sure why I was so confident there, but I picked the upset and got it. Weird.
Villanova 75, Clemson 69. Here's one where I was just dead wrong. Seriously. How could I have seen #25 Clemson (109.8635) losing to #58 Villanova (104.6078)? The mind boggles. I just don't get this one.
Overall, the SPI performed significantly worse today, only going 9-7 for a total of 24-8. Not a good day. For comparison, my wife's bracket is 25-9. Also, she wishes to inform the loyal readers of the Worb that her bracket was not solely determined by name recognition, but by an elaborate process consisting of looking at both the team's seed and their win-loss record. I thought you should know out of duty to accuracy (which means I had to tell you on penalty of no sex until I'm 75).
Only one significant upset on the first day of the tournament.
There were some nail-biters, but my bracket is nearly perfect after day one of play. Here's a rundown of the three misfires I had:
UNLV 71, Kent State 58. The SPI had this one right, actually. #41 UNLV rates 107.2224 compared to #43 Kent State's 106.7044. When the scores are that close, though, I look to the last ten games, and Kent State had a much higher scoring differential. I yielded, and I picked the upset. I paid for it, too.
Pittsburgh 82, Oral Roberts 63. Same problem here. #37 Pittsburgh rates 107.7219 to #39 Oral Roberts' 107.3852. That's even closer than the first matchup. Oral Roberts had nearly eight points higher in the scoring differential category, though, so I went with the big upset. Now I just have to hope Pitt doesn't go on a huge run. They shouldn't, though, since Michigan State (their next opponent) has an SPI of over 109.
Texas A&M 67, BYU 62. Anyone watching this game knows that it was close. #17 BYU had a much better SPI (111.7292 compared to 109.0498), but I forgot to apply the smell test here. BYU plays in the Mountain West Conference. Their biggest win was over Louisville, yes, but it was a team riddled with injuries. I should have known better. It's just hard to pick against your alma mater.
So through the first day, the SPI is 15-1. My bracket is 13-3, but that's just because I'm an idiot and ignored my own system. For the record, my wife, who more of less filled out her bracket on name recognition (I love you, honey), is standing at 14-2, only missing BYU-Texas A&M and USC-Kansas State. That's a little humbling.
Mount St. Mary's 69, Coppin State 60.
I promised regular updates, and I'm here to provide the first. The play-in game was tonight, and on Monday I said that I expected Mount St. Mary's to win over Coppin State. The score is in, and the Mountaineers won by nine points. It was pretty close for most of the contest, but Mount St. Mary's pulled away in the clutch. Aficionados of the SPI will remember that the Mountaineers have a significantly higher rating than the Eagles. Mount St. Mary's boasts an SPI rating of 103.8015 compared to Coppin State's 97.3570. A difference of six and a half points is not to be ignored. Thus far, things are looking good.
For the record, Mount St. Mary's next opponent, North Carolina, has an SPI rating of 117.2178. If a six and a half point difference is huge, then a thirteen and a half point difference is even more so. Stay tuned.
Retooling the ratings metric of the future.
Since I last wrote, I've done a substantial retooling of the Sam Power Index (SPI). The previous metric was essentially a product of the team's RPI rating, their strength of schedule, and their scoring differential. 
I realized, however, that this isn't the best way to go about it. For one thing, the RPI is a metric determined by a team's strength of schedule. Including that factor twice doesn't make much sense. Second, I realized that I was unfairly punishing teams that did poorly earlier in the season, but came alive down the stretch. A team being outscored by an average of seven points throughout the season but winning their last ten games by an average of fifteen points will be a lot more dangerous than their original record might indicate.
So I overhauled the metric, and my Excel spreadsheet now has close to four times as much data on it. I weighted in each team's strength of schedule and scoring differential for the last ten games, and I found some remarkable differences in the rankings. Even better, now that the official seedings have been released, I can compare what the NCAA selection committee and the SPI came up with. Here are the major differences:
Davidson. The Wildcats went 26-6 this year, winning their last 22 games. They're outscoring their opponents by over fifteen points a game, and outscored their last ten opponents by nearly twenty points a game. Yet the NCAA dropped them to a 10 seed for playing in the fairly soft Southern Conference and for not finishing ambitious nonconference opponents like Duke. The SPI has them listed at #4. This is one of the teams that is going to surprise a lot of people.
Georgetown. They rolled through the incredibly deep Big East conference this year, only losing to Pittsburgh in the conference final. They have a scoring differential of 11.5 points and a record of 25-4, earning them the 2 seed in the Midwest region. The SPI puts them at #20 overall, however, since they slowed down in the stretch. Their scoring differential fell to 5.6 points during their last ten games against a schedule that was only marginally more difficult than the rest of their season. The Hoyas aren't a bad team, but they aren't the dominant force everyone thinks they are. They aren't even the top-rated team from the Big East - Marquette (10), Louisville (11), and West Virginia (15) all finished higher.
Western Kentucky. Remember how I said this was the team to watch last time? Their SPI actually rose when I retooled it. They're rated at #14, which, for comparison, is only just below Texas. This is a team that beat their (admittedly weak) last ten opponents by nearly 15 points a game. That's pretty impressive, even for the Sun Belt Conference. Their first two opponents are lower in the SPI (#16 Drake and #24 Connecticut), so don't be surprised if you see the 12th-seeded Hilltoppers in the Sweet Sixteen.
Marquette. Here's a team that hasn't generated much attention this year. The Golden Eagles play in the insanely talented Big East, so they've been overshadowed by teams like Georgetown, Notre Dame, and Louisville all year. Yet they have close to the top scoring differential in the conference at 12.7 on the season. That fell to 11.5 in their last ten games, but that was higher than anyone else in the conference came up with. Even though the NCAA stuck them with a 6 seed, they have the advantage of playing in a soft region. They draw #60 Kentucky (an 11 seed) in the first round and overrated Stanford in the second (3 seed, #27 SPI). They have the guns to get past #13 Texas and even #3 Memphis, I think. Marquette's SPI is nearly three and a half points lower than Memphis', but the benefits of playing in the Big East will have them ready for such a big game. Marquette will be the surprise of the tournament.
I'm ready to predict the play-in game, for those of you looking for proof that this metric will work. The opening round game pits 16a Mount Saint Mary's against 16b Coppin State. These games are usually a toss-up when filling out brackets, but I feel pretty comfortable going to Mount Saint Mary's. Coppin State has had an amazing comeback from a 4-19 start (they finished the season on a 15-1 run) and had a 9.2 scoring differential in their last ten games, but their SPI is only 97.3570 compared to Mount Saint Mary's 103.8015. The fact is that Mount Saint Mary's has been playing all season at the level Coppin State has over the last ten games, and against similar competition. Go ahead and fill that in your bracket in pen.
I'll be posting regular updates as the tournament progresses. Feel free be amazed when I correctly pick all 64 games or mock me mercilessly when I get them all wrong.
Dissecting America's most involved sporting event.
On Sunday, March 16, the NCAA will announce the field of 65 for the NCAA Division I Men's Basketball Tournament. You and I know it as March Madness. More worktime productivity is lost to this event than any other in the world of sports. Brackets are filled out, office pools formed, and millions of dollars are won and lost in bets. Teams like the George Mason Patriots pull off improbable upsets, delighting some fans and infuriating others. An incredible 32 games are played in the first two days of the tournament. It's insanity. It's chaotic. It's as close to perfection as we can get in sports.
In preparing to fill out my own bracket, I've been analyzing the strengths and weaknesses of teams likely to make it into the tournament. There are a number of statistics one can look at to get a sense of the strength of a team. The most readily available one is the Associated Press poll. Sportswriters across the nation rank the top 25 teams, and the final point totals are posted in the official poll. The rankings are used when publicizing a game (a recent game between #1 Memphis and #2 Tennessee comes to mind), but are in essence subjective. There aren't any solid statistics or data that go into the rankings. For a more precise measurement, some turn to the Ratings Percentage Index, which is calculated based on wins and strength of schedule. The RPI is used by the NCAA selection committee in determining which teams should be given at-large berths in the tournament. This gives us a better look at which teams are more likely to do well during March.
Yet this still fails to take into account a statistic I think is a better indicator of future performance than wins and losses - scoring differential. Simply put, this is a statistic that shows, on average, how much a given team is winning its games by. A team winning games by an average of ten points should be better than a team winning by only five points. However, you can't stake everything on scoring differential. There's a big difference between beating Duke by twenty points and beating 0-29 New Jersey Institute of Technology by that same margin. To account for that, I added both the team's RPI and their strength of schedule into the mix to arrive at a statistic I called (for lack of a better name) the Sam Power Index (SPI).
The SPI holds up pretty well with the projected seeding I got from ESPN's Joe Lunardi. Take a sample matchup between West Virginia and Baylor. Lunardi has the Mountaineers taking the higher seed, even though they've only won one more game than Baylor. A look at the SPI shows why: West Virginia has a scoring differential of 13.1, compared to Baylor's 7.4. This gives them an SPI of 4.279, nearly two points above Baylor. The seedings bear this out, putting West Virginia on top. (The Mountaineers actually have the 12th highest SPI, which would likely put them in Elite Eight territory, were it not for third-ranked North Carolina looming in the second round.)
I'm not going to claim that the SPI is perfect. Far from it, actually. According to the SPI, Ohio State should have won last year's finals. Comfortably. Ohio State had an SPI of 7.801, which is significantly higher than any teams's SPI this year. Florida had a paltry 6.406. A gap of over a point is cause for alarm. Yet Florida won by nine points. That's why I include a smell test. Occasionally, there will be predictions of one team beating another that don't seem quite right. An example is a projected third round matchup of 26-5 Xavier and 25-4 Georgetown. Xavier has the higher SPI at 5.119, nearly a full point above Georgetown. Seems like an easy win for the Musketeers. Yet Xavier has faltered down the stretch, while Georgetown has remained constant and won the Big East, the country's deepest conference. Even though the math has Xavier winning, I'm going to trust my gut and pick Georgetown. (And I'm still picking them to lose to North Carolina.)
Of course, anyone reading this column will be wondering who I've picked to win it all. I'm not going to spoil my advantage by giving that away, but I will give you my Final Four: Kansas, North Carolina, Duke, and UCLA. And lest you worry that I'm trying to dodge accountability by not posting my predictions publicly, rest assured that I'll provide updates throughout the tournament. I will say this, though: if you're looking for an upset pick, look no further than Western Kentucky. They're projected to land a 13 seed, but they have an SPI of 3.410. That's better than one of the 3 seeds. Look for them to do some damage this year.
Photo from Associated Press
