<
>

Field Notes, Vol. 2: Final toolbox assembly

3/14/2012

Still agonizing over your bracket? Field Notes, Vol. 2 (word to 2011's Vol. 1) is one college hoops writer’s attempt to guide you through the process as the Thursday deadline looms. Note: Said writer may or may not have a horrendous recent tourney history, which is why he’ll rely so much on advice from others in this series. Consider it a thinking fan’s guide to the bracket.

(And speaking of the bracket, be sure to join our College Basketball Nation Tournament Challenge group, wherein you can test your bracketing skills against yours truly and the rest of the ESPN.com college hoops writing staff. Imagine the bragging rights. So many bragging rights, you guys.)

Part 1 of the stats-based portion of the proceedings is available here. Now, in the final installment of the advanced lesson: How you can use tempo-free and other rankings systems to find value in your bracket -- and why you should.

As you know, conventional rankings like the AP and ESPN/USA Today coaches' poll can be deceiving. They're meant as snapshots of a season, but they're scattershot at best for anything else. (Murray State, for all its wins, is not the ninth-best team in the country. To name just one example.) But besides the KenPom rankings (which are best used as an in-depth stylistic scouting tool), there are other systems that rank and sort teams based on their statistical performances throughout the season. In no particular order, here are the three of the most important:

  • The BPI: Developed by aforementioned O.G. Dean Oliver (who now works as ESPN's head of analytics, which is pretty awesome) and his team, the Basketball Power Index is a ranking system that counts among its missions discovering which teams should be selected for the tournament, as opposed to the ones that are. It is an attempt to provide an advanced rankings system alternative to the RPI, which the NCAA still uses to calculate the measure of teams throughout the season, which is how you end up with Southern Miss as a No. 9 seed.

  • LRMC: Developed by engineers at Georgia Tech, the LRMC "is a college basketball ranking system designed to use only basic scoreboard data: which two teams played, whose court they played on, and what the margin of victory was. LRMC stands for 'Logistic Regression/Markov Chain', the two primary mathematical techniques that were a part of our system. We've subsequently added a Bayesian component as well." Does that last sentence make any sense to you? Me neither. I took a collegiate pre-calculus class pass-fail in college, enough to get me in the journalism school, and subsequently waved farewell to math forever. Best decision of my life. But we'll explain why you don't need to know what that means below.

  • Jeff Sagarin Ratings: Sagarin's ratings are based on the Elo rating system, a mathematical method originally "designed to determine the relative skill levels of players in two-player games such as chess." It also happens to do pretty well for basketball, too, and if you're familiar with college sports at all, you're probably already familiar with Jeff Sagarin.

There are even more complicated ways to break these ratings down. For example, SI's Luke Winn recently built a weighted ranking system comprised of all three of the above plus KenPom, which he used to rank a handful of tournament contenders. (News flash: Kentucky is really, really good.) You should also look at New York Times electoral analyst Nate Silver's methodology, which combines many of these statistical systems (and other factors) into one big list of tournament odds.

You don't have to go that far, of course. But you should take a gander at each of these rankings, because they're a great way to figure out which teams are likely overrated and underrated.

This is where you find value, and value is exactly you need.

Why finding value in the bracket matters so much. If you've played a tournament challenge game before, you already know: Because while at the end of the day, you want to get picks right, you also want to get picks right where other players are wrong. If you pick all four No. 1 seeds to go to the Final Four, there's a pretty good chance that if you're right, lots of other people will be right, too. If you pick all chalk in your bracket, you'll probably be in good shape. But you won't separate yourself from your competition, because a lot of people in your bracket will be doing the very same.

Again, this is simple stuff. So how do you avoid it?

Chris Wilson reprises a column about this very topic each year at Slate. As Wilson notes, the data available in Internet-based bracket games (like ESPN.com's Tournament Challenge) is incredibly valuable, because it gives you a great look at exactly how most people are viewing the bracket going forward:

Your overall strategy should be to look for situations where the national bracket values a team much higher than the objective statistics. For example, at the moment only 4 percent of all the participants in ESPN's Tournament Challenge have picked Ohio State to win the tournament—the right-most column on this table. Pomeroy's log5 analysis of the tournament, by contrast, gives the Buckeyes an 18.7 percent chance of winning it all, making them the second favorite in the field behind Kentucky. This makes Ohio State a fantastic bargain—while cold-blooded, numerical analysis gives Ohio State a roughly 1-in-5 shot at the title, only one in 25 people have picked them to win. As such, the Buckeyes are the most undervalued asset in the 2012 NCAA Tournament.

You also want to rate your upset risk carefully. Again, from Wilson:

Biostatistician Bradley Carlin, who co-authored a 2005 paper (PDF) on contrarian strategies in NCAA brackets, suggests a "champion-only" technique. While most people spend a lot of time puzzling over potential first-round upsets, the mathematical reality is that it's difficult to win a pool without securing those boffo championship game points. The payoff for risk-taking also increases in later rounds. Consider the first round game between 14-seed St. Bonaventure and three-seed Florida State. Less than 4 percent of ESPN players predict that the Bonnies will pull off the big upset, while Pomeroy gives St. Bonaventure a 33.8 percent chance of knocking off the Seminoles. On paper, that differential looks like a great bargain. But consider that this upset will reward the lucky St. Bonaventure backer with a mere one extra point in a standard office pool. If Florida State wins, you're suddenly missing an important player in the bracket.

This is how you build a bracket. You don't have to take a radical view. You don't have to bet against Kentucky, especially if you think the Wildcats -- which are rightly viewed as the best team in the country -- are going to win it all. But you can weigh the costs and benefits of each championship pick, and you can weigh the costs and benefits of each of those first-round upsets. Is a one-point win and a small chance to scream "I had the Bonnies!" with delight worth losing FSU if the Seminoles advance? Probably not.

Is it a lot of fun? Oh yeah. So don't dismiss it out of hand. But if you're picking your bracket with anything remotely close to a data-driven, probabilistic approach, it's important to identify value in teams that everyone isn't going to pick. It's a risk, sure. But it's a risk that could take you to the top.

A few other tips and tricks worth knowing:

  • Las Vegas lines. This is pretty obvious, of course, but it is useful, if only to get you past the psychological difficulty of picking your bracket with the seeds next to each name. Does Vegas have a particularly tight line in a No. 2 vs. No. 15 game? Why or why not?

  • Don't overrate recent performance. This tidbit comes from L. Jon Wertheim and Tobias J. Moskowitz, the authors of "Scorecasting: The Hidden Influences of How Sports Are Played and Games Are Won," and it's a good one: "Almost half of the last 48 Final Four teams won their conference championship, but don't be completely seduced by recent performance, both good and bad. Just as investors can get burned "chasing returns," don't put too much stock in how a team has played over the last few games (Unless there's an obvious factor: a major injury, a star player returning, etc.). The bigger the sample size, the better our ability to forecast performance."

  • Look hard at the No. 12 seeds. You've probably heard this before: The No. 12 seeds always seem to come through with upsets. But here's some information -- again from Wertheim and Moskowitz -- that backs up the general consensus: "Pick a 12 seed. Or pick all four 12 seeds if you're so inclined. For whatever reason, there's a real payoff here. Almost one-third of the time -- and in 19 of the last 22 tournaments -- at least one 12 seed has advanced to the second round. (Last year it was Richmond beating Vanderbilt.) What's more, historically, 12 seeds have beaten 5 seeds more often than 11 seeds have beaten 6 seeds. Nail this pick and it's a nice gain; miss it and your bracket isn't likely to combust."

  • Geography matters: From Nate Silver, former Baseball Prospectus editor, current New York Times Five Thirty Eight electoral maestro, and all-around statistical authority: "Since 2003, for instance, teams playing an N.C.A.A. tournament game within 50 miles of their campus are a remarkable 24-2. One of the two losses came in last year’s championship game, when Butler — playing just miles from its campus in Indianapolis — came within 2 points of defeating a heavily favored Duke team. By contrast, teams travelling at least 1,000 miles to play their game are 121-174, having won just 41 percent of the time." If you heard Gonzaga fans wail and gnash their teeth after hearing they, as a No. 7-seed, would be traveling across the country to play No. 10-seeded West Virginia in Pittsburgh, well, that's why. It's hard to win that far away from home. It doesn't guarantee defeat, but it's one more factor to take into consideration.

  • Check out the preseason poll. I know what you're thinking: "You said polls were garbage? What the --?!" Polls are garbage. Good listening skills, guys. But here's the thing: The preseason poll, believe it or not, actually has some value as it pertains to evaluating teams over the long haul. Click here to find out why.

If it seems silly to think that a late-October poll could help you figure out the tournament in early March, well, you're right. But the NCAA tournament is a fickle beast. We've come a long since the start of the season, and you never know what piece of data might one day to lead to bracket success. The above are a fantastic place to start. You take that knowledge in, you commit it to memory, you look at the bracket, you back up, and all of a sudden you're back in the fold, making picks, deciding games, but doing so with an expanded information toolbox. This is only ever a good thing.

Preparation is key -- in more ways than one. You're prepared, or as prepared as you're going to be. So here's another thing to get prepared for:

Failure.

Chances are, your bracket is going to stink. I hate to break it to you, but it's much easier to pick a really bad bracket than it is to pick a good one. Sometimes -- OK, all the time -- the NCAA tournament makes no sense whatsoever. Crazy things happen. See: 2011. Or 2010. Or any number of other seasons in which top teams fellow to putatively overmatched underdogs. It happens all the time, and there's no way to prepare for that.

The NCAA tournament defies logic. That's why we love it. That's why we try like hell to get it right every year, and that's why we keep coming back. The notion that the archetypal tournament dilettante -- the sweet 80-year-old woman filling out her bracket in a newspaper (a newspaper! can you imagine!) -- can outpick you is absolutely infuriating. Sometimes, and I tell my friends this all the time, it seems just as helpful to go in with absolutely no prior preparation at all.

You're welcome to take that tact if you like. But me? I'll take the knowledge every time. And if you read this far, I'm betting you will, too.