Four facts about the tourney field

Now that the final 64 teams have been set for the dance, I looked at the averages for the tourney field. Four things stood out:

  1. The average coach has been to the Elite Eight 1.52 times. That’s the highest in the 29 years of the 64-team era.
  2. The average team has come into the tourney with a 2.44-game winning streak. That’s the highest since 2002. This number is greatly influenced by the long streaks of Wichita State (34), Stephen F. Austin (28), Florida (26) and North Carolina Central (20).
  3. This is the highest scoring tourney field (74.4ppg) since 2003. So the NCAA got their wish when they changed the rules to favor offenses.
  4. This field hasa margin percentage (average margin/points allowed) of 12.6%. That’s the lowest since 2004, which means the field plays tighter games.
Posted in Tourney Trends | 1 Comment

Cheat sheet for making last-minute toss-up game decisions

The toughest games to pick in the tourney are what I call toss-ups. Toss-up games pit teams within three seed positions against each other. About one-third of all tourney games in the 64-team era (615 of 1827, 34%) involve toss-up pairings. The higher seed tends to prevail in these match-ups, compiling a 337-257 record for a 57% success rate (21 games involved same-seeded teams).

That’s not exactly a percentage you want to take to the bank. So if seeding can’t provide solid guidance on who will prevail in a toss-up game, are there any other individual stats that can? Does having a more experienced coach increase a team’s likelihood of winning a toss-up game? What about being higher scoring…or having a tougher schedule…or having better momentum?

We examined each of the 615 match-ups across 20 statistics to identify the advantages that pointed to the largest separation between winners and losers. If you could only crutch on one chart to make your toss-up picks, this would be it; here are the round-by-round records of teams with certain advantages over their opponent:

13-14_tossup_standard

Posted in Bracket Tools | Leave a comment

FINAL MODELS AVAILABLE!

I just posted the final 16 models, including my new Keeper Small Pool and Billion Dollar brackets.

There has been one HUGE CHANGE overnight based on the results of last night’s Tennessee win. Louisville leapfrogged Arizona in KenPom Pythag efficiency ratings. It has affected the outcome of model #1, but none of the other models that factor in KenPom numbers.

The final models are under the TIPS+ section and they’re called “2014 Models FINAL.”

Posted in General News | 34 Comments

Cranking up the Madometer to measure the 2014 tourney madness

Veteran members are no doubt familiar with the graphic below, but unless newcomers have dug deep into the blog, they won’t recognize the Madometer gauge at the top of the blog.

The Madometer is a simple metric I devised to measure the madness. It works by calculating the seed-position differences between actual winners and perfect high-seed success or failure throughout the dance. If the higher seed advanced in all 63 games (perfect sanity), the cumulative seed value of the winners would be 203. If the lower seed always advanced (sheer madness), their cumulative seed value would be 868. The difference between the two—665—is the predictability range.

Let’s take a closer look at the 2013 dance, the craziest in the 29 years of the 64-team era. If you added up the seed positions of all the teams that advanced through the tourney, the number would come to 341, certainly closer to perfect seed dominance (203 positions), but still 138 positions toward madness along the 665-point predictability range. That works out to a Madometer reading of 20.8%. To put that number in context, the average tournament in the modern era has deviated from by-the-seed results by 14.3%.

The 20.8% Madometer reading makes 2013 the most unpredictable tourney since the field expanded to 64 teams. Before last year, the craziest dance occurred in 2011, when tourney advancers deviated from perfect high-seed dominance by 19.8%. You have to go all the way back to 1986 for third most unpredictable tournament, when the Madometer hit 18.8%.

You could pass it off as coincidence that the two wildest dances have occurred in the last three years. But the fact is, we’ve seen four straight dances that were well above average madness. 2010 and 2012 saw 17.1% Madometer readings, tying them for the eighth craziest tournament out of 29. Take a look at how the last four years stack up against the other 25 dances on the Madometer:

13-14madometer

Once the tourney tips off on Thursday, I’ll be posting the Madometer gauge at the end of every evening. Generally speaking, the Madometer shoots up in the first couple rounds, then settles down. Unless this dance is like last year’s craziness. Can’t wait to see.

Posted in Measuring Madness | 3 Comments

Which region is most likely to blow up?

Yesterday, I attempted to identify which past dance the 2014 tourney most resembled. Based on the seed quality curves, it looked like 2006 and 2011 were the closest comparisons. Both of those dances were particularly mad. 2006 featured the George Mason Cinderella run with no top seeds in the Final Four. 2011 was 19.8% crazy on the Madometer scale and tied the record for upsets with 13. That should scare everyone.

Today, I want offer some insight into which of the four regions is likely to blow up like last year’s West region. That was the quadrant where Wichita State upset Gonzaga on their run to the Final Four. It included five of last year’s 11 upsets, including the 3v14 Harvard win over New Mexico, the 4v13 La Salle win over Kansas State, the 5v12 Ole Miss victory over Wisconsin, along with the Shockers 1v9 and 2v9 surprises.

I took the Pythag values of the top 14 seeds for every region in the 2014 tourney, then I overlaid the same numbers for the 2013 Shocker bracket. Here’s a handy animation, cycling through the curves every five seconds:

2014_Regional_Curves

First, I’ll make a few observations about each region, then I’ll compare them to the havoc that was the 2013 West:

2014 South: The top seven seeds are slotted in fairly orderly fashion. Five seed VCU is a tick better than UCLA, but nothing is really glaring here until the whipsaw that is #8 Colorado versus #9 Pittsburgh. After that, the 10-14 seeds descend in keeping with their efficiency numbers. Maybe that’s why I’m having trouble pulling the trigger on any 4v13, 5v12 of 6v11 upsets in this region. I’ll say this: you do have the second weakest four seed and the nearly the strongest 13 seed (Manhattan’s Pythag is .0003 better) in the UCLA/Tulsa match-up. And I might consider a 1v8 upset, given Pitt’s unusually high Pythag. But then again, it’s Pitt. And it’s Jamie Dixon. And I said I would never get burned by the Panthers again.

2014 East: This region features the second weakest one seed and the second strongest four seed. No wonder people are leaning toward Sparty in a potential 1v4 match-up. Villanova rates out as the toughest two seed, and they could be on a collision course with the weakest three seed in Iowa State. UConn is the toughest of the four seven seeds, so they might be an intriguing second-round upset pick. As for the 5v12 and 6v11 match-ups, unlike the South, this region may be ripe for surprises. While Cincy and North Carolina are the second strongest five and six seeds, Providence is the second toughest 11 and Harvard is far and away the best 12. Looking for a big shocker? North Carolina Central is easily the toughest 14 seed and ISU is the weakest three. Hmm…

2014 West: With the strongest one seed and the second weakest two in Wisconsin, it’s no wonder people can’t figure out anyone else to advance here but Arizona. On top of that, you have the weakest four and second weakest five in San Diego State and Oklahoma. Trouble might loom on the other side of this region, where Creighton is just a hair behind Duke for the mantle of best three seed. And the Badgers’ weakness as a two may open the doors for Baylor or Oregon. They aren’t the strongest six and seven seeds, but they’re right in the mix. The unsual strength in this region comes at the eight and nine positions. Both Oklahoma State and Gonzaga look to be formidable second-round opponents for Arizona. Maybe that—and Marcus Smart—is why so many people are wondering about a Cowboy shocker over the Wildcats.

2014 Midwest: The weakest one, the weakest two, the strongest three and the strongest four. That just begins to explain the craziness that is the Midwest region. The majority of pundits are tabbing Louisville to beat Wichita State and reach the Final Four. And why not? Look at the gigantic disparity in numbers. We’re talking about the second most efficient team in the country. But that might not be the only craziness to ensue in this region. First of all, Kentucky’s the best eight…so a second-round match-up with the Wildcats is no cake-walk for the Shockers. And look at how pathetic the five through seven seeds are. Yes, NC State is a weak 12…but they looked pretty dang good last night. And Tennessee is off the charts are an 11. The 7v10 game looks like a toss-up too. In the second round, should the Vols advance, they could pose problems for either Duke or Michigan, two teams that don’t defend well. Tennessee’s offensive efficiency isn’t great (29th in the country), but they’re the 13th best defense. Could be scary for Blue Devil and Wolverine fans.

Comparing the regions to the 2013 West Shocker: What distinguished last year’s West explosion was a set of great 11-13 seeds and better than average 8-10 seeds. No single region can claim the same make-up in 2014. The East may have the best 11-12 pairing in Providence and Harvard and it also features a weaker one and soft three seed. The Midwest has decent 11-12 seeds and woeful fives and sixes. Then there’s the whole “one-seed-in-disguise” factor with Louisville and the unusual strength of Kentucky. The other region that could blow up, not so much in round one round two, is the West. I think Wisconsin is a weaker than average two seed, with a solid three in Creighton and intriguing six and seven seeds in Baylor and BYU.

Posted in Bracket Tools, Measuring Madness, Tourney Trends | 16 Comments

MAILBAG: Answering some key reader questions…

Running a website about tourney bracket prediction is not a good business model. About two-thirds of my sign-ups come in a four-day period, from Selection Sunday to Wednesday. And with the sign-ups come a ton of questions and comments. I have to admit that I can’t keep up with all the comments on the blogposts, but I’m happy that so many Insiders are picking up the slack with great dialogue.

As for the e-mails, I do try to respond to every one—even the guy who asked me to fill out his billion-dollar bracket for him (just a tad busy for that level of service). Some of the e-mails are topics that everyone should know about. Here are some of the questions that have rolled in over the last couple days?

A BUNCH OF READERS: When are you posting your Keeper bracket?

I will post two Keeper brackets this year—one for small pools and another for “shoot-the-moon” contests. (Gimme that billion dollars!) They’ll probably come later tonight, after the results of the play-in games. That Tennesse/Iowa tilt affects a number of the bracket models. In fact, when I post the next wave, that will be the final PDF you’ll get on the models. NOTE: If you haven’t seen the message late last night on an important Outcome Matching model change, make sure you’re up to speed. Duke and Michigan fans will not be happy.

DON N: KenPom adjusted his rankings after last night play-in games.  I assume he will again after today’s games. Will you produce a revised chart for his bracket?

Yes. I will revisit the KenPom bracket model and update it in case the reordering has changed picks. However, I’m not going to readjust the Excel stats sheet. You’ll have to do that on your own after tonight’s games. That said, when I load the data into my main database, I will use KenPom’s most recent numbers before the “true tourney” tips off tomorrow. It just makes sense to use the latest numbers in my analysis.

DENNIS: Where can I find your bracket models?

In the old days, I used to spend more time on fancy graphics for the home page, which made the location of the models butt simple. Not so much anymore. They’re under the TIPS+ section. Look for the link that starts “2014 Models.”

ANDREW: Where can I find your Pulse Check of the top 13 teams?

One of the problems with the WordPress blog application is that it shoves old posts onto other pages than the home page. I’ve had a couple of requests for the final Pulse Check. So here’s the link.

COMISHKAUF: Is there a glossary for the stats spreadsheet?

I posted on this last year, and should’ve pointed you to it. Here’s what the columns in the stats sheet mean.

Posted in General News | 20 Comments

Pretender candidates at every seed position

A few members have asked for me to identify teams that fit the Pretender criteria in the “Contender|Pretender” feature piece. I thought I saw that someone had done this in a blog comment, but couldn’t find it through the forest of recent posts. So, while watching T.J. Warren dismantle Xavier, I assessed the teams against the Pretender traits–and here’s who history says could fall short of seed expectations. (I’ve included the first characteristic where the team failed; they may have failed others too.)

1 SEEDS

  • Virginia (Inexperienced coach, no All-Americans)
  • Wichita State (SOS)

2 SEEDS

  • Kansas (W%)
  • Michigan (Guard scoring >77%)
  • Wisconsin (<51% Assists/FG)

3 SEEDS

  • Syracuse (<6 wins in last 10)

4 SEEDS

  • San Diego State (3/FGA <27%)
  • Michigan State A/FG > 60%

5 SEEDS

  • Saint Louis (senior laden starters)
  • Oklahoma (weak Pythag)
  • VCU (OE <1.09)
  • Cincinnati (OE <1.09)

6 SEEDS

  • Baylor (weak Pythag)
  • UMass (weak Pythag)

7 SEEDS

  • None

8 SEEDS

  • Colorado (A/TO <1)
  • Kentucky (A/TO <1)

9 SEEDS

  • Kansas State (met all three criteria)

10 SEEDS

  • All have historically weak Pythags (but bear in mind, that I’ve yet to put the adjusted KenPom numbers into my database). BYU and St. Joe’s have the lowest values.

11 SEEDS

  • Iowa (snake-bit coach, tourney novices)
  • Nebraska (<.640 W%)
  • Tennessee (<73ppg)
  • Providence (margin<5ppg)

12 SEEDS

  • Only Harvard meets overachiever criteria

13 SEEDS

  • None of these teams meets overachiever criteria

14 SEEDS

  • None of these teams meets overachiever criteria

15 SEEDS

  • None of these teams have the numbers to spring an improbable upset.
Posted in General News | 9 Comments

ALERT! Major change to Outcome Matching #3 and two new models

I hate to do this to everyone, but Insider Scott pointed out a mistake in the Outcome Matching that has a big impact on the upset victims in the second round and Sweet 16.

Remember when I left Tennessee out of the Cinderella group in round one? Well, I fixed that problem quickly…and didn’t think to analyze if they were a good candidate for subsequent rounds. It turns out they were–the best candidate in fact. And Iowa was also a strong Cinderella.

So…that means either the Vols or Hawkeyes should be tabbed as beating both Duke and Michigan. This changes the fortunes of Iowa State, who now escapes Providence in round two, and Creighton, who gets by Oregon in the Sweet 16.

My apologies for the back-and-forth on this model. But I want to get it right…and I appreciate Scott’s thoroughness.

On the plus side, there are two more models: my Contrarian bracket and the one spawned by the new Brack-o-Matic tool. Tomorrow, I’ll complete the set with two more brackets, Keeper for Friends and Keeper for the World.

Posted in Bracket Tools | 3 Comments

Brack-o-Matic is here!

During last year’s dance, I fiddled around with an Excel model that order the teams by 10 key performance factors…then apply some level of “chaos” to the numbers as a way of simulating the crazy stuff that can happen in the tourney.

This was inspired by a couple circumstances. First, I’ve got a bartender friend named Josh (he’ll never read this!) who asks me every tourney whether I’ve created an “automatic pick” bracket. Secondly, many of the stats models spit out the same picks…and they tend to bias my personal choices. By adding the element of chaos, I get out of some bracket picking ruts that I would otherwise remain in.

So the Brackomatic Excel sheet solves both those problems. It gives Josh the automatic picks he needs (sorry…it doesn’t plop them into a bracket) while enforcing a certain amount of randomness to the teams’ fates.

How much randomness? There’s the rub. Based on some of the work I did to normalize the impact of coaching against team performance, I’ve weighted overall Pythag efficiency to be 67% of the equation, coaching to be 11%–and chaos to have a 22% impact on outcomes. The first two figures are based to an extent on real numbers. The third is a guess. You’re welcome to adjust the Excel model up or down to accommodate your impression on the role of chaos in the tourney.

Using the Brack-o-Matic is easy. Insiders can find it in the TIPS+ section. Just open the Excel file and filter by the “Prospect” column. The teams will be rank ordered from the champion on down. If you filter again, a new order of teams is created. Pay no attention to the final number in the Prospect column, since once you okay the filter, it generates a new value in the Chaos column.

When I publish the new wave of bracket models, you’ll see a new model, #13, called Brack-o-Matic. It’s based on the first time I filtered on this Excel sheet.

Posted in Bracket Tools | 18 Comments

Which past tourney does 2014 most resemble?

Late last night, well into my second glass of wine, my wife gave me a bracket science challenge. When Michelle asks a question about tourney analytics, I listen—not just because she makes me a big pump pot of coffee every morning and everything I need to eat throughout these crazy days, but also because she won our big pool a few years back.

As she was combing over the data last night, Michelle asked, “So which past tourney does this one most look like?”

It’s the sort of question you don’t want to get at midnight, when you’re just about to climb into bed. I mulled it over while tossing and turning and decided on a simple way to answer. I compared the average Pythag of the top 13 seeds with their historical counterparts. Here, in one handy GIF animation, if what I discovered:

2013_Q_Curves

The “eye test,” as well as calculations on the standard deviation between the numbers, comes up with the answer that 2014 is most like 2003, 2006, 2011 and 2013. That should scare the willies out of everyone. Those four years had an average Madometer reading of 17.3% with an average of 10.5 upsets. For comparision, the average dance is 14.3% mad with 8.7 upsets.

When you look deeper at those tourneys, their Final Fours were remarkably low seeded. The average Final Four has a cumulative seed value of 10.8. Three of the four dances most resembling 2014 are well above that.

  • 2003 – One 1, one 2, and two 3 seeds = 9 seed positions, #3 Syracuse
  • 2006 – One 2, one 3, one 4 and one 11 seed = 20 seed positions, #3 Florida
  • 2011 – One 3, one 4, one 8 and one 11 seed = 26 seed positions, #3 UConn
  • 2013 – One 1, two 4’s and one 9 seed = 18 seed positions, #1 Louisville

Interestingly, all of these dances included a seed lower than eight…and three of the four were won by a three seed.

If we look more closely at how the four tourneys compare to 2014, we may get some insight into how the top four seeds will performance—and how crazy the upset games might get. Check this out:

2014_vs_Top4

Some observations by year:

  • 2003 vs. 2014: The 2003 two seeds were considerable stronger than this year’s crew, but the four seeds weren’t nearly as strong. Five and six seeds, on the other hand, were stronger in 2003, while 11 and 12’s were weaker overall. That may suggest that we’ll have more upsets at those positions. For comparison’s sake, only two 5v12 or 6v11 upsets got sprung in 2003.
  • 2006 vs. 2014: The shape of the first eight 2006 seeds is eerily similar to this year. But the nine through 12 seeds are considerably stronger. In 2006, the 11 and 12 upset seeds split their games with favorites.
  • 2011 vs. 2014: If I had to pick one tourney that most resembled 2014, it would be 2011. That year the Madometer was its second craziest (19.8% mad), eclipsed only by 2013—and there were 13 upsets, tied for the most with four other dances. Look at the similarity in the curve all the way to the ten seed. Yes, there are slight differences from seed to seed, but it’s an amazingly similar shape. The fours were even stronger in 2011 than this year. The big differences come in the 11-13 seeds. 11 seeds are up, 12 seeds are down and 13 seeds are bad. In 2011, these seeds were 5-7 in the opening round.
  • 2013 vs. 2014: Last year’s two, three and eight seeds were stronger than they are this year. But the 4-6 seeds were similar to 2014, as were the 11-13 seeds. So those match-ups may generate about the same number of upsets. Last year, the underdog went 5-7 in slaying the giant.

So there’s your answer, Michelle. When you win the billion dollar bracket, will you still make me ham sandwiches?

Posted in Bracket Tools, Measuring Madness, Tourney Trends | 21 Comments