2017 Quality Curve - February Edition

It has been exactly one month since the last Quality Curve (QC) analysis, and that can only mean one thing: It's time to do another one. With half of the conference schedule completed and more than 97% of the non-conference schedule also completed, we are somewhere between the 2/3- and 3/4-mark of the 2016-2017 pre-tournament season. Most of all, our data is better now than it was in January, and it is a little bit closer to its eventual mark (which occurs on Selection Sunday). Without further ado, let's get started with some preliminary factors before we hit the main event.

Preliminary Factors

You are about to enter the February Edition of the Quality Curve Analysis. Please follow these simple instructions before proceeding.

1. If you haven't read the January Edition, follow the link and read the section on "Reviewing The Changes." The efficiency data typically used for this analysis changed formats between this season and the previous season. Understanding these changes is important to understanding the data.
2. In typical QC analyses, the current data is compared to pre-tourney data from previous years in order to maximize predictive value. Since the method of calculation changed (see Step 1), I do not have the pre-tourney data for any of the previous years. Instead, I am approximating those years using their post-tourney data, which happens to be readily available. However, I have noticed sizable movements from pre-tourney to post-tourney data, so if anyone has the post-tourney data using the old methodology (Pythag), I possibly could make a workable substitute.
3. The QC uses the efficiency ratings for the Top 50 teams, which is a close approximation for the 1-12 seeds in the tournament. More often than not, teams in the Top 50 efficiency ratings do not make the tournament and teams not in the Top 50 do make the tournament. The QC gives us a picture of the college basketball landscape, whereas the seed curve (not produced until the Bracket is revealed) gives us a picture of the NCAA tournament quality.

Analysis

Let's start where most of these analysis usually start and look at the current year compared to the sanest (2007) and craziest (2014) tournament years.

While this chart does not tell us much, there are some details that should be pointed out.

• From the 1st to 8th ranked teams, the 2017 curve is closer to the 2007 curve, which suggests strength in the 1 & 2 lines.
• From the 9th to the 16th ranked teams, the 2017 curve is closer to the 2014 curve, which suggests weakness in the 3 and 4 seeds.
• From the 17th to the 32nd ranked teams, the 2017 curve approximates the 2014 curve, which is downright scary.
• From the 33rd to the 47th ranked teams, the 2017 curve splits the difference between the two curves.
• For the final three teams in the Top 50, the 2017 curve approximates the 2014 curve, but again, there is no guarantee that these three teams will be in the field of 68.

Breakdown by Mad-o-meter®

Since we started by comparing 2017 to the best and worst years, let's take this approach one step further and look at 2017 compared to groups of years organized by the Mad-o-meter® Rating. Since 2017 more resembled the 2014 curve, we'll start with the crazy group (2011, 2013, and 2014).

Do your eyes see what my eyes see?

• 2011 (19.85% MOM) featured a 5-4-2-2-0-0 upset-by-round result with a 8-seed and 11-seed in the Final Four. In other words, the craziness happened as the tournament progressed.
• 2013 (20.75% MOM) featured a 7-3-0-1-0-0 upset-by-round result with a 9-seed in the Final Four. In other words, most of the damage was done early and 10 additional points was added to the final total with a 9 over a 2 and two 4s over 3s in the Elite 8.
• 2014 (21.35% MOM) featured a 6-4-2-1-2-0 upset-by-round result with a 7v8 National Championship game. In other words, all hell broke loose and the first round was probably the sanest of the six rounds.
• Excluding 2017's relative strength among the 1st- through 16th-ranked teams, the striking resemblance of 2017 to the other three years from the 33rd- to 50th-rank is stunning, especially seeing how well it approximates the 2014 curve from the 17-32 ranks. Hopefully, the strength of the 1-16 ranks is legitimate and it holds up for another six weeks.

Speaking of strength, let's now look at 2017 compared to the sane group (2007, 2008 and 2009).

I'm not sure what to make of the comparisons, so let's walk through it.

• 2007 (4.06% MOM), 2008 (12.33% MOM) and 2009 (9.17% MOM) featured the sanest three-year span in NCAA tournament history, and ironically it happened following the implementation of the one-and-done rule. You have to go back to the 1993-1996 span to find anything close to this level of sanity when the tournament had 4 consecutive years of a MOM reading in the 10% range, and ironically enough, that happens to be right before the straight outta high school movement.
• 2008 was the craziest of the three years, featuring a 5-3-0-0-0-0 upset-by-round result, yet it was also the only year in tournament history where all four 1-seeds made the Final Four. Strangely enough, the 2017 curve approximates the 2008 curve from the 1-4 ranks (which would be the 1-seeds if seeded in that fashion).
• Interestingly enough, the 2007 curve features the strongest teams from the 32nd rank and on. These are the teams that cause the upsets because they are stronger than average, yet 2007 had the fewest upsets ever! Maybe that had something to do with unusual strength at the top, and 2017 definitely has this feature.

Let's look at another group of relatively stable years (2003, 2004, 2015).

When you consider the comparisons to 2008 in combination with this group, we finally have something to work with for the upper-half of our QC.

• 2003 (11.13% MOM) featured a 3-3-0-0-0-0 upset-by-round result. Its F4 was 3v2,1v3 and its E8 was 1v3,1v2,1v7,1v3.
• 2004 (11.28% MOM) featured a 2-4-1-0-0-0 upset-by-round result. Its F4 was 3v2,1v2 and its E8 was 4v3,1v2,1v7,8v2. Out of these three years, I wish I had the pre-tourney data for this year simply because I think the pre-tourney data would rule this year out from being related to the other two.
• 2015 (12.03% MOM) featured a 4-3-1-0-0-0 upset-by-round result. Its F4 was 1v1,7v1 and its E8 was 1v3,1v2,4v7,1v2. Outside of an overrated 1-seed, a badly matched 2-seed, and an injury-bugged 2-seed, 2015 would have possibly have been the year to rival 2007. It is the only year so far that outpaces 2017 from the 1-8 ranks.
• For the most part, 2017 is stronger from 9 to 46 than any of these years. The sheer strength in the 1-8 ranks for 2015 was a big reason why so many 1s, 2s and 3s made the E8 and so many 1s made the F4. With 2017 being an inversion of 2015 from the 1-16 ranks, maybe we will have a lot of 1- through 4-seeds make the S16, and then some 4s (or maybe 5s depending on seeding) toppling 1s in the next round. Nonetheless, 2008 from the previous group and 2015 from this group do a good job of approximating the 2017 curve from the 1-16 ranks.

Just for some fun, let's take a look at a few more crazy years (2006, 2010, 2012, 2016) to see how 2017 matches up with them.

To begin with, I don't like putting these four years together, but since their MOM ratings are so close, I thought I should just so we can see the visualization. In the analysis, I will separate them according to their similarities (2006 and 2010 together and 2012 and 2016 together).

• 2006 and 2010 featured a relatively sane R64, but after that, lower seeds started winning and kept winning for the next three rounds. I think I can say with a deep sigh of relief that 2017 looks nothing like the 2006 curve. I won't waste your or my time talking about it. From the 31 to 43 ranks, 2017 closely resembles 2010, so maybe some smaller seeds will make some noise in the R32, S16 or E8 rounds. It also matches up at the 1-3 ranks, but I'm not sure how much those three data points foretell for the future for 2017.
• 2012 and 2016 feature almost identical tournament profiles.
• 2012 (17.14% MOM) featured a 7-1-1-0-0-0 upset-by-round results. It had aggregate values of 9, 21, 73, and 210 for the F4, E8, S16, and R32, respectively. It had a 1v2 National Championship game.
• 2016 (18.35% MOM) featured a 8-2-0-1-0-0 upset-by-round results. It had aggregate values of 15, 24, 66, and 215 for the F4, E8, S16, and R32, respectively. It also had a 1v2 National Championship game.
• In a strange twist, their curves begin to separate at the 21st ranked team. From that point on, 2016 looks more like 2010 and that's probably why 2016's tournament profile looks a little wilder than 2012. 
• In fact, one could make the argument that 2016 matches up best with 2017 after the 15th rank (and from 1-4 as well). If this is the case, then it seems we are back at square one: the unusual strength of 2017 at the 5-16 ranks and its implication for the bracket's results.

I hope the mystery of the strong 5-16 ranks has more to do with comparisons to post-tournament data. If that is the case, then exits by 3-, 4- and 5-seeds (not necessarily early ones) would correct this issue and bring their strong numbers more in line with their counterparts listed above. If it is unusually strong teams in the 5-16 ranks, then I expect what I said earlier about 4s or 5s toppling a few 1s in the S16 (maybe even an under-seeded 6-seed toppling a 3- and a 2-seed). Fortunately, we have four more weeks of games and data before the March Analysis and six more before the Final Analysis on Selection Sunday (which will include the 2017 seed curve). By those times, we should have more clarity, and with that, more understanding and hopefully better predictability. Until then, thank you for reading.