Quality Curve Analysis (January Edition)

I thought I would start Project: Perfect Bracket with a very familiar analytical tool: the Quality Curve (QC).

The Idea Behind Quality Curves

The theory behind the curves goes as follow: the quality of the teams in the tournament can predict the quality of the tournament results. In layman's terms, higher quality teams in the higher seeds will result in a stable, predictable tournament (much like 2007) and lower quality teams in the higher seeds will result in an insane, unpredictable tournament (much like 2014).

A Word on Methodology

Before we move into the analysis of these QCs, let's first look at the methods behind the madness in creating the QCs.

The QCs are drawn using Ken Pomeroy's Pythagorean Ratings (pre-tournament). A QC is drawn from a top-down perspective, meaning teams 1-4 in the KenPom ratings are 1-seeds, 5-8 are 2-seeds, and so forth until the 11- and 12- seeds are ranked, which I used five teams instead of four to account for play-in teams. It is important to note that when the final QCs are drawn after the bracket has been announced (as is the case with 2007 and 2014 above), the seeds do not follow the KenPom rankings in a top-down fashion. These are known as Seed Curves (SCs) because the curve is calculated from the 4 teams (or 5 teams for play-in seeds) at that specific seed.  Using 2014 as an example, the #2 ranked team in the KenPom ratings (Lousiville) was given a 4-seed in the tournament as was the 10th-overall ranked team (Michigan St) when 4-seeds should have went out to the teams ranked #13-#16 in the KenPom ratings. This definitely had something to do with the extremely strong 4-seed average in the 2014 SC, as the 4-seeds averaged higher than both the 2- and 3-seeds.

What does it all mean?

From a general analysis of the 2016 QC, it is unusually strong from the 1-5 seeds, steady 6-9 seeds, and slightly below average 10-12 seeds. Looking at the actual data itself, the decline between seeds (from 1 to 2, 2 to 3, and so on):

• From 1- to 6- seeds, the decline ranges from 0.013 to 0.024 per seed.
• From 6- to 9- seeds, the decline ranges from 0.006 to 0.010 per seed.
• From 9- to 12- seeds, the decline ranges from 0.014 to 0.019 per seed.

My take: At this point in the season, all but 2 conferences have fully begun conference play (Atlantic Sun and Ivy League). As teams began conference play, I expect the 2016 QC to bend downward as power conference teams face higher quality competition on a night-in, night-out basis. I also expect the 11- and 12- seed teams to slightly increase as these seeds may be composed of mid-major and minor conference teams whose conference play features teams that power conferences play against in their non-conference schedule. Thus, the better teams in these mid-major and minor conferences should improve their ratings as they separate themselves from the weaker teams in their conference. When I look at the 2016 QC again in February, we'll see how if this prediction comes true.

From a seed-by-seed analysis of the 2016, it predicts a very mixed bag for the tournament.

• 1-seeds are much weaker than their 2007 chalky counterparts. A comparison of other strong 1-seed years such as 2008 and 2015 shows that the 2016 1-seeds fall short of their quality too. It could be another year of 1-seeds failing to meet their seed expectations of a Final Four.
• 2-seeds are in line with their 2007 counterparts. They are probably closer to their 2009 or 2012 counterparts, but definitely well-short of the 2015 counterparts.
• 3-seeds are ungodly stronger than their 2007 and 2014 bracket-mates. In fact, the years that best correspond to strong 3-seeds are 2004, 2008 and 2013. For 3-seeds in their respective years, 2004 produced 3 Sweet 16s and 1 Elite 8, 2008 produced 4 S16s and 2 E8s, and 2013 produced 3 S16s and 2 E8s.
• 4-seeds are right in-line with their 2014 insanity counterparts. If 4-seeds draw good match-ups in their first two rounds (13s and 5v12s), these 4-seeds could play the role of 1-seed spoiler in the S16 round, that's assuming the weak 2016 1-seeds make it this far. 2011 had the strongest 4-seeds of all, surpassing 2014 and 2016. Other comparable but weaker years include 2005, 2006, 2009, and 2013. (Note: 2006, 2011, and 2013 were all heavy upset years with 12, 13, and 11 upsets respectively.) Of the six comparable years listed, only 2009 did not produce a 4-seed in the E8, all others produced at least one with 2013 producing two.
• 5-seeds are head-and-shoulders above their Insanity matches and ridiculously above their Chalk matches. 2005 was the strongest year for 5-seeds, but it only produced 2 S16s and 1 E8, but that may be a result of 2005 having stronger than usual 4-seeds and 12-seeds. 2010 was also a strong year for 5-seeds, producing 2 S16s, 2 E8s and 2F4s, and 1 National Runner-up (guaranteed because both 5s in 2010 were on the same side of the bracket). We'll have to see how the teams are actually seeded in March before we know if 2016's 5-seeds are strong or if they get misplaced into other seeds.
• 6-seeds are below their Chalk comparisons but above their Insanity comparisons. In KenPom rating terms, 6-seeds are .020 weaker than 2016 5-seeds and more than .040 weaker than 2016 4-seeds, which is a very steep decline from those stronger-than-average seeds. With average 6-seeds going against stronger-than-average 2s and 3s, it may not be a year to be a 6. 2009 and 2004 were strong years for 6-seeds. Unless higher quality teams get lower seeds this year, don't expect much from this group.
• 7-seeds are in-line with their Insanity bracket-mates and slightly above their Chalk bracket mates. Could one or two strong 7s get lucky in the 2nd round and take down this year's Chalky 2s?
• 8-seeds are below the 2007 Chalk and above the 2014 Insanity. They do draw weak 1s this year, but an average year in the 8s doesn't convince me of an upset.
• 9-seeds are below both the Chalk and Insanity years. With average 8s and weak 1s, weaker-than-average 9s do not strike me as a Cinderella.
• 10-seeds are slightly below Chalk and well above Insanity. Unfortunately, they draw strong 7s this year, and if they can surmount a strong 7, they are likely to meet a very strong 2. If I was using dunk-contest scoring, I cannot give a 10 to any of the 10s, and I would probably go so far as to give them either 1s or 0s.
• 11-seeds are stronger than their Chalk matches but drastically weaker than their Insanity matches. Their insanity matches in 2014 made 2 S16s and 1 E8 and that year also featured weak 6s and weak 3s, which contributed to their success. With average 6s this year, penciling in one or two 11-seeds to spring an upset may be a safe pick, but I don't know if I can trust them yet to take down this year's strong 3s in the next round or strong 2s in the S16.
• 12-seeds are right in-line with both 2007 and 2014. 2007 saw all four 12-seeds lose to 5s but 2014 saw three 12-seeds spring an upset. Ironically enough, 2007 had weaker 5s than 2014, but all of the 5s in 2007 advanced, and experience with the 5-12 match-up tells me to expect the opposite of logical. For now, I will take a wait-and-see approach with this seed. I think some of these 12-seeds will get stronger ratings as they play inferior competition in mid-major and minor conferences, and there is also the possibility of strong teams being seeded-down into the 12 slot.

This concludes my Quality Curve Analysis.

1. I will do this again on Feb 1 to account for all games through Sun Jan 31 (the February Edition). Then, probably again on Feb 29 to account for all games through Sun Feb 28. If I feel like the QC is taking shape or making drastic changes in certain seeds, I may do interim updates. Finally, I will do the most important 2016 QC on Mon Mar 14 after the brackets are revealed on Sun Mar 13 (and all regular season games have completed).
2. Expect the next article on Project: Perfect Bracket to be up around Sun Jan 10 or Mon Jan 11.