2019 Quality Curve Analysis - Final Edition

Well, the bracket has been released, the match-ups are set, and now it is time to see how the next three weeks will likely play out. If you have read the previous three editions (and if you haven't, here are the links to three very good reads: Jan, Feb and Mar), you will recognize the chart below. It is the Final 2018 Quality Curve.

From the looks of it, there's a lot to talk about, so let's see what we can learn.

What is the QC Telling Us?

If I summarize this chart in one phrase, it would be the title of a Beach Boys hit single: Good Vibrations. From January to February, we see a lifting of the entire QC (except for a few spots), which is a good thing. From February to the present, we've seen the QC vibrate up and down, but roughly in the same range. Let's breakdown when parts of the range reached their peak:

• February Peaks: #1, #4, #11-13, #30-36, #38-43
• March Peaks: #2, #3, #7, #25, #27-29, #37
• Final Peaks: #5, #6, #8-10, #14-24, #26, #44-50

For the most part, the front-half of the curve is playing at or near peak efficiency during this time. As for the back-half of the curve, the #25-43 spots are playing away from peak efficiency, while the tail-end of the curve is playing at peak efficiency. When the top teams are playing at or near peak efficiency and the bottom teams are playing away from peak efficiency, that is indicative of a steepening curve, which is a good thing for sanity. If you are worried about the last seven spots in the QC pulling off an upset, five of the spots are one each of the 8- thru 12-seeds (and all five are over-seeds) and the remaining two did not qualify for the NCAA tournament.

Now, let's look at how 2019 compares to the last two years, and this is a valuable comparison since the last two years provide an excellent QC dichotomy.

If the last chart showed signs of steepness (and likely bracket sanity), this one screams it even louder. Let's break it down into sections.

• The 2019 #1-10 spots are stronger than either 2017 or 2018. For 2018, that statement doesn't mean much, but for 2017, it speaks volumes. Judging that 2017 and 2018 look similar from the #1-6 spots, it is ironic that both years saw a 1-seed and two 2-seeds lose in the R32 (not to mention the 1-seed that lost in the R64). If 2019 is far stronger than either of those years, I think it is safe to predict that no more than two upsets will happen to 1- and 2-seeds collectively before the S16 (I do have my eye on two potential upsets, but I can't make those based on the QC/SC analysis).
• The #11-27 spots are literally stuck in the middle of 2017 and 2018. The sheer strength of this range in 2017 saw four upsets in the R64 (one for a 5-seed and three for 6-seeds) and one upset in the R32 (a 3-seed to an 11-seed). The sheer weakness of this range in 2018 saw four upsets in the R64 (two 4-seeds and two 6-seeds) and two upsets in the R32 (both were 3-seeds to 11-seeds). These totals look like safe targets, but we need to review the SC before making it official: Four upsets in the R64 and either one or two upsets in the R32 for for 3- thru 6-seed. 
• The 2019 #28-45 spots are stronger than either 2017 or 2018. This range of teams includes one 7-seed, all four 8-seeds, two 9-seeds, two 10-seeds, two 11-seeds, a 12-seed, and six NIT-eligible teams (lol, just seeing if you are paying attention). This is interesting because the two potential Cinderellas I mentioned above are in this group, but I've given too much information for anyone to decipher the identity of these two specific teams.
• The 2019 #46-50 spots are mostly weaker than 2017 or 2018, but as I mentioned above, three of these teams are over-seeds and the other two are not in the NCAA tourney. I'm not too worried about these five teams.

As usual, the Selection Committee never seeds teams according to advanced metrics (and I believe that is a good thing), so we need to see how the QC differs from the Selection Committee's evaluation of the tournament teams. To do this, we will construct the 2019 Seed Curve.

The 2019 Seed Curve

Wow, what a thing of beauty! Okay, platitudes aside, let's get down to business. There's only two things I'm going to point out from this chart.

1. The same steepness that defines the QC also can be seen in the SC. There are slight curvatures from the 2- to 5-seeds and from the 7- to 10-seeds, but only the latter seems to have significant bowing. Bowing in the SC helps identify areas of over-seeding and under-seeding (if teams were seeded directly by advanced metrics, the SC would have linear properties instead of quadratic). There is also a spike at the 12-seed, which either indicates severe weakness in the 11-seed group or under-seeded (and potentially Cinderella) 12-seeds.
2. The second point I want to emphasize is lack of dispersion within the seed-group. I've never mentioned this before, but I have kept an eye on it over the last few years. When the four members (sometimes more for seed groups with play-in members) show little dispersion from the seed-group average, good things happen for the seed-group. In 2017, the 3-seed group was tightly centered around the mean, and one 3-seed made the F4 and two more met seed expectations by reaching the S16. Also in 2017, the 9-seeds were tight around the mean, but they were also the trough of a major dip, so only one 9-seed won in the R64. In 2018 (Link), 5-seeds and 9-seeds were centered tightly around the mean: All 5-seeds defeated their 12-seed counterparts with three of the four reaching the S16, and three 9-seeds defeated their 8-seed counterparts with two of them reaching the E8. Also in 2018, 8-seeds were tightly centered around their mean, but unfortunately both 8-seeds and 9-seeds cannot advance simultaneously. In 2019, we see the 6-seeds, 8-seeds, and 9-seeds centered tightly around their mean. The only question that remains is what the result will be: A deep run by one of the members of the seed-group or all four members of the seed-group advancing past their counterpart. I present both options to you because I'm not certain which will happen, nor do I have any tools to determine which will result. I am fairly confident that good things should happen to the 6-seed group and either the 8- or the 9-seed group.

Now, let's look at the QC-SC overlay in the chart below.

As troubling as last year's QC-SC overlay, this year looks even tighter than last year. The Top-28 teams in both the QC and the SC share 27 teams in common, meaning the 1- thru 7-seeds are approximately close to their true seeding (this can also be seen on the OS/US model). In fact, all 1- thru 7-seeds are within two seed-lines of its true seed (determined by advanced metrics). The assumption made last year is the tightness of the QC-SC overlay means upsets are less likely as seeds are indicative of (or approximate to) true team quality. The locations along the curve where the SC rises above the QC are high-probability targets for out-performing seed expectations. Last year, 2-seeds, 3-seeds and 5-seeds were high-probability targets for out-performance, and only the 2-seed group failed to exceed expectations (Expected 12 wins vs Actual 9 wins). According to the 2019 SC, the 2-seed, the 5-seed and the 7-seed group look to be high-probability targets for out-performing seed expectations.

Concluding Thoughts

Typically, I provide my full predictions for round-by-round upsets and the Aggregation Model in this section of the article. Instead, I'm going to save that for a final article to be published late Wednesday night or early Thursday morning. I want to get the results of a few more models before I make a final call. If you don't want to wait for those numbers or want to make your own attempt at a prediction, I do like the 2015 SC and the 2003 SC models (Link) to be close approximations for the 2019 tourney. As always, thanks for reading my work, and I am still committed to my BCT schedule.