2018 Quality Curve Analysis - March Edition

Welcome back PPB-readers! Unless you are in the B10 Conference, we have exactly two more weeks of games until the Selection Show on Mar 11, so that means it is time for the March installment of the QC Analysis series. If you missed the first two installments, here are the links (January and February). While I hope it won't take you two weeks to read this installment, there are a lot of issues that I want to discuss in this installment and I doubt that I will get to address them in the Final Edition. So let's get started.

The Evolution of the 2018 QC

The obvious starting place for the analysis is the change of the QC over the duration of the season, which is displayed in the chart below.

I suppose the easy way to breakdown this graph is by groups. Starting with the Top 10 teams in the QC, the March curve seems to parallel the February curve and show differentiation from the January curve. Considering how bad things were looking in January, the strengthening of the Feb and Mar QCs is a welcome sign, meaning you don't have to eliminate all four 1-seeds before the Sweet 16. There has been a noticeable weakening at the top 3 spots, but on the whole, the QC is stronger now than it was in Jan. From the 11th- to the 18th spots on the curve (potential 3- to 5-seeds), the Jan and Feb QCs seem to parallel each other with the Mar QC being weaker at all points in the range except at the 11-spot. Again, it will depend on where these eventually get seeded by the committee (as the teams ranked in these spots on the QC curve may not get the corresponding seeds), but it does point to potential upsets in these areas. From the 19th- to the 31st-spots, the Mar QC is primarily the lowest of the three curves. This is the most interesting aspect of the QC to this point in the season because I would expect a lot of S16 Cinderella candidates (teams that could topple 1- and 2-seeds) to be found in this range (potential 5- to 8-seeds). Given that the efficiency drop-off per spot is much smaller compared to that of the 2017 QC, the fact that this portion of the curve is still sinking at this time of the year makes me question the potential insanity of the 2018 bracket. I guess it still depends on where the teams in this range get seeded. From the 32-spot to the 50-spot, the Mar QC runs in line with the other two curves until the 45-spot, where the Jan QC falls off. The teams in the 45- to 50-spots would translate to potential 12-seeds.

The Past vs The Present

The next step in our analysis is a comparison of the 2018 Mar QC to the 2017 Mar QC, shown below.

We know two things from last year. First, 2017 was top-heavy and predictable (a 4-4-2-0-0-0 upset road-map where yours truly predicted 4-3-2-0-0-0 upset road-map). Second, the teams at the top of the 2017 QC actually improved their quality over the remaining two weeks of the season. Knowing these two facts and looking at the chart, it still looks like we can expect a crazier tournament than last year. The top 6 teams of 2018 are right in line with their 2017 counterparts, but then the freefall begins. From the 7th-spot to the 29th-spot (potential 2- to 8-seeds), there is nothing but an erosion of quality. The mere thought of one of these teams being considered for a 1-seed speaks to the potential craziness of 2018, until I reveal that two 1-seeds may be comprised from teams in this range (XAV and KU according to many bracketologists). If that happens to be the case, then four of the teams from the 1- to 6-spots could fill the 2-seed line, and maybe even the 3-seed line (talk about sleeper value). We will see the full extent of the damage when the seed curve is analyzed in the Final QC Analysis, but it seems like it could be the end-result. The 30th-spot to the 50th-spot sees 2018 being stronger than 2017, which points to more lower-seeds winning in the opening round, especially given the glaring weakness of the 7- to 29-spots. If 2017 produced 4 upsets in the R64 with all of that strength, a minimum of 5 upsets seems doable in 2018 with all of this weakness. As always, a lot will be determined by seedings on Selection Sunday.

Qualifying the Quality Curve

If you remember from (or re-read) the January QC Analysis, I emphasized the difference one day makes in the statistics. When college basketball went crazy on Saturday Dec 30, it led to sizable one-day dips in the ratings of losing teams. In the February QC Analysis, I began qualifying the QC by displaying the QC within the confines of its range over the month under study. I'm going to do that again for the Mar QC just so we can see how the last month has progressed, but I doubt I do it for the Final QC for various reasons. So, let's take a look at how the QC has progressed over the last month.

The obvious detail in this chart is how often the Mar QC (the yellow line) follows in line with either its maximum value for the month or its minimum value at the month. This does not mean that the specific team located at that spot is playing at its best or at its worst, it means that part of the curve is near its strongest or near its weakest. The specific team at a specific spot on the curve can change on a daily basis throughout the course of the month. For example, just in the last 20 days, TENN has held the 7th, 10th, 11th, 12th, 14th, 15th, and 16th spots, and they have changed spots for various reasons. The point I'm trying to make is this: It has more to do with the curve itself than the specific team at the curve. Let's try to look at the movements of the curve more closely, like the chart below.

This chart shows the maximum-minimum ranges for both the February Analysis (in blue) and the March Analysis (in pink). To begin with, the pink ranges are tighter than the blue ranges. In my interpretation, this is the result of measurement, not a result of team quality. In statistics, the larger you make the size of the sample, then the result of the sample will be closer to the actual value. In March, we have more data points than in February, so March's range should be closer to the actual measurement of team quality than February's ranges. Thus, March should have tighter ranges than February. Second, the movement of the March ranges from where they were in February tells us the same thing the movement of the QC told us (described in "The Evolution of the 2018 QC"). The same movements in the ranges happened at the approximately same intervals of the QC. The only real outlier I see is around the 15- to 17-spots, where the Mar range eclipses the Feb range even though the Mar range is below the Feb range along most of the curve. It is also the same part of the curve where the range widened instead of narrowed like the majority of the curve. I honestly do not know how to interpret it, but I wanted to make note of it and I will be keeping my eye on it for the next two weeks.

As far as the QC Analysis goes, this is pretty much in-line with what I've seen over the course of the season. You may have also noticed the lack of Sagarin QCs. To avoid redundancy, I opted to not implement them in this analysis even though they illustrate the same phenomenon. Also, it gives me more room in the article to discuss another issue, as I predicted in the opening paragraph.

Extra Stuff

In my countless articles on the Quality Curve, I reiterate one important point: It all comes down to how the teams are seeded. The Field of 68 is never seeded team-by-team according to the efficiency ratings. Some teams are over-seeded (usually these teams get upset) and some teams are under-seeded (usually these teams become Cinderellas). Nonetheless, it matters where each and every team is seeded, and this year is no different.........WELLLLLLLL!!!!! Let's just take this opportunity to see exactly how the teams will be seeded in 2018.

In the last two "Welcome to the [.....] College Basketball Season" articles, I have emphasized a critical talking point: The Mind of the Selection Committee (2017 and 2018). For the past two years, the seeding of teams outside the 1- through 4-seeds was greatly influenced by conference affiliation. The stronger the RPI of your conference, the higher the team's seed (even if the team's individual resume was inferior to another team's resume seeded lower than them). A lot of those over-seeded teams were accordingly upset in the tournament or didn't achieve seed expectations (BEC and P12 in 2017 and ACC in 2018). In 2018, the Selection Committee has taken bracketing principles to a whole new level (Sarcasm mine). You may have either read online or heard the broadcasting crew on television discuss the Committee's new approach to evaluating teams and their records: The Quad System. Previously, they would look at a team's record among four categories: Vs RPI 1-50, Vs RPI 51-100, Vs RPI 101-200, and Vs RPI 200+. All of this information would be printed on a team's Nitty Gritty Report along with various other data points. Now, this information is categorized into quads that factor in game location.

• Quad 1 Record: Home vs RPI 1-30, Neutral vs RPI 1-50, and Away vs RPI 1-75
• Quad 2 Record: Home vs RPI 31-75, Neutral vs RPI 51-100, and Away vs RPI 76-135
• Quad 3 Record: Home vs RPI 76-160, Neural vs RPI 101-200, and Away vs RPI 136-240
• Quad 4 Record: Home vs RPI 160+, Neutral vs RPI 201+, and Away vs RPI 241+.

For starters, I would love to know the rationale for the cut-off points for each range. I seriously doubt these cut-off points were determined by back-testing previous years selections. Not to mention, a current look at the RPI numbers shows the sheer foolishness of these cut-offs points. The current #1 RPI team is UVA and the current #75 RPI team is TOL. I seriously doubt anyone in their right mind would claim that a road win at TOL is in the same class as a road win at UVA. We are talking about a TOL team with already 3 losses at home this season (all to teams that -- like TOL -- probably won't make the big dance) and a UVA team with 2 total losses this season (both to likely tournament-bound teams with one of those losses coming at home). I am sorry, but these two teams are not in the same class. There should be a Quad 0 category that factors only road wins at the RPI 1-30, and that change would only do some minor work at separating a great road win from a good road win. Yet, it is still an arbitrary cut-off point, and the use of RPI data is still subject to the "garbage in, garbage out" conundrum.

I could berate this new system enough to make this article take two whole weeks to read, but I will spare you. I am merely bringing it to my reader's attention for informative purposes only. I hope (and it is a really big "hope") this methodology is only implemented for selecting the teams and not for seeding the teams. For reference, here is a link to a bracketology that features the quad system. (Note: This is not my endorsement for this particular bracketologist; he just happens to be the only one that I can find that displays the Quad System). If anyone finds a good link showing the Quad System applied to all 351 Division teams, please send me a link in the comment section and I'll update this article with that link. Nonetheless, I do want to try to pre-emptively get into the mind of the Selection Committee for the 2018 tournament before it starts. The last two times I have done it ex post facto (in the two season-opening articles), and it has proven to be insightful. I don't think it will perfectly pick every upset, but I do think it will be very helpful in spotlighting where to look for upsets.

Anyways, I hope you have enjoyed this rather lengthy article, and if not enjoyed it, I hope you at least found something informative in it. Since we are getting closer to Bracket Crunch Time, I plan on ramping up the production on the blog to a weekly article instead of the current bi-weekly schedule. I still have another topic on my mind, and two weeks from now would be the Final Edition of the QC Analysis, so I wouldn't get to discuss it at that point. As usual, thanks for reading my blog.