2018 Quality Curve Analysis - February Edition

As we approach the completion of another month in basketball, it means that it is time for another installment of quality curve analysis. If you haven't read the January Edition of QC Analysis, here is the link to that article. It is definitely worth the full read, as I believe it will put you ahead of your bracket-picking competition in March. Nonetheless, the key points in that article:

1. Parity in the 2018 tournament (the key ingredient for upset soup)
2. One of the wildest days in college basketball and its distorting effect on the QC
3. Where 2017's tournament contenders were located in the 2017 Jan QC Analysis

At the risk of appearing lazy, I think I'm going to follow these same themes in the Feb Edition. If you took the liberty of reading (or even re-reading the January Edition), the same thematic structure to the article will provide continuity between the two reads. Likewise, I don't think much has changed in the basketball universe since the January edition, so keeping the same themes is fitting. Let's see what the February QC predicts for us.

What a Difference a Month Makes.....Or does it?

As usual, let's start with a chart of the KenPom QC and go from there.

The first and obvious detail in the graph concerns the teams from 1st to 6th. Since January, these teams have seen their efficiency ratings improve significantly. For reference purposes, these teams, in order:

• January: NOVA, MIST, PUR, DUKE, UVA, KU (9th in Feb)
• February: NOVA, UVA, PUR, DUKE, MIST, CIN (9th in Jan).

At this current moment in time, it seems as if the teams at the very top are trying to separate themselves from the pack. This would resemble years like 2008 and 2015 when three or more 1-seeds reached the F4. However, the 1-seeds in those two years had much better efficiency ratings than the teams in this list, in both absolute terms (ratings value vs ratings value) and in relative terms (ratings value compared to the rest of the field). We will need the data from the final six weeks of college basketball before we accept this conclusion, but it is something we need to keep watching. The other detail in this chart worth noting is the overall lack of movement in the remaining teams. There are pockets of deterioration (from ranks 18 to 29) and there are pockets of improvement (from ranks 8 to 11 and from ranks 45 to 50). Let's take a look at the Sagarin QC and see what it has to offer.

This Sagarin QC shows a slightly different story, so let's see what we can make of it. The top 10 Sagarin-ranked teams show a mixed picture of improvement and deterioration, with the magnitude of the changes being very wide. Keep in mind that the January Sagarin QC was taken immediately after one of the wildest days in college basketball. Thus, the differences between the Jan QC and Feb QC may be mostly attributable to the distortions of the Jan QC from that wild day of hoops. From the 11th ranked team to the 22nd ranked team, there was drastic degradation in the ratings. These would be teams with 3- to 6-seeds in the tournament. From the 23rd ranked team to the 50th ranked team, the Sagarin QC resembles the KenPom QC, so the consistency between the two curves is comforting. It would be a lot more troublesome for predictive purposes if the two QCs showed two completely different stories. (This may have something to do with differences in the way the two ratings systems are calculated, but I'll discuss this more in the Appendix at the bottom of the article.)

For the most part, the two charts held to its January form: Steepness in the front of the curve and flat-lining to the end of the curve. This signals a tournament of parity and an above average count of upsets. Even though the teams at the top may be showing separation, they are still weaker than previous years (2008 and 2015) when there was wide-ranging gaps between the teams at the very top and the rest of the field.

What a Difference a Month Makes.....An Entire Month

As I've stated in both QC articles, the January QC happened one day after one of the wildest days in college basketball. Essentially, every QC is a snapshot in time -- the efficiency ratings of all teams at that current day in the season. I want to try to give the QC some qualitative perspective by showing its current state in relation to its movement over the entire month under review.

What this chart shows is the KenPom Feb QC (pink line) encapsulated by its minimum and maximum values at each rank over the duration of the month (since the Jan QC). For example, the current top-ranked KenPom team is compared to the range of efficiency ratings for all top-ranked KenPom teams over the month (Jan 1st to Jan 28th), and so forth for each rank all the way through the 50th-ranked team. Though it may be obvious, I will detail the two points of this chart.

1. The most obvious detail in this chart is how the current QC mirrors the minimum QC. With the exception of a few ranks (the top 6, 15-17 and 42-50), the majority of ranks are at or near their lows for the time period under review. With six weeks left until the Selection Committee unveils the bracket and many teams currently playing at their lows, it does not bode well for a calm and chalky 2018 tournament.
2. The second detail that should be discussed is the maximum-minimum spread. For the top 10 ranked teams, the spread between the maximum and minimum curves is anywhere from 2-5 points. From the 11th- to 50th-ranked teams, the max-min spread is only 1 point. This confirms a few things. First, it confirms that the teams at the top (with larger max-min spreads and finishing the month near the highs) are indeed creating separation. Second, the 1-point max-min range for the entire month for the bottom 40 teams shows the magnitude of craziness on that day of upsets at the end of December. Forty teams moved along a one-point range for the duration of a month, yet on that one crazy day, several teams made a 1-point move from the previous day.

Let's now take a look at the same thing for the Sagarin QC.

Without a doubt, this chart shows approximately the same phenomenon. Most ranks are near their lows for the month, and the higher-ranked teams experienced wider-ranging movements than lower-ranked teams. Needless to say, I'm already looking forward to the March QC Analysis to see what these exact same charts show after the upcoming month of basketball.

What a Difference a Year Makes.....February Edition

As usual, let's start with the KenPom QC.

Simply put, the top 3 teams in 2018 are approximately similar to or better than their 2017 counterparts. However, 2018 never gets close to 2017 until you reach the 30th-ranked team. These are potentially 2- thru 7-seeds in this range. With weaker teams at the top, I think it is very unlikely that all sixteen 1- thru 4-seeds advance past the first game this year. Though the gap between the two curves will shrink, I seriously doubt the gap will disappear in the remaining six weeks. We can only wait and watch, but we need to look at the Sagarin QC and see how it compares to this one.

The Sagarin curves show the same phenomenon. The only real difference between the Sagarin chart and the KenPom chart is that lower-ranked teams in the Sagarin chart are better in 2018 than 2017 whereas in the KenPom chart, most were along the same efficiency levels. I would like to see full confirmation of stronger teams at the bottom from both charts before making any rock-solid predictions, but I'm fairly confident in expecting the 2018 tournament to produce more than the 10 total upsets that we saw from the 2018 tournament.

With that concluded, I hope you enjoyed the February Edition of the 2018 QC Analysis. If the charts are not large enough, please let me know in the comments section so that I can enlarge them. The next article is due on Feb 12, and [teaser] some of the material from this article can be found in 2016 articles. Until then, thank you very much for reading my work, and I have one task left to complete for this article, which I have done below.

APPENDIX: What a Difference A Year Makes.....in the Mind of the Committee

Earlier, I made a fleeting reference to the differences in the way the two ratings systems (KenPom and Sagarin) are calculated. The issue that I wanted to discuss (and will probably emphasize in either the March Edition or the Final Edition) is the recent revelation that Efficiency Ratings will be provided to the Selection Committee. I want to point out that Selection Committee members have had the liberty to use any criteria to evaluate teams, but only the RPI was "officially provided." Now, these ratings systems that we use to evaluate teams will be officially provided to committee members. This will have a profound impact upon us bracket pickers.

1. We have always exploited the lack of understanding of the committee. Where the committee thought a team deserved a 3-seed and efficiency ratings suggested that team should be a 6- or a 7-seed, we have exploited this lack of understanding in identifying potential upsets. If the selection committee has access to this same information, our ability to identify potential upsets is greatly reduced.
2. The use of these advanced metrics by the Selection Committee introduces a new 'uncertainty factor' into the bracket. When the Selection Committee looks at these ratings systems, do they really know what they are looking at and will they evaluate and employ these new metrics appropriately? For the past few years, I've theorized how the Selection Committee evaluated and seeded teams (for 2016 and 2017, conference affiliation received undue priority when seeding). With the supplication of advanced metrics, will they use them in the selection process, but not the seeding process, or avoid them in the selection process, but rely on them in the seeding process (which would really hurt us bracket scientists)? To turn a phrase on the phenomenon, it used to be: "We bracket scientists know what the Selection Committee doesn't know." Now, it is best described as "We both know the same thing, but do bracket scientists know what the Selection Committee understands?"

This is one of the biggest reasons why I think the methodologies and the calculations of these ratings systems should be available to the Committee. They need to know what these ratings systems are measuring and how they are measuring it. I know for a certainty that every ratings system, mine included, sacrifices something in order to more correctly evaluate something else. For example, BPI caps margin of victory in its calculations in order to reduce the diverging impact of outlier data points. They are sacrificing accuracy of measurement in order to improve the precision of results (yes, precision and accuracy are two different things). It's like a mechanic telling me to press the red button on the newly installed NOS (Nitrous Oxide System) in my car without telling me what it does or the appropriate times to use it. So this doesn't get taken out of context, I'm perfectly fine with them having the information; I want to make sure they know what it is and the correct way to use it. For those interested, I did an article last year on this very topic, and this particular subject matter can be found in the final section (labelled "To Infinity and Beyond") of that article.