"My rule was I wouldn't recruit a kid if he had grass in front of his house.
That's not my world. My world was a cracked sidewalk." —Al McGuire

Monday, March 08, 2021

Championship Week & The Bubble: Does It Matter?

Every year around this time, bracketologists and college basketball experts talk about what teams need to do during Championship Week to "punch their tickets" to the NCAA Tournament. The more time I've spent looking at brackets, the more I've wondered if what teams do this week actually matters, or if the Selection Committee opinions are already formed at the start of the week and the only results that matter during Championship Week are the ones that decide an automatic bid.

Darius Johnson-Odom led the 2011 Marquette "Team Bubble Watch" to the Sweet 16

Photo by Andy Lyons | Getty Images

I evaluated Championship Week results of 100 teams from the past 5 NCAA Tournaments. I looked at the last 12 at-large teams in the NCAA Field and the first 8 teams in the NIT Field as the sample of the bubble. Factors included were the Conference Tournament win/loss records, the Quadrant result of those games (neutral court quadrants have not changed with NET), the Effective Seed before and after Championship Week, and the change in that seed.

Effective Seed is a new term to describe a team's placement before and after Championship Week in relation to the at-large cut-line. Effective Seed does NOT mirror the seed line assignments on Bracket Matrix nor the actual NCAA bracket. It only looks at teams in regards to how close they are to the at-large cutoff. The two primary categories are the Last Four In, which were assigned the Effective Seed of 11 because all of the at-large play-in games of the past 5 years and the NIT 1-Seeds (a.k.a. First Four Out) which were assigned the Effective Seed of 12 because they are the first teams typically not included due to the Committee filling in automatic bids on the 12-line. All other Effective Seed numbers indicate the degree away from those numbers, with each number being assigned to four at-large teams per year. To see all the raw data and get further explanation of that data, look here.

For the Pre-Championship Week Effective Seed, I used the last available projection from Bracket Matrix during Championship Week using the Wayback Machine archive, which the Bracket Matrix site administrator recommended. That was then compared to the Effective Seed of the actual Selection Sunday bracket.

What this article will investigate are common threads with teams that moved in or out of the field during Championship Week, the Effective Seed change of teams that had good (2-1) and bad (0-1) performances during Championship Week, the import of quality (Quadrant 1) wins and presence of bad (Quadrant 3) losses, and any common threads with teams that saw their Effective Seed line improve or decline by 2+.

Let's start by looking at the teams that started the week outside the field and "played their way in" to the bracket during Championship Week:

The first thing that jumps out is the lack of teams in this field. An average of just one team per year went from outside to in. Even more notable are the results. Four of the five teams that "played their way in" did so by adding no more than a Q3 victory, hardly something that moves the dial on a resume. The only team that arguably did enough to play their way in from the outside was Michigan, though with them being the only team to move from NIT 2-Seed to in the field, they seem more like an evaluation outlier than anything else.

What seems more likely is that, in large part, the Selection Committee simply looked at these teams differently than the Bracketologists. 2015 UCLA was 5-10 against the RPI Top-100 with 3 bad losses. In the same categories, 2016 Michigan was 4-12 with 0. 2016 Tulsa was 8-8 with 3, 2018 Syracuse was 7-11 against Quadrant 1+2 opponents with 2, and 2019 Belmont was 5-3 with 2. UCLA, Michigan, and Syracuse simply had bad resumes but seemed to benefit from playing in tough leagues that gave them more high-level competitive opportunities. Tulsa somehow moved up despite a poor Championship Week, indicating a clear difference of opinion with the Selection Committee. Belmont was less a case of moving in and rather a bid theft, as they went from presumptive OVC Champion to at-large candidate when Murray State won the automatic bid. Conclusion: Teams that move into the field do not do so on the basis of a strong Championship Week performance.

Next we'll look at other teams that won at least 2 Championship Week games to judge if Michigan was indeed an outlier:

Michigan again stands out as the only team here to move into the field, but what's even more interesting is seeing three teams that went 2-1 and went from in the field to out despite what would be considered a "successful tournament run." While none added a Q1 win, 2016 San Diego State, 2016 St. Mary's, and 2018 USC all added multiple wins without taking a bad loss and went from in to out. Not only that, but 13 of the 20 teams that went 2-1 saw either no improvement or an actual decline in their seeding position. 2018 would be the outlier as three of the four teams to go 2-1 did improve in the eyes of the Selection Committee, but that's likely a tough pill for USC fans to swallow as their team went from in to out in the same year. In addition, it's tempting to look at 2018 and say "you just need to add a Q1 win" to move up, there are 9 teams on this list that added Championship Week Q1 wins and only 4/9 saw their seed improve as a result. 3 actually added at least one Q1 win and saw their seed decline. Conclusion: Winning multiple Championship Week games does not insure seed improvement nor field inclusion.

Let's get a bigger sample of teams that notched at least one Q1 win:

This is the list of all the teams that added at least one Q1 win. Of the 16, we again see one team that moved into the field (2016 Michigan) and one team that fell two seed lines and out of the field (2019 NC State). 11 of the 16 saw their Effective Seed stay the same or decline despite adding quality to their resume. It's also worth noting that all of these teams lost to a Q1 opponent, so no one hurt themselves as a result of adding a bad loss. Conclusion: Adding quality wins in Championship Week does not insure seed improvement nor field inclusion.

So what do the teams that significantly improved their seeds have in common? Next we'll look at the teams that moved their Effective Seed up by at least 2 lines:

The first thing that jumps out is all 9 of these teams at least won one game in their Conference Tournament. While that looks significant, despite the range of victories being from Q4 up to Q1, I would point out that these 8 represent 66 of the 100 teams to win at least one Conference Tournament game. Why did these 8 improve measurably and the other 56 did not? It can't be the Q1 wins either, because as we noted before, there were 9 teams that added 2 wins with at least one in Q1 and just 4 of them are on this list. What this might indicate is that while Championship Week doesn't help much, it can hurt you, as only Richmond is the outlier of improving their position despite an 0-1 record, though they did remain outside the field. Conclusion: Winning doesn't hurt, but it is at best an erratic predictor of significant Effective Seed movement.

Let's look at the other end of the spectrum, starting with the teams that went from in the field to out:

Looking at this group of teams and seeing that 10 of the 11 won 0 or 1 Championship Week game seems like a strong indicator that you need to put in work to stay in the field or move up, but seeing as 13 of the 20 teams to win 2 games either stayed the same or declined in seed, that's a hard sell. What really seems to have happened is something not included in these charts: Bid Thieves. In 2015, it was the Selection Committee's different evaluation of UCLA and Wyoming stealing a bid for the Mountain West that knocked Colorado State and Temple out. In 2016, it was the evaluation of Michigan and Northern Iowa stealing a bid for the Missouri Valley that knocked St. Bonaventure and South Carolina out. In 2018, it was the evaluation of Syracuse, Davidson stealing a bid for the Atlantic-10, and San Diego State stealing a bid for the Mountain West that knocked Baylor, St. Mary's, and USC out. And in 2019, it was Murray State stealing a bid for the Ohio Valley, St. Louis stealing a bid for the Atlantic-10, and St. Mary's stealing a bid for the West Coast that knocked TCU, Texas, Clemson, and N.C. State out. While the Selection Committee clearly had different evaluations on some of those teams (particularly Texas) the bid thieves explain 8 of the 11 teams not included in the final field.I will also note the excellent work by the Bracket Matrix collective in 2017 as they had all the teams correct that year, with the only possible bid thief being Rhode Island, who was the last team in per Bracket Matrix but won the automatic bid to insure that prediction was correct. Conclusion: Bid thieves have more impact on the bubble than Championship Week "runs" that end in a loss.

So if teams that went 2-1 didn't conclusively move up, did teams that went 0-1 move down? Let's look:

This is the first data that indicates there is some value to Championship Week. Only 4 of 34 teams that went 0-1 saw their Estimated Seed improve while 16 saw it decline and 5 fell out of the field. As noted before, however, those teams that fell out were almost entirely down to bid thieves. There were 6 teams that saw their Estimated Seed fall by 2+. Interestingly, all 6 lost to Q2 or better opponents and 4 of them lost to Q1 opponents, which in theory shouldn't really hurt seeding. Of the 8 teams that took a Q3, or "bad" loss, 1 saw their Estimated Seed improve, 4 saw no change, and 3 saw a decline. Conclusion: Going 0-1 in Championship Week won't likely help you, but it's inconclusive as to how much it actually hurts a resume.

So what do teams that saw significant declines have in common? Let's see:

Of the teams that fell hard, 7 went 0-1, 6 went 1-1, and 4 went 2-1. None of those numbers are so significant that there's any clear thread. 12 of the 17 teams took Q1 losses, which shouldn't hurt the resume, and none took worse than a Q2 loss. Many of them fell out (including 2016 San Diego State and St. Mary's, which were projected auto-bids), but if you start in the 9-11 range and fall multiple lines, there's really nowhere to go but the NIT (only 2018 St. Bonaventure started in that range, suffered a 2+ line fall, and remained in the field). Again, these falls go back primarily to Bid Thieves and there's no consistent thread that connects all of these teams. Conclusion: Teams falling significantly during Championship Week have no consistent weekly trend and appear to be a result of a different Selection Committee evaluation and Bid Thieves rather than poor Championship Week performance.

Why did Andrew Rowsey's 2018 Marquette team go to the NIT?

Photo by Mike de Sisti | Milwaukee Journal-Sentinel

So after all this data comes into view, what are the takeaways? First, while we tried to find commonalities with the teams that made significant jumps up or down in seed as well as in or out of the field, the simple reality is that only 26/100 teams saw any meaningful Effective Seed jump and only 16/100 saw movement in or out of the field. That means that, as 74/100 teams saw their Effective Seed either remain the same or move by no more than 1 line (between 1-8 spots on the S-Curve) and as Bid Thieves account for half the field movement only 8/100 teams saw movement in or out of the field on the basis of perceived merit, none of those results were about what happened during Championship Week.

The idea of a team "playing their way in" is a fun myth, but a myth is all it is. It's a nice argument that 2016 Michigan played their way in with their 2-1 run, but considering that 5 teams from outside the field went 2-1 and didn't get in and another 3 went 2-1 and moved from in to out despite their success shows that Michigan was an outlier that was simply evaluated differently by the Bracket Matrix contributors and the Selection Committee membership. To answer the question about Andrew Rowsey's team, it wasn't Marquette's 1-1 performance in March 2018, it was their inability to garner marquee wins and the bad loss to DePaul that kept them out of the field.

Ultimately, what happens this week is certainly fun, but the numbers show that the only results that will have any tangible impact on the bracket we see come Selection Sunday are the games that decide automatic bids. Because of that, it's far more likely that teams will see their bubbles burst (average 2.2/year) than find themselves unexpectedly in the field (average 1.0/year). So enjoy the games, enjoy Championship Week, but when it comes to definitive statements, realize that the field is set by the results from November to the first week of March, not a small handful of games over the course of 3-5 days.

No comments: