Mid-season Thoughts

First off, I hope everyone is having a wonderful holiday season. Second, I wanted to do another article similar to the Opening Week Thoughts article, just to document what I've seen and thought about current college hoops season up to this point, so it will be a lot of opinion and conjecture. With roughly 95% of the remaining games being conference match-ups, it's perfect timing for such an article, and if anything else, it should have a lot of meaty thoughts for sports conversations around the holiday tables. Instead of reviewing important match-ups like the Opening Week article, I'm taking a conference-based approach in this write-up.

Might Casey Model

If you've ever heard of or read the Earnest Lawrence Thayer poem "Casey at the Bat," then you're on the right track for the subject of the current article. Mighty Casey is batting with two runners in scoring position and facing an 0-2 count. While we all know that Mighty Casey strikes out swinging, I want to know how many tournament teams have faced the same situation (an 0-2 count) and how they performed. In college basketball terms, an 0-2 count will be defined as back-to-back tournament defeats in the R64 and facing the threat of a strike out (a third tournament appearance).

Profile of a National Champion (2022-23 Update)

Since there is very little predictive work that can be done in Nov and Dec, I have found it more productive to use this time to review and improve our tools. As promised in the Welcome article, I wanted to follow-up on the Championship Profile analysis in the Final Article for the 2022 Tournament. Over the off-season, I had a feeling that KU could have been a 1-DQ candidate (and maybe even a 0-DQ if methodology could be tweaked), Thus, I'm going to use this opportunity to do a full investigation into the tool. First, I'm going to review the analysis done in the final article. Second, I'm going to review and update the methodology that went into the analysis. Third, I'm going to redo the analysis and see if my hunch in the off-season had merit. Fourth, I'm going to add another metric to the tool and see how it applies to this year's crop of contenders.

2022 ANALYSIS FROM FINAL ARTICLE

(Essentially, this section is a copy-n-paste operation with my current commentary in parenthesis plus light blue to identify it from the original work.)

Let's start with the obvious rules, and work from there. First, National Champs have never been lower than an 8-seed (1985 NOVA), they have never lost the first game in their conference tournament (italics), and in the Advanced Metrics era (2002-present), they have never had double-digit losses (strike-through).

1. GONZ,   ARI,       KU,       BAY
   
2. DUKE,  NOVA,   AUB,      UK
3. TXTC,   TENN,  WISC,     PUR
4. ARK,     ILL,       PROV,    UCLA
5. CONN,  HOU,     IOWA,   STMY
6. ALA,      COST,    LSU,      TEX
7. MIST,    OHST,    USC,     MURR
8. BOST,   HALL,    SDST,    UNC
   

(No mistakes by yours truly, so far) After removing non-power conference teams (STMY, COST, MURR, BOST, and SDST) and hypocritically non-scientifically keeping GONZ, here's our National Champion contenders and their disqualifiers.

You will notice that I've added one extra column (6W). If you follow this link, it will explain the historical significance of this attribute. (TL/DR: An NCAA title is six consecutive wins against power conference quality teams, so if a team has already done this in the regular season, they should be more likely to do it again in the post-season). (First mistake: A full read of the article gives an alternate qualifying criteria which exempted 2000 MIST, and this exemption would have applied to 2022 KU too. Thus, they would be 2DQs instead of 3DQs.)

UPDATE TO THE METHODOLOGY

The only area of focus for this update to methodology will be the PG production. The criteria for Title, E8-Coach, 6W and Prev Trny are pretty much set in stone, as are the three preliminary criteria that reduced the field of 32 possible contenders down to the listed 17. I am also working on a structural approach to post-production, so it will not be addressed here. Furthermore, this methodology update to PG production will bring it in-line with other similar methodologies used in tools like the Experienced Talent Model and the Return & Improve Model (which I have side-lined due to the growing influence of the Transfer Portal on team dynamics). First, let's look at the updated list of past champions.

I've removed the post-production section of the profile for viewing convenience since it's not a focus for this article. Some of the names of PGs and the values for PT/AST have changed from the original article. For the names that changed, this is due to the updated selection criteria (discussed shortly). For the values that changed, this was due to tournament games being active in my spreadsheet pivot tables. Since we are working with the predictive capacity of statistics, the point totals and assist totals should not include the results of games played in the NCAA tourney. These values would have been unknowable at the time of BCW when the data would have been gathered for the model. These values are only slightly different than the original article. My pivot tables only go back to the 2011 season, so only the 2011-2021 seasons have pre-tourney data (as well as 2003 as I was able to find reliable pre-tourney totals for that season). I've also added an untitled column, which I would term the minimum point impact. It simply doubles the assist value and adds that sum to the point total (for example, 2021 BAY would be 5.1*2 + 15.6 = 25.8). I'm aware that assists can lead to three-pointers, but since I'm only concerned with the minimum, I use two as the multiplier, so this value calculates the minimum amount of points that the point guard accounts for his team.

Now, let's look at the criteria. The first step in identifying a team's PG is starts and assists. Logically, a team's PG isn't going to come off the bench and a team's PG will most likely lead the team in assists. Thus, the "qualifying PG" will start more than 80% of a team's games (trying to account for injuries or a player that takes over the job during the season) and leads the team in assists. However, most teams don't follow the simplistic PG-SG-SF-PF-C archetype, as some run dual-PG offensive sets (i.e.- 2021 BAY) or they have multiple PG personnel on their roster (i.e. - 2014 CONN). So, how do we account for these multi-PG teams? If a second player's stats are within the following values of the qualifying PG, then they also qualify as a PG and their totals are averaged with the qualifying PG's totals:

• Within 75% of the qualifying PG's number of starts.
• Within 80% of the qualifying PG's minutes per game.
• Within 50% of the qualifying PG's assists per game.

I probably should have noted the team-by-team application of the rules, but the only real outlier for these rules is 1999 CONN, as G Rick Hamilton also qualifies under all three criteria. Since I don't have reliable game-by-game data past 2011 and being the third starting guard on the team (which would imply the SG role), I decided to leave him as the sole exception to the rule (Ironically, I had to do the same thing for 2022 ARI below). As we'll see from the application of this updated criteria (which brings this model in-line with the criteria of other models), some of the 2022 teams will have DQs removed and others will have another DQ added.

RE-DO OF THE 2022 CHAMPION PROFILE ANALYSIS

Let's look at the new table. DQs in blue changed from the original table (pictured above).

First, I want to start with all of the outliers. 

• ARI's Mathurin can also qualify, but I gave him the Rick Hamilton exception.
• DUKE Moore led the team in AST, but on the roster, he is listed as a F. Likewise, DUKE Roach lost the starting job down the stretch of the regular season once Keels returned from injury, but regained the starting job in place of Keels for the NCAA tourney. More importantly, by the rules, they both qualify.
  
• TXTC had four player qualify, but between the injuries and the raw numbers, I just settled on two since the ending values showed little deviation.
• PUR Thompson and Hunter failed to achieve starts within 75% of the AST leader Ivey, but in the original analysis, I chose Hunter because he started at least 50% of the games, was the chosen starter down the stretch of the regular season, and had twice as many AST per game as Thompson.
• TENN G Vescovi actually does qualify under the new rules, but in the original analysis, I did not count him because Zeigler was better qualified statistically for PG yet he didn't have the starts.
• ARK JD Notae qualifies by himself since Davis did not qualify due to lacking starts.
  
• PROV G Bynum led the team in ASTs but only started ten games, which rules him out under the new guidelines. Durham's does not meet the threshold for PG production, and as a result, PROV adds another team DQ to their count.
• CONN G Jackson qualifies under all three rules, which brings down PG production below the acceptable threshold, adding another team DQ to their count.
• HOU G Edwards qualifies under all three rules, which adds a SR to PG class and adds points to the production without dropping assists below the acceptable threshold. These remove two team DQs from HOU from 5 to 3.
• IOWA was a weird one. Toussaint was the AST leader, but lost the starting job in Feb. Perkins assumed the starting job but only had 40% of the starts and the worst AST of all three candidates. Bohannon had 100% starts, 100% minutes, but around 60% of the AST level of Toussaint. None of the three cross the PG production threshold, so I just left Toussaint. The only issue is a typo on my part, as I copied Toussaint's MPG for his PPG and his PPG for his APG. This has been corrected and a team DQ has been added to IOWA's total.

Second, with the changes in methodology implemented to the same set of data, we now have three Tier 1 contenders for National Champion: GONZ, KU, and TENN. Of these three, I would say TENN would have been the weakest choice due to lacking post-production. Remember, 2022 was the year of the scoring big, and TENN C Fulkerson didn't have numbers anywhere close to the post production of the 2022 pool of contenders nor the post production of past champions. GONZ would have been the next on the chopping block for two reasons. Not since 1990 has a national champion come from a non-power conference, and GONZ did not have close to the NBA talent as 1990 UNLV. Second, GONZ only returned 45% of their minutes and 45% of their scoring from a 2021 National Runner-up team. Very rarely do teams return that low of a percentage and advance further than the previous year's team, which would need to be a National Title to exceed the previous year's National Runner-up performance. Thus, the Championship Profile tool would have left KU as the odds-on favorite for the title.

NEW CRITERIA FOR THE CHAMP MODEL

Intuitively, the Champion Profile Model tries to quantify various qualitative attributes that a Champion should have and that past champions had possessed. For example, production matters to a champion. It's hard to win games if you can't produce, or at least out-produce your opponent. Another example is winning experience. Teams that win conference titles are more likely to be ready to win national titles. In other words, all of the criteria have purpose. One attribute that I believe national champions probably have is chemistry, but I haven't found suitable ways to quantify it until now. I've examined the impact of team composition from the perspective of home-grown versus transfers. Considering the impactful nature of the transfer portal on college basketball (and other college sports too), is there significance in the thought of teams that play together, then grow together, and ultimately win together? I've looked at the same list of national champions and how they've been impacted by transfers.

• 2022 KU: 5Y transfer Remy Martin was a starter for KU until his injury in Jan. For the rest of the season, he came off the Bench except for 1 start on senior night. His starting spot was eventually picked up and retained by Jalen Wilson. 5Y Jalen Coleman-Lands played for three different teams before joining KU, with his only start coming on senior night as well. 
  
• 2021 BAY: Davion Mitchell and Macio Teague were full-time starters for the title run. They both transferred in and sat out the 2018-19 season and fully started for the tournament-less 2019-2020 season. Thus, they practiced without playing for a full year, then they played for a full year, and finally they brought home a title in the third season.
• 2018 NOVA: Eric Paschall followed a similar route. He sat out the transfer year (ironically 2016 if you know your National Champ history), played a full year with NOVA on 21mpg and 8 starts in 2017, and then became a full-time starter in the 2018 title year.
• 2014 CONN: Lasan Kromah transferred into CONN the same year they won it. He did not get his first start until a non-conference game against HARV after CONN started 0-2 in conference play. Kromah lost the starting job as CONN entered the conference tourney and never regained it. Looking at the team starts, Kevin Ollie may have been trying to shake things up in the roster as three players (Boatright, Napier and Daniels) started every game they played, and five players (Kromah, Giffey, Brimah, Calhoun and Nolan) fought for the other 80-84 starts. In short, Kromah wasn't a primary starter, nor was he the starter when CONN entered the NCAA tourney.
• 2002 MARY: Byron Mouton follows a similar pattern as 2021 BAY and 2018 NOVA. He transferred into MARY in 1999-2000, started 83% of the games on the F4 2001 team, and fully started on the NC 2002 team.
• 2000 MIST: Mike Chappell transferred into MIST in the 1998-1999 season, and then he played 14mpg during the 2000 title season with only three starts, one in the non-conference and two in the conference regular season. He was not a primary starter for the roster and possesses a similar profile to 2022 KU's Coleman-Lands.
• 1998 UK: Heshimu Evans is probably the biggest outlier of them all. He transferred into UK and sat out the 1996-1997 season. In the 1998 title season, he played more than 20 minutes a game, but only four starts. From my research, I've found that one of these was a conference game and one of these was an NCAA tourney game. The other two are unknown, which leads me to believe he was not a primary starter for UK down the stretch. First year head coach Tubby Smith also played a nine-man rotation which helped bolster his Evans's numbers.
• No other national champion had a transfer with any starts, and many didn't even have a transfer on the team.

Based on these accounts, we can build several summaries.

1. Most national title contenders do not have a transfer in the starting five. This accounts for the prototypical champion plus 2000 MIST, for a total of 16 champions.
   
2. The most common exception to Rule #1 is a transfer that sits out one year (practice-only), plays significantly the second year, and wins a title the third year (2021 BAY, 2018 NOVA, and 2002 MARY). For 23 National Title contenders, Rules #1 & #2 account for 19 of them.
3. The next most common exception to Rule #2 is a transfer that randomly acquires limited starts (less than three) but is part of the team's rotation (1998 UK and 2000 MIST, and also explains half of 2022 KU). For 23 National Title contenders, Rules #1-3 account for 21 of them.
4. For teams with transfers that acquire a significant number of starts, these transfers did not retain their starting role by the start of the NCAA tourney (one due to injury and one due to roster incentive moves by the coach). Though this explains the final two outliers, this group is also the most recent (both happening in the last ten years). More on this rule later.
5. The transfer criteria also overlaps with another criteria in the Champions Profile Model: The E8-Coach criteria. Only two NCAA champions have had a coach that had not previously reached at least the E8 before making the title run: 1998 UK and 2014 CONN. This may be something to keep in mind when handing out team DQs during BCW.

The real question is how much will the transfer portal push the bounds of this criteria? The transfer portal started in 2018. If you use that year as a dividing line, then everything that happened in the seventeen year span between 1998 and 2014 has repeated itself in the five-year span from 2018-2022. More importantly, Rule #1 was never broken in the seventeen-year span. Now that transfers are modus operandi in college basketball, have we come so far so fast that this is the year that breaks Rule #1? Many of the teams this year that I could logically identify as title contenders have a transfer on their roster and in a starting role. When we approach BCW and the research for this tool, I'm already curious as to the number of title contenders that will violate this criteria, but we won't know until then. As always, thanks for reading my work and I hope to have one more article between now and the January QC Analysis already listed on the schedule.