Franketology: February 4, 2025
Last week I ranted and rambled about KPI and metrics, when I had set out to talk a little bit about my methodology, because I think transparency is important for metrics, but also bracketologists. I am just one man, with a full-time job, 1 podcast to co-host, and currently attempting to get a 2nd podcast off the ground. Also, bracketology is not easy. At the end of the day, the best Bracketologists might miss some things, even the ones who make a full-time living by allegedly being good at this stuff. For instance, here’s the BracketMatrix.com rankings (179 qualifying bracketologists) of the well-known pundits and major outlet bracketologists—aka the folks doing this for a living—is a lesson in “be careful who you trust”:
71. Mike DeCourcy - Fox Sports
77. Jeff Borzello - ESPN
82. Adam Zdroik - RotoWire
98. Brian Bennett - The Athletic
118. Joe Lunardi - ESPN - This is the big one. Joe is the godfather. No one would be here doing this without him (I mean sure, SOMEONE would’ve started predicting the bracket, but Joe did was the first prominent guy to do it). So it surprises many people to see how many students have exceeded the master, but it’s the truth. If you’re citing a single bracketologists (something that should be avoided, more on that below), it should NEVER be Lunardi these days, unless you’re attempting to point out the errors or bias in his brackets.
139. Kevin Sweeney - Sports Illustrated
140. Bill Bender - Sporting News
164. Jerry Palm - CBS Sports
For comparison’s sake, I am ranked 17/48 in the “Bracket Newbies” list, and finished 90/227 in my first season last year. I’m thoroughly green, and this just mediocre so far.
As noted, I rarely, if ever, cite a single bracketologist. The better option is to cite the consensus seed at BracketMatrix.com. I will cite individual bracketologists if a team is on the rise or the fall. Why? Bracket Matrix updates can sometimes be delayed, and just by the nature of the exercise, it can lag a few days in accounting for very recent games. This is due to the structure of the Matrix. There are currently 65 bracketologists. If Bracketologist X releases every Monday, and there are big results Tuesday and Wednesday, that, in the aggregate, causes the Matrix to lag a bit, as Bracketologist X won’t update his bracket until the next Monday, and until then his dated bracket is pulling the consensus seed in the wrong direction.
For instance, the current matrix includes 65 brackets, of which I am one. Not every Bracketologist gets included in the Matrix. There are certain requirements, one is to have a 5 year average greater than 0, which effectively means you outperformed the average bracket more often than not. So in my 1 year I was above-average, and my bracket now appears on the matrix. Unfortunately, my bracket has not been updated since 1/28, and the Matrix as a whole has not been updated since 2/1 (like many of us in this space, the guy runs the Matrix as a hobby, and it’s a huge undertaking I imagine)
So in other words, if you really feel the need to cite individual Bracketologists I trust:
YAGO Brackets - The current reigning # 1 in the standings, currently his 5th season of bracketology. Very consistent, but has not yet won it all.
1-3-1 Sports - 1-3-1 is even more consistent than YAGO, coming in at 2 in the standings, but having won in 2018. That 2018 championship is not even including in his ranking any more, and the man is still in 2nd place. That’s consistency and track record.
March Madness 25 - This bracketologist changes his site every season, which can be annoying, but god damn he knows brackets. As painful as it is to admit, he was one of the Bracketologists to nail SJU, having us in the First 4 Out with Hall, Indiana St. and Wake Forest, which was realistically where we belonged, fuck the committee.
RCerulo Bracketology - I’m not sure I trust R. Cerulo yet, he’s only 132nd in the standings, but he is the reigning champion from last season. Of course last season might be an anomaly. It certainly took the collective wisdom of Bracketologists by surprise, as it was the lowest average bracket score in the past 5 seasons: 338.6, vs. 356.5 in 2023, 347, 346.4, and 341.9.
Others who I think do a great job but may or may not necessarily have the highest rankings: 801 Bracketology, T3 Bracketology, Making the Madness (Jonathon Warriner), NYK Brackets, JBR Bracketology, TSB Bracketology and Heat Check (Lukas Harkins). Can find most of these folks on Twitter if you’re so inclined.
Back to the point, in other words, I might overlook things. I may have your team seeded lower than you like, or maybe there’s no consistency week to week, i.e. I ranked Team A as a 5 seed this week and a 6-seed next even though they won both their games. Or perhaps the inverse, a team lost both games but didn’t drop as much as expected. A lot of that boils down to my method.
I pull all the relevant data into a worksheet. That will include all records, quad records, SOS figures and metrics for every team.
Check the data, make sure everything imported properly. Address formatting issues. Add in columns for calculating averages and win percentages by quads.
Sort by conference, then by predictive metrics averages to determine conference winners. Some bracketologists use the standings, others use different individual metrics. I chose the predictive metrics average for several reasons: 1) we’re attempting to predict who will win the conference tournament, and so predictive metrics have the most value in that regard; 2) averaging the 3 resume metrics will give you the closest outcome; 3) this can also occasionally result in bid thievery happening organically within my bracketology, as in some middling conferences the team with the best predictive metrics averages isn’t always an at-large team. At the moment, this is not occurring, but it has occurred at points this year in the Big West, where one of UC San Diego and UC Irvine was an at-large, but the other had lower predictive metrics, making them conference champs in my bracketology and stealing a bid. In those cases, the higher predictive average team wasn’t an at-large caliber, but the 2nd place team was (since resume metrics are more important for inclusion, while predictive metrics are only used in seeding).
This also dovetails with the new NIT inclusion rules nicely, as I occasionally dabble in NIT Bracketology in this space. For those who do not know, here’s the current NIT bid rules:
16 “exempt” bids are given top the top 2 teams each in the ACC & SEC that did not make the NCAA Tournament, plus 1 team each from the other top-12 conferences, as judge by the average of all team sheet metrics. The 16 “exempt” bids are guaranteed to host a game
Any team that wins its conference regular season championship, but does not get a bid to the NCAA Tournament, is not in a top-14 league (top-12 + SEC & ACC), and has a Team Sheet Metrics average <=125 gets an automatic bid
The rest of the teams are chosen at-large. Given the emphasis on all team sheet metrics average for exempt & automatic bids, that will be my primary guiding factor in determining NIT at-large bids
Back to proper bracketology, I sort by resume metrics and eliminate every team with a resume metrics averages greater than 60
Next I filter out the auto-bids and lock in a bid for any team better than 30 in the resume metrics average. No team lower than 36 in resume metrics average has been left out in the NET era. While there was much hullabaloo last year with St. Johns at 32 and Indiana St. at 28 becoming the 2 best NET ranking teams to be left out of the NCAAT in the NET era, the same logic does not apply to the same statement as resume metrics, since the committee actually uses resume metrics. The NET was never the sacred cow some people—myself included—thought it was. I think, theoretically, the committee could leave out any NET team if they do not have a strong resume, which is judge mostly by the resume metrics (because the committee doesn’t watch enough ball). But generally if you have a high NET/predictive metrics, you’ll have a solid resume.
With an eye predominantly on the resume metrics, I fill out the remaining at-large bids. While resume metrics remain of paramount importance, the full resume is considered for these teams. I also have some hard-and-fast rules this late in the season:" No Q1 victory, no bid. No victory over an at-large team, no bid. Other factors considered beyond resume metrics include quad records, SOS, and predictive metrics. I’d say I mostly focus on wins over the at-large field (as determined by the Matrix) and Q1 wins and Q1 win %. Win % can be of particular importance. Pulling an example from below, Clemson has a 2-2 Q1 record, that’s not great at this juncture, but they’re still well within the field with a 29 resume metrics average. Normal, I would ding a team a seed line or 2 for only having a Q1 win. However, Clemson has a Q1 record of 2-2 (.500), which keeps them competitive as 28th in my True Seed list, a 7-seed, right behind teams with 4 Q1 Ws (Louisville and Miss. St.), but a worse with percentage (both at .444)
After spot-checking to ensure everything looks good, particularly around the bubble, we filter out all of the teams who didn’t get a bid, leaving the 68-team field.
Next we sort the field by predictive metrics average. This is our jumping off point for seeding. In addition, we use the same factors listed in 6 for determining the bubble teams, to adjust a team up or down from the seed line dictated by its predictive metrics average. For instance, Houston has the best predictive metrics average in the country (1.33). However, they are only 10.67 in resume metrics, and have a Q1 record of just 3-4, with just 4 wins versus the field. The 1-seeds in my bracket have 12-1, 5-2, 6-3, and 5-4 Q1 records, and each team has more Q1A wins than Houston has total wins. Therefore, Houston gets bumped (in this case to a 3 seed). Gonzaga is in a similar boat, with predictives average of 13 and a resume average of 46, and a Q1 record of just 2-6. That dropped them down to a 9-seed.
Once the field has been “true seeded", I review the list thoroughly one more time to make sure I haven’t missed anything simple, or made decisions that in hindsight don’t make sense. Once that’s done, I publish.
Now, I do not know other bracketologists methods. But I think the advantages to my method are a few:
Using the predictives average for auto-bids
I re-do this process EVERY TIME. I imagine some folks are just moving people up or down based on results, and I fear that can be too reactionary. I do a complete analysis every time.
I think the methodology is sound, leaning predominantly on the resume metrics up front for inclusion, but the predictives for seeding, but keeping it loose enough to account for the full resume.
The lone potential downside is that I may be prone to make mistakes where a team can swing a couple seed lines from one edition to the other for seemingly no reason. I actually think this is a benefit. Because it allows me to detect potential bad seedings by me every time I do a fresh bracket.
In any event, cut me some slack. Constructive feedback is always welcome on Twitter. Don’t be a dick (a simple, but pretty solid rule for life).
With that out of the way, let’s get to the current Franketology bracket:
Direct link to spreadsheet here.
True Seed List
Auburn
Duke
Alabama
Tennessee
Purdue
Texas A&M
Kansas
Florida
Arizona
Iowa St.
Michigan St.
Houston
Texas Tech
Marquette
Illinois
Kentucky
Wisconsin
Ole Miss
Missouri
St. Mary’s (CA)
UCLA
Michigan
St. John’s
Maryland
Memphis
Louisville
Mississippi St.
Clemson
Creighton
UConn
Baylor
Oregon
Gonzaga
Oklahoma
Utah St.
West Virginia
Last 4 Byes
Texas
New Mexico (AUTO BID, NOT LAST 4 BYES)
Ohio St.
Vanderbilt
Georgiw
Last Four In
Nebrasketball
SDSU
BYU
UCF
Other Auto-Bids
Drake
UC San Diego
Arkansas St.
VCU
McNeese St.
Yale
GCU
Liebrty
High Point
Akron
Samford
UNCW
Lipscomb
South Dakota St.
Milwaukee
Northern Colorado
Norfolk St.
Southern U.
CCSU
Quinnipiac
Bryant
Little Rock
American
FIRST FOUR OUT
Arizona St.
Wake Forest
UC-Irvine
Pittsburgh
NEXT FOUR OUT
Arkansas
USC
Indiana
North Carolina