Bracketology Post-mortem

In this post, I am going to look back at my bracketology and its underwhelming Paymon score of 353. While it’s difficult to beat the Bracket Matrix aggregate (366) in any given year, I also performed below the median score of 358. I want to analyze where I went wrong, particularly with respect to consensus, why I went wrong, and what to take away. Enough dwelling for now, let’s rip off the band-aid…

Green = correct seed line

Yellow = off by one seed line

Orange = off by multiple seed lines

Red = not in the tournament

Hate the 5-8

More accurately, teams ranked 20-30 on the official seed list, but I am in absolute disgust with how I handled this part of the bracket. I think I took what the committee said last year a little too literally when they claimed to dock the Mountain West for building their resume mostly, if not entirely, through conference play. I thought BYU’s profile was clearly worthy of being a top six seed (if not better), but moved them down a seed line for having such a poor non-conference. Not just a weak schedule, but I expected the committee to care more about not even having a quad 2 victory outside of conference play. I think I can fairly take away that games in November matter equally to those in March without any lean that I may have theorized.

If you thought one 6-seed was bad, my 7-line was even worse with no correct teams. I’m not sure what I was thinking in this general range, with some of my placements being so arbitrary even without considering hindsight. I really should have had both UCLA and Kansas on this seed line considering that the Bruins weren’t better than 25 in any one metric and the Jayhawks weren’t worse than 26 (for what it’s worth, they were 28 on the official seed list). Kansas even had the excellent non-conference wins over Duke and Michigan State, giving me no excuse to put them on the 8-line with my own logic. I couldn’t tell you exactly what I was thinking, but I believe I convinced myself that the committee wouldn’t like their 4-8 true road record that does not include a win over an NCAA Tournament team. Back to UCLA, I thought their balance of quality victories with proof that they could get it done in many environments would propel them to a 6-seed despite no metric loving them. I clearly wasn’t alone in this sentiment with the aggregate also having them as a 6, but I know I could’ve had them as a 7 and I didn’t.

I’ll group Saint Mary’s and Gonzaga together because I fell into the trap of “fear of being very wrong” when it came to this WCC pairing. Last year, they both managed to crack the 5-line despite bracketologists having them a seed line or two below, making me think the committee had some weird level of respect for the WCC. I have no problem with how they got seeded; in fact, I think the committee seeded them properly. I arbitrarily moved these teams up a seed line after comparing their profiles this year to last year and thinking they looked better now. I was terrified that they would both somehow find their way onto the 5-line again, duping myself into having them higher than I should have. Props to the committee for getting the WCC seeds correct, I just didn’t expect it.

Last in this grouping to discuss is Missouri and Marquette. If I was truly scared about poor road records hurting a team’s seed, I honestly should have swapped these two teams. I was higher on Marquette than consensus for most of the bracketology season, only to drop them to an 8 after thinking their numbers didn’t quite stack up with all of these other teams who could easily pass for a 6 or 7-seed (Golden Eagles included). Yes, it was the non-conference again that made me higher on them initially, and I really should have stuck to that, but “fear of being very wrong” reared its ugly head with the possibility of putting Mizzou as an 8. Even with a 3-7 true road record, I anticipated wins at Florida, vs Alabama, and top 16 NET/KenPom/Torvik being too good to be an 8-seed (somehow did not arrive at the same conclusion with Kansas). In this case, I was right! Not right enough, though, as the Tigers found themselves on the 6-line. The way I handled these two teams perfectly demonstrates why I shouldn’t have been so scared with my seeding. If I had Marquette as a 7 and Missouri as an 8, I would have lost a point for being multiple seed lines off on Mizzou, but gained two points for getting Marquette’s seed correctly. It’s better to be correct and very wrong than consistently slightly wrong, and I need to consider this in future seasons.

If I were to have been consistent with myself, I still would have made some mistakes in this range: Ole Miss still would’ve been a 5, Memphis and UCLA still would’ve been a 6 (even without a top 24 metric), Louisville still would’ve been a 7, and I would’ve made an even greater mistake with Missouri and put them on the 8-line. Regardless, I feel like I missed some critical points with Saint Mary’s, Gonzaga, Marquette, Kansas, and BYU that were extremely avoidable.

My Single Biggest Miss

When comparing my Paymon score to the matrix aggregate, I lost all of my points for having seven fewer teams properly seeded. We both had the same number of teams in the tournament, and I actually had one fewer team off by multiple seed lines. The matrix projected #5 Memphis as an 8-seed, and #8 Louisville as a 6-seed. I had these teams as a 6 and 7, respectively, but was two seed lines off on a team the matrix nailed.

While my biggest overseeding stand with Memphis was somewhat successful, my underseeding stand with Baylor was quite the opposite. Yes, I recognized the win over St. John’s, another strong non-conference victory over Arkansas, and predictive metrics in the top 30. The problem was that I didn’t see anything else that popped off the screen on this profile. A home win over Kansas isn’t as impactful as it typically would be, and that’s all of their wins over tournament teams compared to 12 losses. As I discussed ad nauseam in a previous post, I really wanted to have West Virginia over them as their profile boasted better wins overall and demonstrated more ability to pick up tough wins without DeVries than Baylor did without Oijanwuna. I was below market on both of these teams with expectation that these injuries would be considered, but expected the committee to say tiebreaker Baylor as a result of better metrics. Thus, they were both on my 11-line with Baylor just avoiding the First Four and WVU in Dayton.

I really thought the Bears would be a 10 at best, especially considering their losing record against the top three quadrants. Clearly, my process did not pan out here as the analytics trumped racking up quality wins as we moved further down the bracket. This runs contrary to how teams were typically seeded farther away from the bubble, and I’m not sure how to account for this discrepancy in the future. The committee is no stranger to leaving out teams with predictive metrics in the 30s, making me think an 11-seed was more than reasonable. While many of my misses in the 5-8 seed lines resulted from poor process, I stand by this seeding.

To Mid-Majority and Beyond

We hear all the time that the committee disrespects smaller conferences and panders toward the big guys. I felt like this discourse was overblown and frankly not supported by evidence over the past few years. Sure, the Mountain West was collectively underseeded in 2024, but the American, Atlantic 10, and WCC were all given better seeds than predicted by the matrix. Even Drake managed to snag a 10-seed while most bracketologists pinned them as an 11. In the post-COVID era of March Madness, only 2023 FAU sticks out as a team who got an unfairly poor read from the committee. As a result, I expected the committee to continue respecting mid-majors who performed well and took advantage of the opportunities on their schedule.

Enter the 2025 Drake Bulldogs and UC San Diego Tritons: both scoring critical non-conference wins against tournament teams while mostly holding serve in conference play en route to an automatic bid. Drake’s analytics weren’t particularly flashy with all of the close wins against some not-so-great teams, but 6-0 against Q1/2 is a truly elite mark. Their profile simply looked better than the Bulldogs of 2024, making me confident that Drake would be a 10 again. I was honestly a bit scared that the committee would have them even higher with top 35 marks in KPI and SOR, but this profile was treated in a fundamentally different manner compared to last year.

We see a similar story with this UCSD squad: a couple bad losses, but a critical win at Utah State propelling this resume. They were 2-1 against both Q1 and Q2, and, unlike Drake, had a top 40 NET and KenPom suggesting that this team was as good as their resume indicated. Especially compared to North Carolina and Xavier, I’m really not sure what box this team does not check to be ranked below these teams. I was extremely confident that the Tritons would find themselves above a 12-seed considering their overall body of work, but it seems like power conference bias crept its way back in the bottom half of the bracket. I thought VCU was more vulnerable to being knocked down to a 12 without a statement non-conference win, but their analytics, their victory over Colorado State aging very well, and the historical respect given to the A10 made me keep VCU as an 11. As you may expect, I still stand by the process by which I arrived at these mid-major seedings, and I’m genuinely unsure of how to approach this in future years when the year-to-year difference can be rather stark.

At the Bottom and the Top

It was hard to have a true edge over the matrix, especially upon reaching the auto-bid-only zone. The aggregate nailed every team seeded 12 through 16. It makes sense considering that there were noticeable drop-offs between seed lines when analyzing the conference tournament champions. I opted to take a stand with Norfolk State, anticipating that their relatively solid resume metrics and victory at High Point would give them the nod over Bryant, but I was wrong. Something tells me the committee does not watch these games particularly closely, and would rely on the predictive metrics as a crutch even more toward the bottom of the bracket than the top. In the future, if the team sheets are remotely comparable, I should probably give the benefit of the doubt to the team viewed more favorably by KenPom.

Back to the protected range, we collectively messed up by having Michigan as a 4-seed over Purdue. I’m genuinely not sure what the justification is for the Boilermakers to be 14th overall: losing record in Q1, only top 16 in KPI (which was seemingly ignored by the committee in other parts of the bracket), and predictives hovering around 20. Their resume screams 5-seed. Onto my incorrect stand, I expected Oregon’s dazzling non-conference to contrast with Arizona’s poor outing and give the Ducks the edge on the 4-line despite worse predictives. This one is honestly on me and I should have been less aggressive with the thought that teams needed to prove non-conference competency or else their seed would’ve been hurt. With KenPom and Torvik in the 30s while Arizona ranged from the low teens up to 9, I definitely could have been more generous with the Wildcats as their wins were comparable to Oregon. I think I was also hesitant to move Arizona up in case conference tournaments still did not matter much, but it seems like their Friday victory over Texas Tech certainly benefited their seeding while other teams were unaffected by wins or losses from Friday onward. I probably should have moved Oregon down to the 5-line, though I didn’t believe any of Purdue, Clemson, or Arizona had a great case to be moved up if I made that switch. This wasn’t egregious, unlike how I handled the teams seeded 20th-30th overall, but I probably should’ve gotten there with the Wildcats on the 4-line.

Final Reflection

I first want to touch on what I felt the committee did better than expected in this bracket. I am pleasantly surprised that there was not much mention of non-conference strength of schedule worsening a team’s seeding (of course the committee points to NCSOS when justifying North Carolina’s inclusion, though). I’ve always heard them discuss a team’s poor NCSOS as an argument against the team even though it’s already factored into SOR. Glad the committee didn’t continue this blatant form of double counting. I’m also impressed that they demonstrated an ability to look at a team’s body of work and rank them above or below what every metric suggested. Although inconsistently applied, I don’t hate giving Louisville an 8-seed despite a top 28 ranking in every number (and top 24 in all but BPI). I had them as an 8 before moving them up to a 7 after beating Clemson in the conference tournament, but that’s another point on its own. Kansas was outside the top 26 and Arkansas was inside the top 38 without any figure agreeing with those official seedings, and Xavier was ranked above North Carolina in defiance of what every metric was telling us. While I don’t agree with all of these outcomes (particularly Xavier > UNC), it’s good that the committee exhibited some semblance of independent thinking.

To wrap up, I don’t see a world where I’ll ever be able to hit 68/68 teams and seeds, but that will always be the goal. I clearly overestimated the influence of some arbitrary factors in my process, such as overweighing non-conference performance. I failed to anticipate predictive metrics determining seeds toward the bottom of the at-large pool instead of serving as a tiebreaker between comparable resumes. It’s evident that how they interpret resumes in one year may not indicate how they’ll be treated in another year. I think the WCC teams, Drake, and UC San Diego are perfect examples of that. I should also note that, as the matrix gets closer and closer to what the committee actually outputs, it becomes more difficult to deviate and be correct. With many of my stands, I was incorrect. With a few, I was correct. And there were also a couple where I was right, but not right enough. At the end of the day, while the committee can be frustratingly inconsistent, I need to use sound process when seeding these teams and make a concerted effort to be right, not less wrong. I’m never going to be perfect, but I must keep these concepts in mind to hone my craft as effectively as possible.