Image from espn.com
In 2022, Cracked Sidewalks called out a discrepancy in the rankings witnessed on the Basketball Power Index metric. It is one of the predictive metrics used on the NCAA's official team sheets that help with selection and (more so) seeding when it comes to the NCAA Tournament field. One of the things that really stood out at the time was how much BPI seemed to favor teams that were in leagues with ESPN television contracts. The reason that seemed newsworthy is because BPI was designed and is run by ESPN. That metric, along with another ESPN resume metric Strength of Record, are the only metrics on the team sheets that are owned and operated by a media company that has a vested financial interest in which teams are selected and perform well in the NCAA Tournament.
Earlier this week, @JBRBracketology pointed out that it looked like BPI might have been fixed. In recent years, BPI had a strong bias against Mountain West teams compared to other predictive metrics. JBR noted that over the years, BPI compared to kenpom rated MWC teams 39.2% worse (2022), 25.9% worse (2023), and 32.2% worse (2024). But compare last year to this year and it's a stark difference:
When we first looked at this our focus was on the 21 teams in the 2022 NET Top-100 that had BPI as a 10+ spot outlier from both of the other metrics in either a positive or negative direction. These are teams that are typically in the running for NCAA bids, either as at-large teams or the better performing mid-major teams that are able to earn automatic bids through their conference tournament. Here is the original graph we shared, with the "Average" being the average of the three predictive metrics at the time (BPI, kenpom, Sagarin) and the "Change" being the amount that BPI improved or decreased a team's metric average from the average of the other two metrics.
ESPN's BPI, a predictive metric used on NCAAT team sheets, factors/seems to penalize altitude into its metric. Here's the highest-altitude arenas & their team's KP & BPI. A startling 40% diff on avg between them.
— Made For March (@madeformarch) January 9, 2024
Affects teams in the Mountain West, Pac 12, Big 12, WAC & Big Sky. pic.twitter.com/YE4HLkABUu
This certainly explained why teams like Wyoming, Colorado State, and BYU were negatively impacted. This also was a problem that provided a true issue for the teams it impacted. In the 2024 NCAA Tournament, excepting BYU and Gonzaga who were swapped seed lines to adhere to bracketing principles, here are the 6 teams that bracketmatrix.com found to be furthest away from their expected seed from the 227 bracketologists that participated:
Five of the six largest bracketmatrix discrepancies were negative seed impacts on Mountain West teams that play at elevation (Boise isn't in Made For March's tweet above, but they play 2,700 feet above sea level). Considering the league's lack of success in recent years, maybe those Mountain West teams were due to lose regardless, but their paths were significantly more difficult than would have been expected. Nevada, Boise, New Mexico, and Colorado State all lost in the Round of 64 as underdogs. Utah State won their opener in an 8/9 game, but then faced Purdue and lost. Might these teams have fared better if they had been given the expected 6-9 seeds their non-BPI metrics indicated they earned?
The excellent bracketology blog Bauertology also took BPI to task last year. He cites many different sources in his longer article but it all comes to the same conclusion Cracked Sidewalks had in 2022, that BPI was a problem that was an outlier from the other metrics and was negatively impacting teams. Now in 2025, it seems that these cries may not have fallen on deaf ears as the metric has clearly been reworked. First, let's look at a 2025 version of the top-100 NET outliers:
There are once again 21 outliers in the top-100, but only one team (#100 South Dakota State) is at a metric gain or drag of greater than 10. Not a single Mountain West team appears in our outliers, and Big East teams like UConn, Butler, and Providence actually benefit more from BPI than any other metrics, indicating the appearance of a network bias that seemed present in 2022 has been corrected, with eleven of the twelve most negatively impacted teams having ESPN contracts.
For those with Big East interests, eight of the eleven teams in the league have BPI as their best of the three predictive metrics, while just one (St. John's) has BPI as their worst metric. I would still argue this bears monitoring. It seems problematic that any of the team sheet metrics are intellectual property of a media outlet that has a vested interest in their media properties succeeding. I feel the NCAA would be better served still by replacing BPI with another predictive metric like Haslametrics or evanmiya that does not make money on league or team successes. I would much rather see something like BRCT (pronounced "bracket") on the team sheet than SOR. But if nothing else, an improvement to BPI is a welcome change.
Now let's see what BPI and the rest of the team sheet metrics and on-court results have produced with our latest S-Curve and bracket:
Multibid Leagues
SEC: 13
Big 10: 9
Big 12: 8
Big East: 4
ACC: 4
Mountain West: 3
WCC: 2
Big West: 2
No comments:
Post a Comment