KenPom’s Caste System: History Driving Narrative

By: The Boston Orange @TBO44

Around 2000 years ago India solidified into their culture a system of immobility, a system where your family name decided your fate. The Caste System. No matter how hard you worked, no matter how much you changed and improved you were always going to be stuck where you started. This is quite literally the opposite of the American Dream where we reward hard work, we reward self-improvement and champion upward mobility.

Ken Pomeroy loves the caste system. He loves it so much that he created a predictive metric for college basketball so he could live out his fantasy for laughing at the “untouchables”. It’s praised as a predictive powerhouse, but in practice its early-season structure feels less like a true performance index and more like an indictment on past failures.

In that system, KenPom is the caste, and Ken is the king.

The Metrics

When the NCAA Tournament committee makes decisions, they consider a variety of data: results-based metrics, resume-based metrics, and predictive ones. Here, we’re zooming into predictive systems — the kind that analysts, media, and bettors lean on most early in the season.

I want to focus on two metrics (however I will give a quick opinion on the main ones cited most frequently):

• BPI (ESPN) — sucks, black-box, hard to scrutinize because we don’t even know what it is
• NET (NCAA) – Will write another article on this, ruining college basketball scheduling (no Colgate or Cornell on Cuse’s schedule)
• KenPom (Ken Pomeroy) — King Ken’s trusted predictive model
• Torvik (Bart Torvik) — Think American Dream, newer, more transparent, updated for today’s roster volatility

The one you have probably not heard of here is Torvik. I believe Torvik is the most complete tool and leans into modern realities — roster turnover, NIL-era transfers, and the chaos of year-to-year change. We will get into them after dismantling King Ken’s master excel sheet.

The Caste System Problem: Historical Weighting

The issue with KenPom is less about the math and more about the philosophy behind it.

In an era where team rosters look more like a wild west version of NBA free agency than traditional college continuity, KenPom still leans heavily on history — especially early in the season. Based on model commentary and public estimates from my research:

• November–December: 60–70% of a team’s rating is drawn from preseason data, last season’s performance, recruiting, and reputation.
• January–February: historical data still makes up roughly 30–40% of the rating.

It’s not bad to consider where a team came from — returning production, recruiting, coaching pedigree all matter. But when those factors outweigh actual on-court results for weeks, you end up locking in artificial ceilings and floors.

The result? Some teams can’t climb fast enough, while others coast on inflated reputations long after their play warrants it. This has very real consequences on Selection Sunday. Look at last season’s bubble conversation, and the table below. Analysts leaned on KenPom to separate North Carolina and West Virginia — but their resumes told a different story. How can you look at these teams and think UNC should be 20 positions higher in any metric?

However, that is not my main point. King Ken supporters will point out that by Selection Sunday the historical weight is negligible (even though I still doubt that based on the table I showed below) so why does it matter if the early season ratings restrict mobility? NARRATIVE. Humans select the 37 at-large bids, not computers. For months, KenPom posts rankings that are factually incorrect. They have little to no bearing on the actual basketball being played on the court that season. Case and point; WVU vs UNC. UNC was propped up by the previous years accomplishments, whereas WVU was always fighting the system pushing them down. The narrative was set.

Enter Bart Torvik: Our Potential Savior

Bart Torvik’s model is built for today. He doesn’t cling to last year’s data the way KenPom does. Instead, his system prioritizes who teams actually are in the current season, using little to no historical data after the first ball is tipped. Below is a summary of how he weighs games in season:

• Games ≤ 40 days old: 100% weight
• Games 40–80 days old: weight gradually decays
• Games ≥ 80 days old: floor weight at 60%

The result? Rankings that reflect performance in the current system, not projections from before anyone played a minute. A metric that makes sense. Let’s take a look at how Cuse stacks up in KenPom and Torvik across 2 games (sorry I wrote 90% of this prior to Drexal then was too hungover to finish it out before the game):

Anyone that has watched any Syracuse basketball through these games knows one thing for sure – Torvik is an astronomically better representation of this team. For any metric to have our offense BETTER THAN our defense immediately invalidates it for me. KenPom’s historical inputs are what drives this. Now obviously I am not saying Syracuse is the #4 team in the nation overall, but Torvik is admittedly volatile early in the season as it gathers data. The bottom line is, I would rather have a volatile metric based on in-season data than a stagnant metric using data that is irrelavent.

What Needs to Change

Here are the two things that need to happen going forward:

1. Torvik should completely replace KenPom. This isn’t 2010 anymore, historical data in college basketball means basically nothing. The Torvik model is a better representation of the state of college athletics.
2. Hold off on widely publishing public rankings until January 1. Let the real performance data settle in first. This eliminates the volatility issue, and the creation of narratives. Let the AP poll do that (they have about as much credibility as me at this point).

Final Takeaway

This whole article seems like a bash on Ken Pomeroy, and it is. However, that is not because it is bad math but bad data inputs. KenPom has done incredible work over the years. It’s a powerful predictive tool. But college basketball — and the way teams are built — has evolved and therefore Selection Sunday tools need to evolve as well. Bart Torvik has created the evolved tool, we need to embrace it.

If the caste system never truly resets, then real mobility is just an illusion. That’s not how basketball should work. Let’s topple King Ken and let Bart rise.