The following article was originally published on March 9th in TSP Live Insider…
The futures markets for the NCAA Championship are starting to pickup. A few of you sent me a TikTok video and some Twitter commentary how the past 22 years or so the NCAA Basketball champion (I’ll find the video and post to Telegram) has fit certain KenPom criteria and the teams that fit the criteria this year would be Creighton, Kansas, Houston. Alabama, UCONN and UCLA. How do I feel about this system for narrowing down the field?
I don’t doubt it is accurate and 22 years is pretty good. However, these little systems aren’t that good because they are adjusted as it goes along. It wasn’t a system created 22 years ago that was forward looking and predicted the winner each year as it went along. It was created after the results were in. The system was designed to fit the already logged results. Can I explain? Sure!
Let’s pretend it is 2021 and over the last 10 years, the team that went on to win the NCAA Championship was ranked no worse than 8th in offensive efficiency and no worse than 15th in defensive efficiency. Using this idea (I am making up the following teams) I see for 2021 that Kansas, UCLA and Temple are all under 8th in offensive efficiency and under 15th in defensive efficiency. So, this would fit my model. Well, in 2021…Baylor actually wins the NCAAB Championship and they were 18th in defensive efficiency and 9th in offensive efficiency. So, when 2022 rolls around I tell everyone that the last 11 years the NCAAB champion was no worse than 18th in defensive efficiency and 9th in offensive efficiency. Do you see what happened? I updated the parameters as I went along to maintain the perfect record. Even though my “system”, as it stood in 2021, didn’t work to predict the NCAAB champ.
So, even though the system did not predict the winner in 2021, when 2022 came about I was able to say that over the last 11 years the NCAAB champ was in the top 18 in defensive efficiency and the top 9 in offensive efficiency. These parameters are now 11-0 for predicting the NCAAB champion. Technically I am correct, but this is a good system? Do you see the error in such analysis? Be careful betting using “systems” like this which are updated every year based on the results of the previous year.
Now we move to 2022…Kansas becomes the NCAAB champ and let’s say they were 19th in defensive efficiency and 12th in offensive efficiency. Well, in 2023 I can now say that over the past 12 years the NCAAB champion was in the top 19 in defensive efficiency and 12th in offensive efficiency and I can be 100% correct! However, the parameters I used when predicting the previous year again failed to predict a winner. I am merely updating the parameters each year to fit the results.
It is a common and seriously flawed process of trying to predict things by locating parameters to fit the past results. Versus finding parameters that PREDICT the future results with accuracy. Huge difference! If today I said the NCAAB champ will be in the Top 10 for offensive efficiency and Top 10 for defensive efficiency and from now through 2036 it predicts the NCAAB champ…solid. I can’t however go year by year and update my parameters to fit the previous results. That’s useless.
It would be like me saying, I have a system that is 5-0 at predicting the winner of the last five Super Bowls. You ready?!?! The team who won the Super Bowl came from a city with two words in it’s name…
Kansas City, Los Angeles, Tampa Bay, Kansas City, New England. I mean it is 5-0…so are you going to bet Baltimore, Philadelphia, Buffalo, Pittsburgh, Miami, Dallas, etc. this year? Why waste your money…my system is a perfect 5-0! Then Miami wins this coming year for the 2024 Super Bowl and so for the 2024-25 NFL season I say…the last six years the team that won the Super Bowl came from either a city that had two words in its name or two “M’s” in it’s name. So, just look to bet teams with two words or two “M’s” in the name and you are golden! Listen, don’t laugh…it’s a perfect 6-0 at predicting the Super Bowl!
That’s not to say that Creighton, Kansas, Houston, Alabama, UConn nor UCLA will win the NCAAB Championship. I mean that’s a healthy cross-section of the top teams…but don’t think this KenPom system is some Holy Grail…it’s backward looking and updated to fit the previous results. It’s not like 22 years ago someone said I will use these KenPom parameters to predict the NCAAB champion…and it worked moving forward. They looked at it this year and listed every team that won in previous years, saw their offensive and defensive ratings and found the key levels where every previous winner was included. If the team this year is 14th in offensive efficiency, then next year this person will say the last 23 years the NCAAB Champ was 14th or better in offensive efficiency and 18th or better in defensive efficiency and here are the five teams which fit that profile. My system is 23-0…how can you argue with it. It’s misleading and more of a magic trick than an actual method of analysis.
Be careful falling for all these systems which fit parameters to results and hope to be predictive. Sometimes they will work…but it is usually by luck versus quality methodology.
Good luck in your action!