Ran some tests on my Predictive Model vs. Vegas

BigBlueFanGA

Heisman
Jun 14, 2005
26,435
23,456
0
Thought some might find this interesting. Basically there’s a 98% probability that my model was actually better than Vegas at predicting the outcomes from the Sweet 16 on.

Now if someone would write me a check, I would gladly disappear. [laughing]
I have $1.12 in my account, will that do?
 

SmyrnaCatFan

All-Conference
Nov 9, 2012
5,700
1,310
113
After the exit, I've kind of not been here much. How would you describe the success or lack thereof of your model for 2019 tourney?
 

TFCat11

Heisman
Mar 25, 2019
6,480
12,101
108
Thought some might find this interesting. Basically there’s a 98% probability that my model was actually better than Vegas at predicting the outcomes from the Sweet 16 on.

Now if someone would write me a check, I would gladly disappear. [laughing]
Do you also have “an inside lane” to the referee contingent? If not, then Vegas wins again lol.
 

Aike

Heisman
Mar 17, 2002
75,464
46,468
90
After the exit, I've kind of not been here much. How would you describe the success or lack thereof of your model for 2019 tourney?

Pretty strong. I entered 10 brackets at ESPN based on variations of my models. Alternated champions, minor differences based on probabilities, etc.

The 10 ranged from 76.5th to 98.4th percentile, with an average around 88.5.
 

Joerules19

Senior
Sep 7, 2015
1,360
683
0
Pretty strong. I entered 10 brackets at ESPN based on variations of my models. Alternated champions, minor differences based on probabilities, etc.

The 10 ranged from 76.5th to 98.4th percentile, with an average around 88.5.

I used your model with a combination of KenPom and my gut to fill out my only bracket. Relied on your model a bit more than the others, but hit on 14 out of 16 and 6 out of 8. Would have won a cool $300 if Tech beat UVA (I had the Zags). Overall, I would say you did a good job
 

Aike

Heisman
Mar 17, 2002
75,464
46,468
90
I used your model with a combination of KenPom and my gut to fill out my only bracket. Relied on your model a bit more than the others, but hit on 14 out of 16 and 6 out of 8. Would have won a cool $300 if Tech beat UVA (I had the Zags). Overall, I would say you did a good job

Thanks. Glad it helped.
 

StubbornPenny

All-American
Nov 2, 2009
10,562
9,892
0
I was a bit wary of the model because it factored in last tournement success - my statistical brain says today’s team shouldn’t really be influenced by a teams success or failure multiple years ago.

HOWEVER, college basketball may be the exception to the rule, since the major players almost always stay on top. Like, if this same rationale were use for the NFL, I don’t think it’s work that well at all as there’s much more power turnover.

Overall I like the model, and would love for you to run it again next year.
 

Aike

Heisman
Mar 17, 2002
75,464
46,468
90
I was a bit wary of the model because it factored in last tournement success - my statistical brain says today’s team shouldn’t really be influenced by a teams success or failure multiple years ago.

HOWEVER, college basketball may be the exception to the rule, since the major players almost always stay on top. Like, if this same rationale were use for the NFL, I don’t think it’s work that well at all as there’s much more power turnover.

Overall I like the model, and would love for you to run it again next year.

It actually looks at the kind of team that was successful in previous tournaments.

It doesn’t actually look back and say, “Nova was good last year, so they should be good this year.”

More like, “Nova was good last year. Teams that play similarly to Nova should do well this year.”
 
Last edited:

kybassfan

Heisman
Jul 1, 2005
20,032
16,368
113
Pretty strong. I entered 10 brackets at ESPN based on variations of my models. Alternated champions, minor differences based on probabilities, etc.

The 10 ranged from 76.5th to 98.4th percentile, with an average around 88.5.

Considering some of the late round upsets, I’d say that constitutes some excellent modeling work. Well done!!
 
  • Like
Reactions: UKUGA and Aike

natron20

Senior
Oct 11, 2006
3,190
659
62
How about texting a cousin in law a month or so ago with some info? Better yet, just let me send you my money and you can drop it where you see fit.
 
  • Like
Reactions: Aike

StubbornPenny

All-American
Nov 2, 2009
10,562
9,892
0
It actually looks at the kind of team that was successful in previous tournaments.

It doesn’t actually look back and say, “Nova was good last year, so they should be good this year.”

More like, “Nova was good last year. Teams that play similarly to Nova should do well this year.”

Okay that makes more sense, but what dictates that a certain style should be dominant this year versus the next? Regular season performance of similar/top teams?
 
  • Like
Reactions: Aike

Aike

Heisman
Mar 17, 2002
75,464
46,468
90
Okay that makes more sense, but what dictates that a certain style should be dominant this year versus the next? Regular season performance of similar/top teams?

It’s actually only looking at several years of tournament performance. It’s predicting that the kind of team that performed well over the past few tournaments will also perform well in the next tournament.

The whole thing actually stems from a thought I’ve had that there could be slight differences between what makes a successful tournament team vs. what makes a successful regular season team, and I wanted to see if I could quantify that.

Interestingly (to me), Kenpom published an article in the Athletic last month detailing how certain statistics are different in the tournament vs. the regular season. So it seems like even he has realized that his famous regular season predictive model may not do the best job at predicting the tournament, because of these subtle differences that I’m talking about.

If you watch point spreads, you probably notice that you can pretty much use Kenpom’s model to predict where they will be set. So even though I’m somewhat surprised that my model “beat Vegas,” I’m not shocked. Vegas sets lines where the public believes they should be, in order to entice 50/50 bets. The public believes in Kenpom.
 
  • Like
Reactions: UKUGA
KSR

Latest

  1. 01 KSR Football Chops Comes Home: The Man Behind the Meme
  2. 02 KSR Football Playoff Viewing Guide: What games Kentucky fans need to follow in Week 11
  3. 03 KSR Basketball Kentucky vs. Indiana: Preview, Odds & How to Watch
  4. 04 KSR Basketball Sources Say goes LIVE to talk Kentucky vs. Indiana and the latest in recruiting
  5. 05 KSR Football SEC Availability Report: What to know about Thursday's update