This betting thing is indeed hard

March Madness Models - Elite 8/Final 4 Recap

Jesse Fischer • April 4, 2015

As the tourney comes to a close let's review how my model's Elite 8 and Final 4 picks held up. Let me start by saying that I will agree that "betting basketball is hard". After my model started out so hot in the opening rounds it's success has died down as the tourney has progressed.

Part of this has to do with having to "reach" more when making picks. In the opening rounds there were a lot of potential plays to bet on with clear outliers given by the models (which made it easy to choose which games to bet on). However as the tourney has gone on the number of games (as well as betting opportunities) has decreased dramatically. When only given a few games to choose from the models are much less likely to find strong betting opportunities where the models think the spreads are "off". It is also worth mentioning that by the end of the tourney the teams are pretty well figured out by everyone and most people's numbers tend to align.

Let me also add that it is hard to disagree with Mark Few in saying that: "The tournament is a crapshoot." March is a time where even the undefeated go down. If I were to get serious about this betting thing it might be wise to avoid later rounds of the tourney and focus on the regular season. This would allow for more (and better) opportunism to bet on. Maybe in the off season I'll see how the models have done historically throughout the season and maybe start this thing up a little sooner next year…

ATS Recap



  • Notre Dame was a strong play.
  • The Louisville game was a little deceiving as it was tied at the end of regulation. If a bricked free throw wouldn't have somehow went in - then Louisville would have likely fouled and lost by however many FT's the Spartans would have made (which could have been none given their percentages). However it we assume they made at least 1 and Louisville missed their shot then this still would have meant a Louisville win against the 2.5 point spread. Instead the game went to OT and they lost by 6. It's not often you are rooting for the team you bet on to miss FT's down the stretch…
  • The models need to account for it not being possible for the Zags to make the Final 4…
  • Finally the Izzo run came to an end, however it's hard to bet against him at this point - should have trusted the numbers here.
  • FWIW the numbers had the Badgers at +6 (where the spread opened).

Elite 8/Final 4 ATS Record: 1-2
Overall Tourney ATS Record: 10-7

Over/Under Recap



  • After a tourney full of Under's (at least when I was paying attention) the scoring seemed to come out.
  • The 3 picks I got wrong M2 got correct, this was more human error than model error. This boils down to what I was talking about above. Since there weren't many opportunities I forced some picks where the models didn't necessarily align on what to pick. The only pick where they did align the models both got the pick correct.
  • M2 was 5-1! How did I go 1-3!?!?
  • Half the games were within 4 points of the Vegas number, it's hard to beat Vegas when their numbers are so good.

Elite 8/Final 4 O/U Record: 1-3
Overall Tourney O/U Record: 3-7

Let me end with this quote by Bo Ryan after the Zona game which you could say predicted the over in their next game against Kentucky (which somehow I picked the under on?). I think there are opportunities to do sentiment analysis on coaches quotes to factor into models, can you imagine!?

"I've been here as head coach for 14 years and it's just, I think maybe, uh, people don't do their homework, on, uh, how many points we score. Maybe if you're saying on the short side of 50, you're talking about what we give up defensively. Um, because we've lead the Big Ten and we've led the nation a couple times, uh, in '08 and I believe in 2012, uh, in holding teams to the fewest number of points. But offensively, we've always been looking to try to get a quick score, the problem is, the doggone guys on the other side try and stop us."