System Failure: A Look Inside What Went Wrong

Today, the ScoreMetrics Lab is going to pull the curtain back on a system we were working on a few months back. We do this just to show you some of the steps we were taking to build a system that failed, so that, together, we can learn. Of course, to start every system, we begin with a theory and this one involved college basketball.  

College sports fans are passionate. The kind of devotion they display for their favorite teams can be down-right scary. This is especially true when it comes to the better teams. While that love and dedication is fine for fans, it can be disastrous for investors.

Everyone wants to back their favorite team or a winner. But just because a team happens to be ranked, does not necessarily mean they are a good investment. Winning teams do not always win against the spread (ATS), after all.  

Of course, when it comes to ScoreMetrics, we don’t just make blanket statements and call them facts. We form a hypothesis, test it, and derive a conclusion from the data we collect. If something works—fantastic!

But if it doesn’t—well, knowing what not to do can be valuable down the road.

Betting on the NCAA Basketball Top-15

While it may not always be advisable to invest in a top-ranked team, there are times where it is. The trick comes in finding the right system that will pinpoint prime opportunities. To that end, let’s dive into that failed model design to help people learn a little bit of what we do in the ScoreMetrics Lab

In our models, we come up with a set of parameters to guide the data collection process. These are the rules we used for this system:  

  •         the team’s ATS win % is between 0% and 75%
  •         the previous game the team was the Home team
  •         the team’s previous game margin is between 1 and 100
  •         the team’s rank is between 1 and 15
  •         the team is the Home team
  •         the game is a Non-Conference game


This is not considered a rule or parameter, but we only look at games played during the regular season. Too many factors can impact games in postseason play other than the ones we have already accounted for.  

Now, for this system to be considered a winner, we are going to want to see winning seasons in eight of ten years. If it is a losing season, losses must be less than a certain percent. Ideally, every season is going to be a winner. But if the net result happens to be a loss, well— if the loss is large it becomes harder to justify sticking with it.  

So—how did this one perform? Based off $100 wagers, the results were as follows:

 

Season

Record

ROI

Money

 

 

 

 

2019-20

10-13-1

-15.50%

($372)

2018-19

17-8-0

30.90%

$773

2017-18

16-5-1

44.70%

$984

2016-17

14-8-0

24%

$529

2015-16

20-11-1

23.80%

$760

2014-15

17-10-2

20.90%

$607

2013-14

17-16-1

-0.10%

($4)

2012-13

10-12-0

-12.20%

($269)

 

Conclusion

So– yeah, this model doesn’t work for us. No system is ever going to be perfect, which is why we allow for one or two losing seasons over a ten-year period. That five-year run from the 2014-15 season through last season is nice, but the losing season this year and the two before the run is unacceptable, causing too much instability and volatility to ever think this would be something we could use. In other words, this is probably a really good gambling system, but a terrible investment system.

Now, overall, despite the four losing seasons, the system turned a profit over the ten-year test span (6 profitable seasons and 4 losing seasons). But after winning in Year One and then suffering losses for three consecutive years, no one is going to have the patience to give Year Five a shot.    

Where did we go wrong? It would take a deep dive to come up with a definitive answer to that question, although it is not hard to see where some of the problems may lie.

Some of the parameters are too broad. Every team’s ATS win percentage is going to fall between 0-75 percent. Opening it up to the top-25 instead of just the top-15 might be a good idea as well, to allow for a larger sample size.

The previous game’s margin being between 1 and 100 probably doesn’t do much good either. Not too many games (if any) have ever fallen outside of it.

Sticking with non-conference games is probably a good idea since we know most teams schedule weaker opponents during that part of the season. Everyone usually plays better at home, but how often do teams have back-to-back home games?

Okay—so, why are we telling you all this? We could ask you for your blind trust and faith, and if you gave it to us, well—that would be nice. Such blind obedience outside of a cult is completely unexpected and just a little unrealistic (no, ScoreMetrics is not a cult).

But by giving you a look behind the curtain, you can grasp a better understanding of what we do. When you understand how the ScoreMetrics process works, it becomes easier to believe in it—and then we all can make a little money.

Hot Off The Press!

zero