DNA of a Cover Machine

Written by Drew Borland of SportSource Analytics

One of our core focuses at SportSource is assisting coaching staffs in finding strategies that help their team win games. Whether that’s run/pass balance, strength/weakness in formations, or situational tendencies, we cover the full spectrum in analyzing the game. Fundamental to this focus is understanding our most frequently asked question: “Which stats directly correlate to winning?”. Before speaking with any authority to our 60+ FBS clients (programs and coaches), it’s imperative we answer this question. We also have to establish a framework by which we constantly re-evaluate factors critical to winning as the game evolves.

Place all your football bets this season at Fanduel.com/BetTheBoard and start with a FREE $1,000 !

Without going into mathematical details like Pearson Correlation Coefficients etc., we created ways to look at the correlation strength for any statistic pertaining to “winning percentage”. With every statistic you can ask one question. “How much does this statistic correlate to a team’s Winning Percentage?”. Statistics are all relative. But scoring differential is obviously one of the most correlated statistics. Other metrics like rushing defense, turnover margin, 1st Down Efficiency/Success, returning starters, and passing defense are where it really becomes fascinating.

Now, what if we asked our correlation engine a different question? Instead of “Winning Percentage,” what if we tried finding statistics that matter most Against The Spread (ATS)? Knowing the answer to this question helps form the building blocks to discover the DNA in a “Good Cover Team”.

Below we’ve mapped out 35 different statistics and their correlative strength towards covering the spread. Many you have grown accustomed to hearing on Bet The Board’s College Football and NFL podcast each week. Our correlation strength is going to be calculated based on all FBS college football teams from 2005 to 2020. The correlation strength is measured from 0 to 1 with “0” representing no statistical correlation and “1” an absolutely perfect correlation.

Let’s take a look at the results.

Quick legend on how to read these scores.

There are technically positive and negative correlation strengths. We have normalized all of them to be positive for simplicity sake. The 0 to 0.10 range means the statistical category is largely insignificant. At roughly 0.30, value shows a strong positive signal towards being correlated to covering the spread and the relevance of a category climbs from there.

There’s clearly some really interesting insight to glean off of this table. Here are just a few interesting takeaways that we noticed:

  • Returning Starters, a very popular topic during the preseason, really doesn’t matter at all when it comes to the spread historically. This could be a bigger correlation earlier in the season but that’s a topic for another article.
  • Raw Scoring Offense and Scoring Defense was very strong (as expected). We actually have several opponent-adjusted scoring offense/defense metrics. They were very close (within 0.03), but the raw values edged them out.
  • Raw yardage numbers (see passing offense/defense above) are basically insignificant; what matters is efficiency.
  • Turnover Margin (part luck, part talent, part scheme) moves the ATS needle more than people think. When analyzing “Winning Percentage” correlative statistics, Turnover Margin is trumped by over half of the other statistics on this list. Translation: TO Margin matters a lot more to covering than winning in general.
  • Both Yards and Plays Per Point represent the explosiveness and the grind to produce points. Yards Per Point has the edge; is also very easy to calculate and incorporates a lot of in-game factors (field position, special teams, etc). It beats several other more esoteric metrics in our platform.

This is just the tip of the iceberg. There are hundreds of statistics and metrics we can measure and analyze correlative strength. We’ll dive a lot deeper into some of those areas in subsequent articles this season.

Remember, correlative strengths are relative. There is no single metric so strongly tied to ATS Winning Percentage that you would bet it blindly. If there were one magic elixir we wouldn’t be writing about it! Derivative and combinatorial statistics can be very obscure (not much value to coaches) but they definitely have substantial value to bettors.

So what does all of this data actually mean? Every week we hear people from every corner of the internet pick games followed by what amounts to a small legal defense on why they picked that team. They will introduce “Exhibit/Statistic A” and “Exhibit/Statistic B” as justification for their logic. As each season goes on and teams accumulate a statistical resume, every bettor must be able to quickly evaluate this information. A fundamental knowledge of the statistics that actually matter can quickly help sift through more meaningless information allowing bettors of all ability levels to make more informed choices. Earlier this year we introduced the PowerSheet to assist with this type of analysis.

%d bloggers like this: