clock menu more-arrow no yes mobile

Filed under:

Beyond the Box Score: Win Correlations (Part Two)

So in yesterday's WinCorr conversation, I mentioned that while WinCorr could be (and was) used to look at the overall stats that are most directly connected to success on the field.  I also mentioned that the measure could be used to create an individual 'footprint' for each team.  To illustrate this, I'm simply going to look at said footprint for a few teams and see where that takes us.

Team-Specific WinCorr's

We'll start with Colorado since, well, I've already done them.  Here are Colorado's top WinCorr's (and my conclusions) from Tuesday's Colorado piece:

Colorado's Top WinCorr's (any categories that were > 0.800)

  1. Offensive Close-game Passing S&P+ (Correlation: 0.875)
  2. Offensive Close-game Passing S&P (0.864)
  3. Offensive Close-game Passing PPP (0.860)
  4. Defensive Passing Success Rate, Pressure Situations (0.854)
  5. Defensive Success Rate, Pressure Situations (0.843)
  6. Offensive Line Yards on Non-Passing Downs (0.837)
  7. Offensive EqPts gained in Non-Passing Downs (0.829)
  8. Total Offensive EqPts gained (0.821)
  9. Offensive EqPts+ (0.819)
  10. Defensive Rushing Success Rate, Pressure Situations (0.817)
  11. Offensive Rushing S&P, Pressure Situations (0.806)

There's some repetition in there obviously, and it builds a very distinct narrative.

  • When Colorado was able to move the ball through the air, they probably won (especially when they were able to rip off a couple big plays).
  • The Buffs found themselves in quite a few 'pressure' (i.e. 4th quarter, score within two possessions) situations, and when the defense stepped up, they won.
  • Offensive 'Pressure' numbers weren't as important, but to the extent that they were, it was all about moving the ball on the ground and staying out of passing situations.
  • Actually, that went for the whole game.  Staying out of uncomfortable situations was key for the Buffs--if they were able to move the ball well in Non-Passing Downs and stay out of Passing Down situations, they stood a chance.  But with a shaky freshman QB behind center, Passing Downs were murder.

In some cases, you can see how this would be useful in making your 2008 predictions/rankings.  Obviously the basic criteria--returning starters, etc.--are still useful, but in discussing Colorado we can see that how they deal with things like finding a more consistent passing game is a little more important than other things.

Anyway...let's take a look at some other teams to determine a) how different one team's 'footprint' is to another's, and b) what all this Team WinCorr concept can tell us.

We'll start with Phil Steele's preseason #1 team, Florida.

Florida's Top WinCorr

  1. Defensive Passing Success Rate, pressure situations (0.960)
  2. Defensive Success Rate, pressure situations (0.928)
  3. Offensive Rushing Success Rate, redzone (0.899)
  4. Offensive S&P (0.888)
  5. Defensive Rushing Success Rate, pressure situations (0.869)
  6. Defensive Rushing S&P, pressure situations (0.861)
  7. Offensive S&P+, close games (0.856)
  8. Offensive S&P, close games (0.849)
  9. Offensive Success Rates (0.847)
  10. Offensive Passing Success Rates, close games (0.845)
  11. Offensive PPP, close games (0.843)
  12. Offensive Passing S&P, close games (0.837)
  13. Defensive Rushing S&P+, Non-Passing Downs (0.832)
  14. Defensive S&P, pressure situations (0.827)
  15. Offensive Passing S&P (0.827)

Tie this with what we already (think we) know about Florida, and you can reach the following conclusions.

  • Florida found themselves in a lot of pressure situations (kind of like Colorado), and their defensive performance was key (as with Colorado).  We kind of already knew this--the offense had no problem generating points against Georgia and Michigan in particular, but they couldn't keep points off the board.
  • Even though we discovered yesterday that Points Per Play was more important than S&P and success rates overall, for Florida it was all about Success Rates.  The explosiveness was always there, but the efficiency broke down at inopportune times.
  • '+' numbers weren't as strong in correlation than the more raw statistics.  This most likely tells you that their performance didn't have much to do with the opponent they were playing.  Offense was pretty much good no matter what and defense was average(ish) no matter what.
  • The importance of redzone success running the ball was very much unique to Florida.  Either it tells you that the predictability of "Tebow right, Tebow left, and Tebow up the middle" held them back at times, or that their rushing game down there was bad until they instituted "Tebow right, Tebow left, and Tebow up the middle."  I'm leaning toward the latter.

So...being that Florida was breaking in a lot of young players on defense, and they've now gotten a year's worth of pressure situations under their belt, you figure the defense will respond better to pressure than they did last year.  That's obviously a good thing for UF's title chances in '08.  Meanwhile, we can't say anything definitive about the Gators' consistency/efficiency, but we do know that UF's non-Tebow running game should be in better shape in '08.  Not only will Kestahn Moore have another year of experience, but if the spring is any indication, USC transfer Emmanuel Moody will be more than ready to make some noise, and true freshman Chris Rainey could be dangerous as well.  Having more realistic threats in the redzone should alleviate that issue.

Okay, so the WinCorr concept says encouraging things about Florida.  How about my own team, Mizzou?

Top Missouri WinCorr's

Many of Mizzou's 'Pressure Situation' numbers came into play here with very high correlations (including numerous with 1.000 correlations), but I'm not including them here because, well, they didn't find themselves in many pressure situations.  And beyond that, their correlations weren't as strong as Florida because, I assume, they only lost twice maybe?  Not sure.

(And I'm still enjoying being able to casually mention that Mizzou was better than Florida, Michigan, etc., last year.  Gotta milk it while it lasts.)

  1. Opponents' score (0.950)
  2. Defensive PPP (0.870)
  3. Defensive Passing PPP, redzone (0.859)
  4. Defensive S&P (0.851)
  5. Defensive EqPts (0.840)
  6. Defensive Passing S&P (0.828)
  7. Defensive Passing S&P, redzone (0.824)
  8. Defensive Rushing S&P (0.807)
  9. Offensive Rushing S&P+, Q4 (0.797)
  10. Defensive Rushing EqPts, Non-Passing Downs (0.791)
  11. Offensive Rushing S&P, Q4 (0.787)
  12. Offensive Line Yards+, Q4 (0.783)
  13. Offensive Line Yards (0.782)
  14. Defensive Passing S&P+ (0.781)
  15. Offensive Rushing PPP, Q4 (0.779)

Uhh...see any trends there?  There are blatant trends, and there are blatant trends.

  • The first 8 measures (and 10 of the 15) were defensive--which is, I guess, what happens when your offense is scoring 30+ points in 13 of 14 games.  The Mizzou defense came up big, but when Mizzou lost (to OU), or when a game was a bit closer than it should have been (against Iowa State, for instance), defensive breakdowns (however minor) were the likely cause (aside from the second half of the Big 12 title game, when it was alllll on the offense).
  • Redzone defense was a bigtime key for Mizzou.  If they held you to FGs, you were toast.
  • Where offense did come into play, it was late-game rushing, late-game rushing, and late-game rushing.  They were able to do it against everybody but Oklahoma.

So defense and rushing are key for Mizzou.  Gotcha.  The Mizzou defense returns 10 Coton Bowl starters--everybody but DT Lorenzo Williams--so that has to strike optimism in Mizzou fans.  As for the rushing game...I guess that's still a question mark with the loss of Tony Temple, though anybody who saw sophomore Derrick Washington carrying the ball this spring had their concerns alleviated a bit.  Washington and Jimmy Jackson (along with Earl Goldsmith, I guess) got some experience in the past simply because Temple was always rather injury-prone, and Jackson proved himself to be as efficient a runner as you could ask for.  Washington's the homerun threat, but if these two RBs with presidential names can get some yards when asked, the Mizzou offense will run just as smoothly as last year.

Okay, one more.  How bout the defending national champion?

Top LSU WinCorr's

  1. Defensive Passing EqPts (0.866)
  2. Opponents' score (0.863)
  3. Defensive Passing S&P (0.861)
  4. Defensive Passing PPP (0.858)
  5. Defensive EqPts+ (0.839)
  6. Defensive EqPts (0.834)
  7. Defensive S&P+ (0.831)
  8. Defensive Passing Success Rates (0.831)
  9. Defensive Success Rate, Passing Downs (0.827)
  10. Defensive Passing S&P+ (0.826)
  11. Defensive S&P+, Passing Downs (0.823)
  12. Defensive PPP, close games (0.820)
  13. Defensive PPP (0.815)
  14. Defensive S&P (0.813)
  15. Defensive EqPts, Passing Downs (0.807)
  • Wow.  Okay.  So...LSU's offensive performance more or less made no difference whatsoever in the results of their games (relatively speaking).  They were solid offensively, but their margin of victory (and their two defeats) were decided by defensive breakdowns.
  • Why do I say defensive breakdowns?  Because a) while LSU had a great overall passing defense, 7 of these 15 categories had to do with passing defense, and b) 3 of these categories were related to passing downs...meaning LSU had the occasional breakdown on passing downs (against, predictably, Arkansas and UK in particular), and when they did, their odds of losing went way up.
  • For the record, here were the top 3 offensive categories: 1) PPP, 1st downs (0.677), 2) EqPts, 1st downs (0.655), 3) S&P, 1st downs (0.639).  The #5 and #6 offensive categories were also 1st-down related.  I'd call that a trend, wouldn't you?  Basically that says that the offense's job was, at its most simplified, to spell the defense long enough to keep it dominant.  If they got some yards on 1st downs (and therefore made themselves more likely to move the chains and eat a bit of clock), they were likely going to win.

This actually bodes well for LSU in 2008.  Their biggest question mark is at QB, but...all they have to do is find a QB who doesn't hand the ball to the other team and a RB who can get 4-5 yards on 1st down, and they'll be in good shape.  I know it's not that simple, but still...this bodes well for LSU.

Predictive Potential of WinCorr

When you boil it all down, statistics are fun for three reasons: 1) you can use them to understand the game better, 2) you can use them to evaluate and compare players/teams, and 3) you can derive quite a bit of predictive ability from them over time. At this point, I can say that #1 has been a rousing success so far, and #2 is slowly rounding into shape, but #3 is completely and conspicuously absent so far for one simple reason: I only have one full year of data.

So while it's easy to say that these WinCorr 'footprints' bode well for LSU, Florida, and Missouri (and not so much for Colorado), at some point we run into a bit of a brick wall with the WinCorr concept for now. We don't know for sure how much of the WinCorr idea is arbitrary from year to year, but it's certainly something to watch.  And the fact that the numbers are so dramatically different for every team certainly make you think there's something going on there.  But as we backtrack into the 2006 numbers and, obviously, prepare for an onslaught of 2008 numbers, we'll have to see what becomes of this.