Downs And Success: An Attempt At A Different View

Submitted by LSAClassOf2000 on August 1st, 2013 at 1:56 PM

DOWNS AND SUCCESS: AN ATTEMPT AT A DIFFERENT VIEW

So, I asked myself a question the other day. Many of us probably assume – as I do admittedly – that sustained success on third down conversions has something of a cascade effect in that drives are sustained and the chances of scoring and, by extension, winning are drastically increased. The question I had is this – in the numbers, how strong is the relationship between some of the statistics involved in possibly sustaining drives?

I decided to go five seasons back to see if we could at least begin to see relationships. Clearly, the farther we go back, the clearer the picture could become, but this is simply an initial delve into this particular subject, so this will do. I figured that if I did begin to see things, an expanded view might be in order in the future.

As I normally do, to keep it somewhat topical for the board, I confined this to the Big Ten, although I do have some data from all of Division I. That will decrease the strength of any conclusion perhaps, but as I mentioned, this was more about scratching the surface.

Like other diaries, this is mainly about confirming relationships that many of us already may intuitively surmise.

SOME BASIC BREAKDOWNS:

I did something here that I normally do in these situations and come up with some arbitrary way to divide win totals and then create a table to see if there was a basic relationship 3rdand 4thdown conversions and relative success (this includes 2008-2012 data for the Big Ten):

 WIN TOTAL 3RD DOWN ATTEMPTS 3RD DOWN CONVERSIONS 4TH DOWN ATTEMPTS 4TH DOWN CONVERSIONS 3RD DOWN CONV. PCT. 4TH DOWN CONV. PCT. 11 OR MORE 1913 861 131 76 45.01% 58.02% 10 1136 490 69 41 43.13% 59.42% 9 1123 464 83 42 41.32% 50.60% 8 941 413 95 54 43.89% 56.84% 7 1859 761 160 83 40.94% 51.88% 6 1090 425 111 58 38.99% 52.25% 5 674 282 64 23 41.84% 35.94% 4 912 334 108 53 36.62% 49.07% 3 OR LESS 1220 430 144 73 35.25% 50.69%

So, on a most basic level, we can see a couple things – there is some relationship between your relative record and your success at converting third downs, but there doesn’t seem to be any significant relationship between relative record and 4thdown conversions, or at least not nearly as significant as the one that may exist on third downs.

Here is some very basic data for first downs, including the number of first downs for each group and the percentage relative to total offensive plays:

 WIN TOTAL FIRST DOWNS TOTAL OFFENSIVE PLAYS % OF FIRST DOWNS / TOTAL PLAYS 11 OR MORE 3032 9614 31.54% 10 1603 5524 29.02% 9 1574 5408 29.11% 8 1349 4517 29.86% 7 2348 9071 25.88% 6 1487 5253 28.31% 5 988 3374 29.28% 4 1143 4231 27.01% 3 OR LESS 1499 5658 26.49%

Again, on a basic level, there is some relationship here – the better the relative record, the better you probably are at getting first downs. Not Earth-shattering stuff at all, right? It just happens to be somewhere in the numbers as well, that’s all.

How strong are those relationships? Well, the R-value for the relationship between 3rddown conversion percentage and winning percentage for the Big Ten sample is 0.542, so there is a mild relationship there, with “mild” perhaps being an operative word here. For the relationship between 4thdown conversion percentage and win percentage, that value drops to 0.277, so there is not nearly the same linkage between the two sets of numbers.

It may be worth noting, however, that for all of Division I, over an expanded eight years of data that I collected, the R-value between 3rddown conversion percentage and win percentage is only slightly higher, sitting at 0.556. However, the R-value for 4thdown conversion percentage in the same relationship is noticeably higher – 0.355. That could be the sample size, of course. It could also be that this relationship might be somewhat different in other conferences or historically for other teams.

WHAT YOU KNEW ANYWAY, BACKED WITH NUMBERS:

I decided to look at some different relationships after the previous exercise produced a less-than-thrilling cursory look at the matter. The obvious thing that does come out of these many possible correlations is the relationship between 3rddown conversion percentage and the percentage of plays that are first downs. The R-value in this case is 0.806, indicating a fairly good correlation, showing you in statistical form something you already knew all along – the more successful you are at converting third downs, the more first downs you will get.  Now, however, you can tell your friends just how closely tied the two things are. More directly, the R-value for 3rddown conversion percentage and first down totals is 0.771, so there’s the same thing in a slightly different set of variables.

WHAT MAY NOT BE SO APPARENT:

On third down conversions from 2008-2012 (the five years here is to accommodate restrictions on finding first down data ) the Big Ten actually averaged 41.04%, with Michigan State and Minnesota actually being the worst at it overall. Neither team reached that number once in five years. Indiana only did so once, as did Purdue. Michigan stayed above that average three times, but of course, Michigan has the dubious distinction of having the worst percentage in any given year in that time period – 27.27% in 2008. Indiana managed 27.65% that same year.

For all of Division I in the past eight seasons ,  here were the least successful  teams on third down:

 Year Name 3rd-down Attempts 3rd-down Conversions Pct 4th-down Attempts 4th-down Conversions Pct 2010 Vanderbilt 171 46 26.90% 19 4 21.05% 2008 Washington St. 179 48 26.82% 18 9 50.00% 2005 Ohio 155 41 26.45% 5 0 0.00% 2007 Army 175 46 26.29% 26 10 38.46% 2010 San Jose St. 172 45 26.16% 13 7 53.85% 2005 North Carolina St. 165 42 25.45% 16 7 43.75% 2006 Temple 155 38 24.52% 23 4 17.39% 2007 Florida Int'l 173 42 24.28% 28 9 32.14% 2009 Washington St. 168 40 23.81% 15 7 46.67% 2006 Florida Int'l 166 39 23.49% 24 7 29.17%

For some contrast, here are the ten best:

 Year Name 3rd-down Attempts 3rd-down Conversions Pct 4th-down Attempts 4th-down Conversions Pct 2006 Hawaii 133 77 57.89% 20 8 40.00% 2010 Stanford 172 99 57.56% 16 11 68.75% 2008 Tulsa 193 110 56.99% 21 9 42.86% 2006 Brigham Young 168 94 55.95% 11 6 54.55% 2008 BYU 176 98 55.68% 10 7 70.00% 2009 BYU 169 94 55.62% 13 5 38.46% 2005 Southern California 167 92 55.09% 30 18 60.00% 2008 Texas 173 95 54.91% 16 13 81.25% 2012 Texas A&M 195 107 54.87% 12 7 58.33% 2011 Wisconsin 170 93 54.71% 10 7 70.00%

BACK TO THE FOURTH DOWN THING:

The only meaningful way to see if there was anything actually going on with fourth downs, I felt I should try to break this admittedly small sample up by team, so again, I used the Big Ten.

One thing that I can say here for all of Division I across eight seasons of data is that the overall success rate for all attempts on 4thdown made in that period id 50.59%, so in essence, it is an even-money proposition to go for it on 4thdown and the numbers bear that out as well.

The Big Ten, from 2008-2012, averaged 52.12% on 4thdown conversions. Iowa and Indiana actually happen to be the least successful in the last five years at converting on 4thdown – Iowa has never hit the average, and Indiana rose above the average only once in four years. Michigan is in a wide field of teams that have converted better than 52.12% in three of the past five years.

As for the whole of Division I in the past eight seasons, here are the ten worst instances of 4thdown conversion success:

 Year Name 3rd-down Attempts 3rd-down Conversions Pct 4th-down Attempts 4th-down Conversions Pct 2010 Troy 200 86 43.00% 13 2 15.38% 2006 Colorado 159 53 33.33% 13 2 15.38% 2011 Kent St. 190 54 28.42% 20 3 15.00% 2009 Virginia 177 60 33.90% 14 2 14.29% 2005 Stanford 146 48 32.88% 7 1 14.29% 2010 Memphis 164 53 32.32% 7 1 14.29% 2012 Connecticut 177 64 36.16% 7 1 14.29% 2012 Kentucky 178 71 39.89% 22 3 13.64% 2009 Hawaii 149 67 44.97% 17 2 11.76% 2009 UAB 141 52 36.88% 10 0 0.00%

Again, for contrast, the best on fourth down:

 Year Name 3rd-down Attempts 3rd-down Conversions Pct 4th-down Attempts 4th-down Conversions Pct 2011 Notre Dame 174 81 46.55% 6 6 100.00% 2005 Penn St. 166 73 43.98% 6 6 100.00% 2011 Stanford 175 92 52.57% 12 11 91.67% 2010 LSU 183 70 38.25% 12 11 91.67% 2011 TCU 169 88 52.07% 11 9 81.82% 2010 Illinois 177 72 40.68% 11 9 81.82% 2008 Texas 173 95 54.91% 16 13 81.25% 2011 Air Force 181 85 46.96% 32 26 81.25% 2007 LSU 223 104 46.64% 16 13 81.25% 2006 Boise St. 154 67 43.51% 21 17 80.95%

PUTTING THIS STUFF ON GRAPHS:

This is what the relationship (or lack thereof) for win percentage, third and fourth down conversion percentage and the relative percentage of plays which were first downs looked like over five years for the Big Ten:

TL;DR CONCLUSION:

So, other than some obvious relationships, I was admittedly a little disappointed that I didn’t find much more approaching the issue from this direction. In the future, I will probably try to go at this from a different angle - what that might be is up for debate. The sample for the conference is admittedly small, so there is that. Perhaps I am missing something which warrants a more careful examination, and not being an expert, that’s entirely possible.  If I have done anything here, it is confirm one or two things statistically perhaps that most of us knew anyway, and I suppose I can count this as a minor victory. Rather than rail against the “eye test” in this case, I seem to have confirmed an instance where it works.

OBLIGATORY:

I hate you, 2008.  I hate you so, so, so very much.

Well at least the US Olympic team did great that year.

Oh wait that was in the offseason before Michigan vs. Everyone took place. Carry on then.

They also look to have one of the highest attempt rates on 4th.  Wonder what is going on there? There is evidence that you should always go for it on 4th down especially any time you are across the 50.  The yardage gained by punting is rarely worth the lost opportunity to score if you do convert the 4th down in your opponents territory.

for a comparison, as that also depends on the defense and other stuff that has nothing to do with the offense.  A team can convert 3rd downs all afternoon and still lose if their defense is bad.

You might find a better correlation with points scored, or touchdowns specifically, or maybe an inverse correlation with field goals (maybe teams that don't convert 3rd downs tend to kick FGs more).

There's more to winning than just scoring points, so incorporating defensive performance would be necessary (and some special teams).  Converting 3rd downs would be a component of predicting scoring, while that could then be combined with the other game aspects to get a correlation with winning.

This is a timely comment because I was discussing it with someone last night. Limiting it to 3rd down perhaps and then incorporating performance on both sides of the ball might be more useful, or as you mention, correlate it to certain aspects of scoring offense. That's probably what the next step here would be. I was actually hoping someone would help fill in the hole I figured was there (this was a shot in the dark with some of the available data admittedly) - thanks for the suggestions.

EDIT: I calculated the 3rd down conversion differential (offense-defense, in this case) for each Division I team and correlated this to both win percentage and PPG and came out with R-values of 0.752 and 0.659 resepctively. I think that is at least part of the adjustment I was looking for the other day but couldn't quite figure out. Finding help like this is one of the reason I love this place. Seriously.

Case in point...

Here's the link in the event that's too small...

Good stuff!

I am curious if there were any teams had a consistently low winning pct. yet a consistently high conversion differential. It would seem unsustainable except for an academy or a defense so bad they had unusually high numbers of first and second down conversions.