This is a guest post by GraniteStateNole. 

This is the time of the year that we begin to turn our eyes to how we think FSU will do in the fall.

Every year, there are several among us who think that this is the year - FSU going to win all of its games - or maybe all but 1.  Others say, we may not be that successful but surely we're going to win the ACC or at least be in the ACC-CG.

We say that because, well, its spring and also because we feel good about the team.  The 'Noles have new coaches - especially on defense.  They have a potential Heisman winning QB returning and we expect the offense to again be special - but every team feels good about itself this time of the year.

One thing that we all have wondered about is how good will the defense be and if it improves - how many more wins might 'Nole fans expect.  For now, knowing how good the defense will be is strictly a matter of opinion.  But, it is possible to get an idea of how many more wins we might expect assuming particular degrees of improvement in the offense and/or the defense.

If you're not a numbers person, then this probably isn't for you.  You can however, skip ahead to the conclusions piece and perhaps get something out of this piece.

BILL JAMES AND EXPECTED WIN PERCENTAGE

Bill James is the father of sabremetrics in baseball.  If you don't know what that is - google it - but for the purposes of this post - lets just say that James excelled in manipulating statistics to reveal truths that were not evident with the old school stats like batting average or RBIs, etc.

One idea that James created was to figure out a team's projected win percentage based on runs scored and runs allowed.  James applied a version the Pythagorean Theorem to create an expected win percentage (EWP).

The formula is:  Expected WP (EWP) = (RS²) / (RS² + RA²) where RS = runs scored and RA = runs allowed

So, if a team scored 64 runs and only allowed 29 (as Philadelphia has so far this season), then its expected WP would be .830 which is fairly close to Philly's current .875 winning percentage through 8 games.  Over the course of a larger sample size (162 games) the WP predictor is surprisingly accurate for most teams.  There are, however, occasionally teams that far exceed their EWP or fall far short of their EWP - these James attributed to simple variance (i.e. good luck, bad luck) and perhaps an ability to outperform in close games (perhaps the team has an exceptional closer, or perhaps they're stacked with HR hitters who can win a game with one swing).

APPLICATION TO COLLEGE FOOTBALL

We can apply the same formula to college football.  We do need to first understand that college football plays a much shorter season and so variance plays a bigger role.  Secondly, the nature of a football game is such that one big play can be decisive and because football awards different point values for different scoring types (7 points for a TD and 3 for a FG), the score in any particular game can overstate the actual competitiveness of that game.  Finally, the quality of opponents is more varied than in MLB (despite the existence of the Baltimore Orioles - there is no MLB equivalent to a Div II game).

Taking the last point first - one way to eliminate the problem of having vastly inferior opponents from skewing the data is to just confine our analysis to conference games only.  And in this case, I've chosen to focus first on just the ACC.  The ACC lends itself quite nicely to this analysis as it is a very balanced conference.

That makes the sample size smaller - so I'll look at 3 seasons of ACC play (2007 - 2009) which works out to 24 games for each team.  This is discussed later in this article.

Inside, we'll get into exactly what FSU needs to do.

Let's look at the data for last year 2009 for ACC play only.

So, reading the first entry.  Based on UVA's points scored (PS) and points allowed (PA), UVA's Expected Winning Percentage (EWP) was .245.  They won 2 games last year.  Based on their .245 EWP, they should have won 1.96 games.  They "out-performed" their expected win total by 0.04 wins.  In other words, they performed almost exactly as we would have expected.

GT and BC were outperformers - each wining about 1.5 more games than their PS and PA predicted.  Perhaps it was just variance (i.e. luck).  Maryland on the other hand - should have won nearly 3 games in conference rather than the 1 that they did win.  MD was probably a better team than their ACC record gave them credit for (the advanced metrics certainly indicate that their defense was decent).

I did a similar analysis for the 2008 and 2007 seasons (I'll include those as reference at the end for those who want to glean through them).

HOW MUCH IMPROVEMENT IS NEEDED TO WIN X NUMBER OF GAMES?

Next, I wanted to figure out - what EWP does a team need to have if they want to win say 7 ACC conference games.  Or 6 ?  Or 2?

So I compiled all the 7 win seasons in the ACC in this time frame.  Then all the 6 win seasons, etc. (There were no 8 win seasons from 2007-2009 in the ACC) and I calculated the EWP for those teams.  Here's the results:

I probably should gather more data going back more years because you can see that there's not a lot of difference between the EWP needed to win 2 games vs. winning 1 game.  Suffice to say - if your EWP is in the .330 range - you're not going to win too many ACC games regardless.

So, it appears that if you want to go 7-1 in the ACC - you need to be around .750 EWP.  Fall a bit short (.700 EWP) and you're probably a 6-2 team, etc.  Not surprisingly - if your EWP is .500 you are most likely to win 4 of 8 games.

So, you're saying "so what?"  Why did you take us on this crazy trip?  Well, now that we know about where FSU needs performance to be to equate to a particular ACC win total - we can work backwards off of the PS and PA numbers to figure out how much improvement would be needed to reasonably believe that the 'Noles might win 7 games in the ACC.

HOW MUCH DOES FSU's DEFENSE NEED TO IMPROVE TO GET TO 7 ACC WINS?

These were FSU's stats from last year.  The 'Noles scored a pretty good 268 pts and gave up a pretty horrific 278 points in ACC play.  The 'Noles won about the number of games they should have won.  FSU's luck balanced out.

Assuming that the offense turned in the same stellar performance, how much better would the defense need to be to project to 7 ACC wins?  Well, we know that FSU's EWP needs to be about .750 and we know the 'Noles scored 268 points in 2009 and we're assuming that to be constant for a moment.  This is just math applying the same EWP calculation, only working backwards.

The answer is that the defense would need to give up about 156 points in that scenario.  The 'Noles would have to be 122 points better on defense (or about 15.25 ppg).  I think its fair to say that that is unlikely to happen (FSU would have to improve its PA by about 44%).

Now keep in mind that FSU could outperform.  Perhaps catch a lucky break or two .  The 'Noles could get to a record of 7-1 and NOT have an EWP as high as .750.  GT and BC we saw outperformed by 1.5 wins just last year.  Let's assume that Florida State WILL outperform by 1.5 wins - that would mean that FSU would need to project out to 5.5 wins in EW to project out to 7 actual wins.  5.5 wins equates to about an EWP of .636.  So now how much improvement on defense would it need to make (remember we're holding the offense constant for now)?  The answer is 202 points would do it.  FSU would have to be 76 points better on defense (that's still a hefty 27% improvement in PA).

Here's the table for FSU PS = 268 (same as in 2009):

WHAT IF THE OFFENSE IMPROVED?

But, what if the offense, as expected, improves?  Well, how much do you think it will improve?

From 2007 to 2008, FSU's PS improved by 16%.  From 2008 to 2009, it improved by 20%.  Can it improve by as much again?  Let's say that it does.  Let's say it improves by 20%.  So now, instead of having 268 points scored in ACC play, FSU scores 322 (i.e we average 40.25 (that's an enormous and almost certainly unattainable number seeing that the offense was already one of the best in the history of the conference)).

What defensive performance would reasonably get FSU to 7 wins in that scenario?  The answer is FSU would need to hold ACC opponents to 187 points.  For 6 wins, FSU would need to hold teams to 206 points.  And if the 'Noles make no improvement on defense in this scenario - FSU should still expect about 5 wins in ACC play.

Here's the chart for FSU PS = 322:

CONCLUSION

So, I hope that helps put some of this in perspective.  FSU's defense needs to improve substantially if it is to have a good chance to win 7 games in ACC play.  Even a 7 ppg improvement on defense and a 20% improvement on offense only leaves the 'Noles at an EWP of .678, or about 5-6 wins in the ACC (unless the 'Noles outperform their expectations (get lucky)). 

So the next time that someone tells you that they think that FSU is going to win out in the ACC or even take 7 games  and go to the ACCCG - you'll know what's really required to get there in terms of team performance.  FSU needs to improve substantially and primarily on defense to accomplish that.  Certainly it can be done, but the improvement required is non-trivial.

BONUS:  WHAT ABOUT BEING IN THE NC HUNT?

For fun, what kind of performance is needed to be in the NC Hunt?  Well, here's a table of the Top 15 BCS schools from 2009 and the Top 10 from 2008 and their EWPs.  These are for conference games only.  I excluded non-BCS schools because I think their level of competition in conference is quite suspect.  Also note that Pac 10 teams played 9 conference games in this time frame and Big East teams only played 7 so their wins and expected wins need to be viewed in that light.

The above teams averaged an EWP of .807.  To be elite (i.e. in the Top 5), teams needed to average an EWP of .846.  Tellingly, only two teams in the above had PA approaching even 250 points in conference play.

One other point worth making - the two top EWP teams over the last two years were 2008 USC at .943 and 2008 Florida at .928 - both of those teams lost 1 conference game that year.  Which just goes to show you that the laws of variance are sometimes with you and sometimes against you.  So even if you reach a point where you should reasonably expect to be at 7 wins, if luck runs against you - you might find yourself at just 6 or even 5 wins.  Such is the nature of the game.

Here are the 2008 and 2007 ACC tables that I mentioned earlier.

2008 ACC EWP Table

2007 ACC EWP Table

Bonus #2

Above I wrote:

Sometimes a team will far exceed its projection (perhaps the team has an exceptional closer, or perhaps they're stacked with HR hitters who can win a game with one swing).

The corollary in football might be a great kicker.  The 'Noles have a pretty good on in Dustin Hopkins and he improved as the season wore on. 

Bonus #3 (By Bud)

FSU gave up 278 points or 35 points per conference game last year, (or 5 touchdowns per game).  And the 'Noles ACC schedule is almost the same with the exception of one game:  FSU drops Georgia Tech (finally) for Virginia.  UVA's offense is pretty awful and Georgia Tech's is excellent.  FSU gave up 229 points in non-GTech conference games, or roughly 33 points per game. So what if, FSU could hold Virginia to 14 points rather than the 49 it gave up to Georgia Tech.  That would decrease the total from 278 to 243.  And what if, in the remaining conference games, FSU could allow one touchdown less per contest?  That would cut the point total in the non-GT conference games from 229 to 180.  Is that reasonable?  I don't know, but I do think it is reasonable to assume that no opposing offense will improve more than FSU's defense will. 

So, replace GTech with UVA, hold UVA to 14 points, and allow a touchdown less per game in the remaining 7 conference games, and FSU goes from allowing 278 points in conference to 194, an improvement of 30% in a single season!  That 30% improvement would put FSU at 6th in the conference, behind only 5 top-20 defenses. 

Then, because the defense is improved, the offense could score more points without improving its level of play, solely because it would benefit from better field position and more possessions.  If FSU could do all of the above and score one more touchdown per game, it would give itself about an even money shot to win 6 or 7 games.  It's not very realistic, but it's probably how it would work out if FSU were to win 7 games.

A big thanks to GraniteStateNole for this excellent column.