Q&A 2011

Email your questions @ dave@cfbmatrix.com or tweet me @CFBMatrix

10/26/11 Drew in PHX, AZ – Hey Dave.  I am not surprised that USC, under Kiffin, is losing games they should be winning but is it random or is there any pattern within the CFBMatrix for their win vs. loses?

11/16/11 Notes/Update:  Today, on local radio 1080 The Fan @Iropp and @Big_Suke were discussing if Kiffin has done well or not in his brief USC tenure.  I couldn’t get through to talk about the CFBMatrix view, but took the opportunity to update this short article and profile.  Kiffin, in 2011, has broken the 2010 trend by beating CFBMatrix #27RR Cal at Cal and #24RR Washington in Los Angeles.  He is now just -1 game effect on the year for a -6 total for the last 2.  Still the 2nd worst net negative coach effect the last 2 years in the PAC12 but the trend is reversing.  The scholarship limits will now kick in to take the program down for 3 years and slowly back up for the 5-6 year process to be complete.  So for now, @Iropp, who didn’t think he has done well is correct IMO.

 Huge pattern in the CFBMatrix Drew.  In 2010, Kiffin’s wins and loses were very evident that he was coaching them down to a lower level of competition that their talent indicated.  As you know in the CFBMatrix model, I have a national 4 year recruiting rank for every team and it this then adjusted up or down depending on if they are at home or on the road.  When you look at USC’s 2010 13 game campaign (see chart below), they were 7-0 against teams with an adjusted 4 year recruiting rank (RR) over #36.  Conversely, they were 1-5 against teams with a RR better than #36.  The only exception was a win over the worst coached team in the country.  In a CFBMatrix nutshell, he is coaching is team down 36 RR positions.  The national average at home is set a zero and on the road it ranges from 0 to +12 for a majority of AQ teams. 
2011 is , although it is early, following the same pattern.  2011 is a much more difficult schedule (#3 CFBMatrix SOS in the country) and thus the pattern looks worse.  Of the 4 games they have played, all 4 fell into the same Kiffin 2010 coaching pattern.  Beat the patsies, loose to the less talented but better teams on the schedule.  If the pattern holds for 2011 the Trojans will finish 6-6 for 2011.  The coaching needs to improve. Desperately. Now.
2010 Scheduled Games Adjusted 4 yr RR Result 2011 Scheduled Games Adjusted 4 yr RR Result
at UCLA 5 W 28–14 @Notre Dame 1
Notre Dame 20 L 16–20 @Oregon 6
@ Stanford 24 L 35–37 @California 17
 Oregon 26 L 32–53 UCLA 25
@ Oregon State 34 L 7–36 @Arizona State 26 L
Washington 36 L 31–32 Washington 34
California 38 W 48–14 Stanford 36
@  Arizona 39 W 24–21 @Colorado 58
@ Washington State 41 W 50–16 Minnesota 62 W
Arizona State 41 W 34–33 Arizona 63
@ Minnesota 43 W 32–21 Utah 68 W
@ Hawai’i 63 W 49–36 Syracuse 79 W
Virginia 68 W 17–14

7/28/2011 Don in Butte, MT – Love the stats on lowest recruiting rank 4 year composite to win major conferences.  Where does Boise State fit in to the numbers?  I’m guessing they should be close.

In looking at the numbers,  Boise State with a 4 year composite recruiting ranking of #75, doesn’t fit in to the conference break downs very well.  The closest would be the Big East where they have had a conference champion in the last 6 years at a #78 composite (Cincinnati ’08), #75 UConn (’10) and a conf. champion average recruiting rank of #58.5.  However, for the rest of the conferences there really is no fit.  If you take away the ’07 miracle at Wake (#66) and the ’10 effort by #48 Wisconsin (shared), there has been no conference champions in the last 6 years in the ACC, PAC10, BIG12 BIG TEN, or SEC with a 4 year composite recruiting ranking under #38.

So here is the math.  Each year for years for the last 6, there have been 27 AQ teams with 4 year composite recruiting rank average of greater than #38.  This is 162 tries by those teams and 155 failures to win a title.  Boise State, in the last 6 years has had 4 year composite rankings of #93, #83, #74, #72, #59 (’09), and #71. 

Heading in to 2011 the BSU ranking is #75 (winning does not mean better recruiting but that’s another article).  No AQ school ranked in the 70s has won a conference title in the last 6 years or been in the top 20.  Even the lowest ranked team to be in the BCS title game had a 4 year composite of #18 (Boise’s best year was a #36 in ’09).

From the viewpoint of a college football fan, some of the logical choices for a conclusion would be (a) those recruiting rankings and stats for the last 6 years are meaningless, (b) the stats are meaningful but the media/computers are skewed to reward winning, (c) Coach Peterson and his staff are elite talent evaluators and the best coaches of the past decade or (d) some combination of the aforementioned A, B, C points to justify your belief or non-belief of the BSU rankings and ability to compete in the conference of your choice.  Perception is each fan’s reality.   So take from it what you want or don’t want to believe, they are just numbers. – Dave

Lowest Rank Div Winner* Lowest Ranked Conf Champ** Ave Rank Conf Champ***
ACC 66 66 28.8
Big East N/A 78 58.5
BIG Ten N/A 48 16.5
BIG 12 41 12 8.3
PAC12 N/A 32 10.4
SEC 23 9 5.3
*Rankings are  4 YR Average National Recruiting Rankings
**36 of 41 conf champs have had 4 yr rankings of #50 or lower
***Ave of all conf champ rankings for last 6 yrs

7/27/2011 @swagsman84 (twitter) Louisville, KY – So can Kragthrope recruit as well as UK?

In looking at the charts below, one cannot really say that Uk or Louisville outrecruits the other.  Both have been rising steadily since 2002 and the 4 year average national recruiting ranking is now in the mid-40s for both (#44 and #45).  UK has made the bigger gain over the last decade, but Louisville has had two back to back solid years with ’11 being its best class in ten years.

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
L’Ville Recruiting Ranking 76 42 62 45 31 41 54 53 47 30
L’ Ville 4 YR Recruiting MA 58 49 48 43 45 46 44
UK Recruiting Ranking 91 57 47 65 34 59 55 43 48 41
UK 4 YR Recruiting MA 67 52 53 56 54 50 45

7/22/2011 @theghostone (twitter)  From what I understand..Zook is a heck of a recruiter. I guess Illinois keeps him to bring in talent, just not coach it.  He must have been overwhelmed at UF & done little better at Ill.

In looking at all of his numbers at Florida and Illinois I don’t see a lot of difference.  Here is the breakdown of stats from within the modeling of the CFBMatrix programs.

1.  He and his staff are -10 games net coaching effect over the last 4 years (one of the worst in the country).

2.  His national recruiting classes since 2006 are ranked as follows:  #30, #21, #22, #46, #65, #36.  This averages out to an average of #36.66.  Sandwiched around him is Michigan State (next above) and Iowa (next below)

3.  The average recruiting class the 4 years before his hiring at Illinois was #42.5, not much worse than this tenure.  While at Florida from ’02-’05 he averages a #11 class, since #6.67 and he left on the heels of 8, 8 and 7 wins seasons (Meyer did a bit better).

4.  You can see in the win.recruiting trend chart below for him since 2002 the trends are decidedly in decline and have an average number of wins in the regular season at 5.5.

Other than the flash in the pan 2007 season at Illinois (still only won 9 games) it has been a decade of underperformance and disappointment from the perspective of the modeling and trends.

-Dave 7/22/11

7/14/2011 @JimmyV425 (twitter)  I am confused by the Ferentz numbers, help a guy?

Jimmy -Lets take the whole thing apart from the beginning (basics) to get the ‘coach effect’ number you see in the Hawkeyes page.  The whole season win total starts with the recruiting ranking.  I use a 4 year composite of all ranking systems to get a total ‘score’ and 4 year ranking for every team.  This composite helps reduce error and bias in ranking systems like Scout and Rivals and sets a level baseline for the CFBMatrix model.  The first pass is recruiting rank vs. recruiting rank.  In 2011 the 4 year rank for Iowa is #42.   In their first 3 games, they play only one team with a better rank (Pitt #38) so the recruiting layer would say Iowa starts 3-1.  Many teams are very consistent in just using recruiting rankings.  The Oregon Ducks, as an example, have a correct W/L prediction over 80% of the time in the last 4 years just my national composite recruiting ranking.

While recruiting is the most important factor, the second most important is the Home/Away schedule.  For teams with a new coach, their home and away variable is our calculated national average.  For teams with longer trends and metrics, I assign a value to their home and away performance (up or down depending on past performance).  Some teams, no matter who they play at home, have  a large value to adjust their numbers (Oklahoma), some are very good on the road and so their road value is higher (OK State).  For others it is a downgrade in the variable.  Cal is a good example in that they have shown no home field advantage in the last 4 years and thus their value is removed.  Getting back to Iowa, their home variable, when added to their recruiting rank puts them in a likely position to beat Pitt (on the road that would not be the case).

So from 2007-2010 the model gave a season total and game by game expectation of the Hawkeyes record for recruiting plus the home/away value.  Any time this ‘predicition’ is broken, then a value +1 or -1 game is given to the head coach.  For Coach Ferentz, he was -4 games in ’07, -1 in ’08, +2 in ’09 and -1 in 2010 (very inconsistent).  He was -2 at home and -2 on the road for an average of -1 game net coaching effect per year.  In 3 of the last 4 years, he has underperformed in the model.

The end result for Iowa in 2011 is a first pass of 7-5 against the recruiting composite.  their home/away value changes that to 8-4 and the average -1 net game effect from Coach Ferentz puts the Hawkeyes at a most likely 7-5.

This CFBMatrix model prediction is also reinforced in the win/recruiting trend chart at the bottom of their page which puts a flat win season most likely, the Year over Year Winning Odds Matrix shows a 57% chance of 7 wins or fewer and the returning starter matrix gives them less than a 48% chance of win 7 or more games in 2011.

Therefore, to keep the model simple, and able to predict the season most likely outcomes just after the recruiting ends

7/11/2011 – @DavisMets (twitter) I don’t understand why ppl r saying rebulding yr 4 Auburn when they had such good recruiting class 3 years ago.

Within the modeling, Auburn is still very strong.  In any other conference and division, I don’t think you would see the same opinion about a Auburn fall off from 2010.  Emotionally, folks see Newton and Fairley as the two players that carried that team, but it had been an elite recruiter for the previous 4 years.  The Tigers, in 2010, were modeled out for 11-1 (loss at Bama) and were expected to contend for a BCS title berth within my model.  This year the Tigers loose their starting QB and a ton of starters.  There is nearly a 1 game difference between teams that lost their starting QB and did not in 2010 and an even bigger gap for those returning less than 15 starters.  To me, within the model, it is the scheduling that is the biggest issue not the lack of talent.

6/28/2011 – HogsFan in Missouri – I know Arkansas is coming off a great year in 2010, but looking at all the numbers, are they really national title contenders as some (Hogs fans mostly) are saying?

Absolutely not.   Trends and metrics that the Razorbacks would have to overcome in 2011 are enormous.  In no particular order (1) Their recruiting is not good enough.  they would have to be the first team in the last 8 years to have a 4 year recruiting average under #18 nationally, not have a top 10 and not have a top 15 class in the 4 years leading up to a BCS title game appearance (and that’s before even trying to model them winning that game).  (2) They had 10 regular season wins last year and since 2001 only 43 teams have gone 10-2.  Of those, just 5 (12%) won 11 and 1 (2%) won 12 games the following year.   And (3) in 2010, teams that returned no QB and 15 or fewer starters with just 15 returning starters and no returning QB (13 teams) averaged -1.23 fewer games won (-16) in 2010 vs. 2009.  Sure, they have a chance, but not a good one.

6/2/2011  Mike in California (via email)  I’m curious why you don’t have Stanford as a contender for a BCS Title.  There is no team that has outperformed its recruiting ranking than Stanford.  And though Harbaugh isn’t there anymore, if that team can replace NT Sione Fua, I think they could replicate 2010.  The schedule is favorable, and they get Oregon at home (in what could be the biggest game in Stanford football history).

The answer from the CFBMatrix model is very simple – They do not fit the model as a BCS title berth contender.  That does not mean they cannot get there, but the odds are very unlikely.   This year may be better odds than most with all the various issues at the top schools (tOSU, Florida, Georgia, Texas, Auburn, Oregon and USC). Every team that has played (won or lost) in the NC game over the last 8 years has had something in common with recruiting.  All but one had 4 straight years of top 25 recruiting classes before their NC game berth and all had at least one top 10 and one top 15 in that span as well.  They were all also in the top 3 in 4 year recruiting rank moving average for their conference (Oregon & Auburn both fit the model and I had $ on both plus 6 others to win it in May 2010).   Stanford meets none of those criteria.  And even though the model doesn’t consider it for the BCS berth, they also have a new coach.Oregon, on the road under Kelly, is 10/12 within the model for road games.  CFBMatrix model says Oregon (ML will be +350 to +500).  I agree that game should be for the PAC12 North rep in the 1st PAC12 championship game.

5/30/2011 Kelsey in Maryland  (via Twitter)  Dave – How does the CFBMatrix model project the potential effects on recruiting of the scholarship losses on the USC Trojans?

Pending response

4/29/2011   Sandy in Orem, UT   How do you calculate recruiting rank composite? Is it simply the average of their relative year to year rankings? Or is it based on a ranking of total accumulated points across those years? The latter would be more accurate as it is based on absolute numbers vs relative rankings.

It is what you suggest and and weighted formula based on the value of the recruiting rankings of each site as they are not equal in value but less valuable than a combined model.  You are correct in that a #1 class from one year may not have the value of a #1 from the previous year.  Some sites do not offer a ‘point’ total and each calculates it differently via their mathematical models.   And if there is an extreme case of non-qualifiers or dismissals in a year I will adjust it but or the most part the time involved levels the averages well.

4/16/2011  @NotJoeLeeDunn State of MS (twitter) @cfbmatrix you know i love your stuff but you need to add major attrition to Ole Miss. They look horrible. four wins max this year.

This is a common question when a team has a handful on non-qualifying high-ranked recruits that never make it to campus.   While many kids may make an impact on a class, that impact is for just one of the previous 4 years.  I designed the CFBMatrix to be very simple from both understanding and the lack of constant change.  In extreme cases, if fans help me out and point to the attrition, I can make adjustments.  But I don’t have the time to go through even team and every player as all programs have these kinds of issues.  Those non-qual problems tend to balance themselves over time and across the country.  While most models and ‘experts’ like to make adjustments each week going forward, I like the fact that the CFBMatrix model just sticks with the numbers available and makes no excuses or exceptions throughout the CFB season.

However, is this a recruiting issue or a coaching effect issue?  The net coaching effect is not just on the field, it is coaching, it is finding and recruiting the right kids for the vision of the program.  I prefer to add or subtract it from the coaching effect over time as coaching starts with targeting the right, or in your example wrong, kids for the program to be successful on the field. 

That all being said, which kids and what year(s) should there be an adjustment in your opinion as I have heard this a couple of times this month from OM fans.


4/11/2011  James in Oklahoma – I was reading an article about FCS vs. FBS teams I found on Twitter here http://tiny.cc/nn92j .  I have been reading all your new stuff and figure you would have some more charts or information.

I don’t have a lot of information on the subject specifically as it relates to the article.  As games get added to the schedule it is only natural to have more games between the two divisions.   The current system really drives it as winning is the most important aspect in the system.  One can complain about these types of games, but what is worse, a national power scheduling one or two cupcakes or a top 10 team being ranked there by beating 9 to 12 sub top 50 teams?

You can see in the chart below that in 2003 there were more of these games played than in 2010 (blue line).

The winning ratio of these games (purple) has been as high as 14% and as low as 2%.  It is not a 99% sure thing.  In this time period it has actually been a 91.4% sure thing.

As a percentage of total games played, the percentage in 2010 were lower than the previous 4 years and were at a level below 2002 and 2003.  Only the ’04 and ’05 seasons had fewer games played  as a percentage of all games than in 2010.

Data Source: Wikipedia.org and Stats Inc.

3/16/2011  @HujoTDs – I listened to your podcast w/ @BlatantHomerism. Do you use others recruiting rankings inputs or create your own?

The team recruiting ranking is a weighted average of all available ‘expert’ data for any given year.  There are a few large bureaus that compile stats, stars and rankings.  They tend to be in the neighborhood of each other, but rarely to they agree.  To solve the question of which one is the best, I took 3 years of rankings from all available data and compared it to the baseline recruiting rank model.  The result was inconclusive and therefore not one ranking system was better than the other over time and teams.  What I did find is some potential bias and that in certain segments (ie higher ranked/mid-level, lower ranked teams) one model may have been better than the others.  To solve the bias and strength issues, the Composite Team Rank is a weighted average of all available national data.  So you can look at each one as it is interesting and each thinks theirs is the best.  But be sure to see our annual rankings and know that they are a representation of all information and as close to a non-biased average as is available no where else today.

3/2/2011  Lisa in California – How does Mack bring down the Horns a game? He’s one of the best coaches out there.

It is simply what I read from the CFBMatrix model as a profile for Coach Brown.  To start let me throw out last year.  As far as talent, SOS and EOS for Texas in 2010 versus their win total, last year was horribly unpredictable for the Longhorns.  Coach Brown was pushing the rare air into the realm of Weiss and Willingham for negative performance in a season.  He has always recruited on a very elite level, such that 11+ regular season wins is the baseline expectation in the model.  I understand that is a difficult accomplishment every year.  Nevertheless the model shows that he has the consistent ability to lose one game per year that he should not be losing.   The CFBMatrix still shows  them with elite talent but the Tier III rating is due to all the coaching changes.  Otherwise, they would be Tier II for a BCS title run in 2011 due to his annual propensity to lose one to many games to get into the BCS title game.  He models out very similar to Lloyd Carr.  Elite talent levels, 10+ wins, one lightning in a bottle National Title and the feeling of consistent underachievement.  That is how I read the CFBMatrix model.

3/1/2011 Joe in Michigan – For coaching & “net game effect” how do U determine this + or – ? Do U assume elite talent teams go undefeated and adjust? Thanks
Thank you very much for being the first person to ask a question about the CFBMatrix.  With regards to the ‘net game effect” the model starts with the baseline of recruiting ranking as to set the expectations.  The baseline model is adjusted for home and away games and then provides a level of win/loss expectations.   The amount a coach is +/- from the baseline starts to establish their net game effect on a season.   the longer the tenure of a coach, the more accurate the CFBMatrix model will be for predictions.  The real issue is new coaches and expectations.  Do you get a Rich Rod or Urban Meyer in the first year?  That is always impossible to really predict.  I take the past results of new coaches as a base expecation.  It is usually never more than +/- 1 to 2 games from the previous year’s finish.
The model does not ‘assume’ elite teams to go undefeated.  Most will have one road loss depending on their SOS and EOS of the conference schedule.  It would be very skewed if everyone had a WAC or other non AQ type fo schedule as there would be 50+ teams predicted to loose one or no games.  A good example would be the Georgia Bulldogs.   Elite talent with a long time coach that is a net negatvie game effect.  The model will likely predict a baseline of 10-11 wins for Georgia, but then it will factor the coaching and scheduling which will take them down in the predicted win total.  The last piece is a win/recruiting rank ratio trend adjustment which will narrow the expectations for wins to its final prediction for 2011

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s