2009-2010 Season Outlook and Some Magic Numbers
When people give their "keys to success," they'll often say something like "Team needs to not turn the ball over," or "Team needs to shut down the opposing runningback and make the offense one-dimensional." While these comments are somewhat insightful, I also find them sort of obvious too. In other words, it's sort of like "duh." Of course teams should not turn the ball over to win the game. Of course teams should try to limit the offense's offensive production.
To me, I don't really care how a team does it, so long as the goal (winning the game) is reached. Regardless of how a team reaches success, their success can often be seen in their statistics. I do admit, I've been a big critic of statistics because they can hide the true story (for example, and perhaps the biggest statistical myth of recent Cal Football fame is the whole Longshore 4th quarter touchdown to interception ratio - a lot less can be derived from that stat line than the typical lay fan will think). But when you look at statistics over the entire year, as opposed to a single game, the disparities are averaged out, and the team's overall success is fairly apparent. For example, teams who only average 50 rushing yards a game typically only have 3 wins a season. And teams who average 200 yards rushing a game typically have 10 wins a season.
So how many wins will Cal have in 2009? In most years, when I think about Cal's season outlook, I generally just made a prediction on how many games Cal will win because I can guesstimate the team's rushing abilities, passing abilities, and defensive abilities. But this year I am troubled because I cannot make that same prediction. I can't predict how many games Cal will win this year because frankly I just don't know how good Cal will be this year. The big question in my mind is "how will Cal's passing game develop?" I think that same question is the big question in 99% of Cal fans' minds. Without answering that question one can't really predict how the team will perform.
So I think the only way to really make any sort of prediction is to make the predictions couched in conditional phrases, such as "if Cal's passing game turns out to be great, then Cal will have a 10+ win season," or "if Cal's passing game turns out to be dreadful, then Cal will have a 6 win season." (Note, those aren't my real predictions, I'm just giving examples of conditional statements).
But how do you define "great" and "dreadful" and everything inbetween? Well, that's where I like to use stats and magic numbers.
In making my 2009 season outlook predictions, I'm going to use two assumptions. First of all, I'm going to assume that the Cal defense is going to stay just about as dominant as it was last year despite the loss of three starting linebackers. I'm thinking that our new linebackers are talented enough, and the eventual starters have gained enough experience from playing time last year to fill in adequately. Second, I'm going to assume that Cal's running game is going to stay at about the same level as it was last year. In other words, as crazy as it might seem to say this, but I still expect Jahvid Best to average approximately 7.5 to 8.0 yards per carry (which is a ridiculous number). Perhaps Jahvid Best's numbers will drop down towards the 7.5 yards per carry range since defenses will clearly be focusing on him rather than Cal's passing game, but for the most part I don't foresee it dropping any lower than 7.0 yards per carry.
Now, about that passing game. As I said earlier, I believe this is the biggest question to be resolved of the 2009 Cal Football team and its the performance of the passing game that will be determinative of Cal's season. It's because I am unsure as to how good (or bad) Cal's passing game will be that I must use conditional statements. But before I get to the conditional statmeents, let me briefly describe what constitutes good numbers and bad numbers for a QB.
When looking at a QB's statistics, the three most important numbers are his (1) completion percentage, (2) yards per attempt, and (3) interception percentage. Some hardcore statisticians believe that the yards per attempt is the king of the QB's statistics because it incorporates not only completion percentage, but yards gained per completion. Nevertheless, I still like looking at all three of the categories.
Completion Percentage
So what constitutes a good completion percentage? Par for the course is about 65%. If your QB throws a 65% completion percentage over the year, that's darn good! That's a good number. If he throws for 67%, that's great! If he's up in the 70%+ range, well he's like Tom Brady and that's freakin fantastic. On the down side, a completion percentage of 60% to 64% is okay. Anything less than 60% is unacceptable.
Just to throw some comparisons out there:
In 2002, Kyle Boller threw 53.6% (although perhaps a good 5% of those passes were drops by Ward).
In 2004, Aaron Rodgers threw for 66.1% (!!!).
In 2005, Joe Ayoob threw for 49.2%.
In 2006, Nate Longshore threw for 60.2%.
In 2007, from the start of the season through the Oregon game, Nate Longshore was throwing at 63.8%.
In 2008, Texas' Colt McCoy threw for 76.7% (!!!!!!!!!!!!!!!!!!)
(Note: I am not including Kevin Riley's stats from 2008 because due to the QB switching and substitution into must-pass situations such as the USC game, his stat line is not accurate)
Yards per Attempted Pass
What makes a good yards per attempted pass? A yards per attempt of about 7.0 is par for the course. Once you get into the 7.5 yards per attempt range, that's certainly above average and pretty good. Anything above 8.0 yards per attempt is ridiculous. On the flip side, 6.5 yards per attempt is sub-par, and anything less than 6.0 yards per attempt is fairly pitiful.
To throw some comparisons out there:
In 2002, Kyle Boller threw for 6.68 yards per attempt.
In 2004, Aaron Rodgers threw for 8.2 yards per attempt (!!!)
In 2005, Joe Ayoob threw for 6.72 yards per attempt.
In 2006, Nate Longshore threw for 8.01 yards per attempt (thank you Desean "deep threat" Jackson!)
In 2007, from the start of the season through the Oregon game, Nate Longshore was throwing at 6.98 yards per attempt.
In 2008, Oklahoma's Sam Bradford threw for 9.8 yards per attempt (!!!!!!!!!!!!!!!!!!!!!)
Interception Percentage
What makes a good interception percentage? A low one. As low as possible. But generally speaking, about 3% is acceptable. Anything more gets to be not good very quickly. When you're at about 2.5% interception percentage, that's great. Less than than 2% is ridiculous.
To throw some comparisons out there:
In 2002, Kyle Boller's interception percentage was 2.37%.
In 2004, Aaron Rodgers' interception percentage was 2.53%.
In 2005, Joe Ayoob's interception percentage was 5.51%.
In 2006, Nate Longshore's interception percentage was 3.45%
In 2007, from the start of the season through the Oregon game, Nate Longshore's interception percentage was 1.23% (!!!!!!!!!!!!!!!!!!!!)
In 2008, Oklahoma's Sam Bradford's interception percentage was 1.66% (!!!!!!!!!!!!!!)
Summary Thoughts
As you can see, the stats (more or less) don't lie. When the Cal QBs had a good stats, such as in 2004, 2006, and through the Oregon game in 2007, the Cal teams did great. When the stat lines suffered, such as in 2002 and 2005, the teams didn't do so well.
2009 CAL FOOTBALL SEASON OUTLOOK
So now that we've had a look at some examples of ideal statistics, here is my personal 2009 Cal Football season outlook couched in conditional statements.
I'm only going to use completion percentage because, while it certainly doesn't include yards per completion like the yards per attempt stat, it also isn't subject to skew such as the yards per attempt stat by an offense that completes really long passes. In other words, a team can have a pretty good "yards per attempt" stat by completing a couple of really long passes more often than not and hiding a low completion percentage. A perfect example of this is Georgia's Mathew Stafford in 2008 (#1 overall draft pick in the 2009 NFL draft). He had a fantastic yards per attempt of 9.0 yards per attempt, but had a very mediocre 61.4% completion percentage. In plain English, his completion percentage wasn't that good, but his yards per attempt numbers were great because he had great down-field threat WRs, and completed more longer passes than most QBs. But what matters to me the most is the completion percentage stat because I want to know if the QB is completing passes. If he's completing passes, I'm going to assume he's also gaining at least the average amount of yardage per pass too.
So finally, here's my outlook on the 2009 Cal Football season:
If Cal's starting QB can complete greater than 65% of his passes, Cal will most likely have an 11+ win regular season.
If Cal's starting QB completes between 62%-65% of his passes, Cal will most likely have a 10 win regular season.
If Cal's starting QB completes between 60%-62% of his passes, Cal will most likely have a 9 win regular season.
If Cal's starting QB completes between 58%-60% of his passes, Cal will most likely have a 8 win regular season.
If Cal's starting QB completes between 55%-58% of his passes, Cal will most likely have a 7 win regular season.
If Cal's starting QB completes less than 55% of his passes, the 2nd string QB should probably be starting.
For the past few months, I've been saying in the DBDs that Cal's magic number is probably 60% completion percentage. By that, I mean that if Cal's QB can throw greater than 60% completion percentage, Cal has a great chance at a 10+ win season. But upon further thought, I think that magic number has to be increased from 60% to something a little higher, such as 62%. The following is my reasoning why.
Months ago, when I first was pondering Cal's 2009 outlook, I did a little stat comparison with previous Cal QBs. Most notably I looked at the 2006 team. During that year, Nate Longshore passed for 60.2% completion percentage and Cal almost went to the Rose Bowl. The 2006 team and the 2009 have a lot of similarities so I began thinking that 60% seemed to be the magic number, hence that is the number I've been reciting for months now.
But upon further thought, I began realizing that the reason why the 2006 Cal Football team's passing game was so successful was because of Desean Jackson. He provided a great deep threat for Nate Longshore, as well as being a superb special teams threat. Defenses truly couldn't really focus on Marshawn Lynch without giving up on pass defense and giving up scores to Desean Jackson. But this year, Cal doesn't really have a wide receiver to be Desean Jackson. Cal doesn't really have a really speedy deep threat that will command the attention of the defensive secondary. So to make up for the lack of deep gains and quick touchdowns that Desean Jackson was able to give the team in 2006, the Cal QB and WRs are going to have to complete more passes than normal. In other words, in 2006 Longshore could get away with a 60.2% completion percentage because Cal had a superb deep threat (evidenced in Longshore's 8.01 yards per attempt), but in 2009 since Cal has no deep threat, Cal's starting QB will h ave to complete more pass attempts to make up for the lack of yards per attempt.
Conclusion
By no means are these numbers a 100% indicator of things to come. Cal's QB might be able to throw a 60% completion percentage and Cal could still go 11-1 on the season with some luck and superb play from the other parts of the team. But for the most part, I do see Cal's season riding on the hand of the starting QB.
Some of you may be wondering what the heck the difference is between 60% and 62%? Well, in 2006 Nate Longshore threw 377 attempts on the year (I'm using 2006 because I think the 2009 team most resembles the 2006 team). Assuming Cal's starting QB this year throws for the same amount of passes, then 60% of 377 is about 226 passes completed. Alternatively, 62% of 377 is about 234 passes completed. That's only a difference of about 8 passes. So some of you may be thinking that isn't a significant difference? Well, perhaps it's not. But sometimes, football games come down to one play - one passing play. Perhaps it's a touchdown pass to seal the game, or a crucial first down which eventually leads to a go-ahead field goal to win, etc. But while it's just one pass, it's still a pass. Those eight passes may not seem like a lot, but perhaps three of those eight passes are against USC. Perhaps those three passes are passes that most QBs won't make, but somehow Cal's starting QB nails the throws and Cal gets two critical first downs to burn the clock and one touchdown. That's all Cal may need to win the game. So while a mere eight passes extra might not be a lot - I contend that it could be the difference between winning and losing (theoretically) up to eight football games. Every throw counts.
In sum, I believe that Cal's starting QB needs to have at least a 62% completion percentage for Cal to have a good chance at a 10+ win season.
4 recs |
31 comments
|
Comments
Completion Rate Analysis
Hydrotech has an interesting analysis, and I appreciate the thought and work behind it. There is no question that Riley’s completion rate needs to increase, but it’s hard to factor in the difference that a big play back like Best makes in the calculus. While the other teams had great RB’s, Best’s big play potential is huge and may offset a lower QB completion rate. In either case, given Riley’s history and the quality of his receivers, I am doubtful whether a 60%+ completion rate is achievable. If it is, I agree that the team has a better chance of a 10+ season. We’ll have a lot more data, and a clearer picture, in a few weeks. Keep up the good work.
by oldbear on Sep 3, 2009 5:16 AM PDT reply actions 0 recs
Good thoughts. Thanks for commenting!
Contact if you want to chat: bearsnecessities@gmail.com
by Avinash on Sep 3, 2009 2:33 PM PDT up reply actions 0 recs
thanks hydro!
I agree that we need something above 60% completion percentage. And also agree with oldbear that Bests big play potential is a factor in the passing game as well. I would love to see Riley up in the 65% range, even if the YPA is not real high.
Go Bears Go
by Rocksanddirt on Sep 3, 2009 7:40 AM PDT via mobile reply actions 0 recs
Who are these ridiculous people voting we need 68%+ for a 10 win season? We had a 10 win season with Rodgers when he put up 66%.
Perhaps Jahvid Best’s numbers will drop down towards the 7.5 yards per carry range since defenses will clearly be focusing on him rather than Cal’s passing game, but for the most part I don’t foresee it dropping any lower than 7.0 yards per carry.
I mean, Marshawn averaged 8.8 ypc as a freshman, 6.4 as a sophmore, and 6.1 as a junior. Best could drop down to 6.5 or so and still have a very successful season. Adrian Peterson at Oklahoma never averaged over 6+ ypc, even the year he rushed for 1900+. So what I’m getting at is the expectation for Jahvid to put up 7+ ypc is a little extreme.
by Missing Barry on Sep 3, 2009 8:37 AM PDT reply actions 0 recs
It seems Jahvid has set the expectations with his sophomore campaign.
Contact if you want to chat: bearsnecessities@gmail.com
by Avinash on Sep 3, 2009 9:15 AM PDT up reply actions 0 recs
And I definitely think he’ll still be on an elite level, it’s just the whole “regression to the mean” issue. Outstanding performances like his year last year are likely exactly that – outstanding – and it’s more likely that his future performance will be more ordinary for his talent level (which should still be among the nations best, but probably not where it was last year).
by Missing Barry on Sep 3, 2009 11:37 AM PDT up reply actions 0 recs
There’s an interesting philosophy on play-calling that would suggest Best’s ypc were actually too high last year, and it would be better if he was down in the 6-7 ypc range. The idea is that, if your quarterback is throwing for only, say, 6 yards per attempt, and your tailback is running for 8, then you’re not running enough. Especially because running is a lower-risk proposition, you should keep running and running until the defense adjusts to bring the yards per carry down, thus opening up more opportunities for the passing game. A theoretical equilibrium would exist when yards per carry and yards per pass attempt were more or less the same, perhaps adjusted for the higher risk associated with throwing the ball.
One way to help Riley become more efficient might be to run the ball more. This would almost certainly bring down Best’s (and/or Vereen’s) per-carry average, but perhaps to the benefit of the passing game and the overall effectiveness of the offense.
Even if you don’t buy that theory about play-calling balance, any back averaging over 8 yards per carry is not just good but is also probably not touching the ball enough.
Go Bears!
by California Pete on Sep 3, 2009 4:21 PM PDT up reply actions 2 recs
Rec'd for insight
Cal did run the ball more than they passed it. They had 38 more run attempts than pass attempts (435-397), as opposed to 2007 where they were nearly identical (441-443). The offense ran about 50 less plays in 2008. I guess the problem is that (a) Best got injured in several games (CSU, USC and Miami), and it limited his carries throughout the season, thus inflating his stats over what they would’ve been if he’d played the whole season.
I’d suggest looking back over the past few years and see how the running and passing ypc’s compare for Cal and the Pac-10, maybe do a fanpost on the topic. It’d be great stuff to look at.
Contact if you want to chat: bearsnecessities@gmail.com
by Avinash on Sep 3, 2009 4:41 PM PDT up reply actions 0 recs
I voted 68% and what!?
…but I just always vote for the top selection, go figure.
by CaliforniaCMB on Sep 5, 2009 11:37 AM PDT up reply actions 0 recs
Good post. Looking forward to the open thread for tonight’s Oregon v. Boise St. game.
by The Hombre on Sep 3, 2009 9:35 AM PDT reply actions 0 recs
A well thought-out post, Hydro. Re<’d.
I like this but I really think Best goes for 2000 yards this year. He had 1600 and he was hurt/ineffective/taken out of some games by the score. This year, I see him “putting it all together.”
The key isn’t how many passes Riley completes or yards Best gains. It’s O-line play. A QB with all day to throw (like any offense versus the niners pass rush) will be effective. If Cal’s line plays like ’04, Best might single handedly beat 7-8 teams.
What I really hope the offense has captured this off-season is the following:
- well timed deep outs/ins and slants.
- some Sofele fly sweeps/WR screens to add another element for teams to think about and/or occasional spread option runs (both used less than 5 times/game).
- Use Best more in the passing game.
- know how to find the TE when backs/receivers get a lot of attention.
- no-huddle. Ludwig’s done it his last, what, 3-4 games? If used properly and executed well, it can be a game changer against any foe..
"We lose to Stanford in many sports, but if you want to make a Cal team quit, bring a weapon."
--Coach Clark
by carp on Sep 3, 2009 9:56 AM PDT reply actions 2 recs
I agree that the no-huddle could be very valuable at times, in particular against Maryland. It seems like Maryland’s new defense like to sub a lot and call exotic blitzes. If we no huddle, I think we could neutralize this to a certain extent and limit their packages.
Are we on to something? Could this by why Tedford closed practices???
GOLD OUT MOZAMBIQUE!
by OskiMonsta on Sep 3, 2009 2:27 PM PDT up reply actions 0 recs
I put the chances of Cal using a no-huddle as a part of its regular offense at around 0.0005%. It just doesn’t seem like Tedford’s style.
www.californiagoldenblogs.com
by HydroTech on Sep 3, 2009 3:09 PM PDT up reply actions 0 recs
I think Best will run for fewer yards this year, in part because Riley will be better, defenses will be keying in on stopping him, and Cal played some epically, epically horrid teams last year. There is no way UW and WSU are giving up 500 yards to Best again.
What happens in Vegas stays in Vegas. What happens in California makes the world go round.
by Spazzy Mcgee on Sep 3, 2009 2:56 PM PDT up reply actions 0 recs
Binomial statistics
Suppose this win prediction model is correct. When will we know with statistical confidence which track Riley is on? I played around with some binomial statistical error estimation to find out: the standard deviation of a binomial distribution with mean eff (completion ) and sample size N (# of passes) is sigma = sqrt(eff*(1-eff)/N). If eff = 0.62 and N = 377, then sigma is 2.5 percent: this means we will never know with statistical certainty, at any point in the season, whether Riley is inherently a 62 passer, or merely a lucky 59.5% passer, or an unlucky 64.5% passer. So 2% increments simply do not have predictive usefulness.
But how many passes will it take to know he is definitely a bad passer? If his inherent % is 55%, then we will know with 68% confidence level that he is worse than 62% when N = 50 or so (sigma = 7%). Similarly, if he is inherently 62% then we will know he is definitely better than 55% also after about N=50.
So I would say that after 3 games or so, we will know that if he is playing very well (>69%), that a >62% season is likely (>68% chance), or if he is playing very badly (<55%), that a >62% season is unlikely (<32% chance). However, he could still be throwing <60% after 6 or 7 games and still be indistinguishable from an unlucky 62% passer (and after the first 7 we will already know whether we are on track for 9-10-11+ win season just by looking at our W-L record). As N grows, the picture becomes a little sharper, but not dramatically so, due to the square root dependence on N.
Not so surprisingly, this coincides with one’s intuitive expectations about evaluating QB’s. If he looks good for one game, or even two, that could be “a fluke”, but after three good games one is fairly confident he is not a bad (or good) QB who is on a lucky (unlucky) streak. And you don’t fully understand his greatness or ineptitude until after a full season, and even then there are lingering doubts.
All this assumes all other factors are inherently constant over the course of the season, such as team performance, opposing team performance, etc. These kinds of assumptions hold up surprisingly well for things like FG% or FT% in basketball, no idea if they apply to football. One has to worry that a particular game, or even portions of games, are biased estimators of mean performance, because of injuries, weather, game plan, the score, etc. You probably have to throw out the EWU game from the get go and use the passing stats from Maryland-Minnesota-Oregon to get the first crude performance snapshot.
by the bear facts on Sep 3, 2009 10:01 AM PDT reply actions 7 recs
fascinating stuff
rec’d!
So, basically, you gotta Go Bears!
by ragnarok on Sep 3, 2009 11:22 AM PDT up reply actions 0 recs
Yes, very fascinating.
www.californiagoldenblogs.com
by HydroTech on Sep 3, 2009 11:23 AM PDT up reply actions 0 recs
One of the biggest problems, which you kind of addressed, is the talent disparity issue in college football. There’s such a big range of competition that I imagine it has a significant effect on the assumption that “other factors are constant”….
by Missing Barry on Sep 3, 2009 11:40 AM PDT up reply actions 0 recs
nice evaluation!
any chance you can jack up the confidence interval to 80 or 90%? and get an idea of the N for good/bad year?
Go Bears Go
by Rocksanddirt on Sep 3, 2009 11:51 AM PDT up reply actions 0 recs
The N size is number of passes, which can’t be increased…which is why any prediction will either have a low margin of error or have a lower confidence level (bear facts, is that correct?).
The other problem, as he also points out (awesome comment btw), is any prediction assumes all other factors to be constant, which is not the case.
by HolmoePhobe on Sep 3, 2009 12:39 PM PDT up reply actions 0 recs
Rocksanddirt’s question is well-posed. CL for the same statistical test can be increased if N is increased, which means you simply have to wait more games into the season before you make your conclusion at the higher confidence level. Of course the longer into the season you have to go the less of a “leading indicator” this whole completion percentage business becomes for a W-L record.
The answer to his question is more or less in my original post. 90% CL corresponds to 1.65sigma, so therefore instead of 50 passes you need 50*1.65*1.65 = 136 passes (about 4.7 games) to make the same conclusions at the higher CL. By the way, I was sloppy in my original post, thinking 50 passes = 3 games. 50 passes is really only 1.7 games (for Cal 2006), so the 68% CL figures I cited above will be realized late in the third quarter of the Minnesota game (discarding EWU), and not at the end of the Oregon game as I originally claimed.
It would be fun (for me) to back test this against the 2004-5-6 Bears: does the completion percentage after 2,3, or 4 games predict the average at the end, and is it to an accuracy consistent with statistical variation alone? Anyone have the stats in a spreadsheet?
Note that, although there can be many variables which affect pass completion, like SOS, if the integration time for our sample is the same time scale (or greater) at which all of these factors average out, then it is an unbiased estimator of the whole season even if they are not constant. As a crude example, if the total season schedule has one half bad teams and one half good teams, then I can make an estimate for the whole season average from a two-game subset which has both a good team and a bad team in it. If the relevant variables are varying strongly over medium time scales, so that no few game subset at the beginning is an unbiased estimator of the average, then statistics cannot help you. It’s not obvious to me that pass completion in football always has that property.
by the bear facts on Sep 3, 2009 7:52 PM PDT up reply actions 0 recs
lol
What happens in Vegas stays in Vegas. What happens in California makes the world go round.
by Spazzy Mcgee on Sep 3, 2009 2:57 PM PDT up reply actions 0 recs
One crucial bit is missing: Third-down conversion rate
Now this doesn’t always have to do with the QB but most of the time it does.
A QB can complete only 1/3 of all of his passes but that 1 pass happens to be on third-down and happens to convert. So he could hover around a 33% completion rate but with a 100% third-down efficency. That team would do wonderfully.
In other words, Go Bears!
by royrules22 on Sep 3, 2009 11:31 AM PDT reply actions 0 recs
if they throw on 1 and 2nd downs....if you run and throw mixed in
than your third down metric is off from your example…..
Go Bears Go
by Rocksanddirt on Sep 3, 2009 11:53 AM PDT up reply actions 0 recs
and i need to try and do fewer things at the same time, as that
comment makes little sense.

Go Bears Go
by Rocksanddirt on Sep 3, 2009 11:56 AM PDT up reply actions 0 recs
I’m just giving an extreme example: no running what so ever, 1st and 2nd down are incomplete passes, 3rd down is pass complete that converts to 1st down or TD.
In other words, Go Bears!
by royrules22 on Sep 3, 2009 1:57 PM PDT up reply actions 0 recs
Another way to look at it:
- last year Cal won 9 games with an overall completion % of about 53%
- you assume that running game and D will be about the same
- don’t forget ST, which I think could improve (think kickoffs in bounds, inside the 5) but let’s assume it’s about the same too
Then, how much better does Cal’s completion percentage have to be to win one more game? I think it might actually be less than 62%.
GOLD OUT MOZAMBIQUE!
by OskiMonsta on Sep 3, 2009 2:40 PM PDT reply actions 0 recs
Good points.
www.californiagoldenblogs.com
by HydroTech on Sep 3, 2009 3:10 PM PDT up reply actions 0 recs

by 





















