There are 119 users in the forums

Remember
Not a member? Register Now!

Really? 7.2 Wins?

  • Lifer
  • Veteran
  • Posts: 950
When they talk about regression to the mean, they're talking about the tendency for any statistic to move towards the norm over time. In other words, if you flip a coin 100 times, the statistics say it's gonna come up heads just about 50 times. Now let's say you start flipping that coin and it comes up heads 15 out of the first 20 tries. Will that trend continue? No. "Regression to the mean" simply says that by the time you get to 100 coin flips, the ratio will settle down and you'll end up with something much closer to 1:1.

Football Outsiders makes some valid points when they say that the 49ers are unlikely to maintain their anomalous advantage in turnover differential, low rate of injury, and victory in close contests. True enough, but that doesn't mean they'll necessarily lose more games. With better production in the red zone, improvement on third down, more explosive plays, and higher point production, they might win MORE games this year. If statistics could accurately predict what was going to happen next, there would be no need to play the games. Truth is, anything can happen... and probably will.

My personal take is that it's gonna be much harder this year. A tougher schedule, we're the hunted instead of the hunter, and we'll probably have more turnovers and injuries. Balancing that, I see potentially large gains in Offensive production, and a relatively weak division. My prediction is that we'll go 9-7, but we'll be much better equipped to make a run in the playoffs. And my prediction has exactly the same chance as the stats-crunchers at FO.
unless we get a lot of injerd players on D , we will have a good year not less than 10 wins !
You wouldn't know it from reading the comments here, but most of the article on the 9ers is spent complimenting Harbaugh and explaining how he outcoached the opposition in key games. Although the computer simulations give the 9ers a mean of 7.2 wins, the writer suggests that they might well do better than that. Green Bay and New England are predicted to do very well, no surprise there.
Originally posted by Lifer:
When they talk about regression to the mean, they're talking about the tendency for any statistic to move towards the norm over time. In other words, if you flip a coin 100 times, the statistics say it's gonna come up heads just about 50 times. Now let's say you start flipping that coin and it comes up heads 15 out of the first 20 tries. Will that trend continue? No. "Regression to the mean" simply says that by the time you get to 100 coin flips, the ratio will settle down and you'll end up with something much closer to 1:1.

Football Outsiders makes some valid points when they say that the 49ers are unlikely to maintain their anomalous advantage in turnover differential, low rate of injury, and victory in close contests. True enough, but that doesn't mean they'll necessarily lose more games. With better production in the red zone, improvement on third down, more explosive plays, and higher point production, they might win MORE games this year. If statistics could accurately predict what was going to happen next, there would be no need to play the games. Truth is, anything can happen... and probably will.

My personal take is that it's gonna be much harder this year. A tougher schedule, we're the hunted instead of the hunter, and we'll probably have more turnovers and injuries. Balancing that, I see potentially large gains in Offensive production, and a relatively weak division. My prediction is that we'll go 9-7, but we'll be much better equipped to make a run in the playoffs. And my prediction has exactly the same chance as the stats-crunchers at FO.
good post and basically sums it up.
Originally posted by Lifer:
When they talk about regression to the mean, they're talking about the tendency for any statistic to move towards the norm over time. In other words, if you flip a coin 100 times, the statistics say it's gonna come up heads just about 50 times. Now let's say you start flipping that coin and it comes up heads 15 out of the first 20 tries. Will that trend continue? No. "Regression to the mean" simply says that by the time you get to 100 coin flips, the ratio will settle down and you'll end up with something much closer to 1:1.

Football Outsiders makes some valid points when they say that the 49ers are unlikely to maintain their anomalous advantage in turnover differential, low rate of injury, and victory in close contests. True enough, but that doesn't mean they'll necessarily lose more games. With better production in the red zone, improvement on third down, more explosive plays, and higher point production, they might win MORE games this year. If statistics could accurately predict what was going to happen next, there would be no need to play the games. Truth is, anything can happen... and probably will.

My personal take is that it's gonna be much harder this year. A tougher schedule, we're the hunted instead of the hunter, and we'll probably have more turnovers and injuries. Balancing that, I see potentially large gains in Offensive production, and a relatively weak division. My prediction is that we'll go 9-7, but we'll be much better equipped to make a run in the playoffs. And my prediction has exactly the same chance as the stats-crunchers at FO.
I don't disagree with you (accept the 9-7 prediction). Statistics are statistics. But I think with hire of JH & Co. the 49er have crated a new norm for themselves, much the same as Bill Walsh created a new norm. So last season's 13-3 was a deviation from an old norm but our standard from here-on-out, I believe. I don't think we win no less than 11 games. We have a target on our backs, that's for sure. But I think 9-7 is a regression. What people must understand is that we seriously addressed the issues holding this team back. So the formula to beat us won't look like anything from last season. Teams are gonna have to be more complete to beat us - from coaching to special teams to backups. (And there aren't many complete teams in the NFL.) So I don't think last season was a fluke (not that you said it was). The missing part to this team was coaching and had been for some time before JH.
Originally posted by Lifer:
When they talk about regression to the mean, they're talking about the tendency for any statistic to move towards the norm over time. In other words, if you flip a coin 100 times, the statistics say it's gonna come up heads just about 50 times. Now let's say you start flipping that coin and it comes up heads 15 out of the first 20 tries. Will that trend continue? No. "Regression to the mean" simply says that by the time you get to 100 coin flips, the ratio will settle down and you'll end up with something much closer to 1:1.

Football Outsiders makes some valid points when they say that the 49ers are unlikely to maintain their anomalous advantage in turnover differential, low rate of injury, and victory in close contests. True enough, but that doesn't mean they'll necessarily lose more games. With better production in the red zone, improvement on third down, more explosive plays, and higher point production, they might win MORE games this year. If statistics could accurately predict what was going to happen next, there would be no need to play the games. Truth is, anything can happen... and probably will.

My personal take is that it's gonna be much harder this year. A tougher schedule, we're the hunted instead of the hunter, and we'll probably have more turnovers and injuries. Balancing that, I see potentially large gains in Offensive production, and a relatively weak division. My prediction is that we'll go 9-7, but we'll be much better equipped to make a run in the playoffs. And my prediction has exactly the same chance as the stats-crunchers at FO.
I certainly agree with this - however, my argument is that there are way too many variables to have this factor into any analytical "mean". In your example of the coin, we know the baseline is 50%. The flip is a fixed action that can only result in heads or tails. I think a closer analogy to what FO is doing is trying to predict a coin flip in a moving car. Here you introduce factors that may or may not influence the action of the flip and thereby affecting the result. The model in which they use to "predict" must be weighted in some way as to predict certain teams being more likely to win more games and other less likely. I understand what they're trying to argue, but assess that "weight" of a team is in itself completely subjective. Why is Green Bay better than let's say Patriots or the Giants, or even the Saints? How can that be expressed as objective data for creating a model in which we can reasonably predict performance? I don't think it's possible personally.

Yes, teams that do exceptionally well tend to perform back down to a "mean" - but again, I ask the question - what is baseline for that team? And where do you begin the evaluation? With a coin it's much easier because the outcome is fixed. The coin is never "weighted" or in abstract, influenced by the process of being flipped (who's flipping, environmental factors, etc).

IMHO, I think this article is written more for press than revealing some actual, hardcore analysis.
This is fine for me. I personally hate it when the media crowns a winner in the offseason. I prefer underdog status...
  • cciowa
  • Veteran
  • Posts: 27,019
Originally posted by MntIdaGold:
This is fine for me. I personally hate it when the media crowns a winner in the offseason. I prefer underdog status...

I am glad the media and the rest of the world hates us again cuz we are good. it reminds me of the championship days
www.youtube.com/watch?v=EzE8a2IfH9c
a couple things:

1. What they also don't take into account is that we easily coulda/shoulda/woulda been 19-0. There wasn't a single game where we were legitimately outplayed and deserved to lose (3 of our 4 losses were by a combined 8 points, 2 of which were in OT). The Dallas game we win if we had Goldson (M. Williams f**ked up several times that game leading to touchdowns) OR if Harbaugh accepts that penalty and takes the Akers 55 yd FG off the board. The Arizona game we win if the refs don't mistakenly disallow that Jon Goodwin touchdown (plus our best player, Willis, was out), leading to a 14 point swing in a 2 point loss. The Baltimore game had us at an egregious competitive disadvantage as we were the only team in NFL history to travel cross country for a Thursday game where we had essentially zero time to prepare (did we even get a single practice in that week?), plus we still likely would have won were it not for the terrible calls negating Ginn's 75 yd touchdown and Tarell Brown's interception. The NYG NFCCG game we obviously win if Ted Ginn played OR Bowman's forced fumble late would have counted. Not to mention our biggest (only?) weakness last year was our 3rd down and red zone conversion rates, which were a direct result of no offseason and should be MUCH better in 2012. Our biggest (only?) positional weakness last year was WR and now we have replaced Brett Swain and Joe Hastings with Randy Moss and Mario Manningham.

2. Some of the examples they list that we were lucky are skewed. For instance, they talked about how few players we had go on IR which is the most overrated stat of all time. If a team puts a backup safety on IR because going into the last game of the season he has a hangnail, that counts the same for that stat as if a team puts their starting QB on IR in August because of a catastrophic injury. So right off the bat I don't put much stock into that. Besides, we were not as healthy as they make us seem on defense - Sopoaga had his staph infection, McDonald was battling his hamstring all year, WILLIS missed a quarter of the season, Goldson missed the first 2 games which was a direct cause of our Dallas loss, our nickel spot was constantly in a state of flux due to injuries for a while 'til Culliver really nailed it down, etc. Also, remember how last offseason EVERYONE was saying how we were the team most at a disadvantage because of the lockout, and they were indeed actually right if you really think about it which makes our 2011 success that much more amazing and now we get our first ever offseason under the best coach in football. Where do they mention that?

Moral of the story? Even if we struggle to replicate our turnover differential or have some more injuries, that should be more than canceled out by the heavily upgraded WR corps and the fact that now we have an offseason under Harbaugh (which, among many other things, should improve significantly improve our 3rd down and red zone offense) and that this will be the first time Smith has the same OC/offense in back to back seasons. Plus, the fact that everyone is a year older really only has the potential to negatively affect just 2 guys, Justin Smith (still highly unlikely though) and Frank Gore, while it will help way more than 2 guys (Aldon, Bowman, Culliver, Iupati, A. Davis, etc.)
I'm counting 10-11 wins just by going down the schedule. That's if we only beat the teams we should beat.
Do they own a betting webiste? I'd really would like to take the over with even money.
  • Lifer
  • Veteran
  • Posts: 950
Originally posted by NinerGM:
Originally posted by Lifer:
When they talk about regression to the mean, they're talking about the tendency for any statistic to move towards the norm over time. In other words, if you flip a coin 100 times, the statistics say it's gonna come up heads just about 50 times. Now let's say you start flipping that coin and it comes up heads 15 out of the first 20 tries. Will that trend continue? No. "Regression to the mean" simply says that by the time you get to 100 coin flips, the ratio will settle down and you'll end up with something much closer to 1:1.

Football Outsiders makes some valid points when they say that the 49ers are unlikely to maintain their anomalous advantage in turnover differential, low rate of injury, and victory in close contests. True enough, but that doesn't mean they'll necessarily lose more games. With better production in the red zone, improvement on third down, more explosive plays, and higher point production, they might win MORE games this year. If statistics could accurately predict what was going to happen next, there would be no need to play the games. Truth is, anything can happen... and probably will.

My personal take is that it's gonna be much harder this year. A tougher schedule, we're the hunted instead of the hunter, and we'll probably have more turnovers and injuries. Balancing that, I see potentially large gains in Offensive production, and a relatively weak division. My prediction is that we'll go 9-7, but we'll be much better equipped to make a run in the playoffs. And my prediction has exactly the same chance as the stats-crunchers at FO.
I certainly agree with this - however, my argument is that there are way too many variables to have this factor into any analytical "mean". In your example of the coin, we know the baseline is 50%. The flip is a fixed action that can only result in heads or tails. I think a closer analogy to what FO is doing is trying to predict a coin flip in a moving car. Here you introduce factors that may or may not influence the action of the flip and thereby affecting the result. The model in which they use to "predict" must be weighted in some way as to predict certain teams being more likely to win more games and other less likely. I understand what they're trying to argue, but assess that "weight" of a team is in itself completely subjective. Why is Green Bay better than let's say Patriots or the Giants, or even the Saints? How can that be expressed as objective data for creating a model in which we can reasonably predict performance? I don't think it's possible personally.

Yes, teams that do exceptionally well tend to perform back down to a "mean" - but again, I ask the question - what is baseline for that team? And where do you begin the evaluation? With a coin it's much easier because the outcome is fixed. The coin is never "weighted" or in abstract, influenced by the process of being flipped (who's flipping, environmental factors, etc).

IMHO, I think this article is written more for press than revealing some actual, hardcore analysis.

Absolutely agree. There are way too many variables and random inputs in football for any statistical analysis to predict future performance in the NFL. It's a chaotic system in the literal sense. But even though statistical analysis doesn't tell the whole story, it can be a useful tool in cutting through some of the things that cloud the reasoning of football fans (ego, emotion, wishful thinking, unfounded assumptions) and provide useful insight. For this reason, I find their work interesting.

One thing we know for certain: Human beings aren't very good at predicting the future. Doesn't matter whether you're using computers to run statistical simulations, or reading goat entrails; no one can see the future. Some methods have a higher rate of success and some scenarios lend themselves to prediction more than others but, at the end of the day, Football Outsiders' track record is about the same as the various NFL experts, pundits, bloggers, and odds makers in Las Vegas. They win some, they lose some.

BTW, I think the mean that they use for turnover differential is league-wide and goes back at least 20 years. So it's a pretty good sample size. It's certainly not as simplistic as a coin toss (and I didn't mean to imply that it was) but the larger your data set, the more accurate your statistical models become. But again, even the most accurate statistical model is not a crystal ball.
Originally posted by drunk49er:
they manage to nail some unpredictable ones, and they manage to drop some real stinkers.

don't really care

I hear that notion, I don't really care either, if I was the Niners I would use that as inspiration.....Make me play to win everything.
Hard to imagine them losing any less than 10 games. Honestly if they went 9-7 there is a big chance they might not even win the division. They will likely need at least 10 wins if not more to make it to the playoffs. It would probably be the biggest let down ever if they missed the playoffs next year. That would be 100x worse than the loss to the giants in the championship.

I actually hate predicting anything because in the end we are all wrong. Who would have said the Giants of all teams would win the superbowl? Or that the packers would go one and done in post season. Or the colts couldn't even win 3 games without Peyton. Or the Broncos making the playoffs. The Lions being relevant? Or of course the 49ers going 13-3 and making it to the NFCC game.

Point is that it doesn't matter because the only thing for sure is that unexpected s**t is going to happen next season.