CFB Matrix - predictions

[bshirt]

well, it's easy to dismiss statistics on the internet as flimsy and worthless...but I have to say, this guy really does a bang up job on his. The problem is that you can't statistic the intangibles.

 

As an example, Lavonte David was an intangible. Kickers that don't miss are intangible. QB's that make HUGE improvements 1 year to the next are intangible. A QB on an opposing team throwing 3 interceptions in one game is an intangible. Coaches not making adjustments from one half to the next is an intangible. No amount of stats will EVER account for all of these things.

 

So while he does have a pretty fair predictor of who the stats show should win...we know, just by the NorthWestern game last year...that it doesn't always happen no matter how confident this guy is.

 

 

Hey, I love that site and it's well thought out and his numbers don't lie. He essentially weights recruiting (their last four years) as a huge factor in projecting a team's chances of success. Much more than returning starters. But there are exceptions. Lots of them. KSU, Texas, Florida, Notre Dame, etc. Mostly he credits that to either excellent or very marginal coaching (he's probably right, imo).

 

He does rip apart the Boise St argument quite nicely that I also fully agree with. Strength of schedule "does" matter a lot in achieving a zero or one loss season.

 

I find it's a fun read. Now, if only the guy wasn't a Duck fan.....

[zoogs]

well, it's easy to dismiss statistics on the internet as flimsy and worthless...but I have to say, this guy really does a bang up job on his. The problem is that you can't statistic the intangibles.

 

It's not that it's on the internet. It's that his methods appear to have no rigor. It's far too generous to call this a statistical analysis.

 

All he is doing from the parts of it I've read are using small data sample sizes to take average, calculate percentages and then draw conclusions. Never mind that correlation doesn't imply causation, he doesn't once appear to even have a standard deviation in there. Not to mention, he has no measures of correlation.

 

You're right that no amount of stats can account for everything. But if he's doing something more than amateur grade school brute forcing, I would really like that pointed out and explained to me. Otherwise it's just a lot of rambly hand-waving (you could say BS). If that is all he's doing, this is a joke.

 

The numbers he's "crunched out" here seem like interesting trivia factoids for a football column. And yet he bills it as much more. For example, one question I have -- bshirt, what does he even mean by "weighting"? How does he "weight"? How does he use "excellent or marginal" coaching as a metric?

[VA Husker Fan]

I haven't looked at this much, but I wonder if he (or anyone) has tried to apply the same criteria to FCS and other divisions. I suspect it will fall apart because of drawing conclusions from a small BCS sample size, but that's just my guess.

[zoogs]

Well, FCS is a different world entirely pretty much, so different correlations in effect would only be natural.

 

An example of the sample size issue I have is, going back to that article about how important returning starters are

 

The 18 teams that have gone back to back 10+ win regular seasons in last 4 years

 

He probably should have stopped right there before he eventually got to

 

So what does all this mean for teams coming off a 10+ win regular season. Well, I see it as follows:

 

(#3 on his list for this second quote was "To stay at a high level the return of the QB and kickers is very important")

[NUance]

Picking 80% is no big deal. I could probably do that most years. Picking 80% AGS would be impressive, near miraculous. Heck, even 70% vs the spread would be sort of awesome.

[bshirt]

well, it's easy to dismiss statistics on the internet as flimsy and worthless...but I have to say, this guy really does a bang up job on his. The problem is that you can't statistic the intangibles.

 

It's not that it's on the internet. It's that his methods appear to have no rigor. It's far too generous to call this a statistical analysis.

 

All he is doing from the parts of it I've read are using small data sample sizes to take average, calculate percentages and then draw conclusions. Never mind that correlation doesn't imply causation, he doesn't once appear to even have a standard deviation in there. Not to mention, he has no measures of correlation.

 

You're right that no amount of stats can account for everything. But if he's doing something more than amateur grade school brute forcing, I would really like that pointed out and explained to me. Otherwise it's just a lot of rambly hand-waving (you could say BS). If that is all he's doing, this is a joke.

 

The numbers he's "crunched out" here seem like interesting trivia factoids for a football column. And yet he bills it as much more. For example, one question I have -- bshirt, what does he even mean by "weighting"? How does he "weight"? How does he use "excellent or marginal" coaching as a metric?

 

 

Help me out here, zoogies.

 

He's using all Div1 teams, their W/L results for the last decade or so, their "average" recruiting ranking (Scout, Rivals, etc) which is he uses for the previous four years. He also factors in "where" the game is played and "coaching" (results relative to recruiting rankings).

 

Why is that such such a small data sample?

[bshirt]

Picking 80% is no big deal. I could probably do that most years. Picking 80% AGS would be impressive, near miraculous. Heck, even 70% vs the spread would be sort of awesome.

 

 

Yes, that's true enough.

 

He claims he was the #1 pre-season game-predicting publication in 2011 and the #1 national publication for picking the order of finish in the SEC, Pac12 & Big10 in 2011. Now if true, maybe that's just a one year fluke. I don't know. But that's fairly impressive I would think.

[True2tRA]

Maybe someone with a stats background who has read more of this stuff can contradict me, but I remain extremely unimpressed with his methods and how carelessly he starts making conclusions.

 

This guy, does not appear to have a stats background. Perhaps I've missed it, but his methodology appears to consist entirely of

 

1) tally up some (small amount) of data

2) use Division. He may also have some subtraction, addition, and multiplication in there.

3) ????

4) ANSWERS!

 

I could be wrong because I didn't really dig deep into everything. Unless I am missing something here, this all appears to be largely worthless. A lot of reading for some flimsy conclusions and meaningless results.

 

well, it's easy to dismiss statistics on the internet as flimsy and worthless...but I have to say, this guy really does a bang up job on his. The problem is that you can't statistic the intangibles.

 

It's not that it's on the internet. It's that his methods appear to have no rigor. It's far too generous to call this a statistical analysis.

 

All he is doing from the parts of it I've read are using small data sample sizes to take average, calculate percentages and then draw conclusions. Never mind that correlation doesn't imply causation, he doesn't once appear to even have a standard deviation in there. Not to mention, he has no measures of correlation.

 

You're right that no amount of stats can account for everything. But if he's doing something more than amateur grade school brute forcing, I would really like that pointed out and explained to me. Otherwise it's just a lot of rambly hand-waving (you could say BS). If that is all he's doing, this is a joke.

 

The numbers he's "crunched out" here seem like interesting trivia factoids for a football column. And yet he bills it as much more. For example, one question I have -- bshirt, what does he even mean by "weighting"? How does he "weight"? How does he use "excellent or marginal" coaching as a metric?

[True2tRA]

Picking 80% is no big deal. I could probably do that most years. Picking 80% AGS would be impressive, near miraculous. Heck, even 70% vs the spread would be sort of awesome.

 

 

Yes, that's true enough.

 

He claims he was the #1 pre-season game-predicting publication in 2011 and the #1 national publication for picking the order of finish in the SEC, Pac12 & Big10 in 2011. Now if true, maybe that's just a one year fluke. I don't know. But that's fairly impressive I would think.

The best suggestion for you I can make is to sell your house, your cars, any gold or silver you own, empty your bank accounts, and go ahead and cash out that 401k, hell, even go get the kiddies piggy banks and invest it all in your trip to Vegas. Bet on everything this guy predicts, bet it all. Surely by the end of the year, you will know if it was a one year fluke, or not.

[bshirt]

Picking 80% is no big deal. I could probably do that most years. Picking 80% AGS would be impressive, near miraculous. Heck, even 70% vs the spread would be sort of awesome.

 

 

Yes, that's true enough.

 

He claims he was the #1 pre-season game-predicting publication in 2011 and the #1 national publication for picking the order of finish in the SEC, Pac12 & Big10 in 2011. Now if true, maybe that's just a one year fluke. I don't know. But that's fairly impressive I would think.

The best suggestion for you I can make is to sell your house, your cars, any gold or silver you own, empty your bank accounts, and go ahead and cash out that 401k, hell, even go get the kiddies piggy banks and invest it all in your trip to Vegas. Bet on everything this guy predicts, bet it all. Surely by the end of the year, you will know if it was a one year fluke, or not.

 

 

[Excel]

This places too large an emphasis on recruiting.

 

How a team recruits this year won't impact them much next year, at least not nearly as much as it will in two to three years. Also, simply having a great recruiting class in no way predicts a great performance by the team...say hi Notre Dame, Florida State and Clemson.