Everyones favorite statman doesn't think much of SU

[HRE Otto IV]

I love Connelly’s efforts. I think he does a great job writing up previews and his formula is good at evaluating past performance. As a predictor, not so much. College football has way too many variables, many of which can never be quantified.

How do you factor in coaching? Especially with all the coaching changes? Some coaches have a great system which makes the players outperform their talent. Other coaches are great at developing players making them better than when they came in. How do you assign value to that?

How do you factor in the roster turnover which happens every year for every team? Recruiting rankings is a decent tool but highly flawed. In addition IMO the stars mean more for predicting a Frosh or Sophomore than a Junior or Senior, which past performance IMO is a better predictor.

I don’t think Connelly is the issue. It is the people who put too much value on the formulaic stats as being black and white. Pro sports is a bit different as there are less variables. So these advanced stats have more value. But even then the formulas are created by humans using a chosen set of variables. If the advanced stats are the be all, then why do they all use different formulas? Shouldn’t Sagarin and Connelly be using the same thing in that case?

 

[sufandu]

I love Connelly’s efforts. I think he does a great job writing up previews and his formula is good at evaluating past performance. As a predictor, not so much. College football has way too many variables, many of which can never be quantified.

How do you factor in coaching? Especially with all the coaching changes? Some coaches have a great system which makes the players outperform their talent. Other coaches are great at developing players making them better than when they came in. How do you assign value to that?

How do you factor in the roster turnover which happens every year for every team? Recruiting rankings is a decent tool but highly flawed. In addition IMO the stars mean more for predicting a Frosh or Sophomore than a Junior or Senior, which past performance IMO is a better predictor.

I don’t think Connelly is the issue. It is the people who put too much value on the formulaic stats as being black and white. Pro sports is a bit different as there are less variables. So these advanced stats have more value. But even then the formulas are created by humans using a chosen set of variables. If the advanced stats are the be all, then why do they all use different formulas? Shouldn’t Sagarin and Connelly be using the same thing in that case?

That's the problem. In this era of fantasy sports and advanced statistics we forget that athletes and coaches are people with immeasurable charactistics. Everything may regress to the mean eventually, but that can't predict any individual game or even a season.

 

[OrangeDW]

Turnovers are certainly flukey, but you can do certain things to increase your odds of being a turnover forcing team.

1) Stop the run = More 2nd/3rd and long passing attempts.

2) Rush the passer = Speed up the QB’s decision making.

3) Score = Put the pressure on the opponent to come from behind

4) Ballhawk ( Safeties with a knack for the ball and a defense that is taught to strip-tackle)

Not sure how our run defense is looking, but I’m pretty confident in the other three points.

I could be wrong, but didn’t Dino’s teams at previous stops typically do pretty well at creating turnovers due to at least some of those factors? I think I remember hearing because they put up so many points and put pressure on the other team to keep up, that they were good at forcing turnovers.

 

[HOFCeluck]

Forcing tos helps but there’s still a lot of randomness to turnover margin.

 

[HRE Otto IV]

That's the problem. In this era of fantasy sports and advanced statistics we forget that athletes and coaches are people with immeasurable charactistics. Everything may regress to the mean eventually, but that can't predict any individual game or even a season.

It makes more sense to look at stats for a sport that plays 162 games than a sport that plays 11-12 (FCS games are even harder to explain).

It is a lot easier projecting a 30 year old hitter with 5 seasons of 500+ at bats or a 30 year old pitcher with 5 years of 200+ innings than it is projecting a 20 year old QB with 87 passes thrown.

 

[sufandu]

It makes more sense to look at stats for a sport that plays 162 games than a sport that plays 11-12 (FCS games are even harder to explain).

It is a lot easier projecting a 30 year old hitter with 5 seasons of 500+ at bats or a 30 year old pitcher with 5 years of 200+ innings than it is projecting a 20 year old QB with 87 passes thrown.

Right. And from a team perspective, there is less turnover from year to year in pro sports. College football loses a large percentage of its best players every year. It's one of the things that makes it unpredictable and fun.

 

[TheCusian]

The idea of a stat nerd making projections is funny. Stats are absolutely meaningless until you have an adequate sample. There is literally no sample yet because no games have been played with the current rosters and there is too much player turnover in college sports. This is stupid.

You can get info from the previous year - and then factor in who has left and the quality of recruits/transfers to get to a decent idea of what will happen next. It’s not perfect and he’s the first to lay out the issues with it.

The hilarious thing is people trumpeting season prognosticators that do none of that and just kind of guess.

 

[sufandu]

You can get info from the previous year - and then factor in who has left and the quality of recruits/transfers to get to a decent idea of what will happen next. It’s not perfect and he’s the first to lay out the issues with it.

The hilarious thing is people trumpeting season prognosticators that do none of that and just kind of guess.

I prefer someone that knows the game and can look beyond the numbers. Nobody is perfect, but I won't buy advanced stats as a predictor before a game has been played and so many changes have occurred. Of course, based on past performance Clemson is going to be elite. It doesn't take a formula or a keen eye to notice that.

I'll pay more attention to that stuff toward the end of the season when comparing two teams that don't play head to head and don't have common opponents. That's where it's most useful. Even then it has to be taken with a grain of salt. There is no way a team with a final SRS of 50 (SU) should take an SRS of 69 (Clemson) to the limit or that a 69 should take a 72.8 (Alabama) behind the woodshed.

 

[SUskibum]

As a general rule advanced analytics/predictive modeling (in any field) are really good at forecasting scenarios that are prone to extrapolation or mean reversion but are really pretty bad at predicting trend inflection points (positive or negative)

Also, the more data available, the better the potential for the model accuracy, almost always.

Here are the typical scenarios that are in the wheelhouse for modeling/forecasting
1. Consistently Good team staying good
2. Consistently Good team had a poor season but likely to bounce back up
3. Consistently Bad team staying bad
4. Consistently Bad team had an unusually good year but likely to come back down

Here are the typical scenarios that are blind spots for statistical models
1. Consistently Good team about to start a stretch of a few (or more) bad seasons
2. Consistently Bad team about to start a stretch of a few (or more) good seasons

 

[TheCusian]

Forcing tos helps but there’s still a lot of randomness to turnover margin.

Yep. But if you’re trying to get more - dominant pass rushers and zone defense playing with a lead helps.

 

[TheCusian]

Stats in CFB are always handicapped by small sample sizes, roster turnover, and young athletes - but you’d rather have your hands on some than not.

 

[KaiserUEO]

FSU to a huge jump???? lololololololol *deep breath* BAHAHAHAHAHAHAHA
That data alone makes me not want to read anything his magic formula comes up with. Yikes

Florida St is NEVER far off.

 

[Chip]

Florida St is NEVER far off.

I agree, but I can see why people think this moreso than any other down year.

Players and fanbase gave up on the coach. OL was a disaster with no answers, young or experienced. It's probably the worst collective performance by a power team OL unit of all time (I said power, not power 5, so I'm not including the GRob years). That position needs such an overhaul. Losing Francis at QB doesn't help.

I'd be surprised if they made too much noise this year, but yeah, hard to imagine they'll be as bad as last year. By 2020, they'll be back on the path. It's kind of like the Miami Hurricanes of 1997 and 1998.

 

[OttoinGrotto]

As a general rule advanced analytics/predictive modeling (in any field) are really good at forecasting scenarios that are prone to extrapolation or mean reversion but are really pretty bad at predicting trend inflection points (positive or negative)

Also, the more data available, the better the potential for the model accuracy, almost always.

Here are the typical scenarios that are in the wheelhouse for modeling/forecasting
1. Consistently Good team staying good
2. Consistently Good team had a poor season but likely to bounce back up
3. Consistently Bad team staying bad
4. Consistently Bad team had an unusually good year but likely to come back down

Here are the typical scenarios that are blind spots for statistical models
1. Consistently Good team about to start a stretch of a few (or more) bad seasons
2. Consistently Bad team about to start a stretch of a few (or more) good seasons

Wow. Well done!

 

[PAcuse]

Florida St is NEVER far off.

Sure but they are not making that big of a jump. New OC, line is still a mess, lost to much on defense in the draft...

 

[Crusty]

The idea of a stat nerd making projections is funny. Stats are absolutely meaningless until you have an adequate sample. There is literally no sample yet because no games have been played with the current rosters and there is too much player turnover in college sports. This is stupid.

Absolutely correct.

 

[sutomcat]

I love Connelly’s efforts. I think he does a great job writing up previews and his formula is good at evaluating past performance. As a predictor, not so much. College football has way too many variables, many of which can never be quantified.

How do you factor in coaching? Especially with all the coaching changes? Some coaches have a great system which makes the players outperform their talent. Other coaches are great at developing players making them better than when they came in. How do you assign value to that?

How do you factor in the roster turnover which happens every year for every team? Recruiting rankings is a decent tool but highly flawed. In addition IMO the stars mean more for predicting a Frosh or Sophomore than a Junior or Senior, which past performance IMO is a better predictor.

I don’t think Connelly is the issue. It is the people who put too much value on the formulaic stats as being black and white. Pro sports is a bit different as there are less variables. So these advanced stats have more value. But even then the formulas are created by humans using a chosen set of variables. If the advanced stats are the be all, then why do they all use different formulas? Shouldn’t Sagarin and Connelly be using the same thing in that case?

Connelly’s model is fine. It is not unlike Phil Steele’s and I woouldn’t be surprised if one borrowed some concepts from the other.

No question the turnover margin was unusually high. But to me, you need to look at the causes for it. Was it mere chance? I don’t think so.

Most of the margin was based on the high number of turnovers the opposition had against Syracuse last season. I think watching the games, a big cause was the very strong pass rush the Syracuse defense was able to muster. back. Strong QB pressure drives turnovers, be it from QBs getting the ball knocked out as they get hit or QBs throwing under heavy pressure, making dangerous, interception prone throws they would never try under ordinary circumstances.

To have a strong pass rush, you ideally need very good DEs, you need a secondary that is good enough to make the QB wait a bit for a receiver to get open and you need a strong offense, that puts points on the board and puts pressure on the opposing offense to score, preferably quickly.

We have 3 DEs returning who consistently had exceptional numbers for sacks and QB pressures.

The entire secondary is back.

And the SU offense looks to be good, maybe very good and possibly great. I think it is reasonable to expect performance similar to 2018. The RBs and WRs are going to be better, the OL should be around the same level and the QB performance should also be around the same level. We will definitely see the rushing yards and rushing TDs go down at B but completion percentage, yards per throw and probably interception completion should be better.

I think the turnovers on offense will be a little lower. The turnovers the defense produces will probably be a little lower too, but I expect the margin to be in SU’s favor. The 2 areas of risk are how Devito does making decisions and throwing the ball, and how well the defense does stopping the run. I don’t think we are going to see a lot of teams hurting us passing the ball...most are going to be afraid to throw it, especially on slow developing plays down the field.

 

[All4SU]

This is a big one. I would expect a regression to the mean in 2019 and I’m not sure how that affects our win total.

Some of turnovers are luck. But most come from pressure and tight coverage. I suspect our pressure will be as good, and our coverage should be better. That suggests to me that unless we are the one turning the ball over, we should still come out on the positive side.

 

[syrBossHogg]

 

[cusefan88]

It's a really good and fair preview. Everyone should check it out.

 

[Fat tire 13]

Interesting observation if you look at this and the "out of the box" thread. We had +13 TO margin this past year, but was -12 the year before that. Then longer term it was much closer to even each year. I'm not sure what to conclude except it has been much more volatile under Dino. The bullish conclusion is as he has brought in more of his recruits, it leads to a better margin on a consistent basis

 

[TheCusian]

Very fair.

That projected drop is driven primarily by two things: the loss of Dungey-to-Custis and the fact that S&P+ doesn’t look fondly at single-year leaps. A team sort of has to prove itself twice before S&P+ trusts it.

If you're down on his projection - here's your hope. Do you think Dungey-to-Custis can be replicated by DeVito-to-Someone-Prob-Harris? And do you think last year was blip created by TO's and a weak ACC or by actual improvement? (My take is a bit of both)

 

[HRE Otto IV]

Very fair.



If you're down on his projection - here's your hope. Do you think Dungey-to-Custis can be replicated by DeVito-to-Someone-Prob-Harris? And do you think last year was blip created by TO's and a weak ACC or by actual improvement? (My take is a bit of both)

His write up is s great but his projection IMO is s joke. Anything worse than 8-4 is a disappointment and he has us at 6.7 wins.

 

[PhatOrange]

The schedule sets up nicely but there is some valid cause for concern. New QB, new tackles and a 3 deep defensive line unit that will have to stay healthy; otherwise #4 is Shaq Gros and #5 is Harper.

 

[upperdeck]

when you are chasing and playing QBs that you dont want to play bad things happen. when QBS try to do too much bad things happen.. when your D is playin 3-4-5th guy at spots bad things happen