Prep Time 2012
- Thread starter beej67
- Start date
beej67
new around here
- Joined
- Oct 18, 2002
- Messages
- 56,552
Unless they pull a Wake I think it helps us.
(recall when Wake played Navy the week before, they switched defenses completely in the week, and we practiced against all the wrong fronts going into that game)
beej67
new around here
- Joined
- Oct 18, 2002
- Messages
- 56,552
Same goes for Duke.
And I'm not entirely sure having all summer helps VT as much as just having two weeks helps for it. Especially since it's VT and they tend to lose focus in early games.
So overall, I think the Prep Time aspect of this year's schedule isn't terrible.
- Joined
- Oct 18, 2009
- Messages
- 61,115
This has not ever been proven once. There are too many variables to account for. Not enough data. Dead horse, etc. Seriously though, this stat is about as valid as any random superstition.
Some things you'd have to consider and account for to do this properly: how much prep time we have in addition to theirs, how much familiarity (consecutive years of play) we / they have with the offense, injuries, weather, home/away crowd advantage (and travel distance for the away team), the defensive style of the other team, the offensive style of the other team (run-based teams will reduce our yardage by reducing TOP), the rank of the other team, stadium directionality and time of day, and so much more. Not to mention, most are just looking at win/loss or straight yardage production when they should be looking at relative performances (how much yards per play does a team generally give up, and how much more/less did we get, versus how much more/less we usually get against teams with only 7 days prep time).
This is not one of those signals that comes out of the noise if you get enough data points, because you need to filter all of those data points several times first in order to determine if the cause of the dropoff is the prep time, or if the drop off is actually any worse for us than it is for other teams. There is just not enough data to do this, I've tried, with 10 years of CPJ data.
Some things you'd have to consider and account for to do this properly: how much prep time we have in addition to theirs, how much familiarity (consecutive years of play) we / they have with the offense, injuries, weather, home/away crowd advantage (and travel distance for the away team), the defensive style of the other team, the offensive style of the other team (run-based teams will reduce our yardage by reducing TOP), the rank of the other team, stadium directionality and time of day, and so much more. Not to mention, most are just looking at win/loss or straight yardage production when they should be looking at relative performances (how much yards per play does a team generally give up, and how much more/less did we get, versus how much more/less we usually get against teams with only 7 days prep time).
This is not one of those signals that comes out of the noise if you get enough data points, because you need to filter all of those data points several times first in order to determine if the cause of the dropoff is the prep time, or if the drop off is actually any worse for us than it is for other teams. There is just not enough data to do this, I've tried, with 10 years of CPJ data.
- Joined
- Mar 28, 2011
- Messages
- 38,174
CPJ could use it to our advantage.
Also I think VPI won't be as out of focus as we expect, both sides know the importance of this game, but I also think our guys will probably play their best game against VPI.
X's/O's: ADVANTAGE (wash). CPJ vs Grantham part 3. Both will have a good feeling for what the other wants to do before either game is played. If this was going to be the first time CPJ went against Grantham I would say advantage GT, but its not.
Timing/Technique: ADVANTAGE (ugay). They get an extra week to adjust to the speed of the 3-O and our blocking techniques.
Personnel game planning: ADVANTAGE GT. We get to see what personnel strengths/weaknesses ugay has first hand against the 3-O. If they have any players that are "cheating" or weak getting off the cut block, you can bet CPJ will see it on tape and try to exploit them.
beej67
new around here
- Joined
- Oct 18, 2002
- Messages
- 56,552
No, actually it's quite proven for CPJ's tenure at GT, vs GT's level of competition. And last year just added more data points to the trend.
Try it.
Legal Jacket
Dodd-Like
- Joined
- Feb 15, 2011
- Messages
- 4,255
The problem is you are comparing, in a vacuum, whether a defense is more prepared when they get two weeks of practice versus one week of practice, as opposed to whether that matters significantly more in this offense versus other offenses.
It's really a pretty dumb argument that people tend to point to data that shows we do worse against opponents with more prep time. Of course we do, they had more time to prepare for us!
What really matters, imo, is does a defense with two weeks of practice before facing us do significantly better than you would expect that defense to do after two weeks of practice for a generic offense. I haven't seen that data.
Legal Jacket
Dodd-Like
- Joined
- Feb 15, 2011
- Messages
- 4,255
Right? Pretty sure everyone agrees that a team with two weeks practice will be more prepared than a team with one weeks practice.
beej67
new around here
- Joined
- Oct 18, 2002
- Messages
- 56,552
Nah, I did that part too. Feel free to do the analysis again if you like, or to search-fu the one I did in 2011 and 2010 and 2009. It's up on Stingtalk somewhere. I'm not doing it a fourth time, you feel free. I think Cyptomcat did one as well. Just strip out non-BCS quality opponents when you run your numbers.
You'd think, but the difference between the prep time benefit vs GT and the prep time benefit vs other teams is huge. The last time I ran the analysis, Butch Davis actually had a harder time beating non-GT teams after a week break than he had beating non-GT teams on only a week prep.
The GT numbers were profound though. Having an extra week really really helps our opponents, a lot more than it helps us.
Legal Jacket
Dodd-Like
- Joined
- Feb 15, 2011
- Messages
- 4,255
Pretty sure I've seen those and the methodology was flawed. No need to redo the numbers, just post the methodology. Iirc, all you really proved is that we did worse against defenses when they had longer to prepare for us.
I'm not sure what this proves other than my point. I did the math. Discounting cupcakes and GT, Butch was 3-7 against teams when he had longer than a week off. Same math, he was 19-12 against teams who had less than a week preparation.
So he did worse against teams that we more prepared? Again, just seems like you are trying to prove the obvious. What that analysis doesn't prove is how much better those teams did against UNC, with two weeks practice, than what they were expected to do, with two weeks practice, against someone else.
The hard part of the analysis is determining the baseline.
Legal Jacket
Dodd-Like
- Joined
- Feb 15, 2011
- Messages
- 4,255
For example, I found a coupe of the articles cyptomcat cited on espn and ajc.
http://espn.go.com/blog/statsinfo/post/_/id/23461/timing-is-everything-for-georgia-tech
The ESPN article is flawed in that it simply looks at record and rushing TDs for opponents with extra time and without extra time. It ignores that, generally speaking, tougher opponents have had byes when facing us - especially when factoring in bowl games (which it does). It also lumps in all varieties of time off without trying to figure out whether more time = better performance.
http://blogs.ajc.com/georgia-tech-s...-extra-time-matter-when-facing-techs-offense/
The AJC article is a little better, but still flawed. It's better because it takes into account the relative strength of the opponent. But again, it fails to address the critical question which is NOT whether a defense will fare better with 14 days off as opposed to 7 days off (clearly it should be the latter) BUT INSTEAD is whether this defense will fare better with 14 days off against this offense compared to how much better it should be expected to fare against other offenses with the same time off.
Legal Jacket
Dodd-Like
- Joined
- Feb 15, 2011
- Messages
- 4,255
Here's a post of yours I was able to find form 2010. Granted its a little out of date.
http://www.stingtalk.com/forums/showthread.php?t=45579&highlight=prep
First criticism, you didn't factor in whether we were expected to win. You didn't fact in why we won or lost (for example, if the hypothesis is that our offense plays worse, a 65-66 loss doesn't really prove that hypothesis).
For example, one of your conclusions was based on comparing the % change in wins with Clemson, Ohio State, and GT. Ignoring that the sample size was ridiculously small, that analysis also ignores whether each team was supposed to win each game and whether the other team could possibly be expected to win with the benefit of more time off. For example, you concluded that because Ohio State only saw an 18% drop, but GT saw a 34% drop, that GT's offense is easier to prepare for. The problem is that most teams should have expected to lose to Ohio State in 2008 and 2009. So, for example, if the expected gap was 20 points, and you should expect a team to play 10 points better with an extra week off, you shouldn't expect the W/L % to be affected as much. Other way with GT, if we expected to win by 5 points, you would expect the W/L % to change. The main flaw in your "analysis" is that you aren't looking at what teams are expected to do.
In fact, any analysis that looks merely to a W/L and tries to conclude something from it is going to be deeply flawed. The main reason is that a win or a loss is just one statistical event - it's much better to look at things like y/c, points, etc., where you can read more than a yes or a no.
The other problem with wins and losses, especially if you are just talking about how our offense plays, is that you inherently include defense, special teams, dumb mistakes (e.g. an ill-advised throw or a bad fumble that has nothing to do with the other team's preparedness). In essence, you are polluting the information that matters (how the defense played versus the expectation of how the defense would play, both after 14 days off) with a bunch of other factors that yield a binary event. The binary W/L analysis suffers from small sample size analysis too.
http://www.stingtalk.com/forums/showthread.php?t=45579&highlight=prep
First criticism, you didn't factor in whether we were expected to win. You didn't fact in why we won or lost (for example, if the hypothesis is that our offense plays worse, a 65-66 loss doesn't really prove that hypothesis).
For example, one of your conclusions was based on comparing the % change in wins with Clemson, Ohio State, and GT. Ignoring that the sample size was ridiculously small, that analysis also ignores whether each team was supposed to win each game and whether the other team could possibly be expected to win with the benefit of more time off. For example, you concluded that because Ohio State only saw an 18% drop, but GT saw a 34% drop, that GT's offense is easier to prepare for. The problem is that most teams should have expected to lose to Ohio State in 2008 and 2009. So, for example, if the expected gap was 20 points, and you should expect a team to play 10 points better with an extra week off, you shouldn't expect the W/L % to be affected as much. Other way with GT, if we expected to win by 5 points, you would expect the W/L % to change. The main flaw in your "analysis" is that you aren't looking at what teams are expected to do.
In fact, any analysis that looks merely to a W/L and tries to conclude something from it is going to be deeply flawed. The main reason is that a win or a loss is just one statistical event - it's much better to look at things like y/c, points, etc., where you can read more than a yes or a no.
The other problem with wins and losses, especially if you are just talking about how our offense plays, is that you inherently include defense, special teams, dumb mistakes (e.g. an ill-advised throw or a bad fumble that has nothing to do with the other team's preparedness). In essence, you are polluting the information that matters (how the defense played versus the expectation of how the defense would play, both after 14 days off) with a bunch of other factors that yield a binary event. The binary W/L analysis suffers from small sample size analysis too.
Legal Jacket
Dodd-Like
- Joined
- Feb 15, 2011
- Messages
- 4,255
And just so we are clear about what I'm saying, the analysis would have to look at the in a vacuum expected change in certain statistical categories for individual teams versus what the expected statistical change would be against us.
For example:
I think you'd first have to take all teams over the history of stat keeping and measure the difference between how they did playing teams with no break (7 day control) versus a longer break. And not win-losses, I'm talking about various statistical metrics like first downs, pass completion, rating, ypc, rushes over 20 yards, rushes over 10 yards, points scored, etc. So you'd have to chart those deltas for days off v. change and see if there is some sort of regression. Throughout the history of the game, it may be that teams average giving up 20 yards less after 9 days prep and 50 yards less after 14 days prep. If we fare 15 and 40 yards less on average, than obviously we are doing better than the mean.
The real problem comes from taking into account differences in how good teams are. The easiest way to fudge around it is to take all historical data and hope that it evens out over time. The practical problem is that the change in expectation for a good team with an extra week prep may be different than a bad team with an extra week prep. So ideally you could also analyze how the deltas of each teams change. Like, with an extra weeks prep over the history of college football, what is the change to ypc of a team that gives up an average of 3 ypc. Is that different than 4 ypc? You may have to factor in formations too. It gets very complicated in a hurry - much more completed than the short shrift it seems you gave it before.
At the end of the day, you'd want to know what was expected on 7 days off versus 14 (or 9 or whatever) days off. So then you'd compare how GT did against that opponent with 7 days off versus its baseline 7 days off (say, gave up 3.2 ypc on average expected 7 days off). Then you'd do the same for 14 days off opponents and see the difference. If you did that for every team across CPJ's tenure here, I think you'd get a nice regression that shows whether there is actually a statistically meaningful difference between prep time against him versus prep time against others.
For example:
I think you'd first have to take all teams over the history of stat keeping and measure the difference between how they did playing teams with no break (7 day control) versus a longer break. And not win-losses, I'm talking about various statistical metrics like first downs, pass completion, rating, ypc, rushes over 20 yards, rushes over 10 yards, points scored, etc. So you'd have to chart those deltas for days off v. change and see if there is some sort of regression. Throughout the history of the game, it may be that teams average giving up 20 yards less after 9 days prep and 50 yards less after 14 days prep. If we fare 15 and 40 yards less on average, than obviously we are doing better than the mean.
The real problem comes from taking into account differences in how good teams are. The easiest way to fudge around it is to take all historical data and hope that it evens out over time. The practical problem is that the change in expectation for a good team with an extra week prep may be different than a bad team with an extra week prep. So ideally you could also analyze how the deltas of each teams change. Like, with an extra weeks prep over the history of college football, what is the change to ypc of a team that gives up an average of 3 ypc. Is that different than 4 ypc? You may have to factor in formations too. It gets very complicated in a hurry - much more completed than the short shrift it seems you gave it before.
At the end of the day, you'd want to know what was expected on 7 days off versus 14 (or 9 or whatever) days off. So then you'd compare how GT did against that opponent with 7 days off versus its baseline 7 days off (say, gave up 3.2 ypc on average expected 7 days off). Then you'd do the same for 14 days off opponents and see the difference. If you did that for every team across CPJ's tenure here, I think you'd get a nice regression that shows whether there is actually a statistically meaningful difference between prep time against him versus prep time against others.
- Joined
- Oct 18, 2009
- Messages
- 61,115
Without saying as much as legal jacket did, having spent days trying to do this properly, it would highly surprise me if we're actually talking about the same trend at this point if you have found something in just the last four years of data.
To be clear, i'm talking specifically about the myth that extra prep time does more damage to our offense than to other offenses. That's really the subject of the rant. It seems like you're saying that the trend you've found is just that extra prep time makes it more likely for us to lose, which should always be true for every team. If that's the trend you're pointing at, cool, I get it. Otherwise I need a little help to get on the same page.
To be clear, i'm talking specifically about the myth that extra prep time does more damage to our offense than to other offenses. That's really the subject of the rant. It seems like you're saying that the trend you've found is just that extra prep time makes it more likely for us to lose, which should always be true for every team. If that's the trend you're pointing at, cool, I get it. Otherwise I need a little help to get on the same page.
