Prep Time 2012

cyptomcat said:
Boom wasn't for you, since you already agreed on most of the main premise. There was still disagreement on that by others even on that part.

Why do teams coming off bye weeks do much worse than when they are coming off a bye week against Tech? I think we also agreed on this before, that extra time against the Tech offense is worth more than extra time against a 'typical' offense.

To compare:


http://www.teamspeedkills.com/2010/4/15/1424019/how-much-do-bye-weeks-matter
Click to expand...

Ha, I was just being a dick. I think we are in dead agreement on the premise. We need to come up with a better methodology to actually figure out the "why."
 
nvastinger said:
Ok, now show the median and average that Tech/opponent was favored by in each circumstance. With these small sample sizes, it's entirely possible that Tech was favored by much larger amounts in the 7-days category than in the extra days category. I'd expect a better record in cases where Tech was favored heavily than by half a point.

Then do the same thing for a group of teams with similar schedules, say the other members of the Coastal Division. Then we might have something meaningful to talk about and stubbornly stick to our prior beliefs no matter the result.
Click to expand...
How am I stubbornly sticking to something when I am doing extra work and sharing it?

I think it's time for you all to do some of that if you are sincere about the discussion.

Spread/differential analysis will be the last thing I will do.
 
If only LSU had had a 5th or 6th week to prep for Bama! Oh wait!

Of far greater importance than an extra week of preparation, is having already played a team during the season (per hard to beat a team twice axiom---i.e. LSU/Bama). How did either and/or both workout for Clemson vs us?

I think the "pro advantage" folks are just trying too hard to force fit a flawed theory. If someone is just determined to make a case via stats, go to the 70s/80s and review OU/NU/UT---warning, you will be disappointed.
 
So with this myth debunked, does that automatically debunk it's ugly twin sister:

CPJ's offense is tougher to prepare for in 3 days than other, more common offenses?
How could it not?
 
cyptomcat said:
yeah that's why many have talked about this before no stats, and then several different sources including AJC and ESPN cited it with stats since.

Opponent strength is indeed accounted for in some of the stats. The numbers are all there, I have made it available for all, why don't you do some number crunching and show us how we are wrong? Talk is cheap.
Click to expand...


First, I'm arguing that prep-time cannot be treated as an isolated statistic, one way or the other, in the way that you are trying to do. I gave you evidence--not just talk--that GT's overtime loss in the Sun Bowl was likely more attributable to key injuries than Utah's prep-time. In fact, I do not think that any reasonable person--when confronted with the the 3rd quarter score and the nature of injuries on the OL and D which occurred during the game--would attribute that win to extra prep time.

Secondly, I want to commend you for your honesty since by referencing the AJC and ESPN articles, at least you're admitting that the issue you were concerned about was prep-time for CPJ's offense. That's where the 2011 VPI data--not just talk--comes in, showing that our D lost us that game not our O.

Finally, I agree with Legal that your whole methodology is flawed because you don't actually deal with game-statistics which might be attibutable to prep, somehow normalized for opponent.
 
ok, how we did compared to the vegas line:
5-9 (.36) with extra prep for the opponent
18-11 (.62) with no extra prep for the opponent

and that takes into account everything we can possibly think of since it's by Vegas. Even the extra week, but I bet their extra week adjustment should be a lot different in our case compared to other teams' cases.

With extra time for opponent, we do on average 7.66 points worse than the Vegas spread.
With no extra time for opponent we do on average 3.39 points better than the Vegas spread.

I rest my case.

Link for data:
https://docs.google.com/spreadsheet/pub?key=0Ar8cqnkh36RRdENBQk1fRTYyRDFIVTdOa1Q0cXNrbGc&output=html
 
Last edited:
cyptomcat said:
yeah 0.05 is usually used, and personally I would even look for 0.01. If we have couple more seasons of the same kind of results, the p value would be about 0.05. (Just ran the numbers).

Regardless, when you have to form an opinion, p value is one stat you bring into the discussion. Remember this discussion didn't start because of some random stats. This discussion started years ago when A LOT of people thought extra time against triple option would be beneficial. It was just a 'myth' then. Since then the hypothesis has materialized in win loss stats and offensive stats. Starting with that hypothesis, my numbers do say there is only ~10% probability the numbers happened by bad testing observations.
Click to expand...

What you say is correct, but I think you are giving people the false impression that there is only a 10% chance that preparation time does not matter. That is not what the numbers say. They say that given the null hypothesis (preparation time doesn't matter), you would expect to see data like you did 10% of the time. 10% is not the probability the null hypothesis is true. 90% is not the probability that the alternative hypothesis (preparation time does matter) is true. A p-value of 0.10 just means, with regard to the hypothesis, that nothing statistically significant is shown.

From Wikipedia:
There are several common misunderstandings about p-values.[5][6]

  1. The p-value is not the probability that the null hypothesis is true.
    In fact, frequentist statistics does not, and cannot, attach probabilities to hypotheses. Comparison of Bayesian and classical approaches shows that a p-value can be very close to zero while the posterior probability of the null is very close to unity (if there is no alternative hypothesis with a large enough a priori probability and which would explain the results more easily). This is the Jeffreys–Lindley paradox.
  2. The p-value is not the probability that a finding is "merely a fluke."
    As the calculation of a p-value is based on the assumption that a finding is the product of chance alone, it patently cannot also be used to gauge the probability of that assumption being true. This is different from the real meaning which is that the p-value is the chance of obtaining such results if the null hypothesis is true.
  3. 1 − (p-value) is not the probability of the alternative hypothesis being true (see (1)).
    In fact, frequentist statistics does not, and cannot, attach probabilities to hypotheses. Comparison of Bayesian and classical approaches shows that a p-value can be very close to zero while the posterior probability of the null is very close to unity (if there is no alternative hypothesis with a large enough a priori probability and which would explain the results more easily). This is the Jeffreys–Lindley paradox.
Click to expand...
I asked someone who knows much more about statistics than me, and they disagreed that a p-value of 0.10 was even "interesting" as I said before. They said a p of 0.10 only means it not unlikely your result is just due to sampling error. They also questioned whether you need to adjust for multiple hypothesis testing since you calculated two different statistics.

Unfortunately, the only thing we can conclude is that it remains to be seen whether preparation time matters.
 
AE 87 said:
First, I'm arguing that prep-time cannot be treated as an isolated statistic, one way or the other, in the way that you are trying to do. I gave you evidence--not just talk--that GT's overtime loss in the Sun Bowl was likely more attributable to key injuries than Utah's prep-time. In fact, I do not think that any reasonable person--when confronted with the the 3rd quarter score and the nature of injuries on the OL and D which occurred during the game--would attribute that win to extra prep time.

Secondly, I want to commend you for your honesty since by referencing the AJC and ESPN articles, at least you're admitting that the issue you were concerned about was prep-time for CPJ's offense. That's where the 2011 VPI data--not just talk--comes in, showing that our D lost us that game not our O.

Finally, I agree with Legal that your whole methodology is flawed because you don't actually deal with game-statistics which might be attibutable to prep, somehow normalized for opponent.
Click to expand...
When you look at specific games, there are many reasons for how it ends. When you look at 50 games together, that's when you see the trends emerge.

I already did analysis on our offense and posted it in this thread, you must have skipped through that part. Again it points out extra time is an advantage and it does use game statistics and normalizes for the opponent.

Last but not least, game stats are accounted for in the Vegas spread (what we can imagine and probably beyond that since it's big money business). That's the last analysis I posted.
 
gtzulu said:
What you say is correct, but I think you are giving people the false impression that there is only a 10% chance that preparation time does not matter. That is not what the numbers say. They say that given the null hypothesis (preparation time doesn't matter), you would expect to see data like you did 10% of the time. 10% is not the probability the null hypothesis is true. 90% is not the probability that the alternative hypothesis (preparation time does matter) is true. A p-value of 0.10 just means, with regard to the hypothesis, that nothing statistically significant is shown.

I asked someone who knows much more about statistics than me, and they disagreed that a p-value of 0.10 was even "interesting" as I said before. They said a p of 0.10 only means it not unlikely your result is just due to sampling error. They also questioned whether you need to adjust for multiple hypothesis testing since you calculated two different statistics.

Unfortunately, the only thing we can conclude is that it remains to be seen whether preparation time matters.
Click to expand...

The p-value for the spread line differentials between the two sets is

DRUM ROLL

0.0101 using the t-test
0.0162 using the ranksum test

Statistically Significant! BOOYAH!

*edited to fix numbers*

Data is already linked, feel free to give it to your friend. :biggthumpup:

(I would also like to hear the multiple hypothesis involved in the previous case of the offensive ratios or in this current case of spread line differentials. Private message would be best for that.)
 
cyptomcat said:
ok, how we did compared to the vegas line:
5-9 (.36) with extra prep for the opponent
18-11 (.62) with no extra prep for the opponent

and that takes into account everything we can possibly think of since it's by Vegas. Even the extra week, but I bet their extra week adjustment should be a lot different in our case compared to other teams' cases.

With extra time for opponent, we do on average 7.66 points worse than the Vegas spread.
With no extra time for opponent we do on average 3.39 points better than the Vegas spread.

I rest my case.

Link for data:
https://docs.google.com/spreadsheet/ccc?key=0Ar8cqnkh36RRdENBQk1fRTYyRDFIVTdOa1Q0cXNrbGc#gid=3
Click to expand...

1) I don't see actual points and Vegas spread points in the data link. What am I missing / misunderstanding?

2) The Vegas line can't account for injuries during the game which I expect is a better explanation of what happened, at least, in the Sun Bowl.

3) Once you bring in another statistic like the Vegas Spread, you need to normalize it for other teams.
 
cyptomcat said:
The p-value for the spread line differentials between the two sets is

DRUM ROLL

2.6852e-04 using the t-test
5.0640e-04 using the ranksum test

Statistically Significant! BOOYAH!

(Even when I take out the outliers, it is still statistically significant, p value less than .01)

Data is already linked, feel free to give it to your friend. :biggthumpup:

(I would also like to hear the multiple hypothesis involved in the previous case of the offensive ratios or in this current case of spread line differentials. Private message would be best for that.)
Click to expand...

Lies, ---- lies, and cyptomcat posts.

Teams played after extra preparation days are typically better teams. That's why it gets scheduled that way. So you're really just saying that we do worse versus the spread when we play better teams. There are lots of explanations for doing worse against the spread when you face better teams besides more prep time.
 
AE 87 said:
1) I don't see actual points and Vegas spread points in the data link. What am I missing / misunderstanding?

2) The Vegas line can't account for injuries during the game which I expect is a better explanation of what happened, at least, in the Sun Bowl.

3) Once you bring in another statistic like the Vegas Spread, you need to normalize it for other teams.
Click to expand...

I think my link might have been broken, but in any case this one should work: (use the tabs on top for different spreadsheets)
https://docs.google.com/spreadsheet/pub?key=0Ar8cqnkh36RRdENBQk1fRTYyRDFIVTdOa1Q0cXNrbGc&output=html

2) The issue here is about 50 games, not a single game.

3) Vegas does normalize for other teams on average, that's the whole point.
 
gtzulu said:
Lies, ---- lies, and cyptomcat posts.

Teams played after extra preparation days are typically better teams. That's why it gets scheduled that way. So you're really just saying that we do worse versus the spread when we play better teams. There are lots of explanations for doing worse against the spread when you face better teams besides more prep time.
Click to expand...
Are you saying vegas does not account for these factors?

Then you should be making a lot of money this upcoming season!

Btw, I totally expect vegas to fix for the GT extra preparation factor sooner or later. They are not dumb.
 
cyptomcat said:
ok, how we did compared to the vegas line:
5-9 (.36) with extra prep for the opponent
18-11 (.62) with no extra prep for the opponent

and that takes into account everything we can possibly think of since it's by Vegas. Even the extra week, but I bet their extra week adjustment should be a lot different in our case compared to other teams' cases.

With extra time for opponent, we do on average 7.66 points worse than the Vegas spread.
With no extra time for opponent we do on average 3.39 points better than the Vegas spread.

I rest my case.

Link for data:
https://docs.google.com/spreadsheet/ccc?key=0Ar8cqnkh36RRdENBQk1fRTYyRDFIVTdOa1Q0cXNrbGc#gid=3
Click to expand...

Nice find.
 
cyptomcat said:
Are you saying vegas does not account for these factors?

Then you should be making a lot of money this upcoming season!

Btw, I totally expect vegas to fix for the GT extra preparation factor sooner or later. They are not dumb.
Click to expand...

Your result is indeed remarkable from a betting perspective. p=0.2652e-04 is much lower than the usual significance cutoff. You'd think if Vegas was going to catch it they would have already caught it, and p wouldn't be so low. If preparation time isn't an input to their statistical models, the oddsmakers might totally miss the preparation time factor... :hsugh:

I think you're making the point though that if there were differences in GT's results against the spread just due to difference in opponent strength, Vegas would have already noticed. That is a good point, but I don't know, maybe we overestimate the oddsmakers' abilities. Are they really looking at GT specifically and finding the factors particular to GT that predict outcomes?
 
The Dawg fans are right: we really are a bunch of nerds.
 
You must log in or register to reply here.
Back
Top