goemaw.com
TITLETOWN - A Decade Long Celebration Of The Greatest Achievement In College Athletics History => Kansas State Football => Topic started by: CHONGS on October 15, 2012, 10:52:50 PM
-
Game-by-game stats of offensive efficiency and defensive efficiency. Also included are the opponent's offensive efficiency and defensive efficiency to help you weight the game-by-game stats against.
*now hopefully fixed! :crossed:*
-
(https://goemaw.com/forum/proxy.php?request=http%3A%2F%2Fi74.photobucket.com%2Falbums%2Fi247%2FKC10Chief%2FFunny%2520Pics%2FVanDammeCantSeeRedX.gif&hash=ac97d281089655a7bd06df21f98fa9761ba26e24)
Sorry chings I can see your data.
-
(https://goemaw.com/forum/proxy.php?request=http%3A%2F%2Fi74.photobucket.com%2Falbums%2Fi247%2FKC10Chief%2FFunny%2520Pics%2FVanDammeCantSeeRedX.gif&hash=ac97d281089655a7bd06df21f98fa9761ba26e24)
Sorry chings I can see your data.
yeah, some kinks to work out I think. You will need to download and install something called the CDF Player though (its free, but if you don't wann dowload and install it I understand).
-
Yo Chingon, I had to save/open in Mathematica to see this, but it's really fantastic work. Great stuff :katpak:
-
JEAN CLAUDE WHAT HAPPENED TO YOUR FACE
Seriously though, 166MB plugin? Woof!
-
I downloaded the plug-in but I still can't see any data. :bawl:
-
Someone who downloaded the plugin should just copy paste all of the info and then download it locally to our server. TIA
-
Okay, tried to download. Like Chico I'm not getting anything.
(https://goemaw.com/forum/proxy.php?request=http%3A%2F%2Fi.imgur.com%2FCGCrH.png&hash=d20d0022a034af95bcef796e81a9109e2f0e58ff)
-
seems "virus-y". I'm out.
-
same issues as others above.
-
I think I fixed the problem....
-
very cool. wish i understood it more. but that's my problem.
-
Pretty cool stuff. What did you use to create it? Mathmatica?
-
(https://goemaw.com/forum/proxy.php?request=http%3A%2F%2Fi.imgur.com%2FBXh2U.png%3F1&hash=14dbf2997328127b29e49a6c9023004615d369da)
for those of you lgcpiiq folk
-
how do we get above one for offensive efficiency? defense doing work or what?
-
PPP seems like a so so determinant for offensive efficiency.
Essentially, it rewards big plays and not turning the ball over - which is all good. Buuut, it does not factor in time of possession (which we clearly just try to run clock sometimes), starting field position (starting further from the end zone means a worse PPP?), whether we are in front on the scoreboard (obvs running the ball and taking more plays)...and other confounding variables that would seem to greatly influence PPP
I think it provides a decent order of magnitude type of comparison...but thats it.
-
PPP seems like a so so determinant for offensive efficiency.
Essentially, it rewards big plays and not turning the ball over - which is all good. Buuut, it does not factor in time of possession (which we clearly just try to run clock sometimes), starting field position (starting further from the end zone means a worse PPP?), whether we are in front on the scoreboard (obvs running the ball and taking more plays)...and other confounding variables that would seem to greatly influence PPP
I think it provides a decent order of magnitude type of comparison...but thats it.
if you look at
Rush YPP
Pass YPP
Total YPP
PPP
TO %
i feel like you have a pretty good idea of an offense (and defense)
-
PPP seems like a so so determinant for offensive efficiency.
Essentially, it rewards big plays and not turning the ball over - which is all good. Buuut, it does not factor in time of possession (which we clearly just try to run clock sometimes), starting field position (starting further from the end zone means a worse PPP?), whether we are in front on the scoreboard (obvs running the ball and taking more plays)...and other confounding variables that would seem to greatly influence PPP
I think it provides a decent order of magnitude type of comparison...but thats it.
if you look at
Rush YPP
Pass YPP
Total YPP
PPP
TO %
i feel like you have a pretty good idea of an offense (and defense)
Agreed. Is TO% calculated as a percent of total plays or possessions?
-
hmm it doesn't seem to be working for me on my ibm do I need to use my original re-boot disc?
-
PPP seems like a so so determinant for offensive efficiency.
Essentially, it rewards big plays and not turning the ball over - which is all good. Buuut, it does not factor in time of possession (which we clearly just try to run clock sometimes), starting field position (starting further from the end zone means a worse PPP?), whether we are in front on the scoreboard (obvs running the ball and taking more plays)...and other confounding variables that would seem to greatly influence PPP
I think it provides a decent order of magnitude type of comparison...but thats it.
if you look at
Rush YPP
Pass YPP
Total YPP
PPP
TO %
i feel like you have a pretty good idea of an offense (and defense)
Agreed. Is TO% calculated as a percent of total plays or possessions?
in the mocat system (or soon to be renamed SNOWBRAHA) it is TO per play, which is obviously a very small number, but just like the other 4 factors, is divided by the average to get a value that is either higher or lower than 1.0
-
Still wish there was data available on points per possession both offensively and defensively.
-
Still wish there was data available on points per possession both offensively and defensively.
Yeah, I like it better. (in theory)
-
well be my guest. have fun compiling that much data. it's much harder to find out how many possessions a team has than simply adding passing attempts + rushing attempts
-
do we get nerd points if we already had the plugin installed.... :peek:
-
well be my guest. have fun compiling that much data. it's much harder to find out how many possessions a team has than simply adding passing attempts + rushing attempts
I know... it would be nice if the stat services had it available though. Not like it would be hard or unheard of, they do it for basketball.
-
well be my guest. have fun compiling that much data. it's much harder to find out how many possessions a team has than simply adding passing attempts + rushing attempts
I know... it would be nice if the stat services had it available though. Not like it would be hard or unheard of, they do it for basketball.
possessions in basketball are essentially the same as plays in football; teams usually average somewhere in the 60-80 range
-
well be my guest. have fun compiling that much data. it's much harder to find out how many possessions a team has than simply adding passing attempts + rushing attempts
I know... it would be nice if the stat services had it available though. Not like it would be hard or unheard of, they do it for basketball.
possessions in basketball are essentially the same as plays in football; teams usually average somewhere in the 60-80 range
Oh boy, tell me more :dubious:
The point is that the possession is the smallest unit you can break a football game into to get the best idea of an offense's (or defense's) effectiveness. It eliminates the pts/play bias that causes a team that averages 12 plays per possession appear to be half as good as one that averages 6 plays per possession when they score at the same rate.
-
Exactly. I'm sure the charts would still have similar results as points per play, but I still think it'd be a better indicator of a team's overall efficiency.
-
well be my guest. have fun compiling that much data. it's much harder to find out how many possessions a team has than simply adding passing attempts + rushing attempts
I know... it would be nice if the stat services had it available though. Not like it would be hard or unheard of, they do it for basketball.
possessions in basketball are essentially the same as plays in football; teams usually average somewhere in the 60-80 range
Oh boy, tell me more :dubious:
The point is that the possession is the smallest unit you can break a football game into to get the best idea of an offense's (or defense's) effectiveness. It eliminates the pts/play bias that causes a team that averages 12 plays per possession appear to be half as good as one that averages 6 plays per possession when they score at the same rate.
yeah i get what youre saying, except, k-state is absolutely off the charts on PPP, even with our 12 plays per possession
-
JFC guys.
http://www.adjustedstats.com/ratings-stats/cfbteams.php?team=Kansas+St. (http://www.adjustedstats.com/ratings-stats/cfbteams.php?team=Kansas+St.)
-
To summarize for the straight to the bottom crowd, K-State is #1 in the country in both raw and adjusted points per possession.
-
To summarize for the straight to the bottom crowd, K-State is #1 in the country in both raw and adjusted points per possession.
Also #1 in adjusted points per play
-
JFC guys.
http://www.adjustedstats.com/ratings-stats/cfbteams.php?team=Kansas+St. (http://www.adjustedstats.com/ratings-stats/cfbteams.php?team=Kansas+St.)
Win probability for Saturday: .823
Score prediction: 47-33
-
I think there is a slight difference of "philosophy" as it were on how to compare teams. My goal is to use as few parameters as possible. I have no doubts a model with 27 fit parameters will fit the data better than a model with 3. I appreciate the limitations of using only 2+1 statistics (points per play scored, points per play given up, and losses), but in fact that is my goal. There is a strong correlation between the Pythagorean win % calculated with OE and DE and the actual win %.
Another limitation is the availability of statistics in computable form. Possessions are not an easy stat to extract for every team for every game, but I agree I would love to have it.
-
I think there is a slight difference of "philosophy" as it were on how to compare teams. My goal is to use as few parameters as possible. I have no doubts a model with 27 fit parameters will fit the data better than a model with 3. I appreciate the limitations of using only 2+1 statistics (points per play scored, points per play given up, and losses), but in fact that is my goal. There is a strong correlation between the Pythagorean win % calculated with OE and DE and the actual win %.
Another limitation is the availability of statistics in computable form. Possessions are not an easy stat to extract for every team for every game, but I agree I would love to have it.
seems like extracting possessions would be easy. Punts, turnovers, scores, or ends of halves signify ends of possession. Is that in the wolfram download thingy?
-
I think there is a slight difference of "philosophy" as it were on how to compare teams. My goal is to use as few parameters as possible. I have no doubts a model with 27 fit parameters will fit the data better than a model with 3. I appreciate the limitations of using only 2+1 statistics (points per play scored, points per play given up, and losses), but in fact that is my goal. There is a strong correlation between the Pythagorean win % calculated with OE and DE and the actual win %.
Another limitation is the availability of statistics in computable form. Possessions are not an easy stat to extract for every team for every game, but I agree I would love to have it.
seems like extracting possessions would be easy. Punts, turnovers, scores, or ends of halves signify ends of possession. Is that in the wolfram download thingy?
Hmmm I would have to see how close the numbers of punts + to + fg attempted + fourth downs not converted + safeties is to the numnet of offensive possessions, it might just be close enough. I will miss on drives stopped by the half and multiple turnover plays might muck it up but should be statistically small.
-
Is a turnover on downs listed as a turnover? What about of a fumble is returned for a TD?
-
I think there is a slight difference of "philosophy" as it were on how to compare teams. My goal is to use as few parameters as possible. I have no doubts a model with 27 fit parameters will fit the data better than a model with 3. I appreciate the limitations of using only 2+1 statistics (points per play scored, points per play given up, and losses), but in fact that is my goal. There is a strong correlation between the Pythagorean win % calculated with OE and DE and the actual win %.
Another limitation is the availability of statistics in computable form. Possessions are not an easy stat to extract for every team for every game, but I agree I would love to have it.
seems like extracting possessions would be easy. Punts, turnovers, scores, or ends of halves signify ends of possession. Is that in the wolfram download thingy?
Hmmm I would have to see how close the numbers of punts + to + fg attempted + fourth downs not converted + safeties is to the numnet of offensive possessions, it might just be close enough. I will miss on drives stopped by the half and multiple turnover plays might muck it up but should be statistically small.
Is there not a timestamp on plays?
And I think a multiple turnover play should probably count as a new possession - the next play is 1st and 10 (or a score) no matter what.
-
Is a turnover on downs listed as a turnover? What about of a fumble is returned for a TD?
I don't think it officially is, but I could be wrong.
The trouble again will be compiling all of these stats. While in principle this could maybe be extracted from the website SleepFighter mentioned, it would take downloading/scraping at least 700+ webpages by the end of year. I think they would greatly frown upon that.
Right now I extract my stats from NCAA and I only have to scrape 5 or so pages. It it through my own record keeping in fact that I can break it down into game by game stats. There is also the process of building the schedule matrix which can be a pain in the ass.
-
Is a turnover on downs listed as a turnover? What about of a fumble is returned for a TD?
I don't think it officially is, but I could be wrong.
The trouble again will be compiling all of these stats. While in principle this could maybe be extracted from the website SleepFighter mentioned, it would take downloading/scraping at least 700+ webpages by the end of year. I think they would greatly frown upon that.
Right now I extract my stats from NCAA and I only have to scrape 5 or so pages. It it through my own record keeping in fact that I can break it down into game by game stats. There is also the process of building the schedule matrix which can be a pain in the ass.
can you link to the pages you scrape?
-
I think there is a slight difference of "philosophy" as it were on how to compare teams. My goal is to use as few parameters as possible. I have no doubts a model with 27 fit parameters will fit the data better than a model with 3. I appreciate the limitations of using only 2+1 statistics (points per play scored, points per play given up, and losses), but in fact that is my goal. There is a strong correlation between the Pythagorean win % calculated with OE and DE and the actual win %.
Another limitation is the availability of statistics in computable form. Possessions are not an easy stat to extract for every team for every game, but I agree I would love to have it.
seems like extracting possessions would be easy. Punts, turnovers, scores, or ends of halves signify ends of possession. Is that in the wolfram download thingy?
Hmmm I would have to see how close the numbers of punts + to + fg attempted + fourth downs not converted + safeties is to the numnet of offensive possessions, it might just be close enough. I will miss on drives stopped by the half and multiple turnover plays might muck it up but should be statistically small.
Is there not a timestamp on plays?
And I think a multiple turnover play should probably count as a new possession - the next play is 1st and 10 (or a score) no matter what.
The trouble is getting this data and being able to compute with it. It's the trouble almost every company has: all this data and no efficient, plausible way to use it. In my opinion, the work required does not impart a big enough benefit. Averaged over a whole game and across a whole season and against all teams I imagine points per play and points per possession will differ merely by a scaling factor. This scaling factor is irrelevant if you normalize the data in a consistent manner and should not affect the overall correlation with actual winning %.
-
This is an example:
http://statistics.ncaafootball.com/merge/tsnform.aspx?c=ncaa-football&page=cfoot/stat/ncaa-team-totaloff.htm
(note only 120 teams are reported, the newest 4 are neglected, but don't really matter much anyway).