Post by prossman on Feb 26, 2012 7:28:14 GMT -7
Statistical Analysis Brainiac Wanted (Plus Possible CB Target)
--------------------------------------------------------------------------------
I've been looking over the potential FA CBs and some of their stats and such at PFF. Trying to use this site as much as possible before the subscription is up......should've just went with the annual for slightly more.
Anyway, I'm trying to find a way to rank the players based on their performance.
Some categories I have include some of the standard stuff. Targets, completions, completion % and other standard stuff.
But I can't really figure out a decent way to rank players. Some of the categories I've just ranked numerically based on where they fall. I'll likely work with others to try and provide more emphasis to certain stats.
How should players be ranked? Numerically it really doesn't make sense to order them in some situations.
Is a guy who allows 51% really an entire rank better than a guy who allows 52%? And some of the rankings are this close. Some come down to partial percentages.
The best I have come up with thus far for % is to stratify percentages. 50%, 55%, 60% and assign point values to each increment.
Also, is a rank difference of 1 really a true estimation of importance of 1 INT? I think that's selling it short.
Same with TDs. Does a player who allows 1 TD really only deserve 3-4 numerical points over players who allow 3 TDs? I think that is really underselling the value of preventing TDs.
For these, I thought it would be interesting to assign a point value per interception and TD to properly award and punish high numbers. What about pass defends? Just sort of off the cuff, I gave a TD a value of 7, INT 5 and PD 3 and Penalties 3 although I might up penalties because each penalty by a DB will most likely give a fresh set of downs AT LEAST. Illegal contact, holding and PI are all we're really talking about. Maybe Unnecessary as well but that's still a 1st down. I'll probably try to up the penalties somehow.
You get a point total by subtracting the negative (TDs / Penalties) from the positive (PD and INTs). Obviously, more negative is better and I'll probably switch the two so the positive plays dominate so a more positive number is indicative of good play.
Anyway, just wanting to get your guy's opinion on how players should be ranked. Straight ranks seem to ignore a lot of context and I'd like to see a ranking that doesn't just say a player had the best percentage against him but ignores that he was only thrown at twice per game.
One stat I haven't seen anywhere that I'm trying to include is opponent quality. Specifically speaking, number of Top 10 and Bottom 10 offenses that each corner played against.
But how much should these impact results. I think they're pretty big and I was thinking about taking [([Top 10 - Bottom 10] / 16) + 1] and multiplying that by the total score to try and account for quality of opponents.
This would give a guy who played 4 of each a multiplier of 1 while a guy who had more bottom feeders on the schedule a multipler < 1 and a guy with multiple top offenses on the schedule a multipler of > 1. I thought this would provide a better indication of relative play than just numbers alone. Playing against a bunch of scrubs kind of sucks when you transfer to the NFCE and have 4 guaranteed games against teams who were Top 10 in offense last year.
Only problem here, some guys would get a great, great boost and some would be absolutely hammered. So be it I guess.
So, I'm kind of stuck here.
I have multiple rankings, some numerical and categorical and no way to standardize anything to a reasonable and unbiased number.
I don't want to put TDs, INTs, PDs and Penalties into ratios in terms of attempts or completions because I think it diminishes their value or damage.
Any thoughts.
One thing I will say just looking at the early stuff is that Tim Jennings might be a good CB option.
He played a VERY tough schedule in facing Ten Top 10 opponents. If I were to use the multipler system he'd have a multipler of 1.43 which seems incredibly high but when you play 10 offensively powered teams, maybe it's warranted.
He was targeted 2nd most at 110 times. Eric Wright is #1 at 120. But Jennings allowed 57% and has a YPA that ranks 5th. Didn't give up a single TD and had a QBR against him of 68, or 7th best amongst the group.
--------------------------------------------------------------------------------
I've been looking over the potential FA CBs and some of their stats and such at PFF. Trying to use this site as much as possible before the subscription is up......should've just went with the annual for slightly more.
Anyway, I'm trying to find a way to rank the players based on their performance.
Some categories I have include some of the standard stuff. Targets, completions, completion % and other standard stuff.
But I can't really figure out a decent way to rank players. Some of the categories I've just ranked numerically based on where they fall. I'll likely work with others to try and provide more emphasis to certain stats.
How should players be ranked? Numerically it really doesn't make sense to order them in some situations.
Is a guy who allows 51% really an entire rank better than a guy who allows 52%? And some of the rankings are this close. Some come down to partial percentages.
The best I have come up with thus far for % is to stratify percentages. 50%, 55%, 60% and assign point values to each increment.
Also, is a rank difference of 1 really a true estimation of importance of 1 INT? I think that's selling it short.
Same with TDs. Does a player who allows 1 TD really only deserve 3-4 numerical points over players who allow 3 TDs? I think that is really underselling the value of preventing TDs.
For these, I thought it would be interesting to assign a point value per interception and TD to properly award and punish high numbers. What about pass defends? Just sort of off the cuff, I gave a TD a value of 7, INT 5 and PD 3 and Penalties 3 although I might up penalties because each penalty by a DB will most likely give a fresh set of downs AT LEAST. Illegal contact, holding and PI are all we're really talking about. Maybe Unnecessary as well but that's still a 1st down. I'll probably try to up the penalties somehow.
You get a point total by subtracting the negative (TDs / Penalties) from the positive (PD and INTs). Obviously, more negative is better and I'll probably switch the two so the positive plays dominate so a more positive number is indicative of good play.
Anyway, just wanting to get your guy's opinion on how players should be ranked. Straight ranks seem to ignore a lot of context and I'd like to see a ranking that doesn't just say a player had the best percentage against him but ignores that he was only thrown at twice per game.
One stat I haven't seen anywhere that I'm trying to include is opponent quality. Specifically speaking, number of Top 10 and Bottom 10 offenses that each corner played against.
But how much should these impact results. I think they're pretty big and I was thinking about taking [([Top 10 - Bottom 10] / 16) + 1] and multiplying that by the total score to try and account for quality of opponents.
This would give a guy who played 4 of each a multiplier of 1 while a guy who had more bottom feeders on the schedule a multipler < 1 and a guy with multiple top offenses on the schedule a multipler of > 1. I thought this would provide a better indication of relative play than just numbers alone. Playing against a bunch of scrubs kind of sucks when you transfer to the NFCE and have 4 guaranteed games against teams who were Top 10 in offense last year.
Only problem here, some guys would get a great, great boost and some would be absolutely hammered. So be it I guess.
So, I'm kind of stuck here.
I have multiple rankings, some numerical and categorical and no way to standardize anything to a reasonable and unbiased number.
I don't want to put TDs, INTs, PDs and Penalties into ratios in terms of attempts or completions because I think it diminishes their value or damage.
Any thoughts.
One thing I will say just looking at the early stuff is that Tim Jennings might be a good CB option.
He played a VERY tough schedule in facing Ten Top 10 opponents. If I were to use the multipler system he'd have a multipler of 1.43 which seems incredibly high but when you play 10 offensively powered teams, maybe it's warranted.
He was targeted 2nd most at 110 times. Eric Wright is #1 at 120. But Jennings allowed 57% and has a YPA that ranks 5th. Didn't give up a single TD and had a QBR against him of 68, or 7th best amongst the group.