New Advanced NFL Stat (1 Viewer)

[Lavachebeadsman]

Link to full article.

After doing the research for this article and getting some feedback, I really wanted some input from the FBG community. What I have created is a way to determine how of a players fantasy points he gets in the redzone. I'm not exactly sure its' application thus far, other than to get a clearer indication of a players usage. Additionally, it would be possible to see how a player did with an excess or lack of redzone looks and compare that to future expectations. Any thoughts about the models application would be cool, or tell me if it is totally worthless. What follows is the first table I made, detailing 2012 RB Redzone Reliance Percentage.

EDIT: I cannot figure out how to post the table in this post. Can any of you help post the table from the link?

 

[ConnSKINS26]

Sweet watch.

 

[ConnSKINS26]

But really, good work. Interesting stuff. Gonna have to delve deeper into it.

 

[BrainDeadGenius]

I like the concept. I'm always interested in reading new perspectives. I'll have to put the players into tiers to determine which players put together similar Redzone Reliance Factor. For now, it appears that you have goal line backs with the highest totals and PPR RBs have the lowest totals. Not sure how it can be used to predict future success.

 

[amicsta]

Really interesting article. It would seem that some baseline would have to be established to try and find a future regression or improvement, but thst would seem difficult since some teams clearly find the redzone more than others. Foster and Ridley were both over 40%, but that could mean something different for both players.

 

[Lavachebeadsman]

Really interesting article. It would seem that some baseline would have to be established to try and find a future regression or improvement, but thst would seem difficult since some teams clearly find the redzone more than others. Foster and Ridley were both over 40%, but that could mean something different for both players.

Indeed, this data set definitely needs more fleshing out, which is why I wanted to bring it to FBG. I'm really looking for some advice or some very general takeaways so I know what to look for. I really understand fantasy sports and find them to be a practical application for mathematics, but in general, I am a writer and don't excel with numbers.

 

[GregR]

EDIT: I cannot figure out how to post the table in this post. Can any of you help post the table from the link?

Probably easiest to use a code block. You may need to add tabs or spaces to some lines depending on text length to get it to stay lined up right. If you need to see what a code block looks like, just hit reply to this post and look at what I show below:

Col 1	Col 2	Col 3Boo	100	200Ya	200	400

 

[amicsta]

Really interesting article. It would seem that some baseline would have to be established to try and find a future regression or improvement, but thst would seem difficult since some teams clearly find the redzone more than others. Foster and Ridley were both over 40%, but that could mean something different for both players.

Indeed, this data set definitely needs more fleshing out, which is why I wanted to bring it to FBG. I'm really looking for some advice or some very general takeaways so I know what to look for. I really understand fantasy sports and find them to be a practical application for mathematics, but in general, I am a writer and don't excel with numbers.Well on it's surface, I'm not sure this tells us anything more than we already know. The fantasy players that have the highest % are the ones who had a high number of redzone TD's. That's why guys like Charles and Spiller stand out on one end (guys who break long runs with low amounts of redzone work) and Foster and Ridley on the other (guys who are in great offenses with many redzone opportunities). You would have to delve into each teams history to see if it is uncharacteristic to have that many redzone TDs for NE and HOU and so few in BUF and KC. Again, this is just my initial reaction, I have not looked deep into it myself.

 

[Otis]

Paging Chase to debunk this

 

[Lavachebeadsman]

EDIT: I cannot figure out how to post the table in this post. Can any of you help post the table from the link?

Probably easiest to use a code block. You may need to add tabs or spaces to some lines depending on text length to get it to stay lined up right. If you need to see what a code block looks like, just hit reply to this post and look at what I show below:

Col 1	Col 2	Col 3Boo	100	200Ya	200	400

I tried it in the code block and didn't work how I thought it would. All of the numbers were all bunched up. It's a massive table and it would take forever to sort out and space each number that needed it.

 

[Lavachebeadsman]

Paging Chase to debunk this

I'm not really trying to prove anything. Is there anything to debunk? I just want research help before I spend time going down a dead road.

 

[Chase Stuart]

Paging Chase to debunk this

I'm not really trying to prove anything. Is there anything to debunk? I just want research help before I spend time going down a dead road.That was my interpretation, too.

 

[big windy]

I think the next thing to explore to determine how useful the stat is would be to test for persistence versus chance. Maybe go back however many years of data that are available, take the average RZR for each season, look at the residuals on a player by player basis, and determine if players tend to have an above average or below average RZR over their careers. You would need to identify players by Team_Player (HOU_AFoster) to account for team changes. It would also be interesting to see how a player's RZR can change team to team after controlling for the variance in team red zone series.You also probably want to see how correlated RZR is with total team red zone series. One would assume "really high" but you don't know until you look. As was mentioned earlier, the higher the correlation, the less interesting the stat, I would assume. But It may also uncover some interesting outliers. A player that is ridden hard in the red zone but plays for a poor offense may get a big boost with small improvements on offense, or the opposite since the new offensive tools will be used more in the RZ. It could vary by position. Trade/FA implications would also be interesting for these outliers. As usual, the questions you answer along the way may be just as or more useful than your original goal.

 

[Lavachebeadsman]

I think the next thing to explore to determine how useful the stat is would be to test for persistence versus chance. Maybe go back however many years of data that are available, take the average RZR for each season, look at the residuals on a player by player basis, and determine if players tend to have an above average or below average RZR over their careers. You would need to identify players by Team_Player (HOU_AFoster) to account for team changes. It would also be interesting to see how a player's RZR can change team to team after controlling for the variance in team red zone series.You also probably want to see how correlated RZR is with total team red zone series. One would assume "really high" but you don't know until you look. As was mentioned earlier, the higher the correlation, the less interesting the stat, I would assume. But It may also uncover some interesting outliers. A player that is ridden hard in the red zone but plays for a poor offense may get a big boost with small improvements on offense, or the opposite since the new offensive tools will be used more in the RZ. It could vary by position. Trade/FA implications would also be interesting for these outliers. As usual, the questions you answer along the way may be just as or more useful than your original goal.

Really, really helpful stuff. Thank you. This data crunching stuff is mind numbing, especially because I'm starting wholly from scratch. That table took me about 2 and a half hours to complete fully, so going back to years past is going to be a pain but it's something I'm going to try and do.

 

[big windy]

Really, really helpful stuff. Thank you. This data crunching stuff is mind numbing, especially because I'm starting wholly from scratch. That table took me about 2 and a half hours to complete fully, so going back to years past is going to be a pain but it's something I'm going to try and do.

No problem. I agree, this stuff is hard. Picking up a relevant programming language (R, Matlab, etc.) helps a ton because once you code something up, you can re-use it with new datasets going forward. If you're doing it in Excel, it's a manual rework process every time. Also, the specific tests I mentioned were the result of some quick spit-balling. They were more examples of how to think about the problem rather than specific solutions. Good luck!

 

[nightmare]

Yards per carry and catch isn't as relavent in the red zone-hard to increase it if you get a lot of 2 yard TD runs. It's good work but I believe Foster got less work between the 20's so he could be fresher for red zone looks at each end of the field.

 

[GregR]

Yards per carry and catch isn't as relavent in the red zone-hard to increase it if you get a lot of 2 yard TD runs. It's good work but I believe Foster got less work between the 20's so he could be fresher for red zone looks at each end of the field.

I would call that accurate. It seemed like the Texans usage of their backup RB often included putting him in around the 40 and then putting Foster back in around the 20. Though not the case on every drive, it happened enough it seemed to me like it was planned.

 

[Lavachebeadsman]

Really, really helpful stuff. Thank you. This data crunching stuff is mind numbing, especially because I'm starting wholly from scratch. That table took me about 2 and a half hours to complete fully, so going back to years past is going to be a pain but it's something I'm going to try and do.

No problem. I agree, this stuff is hard. Picking up a relevant programming language (R, Matlab, etc.) helps a ton because once you code something up, you can re-use it with new datasets going forward. If you're doing it in Excel, it's a manual rework process every time. Also, the specific tests I mentioned were the result of some quick spit-balling. They were more examples of how to think about the problem rather than specific solutions. Good luck!This may end up being my downfall. I am doing all the computations in Excel and the process is quite time consuming. For those who care, I am planning on doing this for wide receivers for 2012 before delving into past stats.

 

[GregR]

I'm not exactly sure its' application thus far, other than to get a clearer indication of a players usage.

It can be useful to just crunch some numbers and see what they show... sometimes you'll find something you never considered looking for.But if I were you, I would feel I was at the point I needed to find an answer to your statement above before I put much more time into it. What use is it? And as importantly, is it better than other stats we currently use for the same thing?We already have stats like number of red zone touches (carries+targets or carries+receptions)... or percentage of team red zone touches. I'd be testing what you made to see if it helps predict touchdowns better than those.Though in doing so I'd be careful of the fact you have built into your result the quantity you would be checking the correlation to. I'd probably remove touchdowns and only work with yard and PPR fantasy points so you don't bias your results.(Edit to add: At least, I'm thinking removing touchdowns would be best before using a correlation coefficient. But I'll defer to Chase on it if he wants to chime back in.)

 

[nightmare]

I'm confused, 8 of A brown's 19 carries were TD's. 1.9 ypc. You remove the TD part and he looks like a lousy short yardage back.

 

[GregR]

I'm confused, 8 of A brown's 19 carries were TD's. 1.9 ypc. You remove the TD part and he looks like a lousy short yardage back.

This is good, this is what I'm talking about. Now you are uncovering what stats are meaningful towards a given end and which are not.So you've decided that (at least for Brown) that red zone ypc is a bad predictor of... what? Ability as a short yardage back? A predictor of touchdowns?So now take your own comment one step further. If the red zone yards per carry were not useful to use for us to gauge touchdowns or short yardage back quality... is it just the result (yards/carries) that isn't useful. Are RZ yards by themselves useful? Are RZ carries by themselves useful? If neither of them ends up being useful for a given purpose, then does including them in a stat like red zone fantasy points/total fantasy points make it a better or worse predictor of touchdowns? You would think it would probably make it a worse predictor.Though I imagine if you actually go through those steps, we'd find that number of red zone touches is a fair predictor of touchdowns, and is probably a far better one used alone than it is including red zone yardage, red zone fantasy points, total fantasy points or total fantasy yardage. But that's just what I suspect, not what I know, so if I were wanting to figure out what stat would be useful I'd test each of those.As I said, sometimes it's useful to just wing it and see what something shows you. But once you've figured out what use it might have then you probably should start testing to see how well it works, and whether the factors you've included make it better or worse than doing it another way. By looking at the different combinations of factors that go into a particular stat like this we may be able to show that some other version of the stat would be better for the use we want to make of it.

 

[Donsmith753]

Really, really helpful stuff. Thank you. This data crunching stuff is mind numbing, especially because I'm starting wholly from scratch. That table took me about 2 and a half hours to complete fully, so going back to years past is going to be a pain but it's something I'm going to try and do.

No problem. I agree, this stuff is hard. Picking up a relevant programming language (R, Matlab, etc.) helps a ton because once you code something up, you can re-use it with new datasets going forward. If you're doing it in Excel, it's a manual rework process every time. Also, the specific tests I mentioned were the result of some quick spit-balling. They were more examples of how to think about the problem rather than specific solutions. Good luck!

This may end up being my downfall. I am doing all the computations in Excel and the process is quite time consuming. For those who care, I am planning on doing this for wide receivers for 2012 before delving into past stats.Might be possible to do it with a macro, depends what you're looking to do?

 

[nightmare]

Holy moly-I don't think stats are a predictor of ANYTHING. we make predictions on the results of the stat. A Brown can't improve on his YPC because half were TD's. His run stops at the endzone. You need everything to make predictions. H Miller was not ranked very high in redraft-along comes a new offense and he's the second leading receiver for Pitt and a great FF starter. Stats didn't predict that. Three year averages didn't predict that either. Was it injuries(O-line) or the new offense? Yes, there will be trends but every year half of the top 12 in every FF starting position do not repeat even with the same opportunity, I would like to see team game logs-not just player. Does Green Bay always throw it into the endzone? Were A Browns runs on 1st or 3rd down? We make predictions on TRENDS. Team game logs show trends.

 

[GregR]

I don't think stats are a predictor of ANYTHING. we make predictions on the results of the stat.

 

[nightmare]

Yeah, I need to be more specific-end of season stats for next year. Everybody picked up A Brown when he got playing time.Is he the goal line back next year? probably. but he might not get as many looks if D Wilson rush/recs 5-15 yardTD's like B Westbrook used to do on a regular basis. Westbrook was good for 5 to 8 rush and 3 to 5 receiving, VERY few of those were from 1 yard out.

I don't think stats are a predictor of ANYTHING. we make predictions on the results of the stat.

 

[Lavachebeadsman]

For those interested, I did these numbers for WR's and TE's for 2012. http://sportswunderkind.com/redzone-reliance-percentage-pass-catchers/

 

[Drop]

Why is being 20 yards or less away from the endzone so significant to this model? Why not base it on something more important to the strategy of the game, like being within field goal range?I feel that maybe at lower levels of competition 20 yards away is a decent arbitrary mark, but in the NFL kickers can score at a high % from further out and QBs can throw accurately from further away.I would be interested to see things from the perspective of being within range of the FG kicker being 90%, 85%, 80% or something along those lines, as well as being within the 10 (no first downs) because it's strategic and not based on a traditional arbitrary zone. Is data from the 21 yard line so different from 19 yard line data?