Sunday, June 17, 2012

Attribute Spotlight - Block Shed (BSH) ratings DE's and DT's in Madden 12.

The idea for this blog came from Clifton Beeker Jr (twitter: @blood9585) who won the first Annual Madden Manniac giveaway contest.  I selected his idea as the winner and preordered him a copy of Madden 13 through Amazon.com.

Here is his idea:

May 28, 2012 7:49 PM
Interested to see really how the block shed rating is determined for top run stuffing and elite pass rushing defensive linemen.

Xbox 360
@blood9585
-----
The rest of this blog post will be my take on this idea.  He wanted top run stuffing and elite pass rushing defensive lineman in this blog, so I decided to use Madden 12's final overall (OVR) ratings to select these players.  Why not use EA's own ratings and see how it plays out. (Please note - I typically don't use OVR rating for anything in my blogs because I believe the formula is extremely flawed.)  In the end, the cut for each position ended up being 84 OVR or better.  I also got the average BSH rating for all LE's, RE's, and DT's in Madden 12.  

One of the reasons I choose this idea as the winner is because I knew it would be a challenge.  If you have followed my blogs in the past, then you know I believe attribute ratings should be as objective as possible.  The block shed (BSH) attribute has a component of subjectivity that can't be ignored, therefore it's a challenging attribute to rate.  

The biggest problem with rating BSH, is how to rate players who get constantly double-teamed.  Many dominate defensive lineman face a lot of double-teams, this has a negative impact on their statistics.  Not to mention, defensive lineman in a 3-4 scheme tend to face more double-teams then those in a 4-3 scheme.  That said, some of the best defensive lineman can even beat double teams from time to time.  In the case of the block shed (BSH) attribute, statistics can't possibly tell the whole story.  Don't get me wrong, I still think statistics play an important role with this attribute, but so does the eye test and that is where the subjectivity (bias) comes in to play.  

To date I have not found a website that tracks the number of times or the percentage (%) of plays that an individual player is double-teamed.  If there was such a statistic I think the BSH attribute could be much more objective and less subjective.  The key to rating the BSH attribute is to watch a lot of game tape and  combine that with useful statistics.  Unfortunately, the "Madden ratings team" isn't big enough to give the BSH attribute that kind of attention, they just don't have the resources.  

My goal was to try and figure out what statistics (if any) were used to rate the BSH attribute.  I would also be taking a close look at the consistency of the attribute rating as well.

First, I collected the BSH rating for every LE, RE, and DT from the final roster update in Madden 12.  This allowed me to come up with an average, mode (the most reoccurring number), and a range for the BSH attribute.  Here are the results:


You can see from the above table that Defensive Tackles have the highest average BSH attribute, followed by Right and Left Ends.  The Modes (most reoccurring number) are very close and the ranges are similar as well.  

Next I created spreadsheets for all three positions which included:  Overall rating (OVR), Strength (STR), Power Move (PMV), Finesse Move (FMV), and BSH attributes.  I also included their sack totals from 2011 and the last three seasons.  The sack statistics where gathered from NFL.com.  

I added two ratings from www.ProFootballFocus.com.  The first is the Pass Rush Productivity (PRP) rating, which is derived from sacks, QB pressures, and QB hits.  Basically, the PRP gives you more information to consider when determining how effective a player is as a pass rusher.  Sacks alone don't paint a full picture.  The second ratings I used from www.ProFootballFocus.com was their Run Stop % rating.  This is their definition of stop %, "Stops constitute a "loss" for the offense so this table shows the percentage of stops per snap played only in run defense."

I choose these two statistics from PFF because I believe they can aid in rating the BSH attribute.  By no means do I think any of these statistics should be the sole resource for rating BSH.  
---
Here is the spreadsheet for LE's:

BLUE = highest or best in that category.

An (*) indicates the player has been in the league less than 3 years.

The average BSH attribute for all LE's in Madden 12 was 75.8.  

The first thing that should jump out at you is that Robert Mathis has a BSH attribute of 69, that is almost 7 points lower than the average BSH attribute for all LE's.  You can look at all the statistics and game tape you want, but I would be embarrassed if I had to take credit for this rating.  Did you know that there is a longsnapper named John Denney who has a block shed of 70?  Mr. Denney has a total of 29 tackles in 7 seven seasons with zero sacks, yet he has a better BSH than Robert Mathis.  Really?

Mathis is not alone, there are several top LE's at or below the average BSH rating.  Why is this happening?  I will address that question later in this blog.
----
Here is the spreadsheet for RE's:

BLUE = highest or best in that category.


An (*) indicates the player has been in the league less than 3 years.

The average BSH attribute for all RE's in Madden 12 was 76.2.  

The players that immediately jumped out at me in this spreadsheet, were:  Peppers, Freeney, Umenyiora, and Dumervil.  Again, these ratings are embarrassing.  Remember long snapper John Denney?  His 70 BSH is higher then that of Dumervil, Peppers, and Freeney.    

Look at Trent Cole; he is among the top players at this position in sacks and was first among this group in both PFF ratings. Yet, he only has a 77 BSH which is just a tick above the Madden average for RE's.
----
Here is the spreadsheet for DT's:

BLUE = highest or best in that category.




An (*) indicates the player has been in the league less than 3 years.

The average BSH attribute for all DT's in Madden 12 was 80.5.  

The defensive tackles have the highest BSH attribute among the defensive lineman.  There is a lot of inconsistency among these 22 players.  Most defensive tackles don't rack up the sacks and in the case of the 3-4 Nose Tackles, they face a lot of double-teams.  Because they play in the trenches, this is probably the toughest position when it comes to rating BSH.  You may have noticed, that many of the stronger DT's have the higher BSH ratings among this group.  Hmmm.

Here are a few Defensive Tackles who have lower OVR ratings, but have high STR and BSH attributes:

----
After reviewing all of the data, I started to see a trend.  It seems like many of the DE's and DT's with a higher strength (STR) attribute, also had a high BSH attribute.  This didn't occur in all cases, but enough to make me want to do a scatter plot to see if there was a correlation among these 63 players.

Here is the scatter plot comparing the BSH attribute to STR:


It's widely accepted that any R^2 value 0.80 or higher is a strong correlation.  The closer the R^2 is to 1.0 or -1.0, the stronger the relationship.  The closer the R^2 value is to 0 the weaker the relationship.    


You can see this data set has an R^2 of 0.7714.  While not over 0.80,  there is still a clear positive correlation.  Some might argue it's a strong correlation even though it's slightly below the 0.80 benchmark. For more information on this type of statistical analysis check out this link:  http://mathbits.com/MathBits/TISection/Statistics2/correlation.htm

-----

So what does all of this information tell us?  Honestly, it's does tell us much.  After going through all of the attributes and statistics I can't find anything truly consistent about how the "Madden ratings team" assigns the BSH attribute.  I do believe the strength attribute plays a part, but it's not as much of a factor when you look at players with a lower overall.  Strength is used by EA to manipulate the OVR ratings of some players, especially DT's.  Both awareness and strength are tied as the #1 factor in the Defensive Tackle overall (OVR) formula.  BSH, PMV, and FMV, are all tied as the 2nd most influential attributes in the OVR formula for DTs.  So, both STR and BSH can be used to reward or penalize a defensive tackle, even if it's not a true representation of the actual NFL version of that player.  For more on the Madden 12 OVR formula, check out this blog:  http://maddenmanniac.blogspot.com/2012/01/madden-12-what-attributes-effect-ovr.html

Can an NFL player be strong, but not excel at shedding blocks?  I believe the answer is clearly, "yes".  Shedding blocks is not all about strength.  With good technique a weaker player can shed blocks.  There are several techniques that defensive lineman learn to increase there chances of shedding a block.
----
Out of curiosity, I wanted to know how much the BSH attribute factors into simulated seasons and gameplay.  I simulated three seasons with the NY Giants (Year 1 for all three, with the same schedule).  First I turned the BSH attribute down to 0 for all of their DE's and DT's.  I didn't change any other attributes.  I turned injuries off, because I wanted the same players to be available for all three seasons.  

After the first season, not one player on the defensive line recorded a sack.  That's right, zero sacks for the defensive line with the BSH attribute at zero.  

I restarted the franchise for the next simulated season.  I turned the BSH attribute up to 99 for all of the defensive lineman.  Again, I didn't change any other attributes.  This time the defensive line accounted for 38.5 sacks.

For the third and final simulation, I restarted the season and turned the PMV and FMV up to 99 for all of the defensive lineman and set the BSH attribute to 0.  Again, the defensive line did NOT record a single sack and they only accounted for 9 total tackles as a group.  WOW.

Some people tell me that attributes don't matter.  I believe several attributes matter and for simulation purposes, BSH plays a huge role in the both sack and tackle statistics.  I know this was a drastic comparison by using 0 and 99, but it was the quickest way to see if there was an impact.

On top of the simulated seasons, I played 3 games against the CPU, one with each of these attribute settings.  I even picked a horrible team to play against (Jacksonville), just to give my 0 BSH defensive line a chance.  In the end, the 0 BSH teams had a very difficult time.  They gave up a lot of big runs to MJD and put very little pressure on the QB.  I even controlled several of my lineman and had very little success getting off the block.  When I played with the 99 BSH team, they not only applied more pressure to the QB, they played the run much better.

The point with this experiment, is that having realistic (accurate) BSH ratings should be a priority for the ratings team.  I believe this attribute is one of the most important for defensive players and it should be as accurate as possible.  Players like Peppers and Mathis are not nearly as effective as they should be, due to having a poor BSH attribute.  This needs to change.

So how should the BSH attribute be rated?  First and foremost, the "Madden ratings team" needs to be expanded so evaluators can watch more game tape.  In my opinion, watching tape and reviewing the best available statistics will produce the most accurate and consistent block shed ratings.  The ratings team needs to be expanded not only because there is limited time between roster updates, but also by having more people involved will reduce bias.  BSH is also an attribute that should NOT change week to week.  I believe a player can improve, but this attribute should not be used just to manipulate a player's OVR rating.  Accuracy and consistency in player ratings should always be the priority, regardless of the impact on OVR.

I would like to thank Clifton for his great idea and I hope I was able to shed some light on the BSH attribute.  Hopefully, the Block Shed attribute can be more accurate and consistent when it comes to Madden 13 player ratings.
------

*** View and vote on all of my Game Changer Ideas here:  http://mgi-gc.easports.com/a/ideafactory.do?id=18503&mode=author&discussionFilter=activeStay tuned for more Madden 13 player rating projections.  Thank you for following my blog.  You can also follow me on twitter @mannmicj

Happy Birthday to my wife, Julie and Happy Father's Day to all the great dads out there (mine included).











12 comments:

  1. this is awesome mike - great work here - keep it up mann (see what i did there?)

    ReplyDelete
  2. Greaf stuff. Keep up the good work.

    ReplyDelete
  3. Very interesting analysis. However, I think you missed a few factors that you should explore. You should explore the correlation between ...

    1. Weight and STR.
    2. Weight and BSH.
    3. PFF's run defense grade per snap and BSH.
    4. a player's Combine weight lift exercise and STR.

    Cheers.

    ReplyDelete
    Replies
    1. Sorry, I can get everything into one blog. I have a day job and do all of this research in my spare time. I like some of your ideas and they would make interesting blogs. I might hit on some of them in the future:)

      Thanks for following the blog.

      Delete
    2. I actually started to look into them last night. I don't have access to all the PFF numbers so I'll focus on what I can. If you want, I can email to you the figures and graphs.

      Delete
    3. Thanks for your interest. I try to do all of my own research. I will look at what you have, but if I do a blog on those things I will do redo all data collection.

      Thanks.

      Delete
  4. do you know of any site that has a list of player ratings and their individual attributes after the final madden 12 roster update? i see some of your blog has the final player attributes after the update and i was wondering if any site had a list of all the changes, not just the overall rating but of all the individual attribute changes. thanks!

    ReplyDelete
    Replies
    1. Not that I know of. I have to compile the information one player at a time by viewing the rosters in the game. It's a pain. I doubt EA would ever release that info in a separate file or blog because it would be even easier to find errors and inaccuracies. There is a reason they only show OVR on the roster update blogs. My goal is to get them to provide more data. Last year, they did release the original rosters and attribute prior to the release of the game, but after that you have to track it yourself. Not to mention, many of the those original attributes change prior to the release of the game, plus every week of the season.

      If I had more time, I would attempt to do it myself. I juggle my blog with my day job, so time is limited. Thanks for following the blog.

      Delete
  5. Michael, I have been a ratings freak for years as well. I am a little lighter on the statistical side. I do believe there is a great system out there to produce great attribute ratings. Have you considered creating one? I am shocked EA sports does not use such an objective system. I would imagine you would always have to have some type of subjective input but the basis of it should be formula based. A formula that could scrub statistical weekly data from nfl stats and make the appropriate adjustments. Another one that boggles my mind is Jason Babin of the Eagles. His attributes basically make him a DT. While he is widely known as a speed rusher... He has like 87 STR and 74 spd. And his struggles against the run are well known.

    ReplyDelete
    Replies
    1. I would need more resources to create a completely new system. Right now, I have trying to help them find ways to make the system they have better (with my blogs). Unfortunately, one person makes most of the decisions and that person doesn't have the attention to detail required to produce accurate and consistent ratings. Much of it is not his fault, he is understaffed and lacks resources. On the other hand, there is plenty that he can change now. He chooses not to. I don't know why.

      The first thing required for consistency and accuracy is to get a TEAM to do ratings/attributes, not just one person. One person can not do it alone. It will also reduce bias to have a TEAM involved. I am sure EA doesn't want to spend the resources ($$$) to do that.

      As long as Madden fans only concentrate on OVR, EA will continue to give us inaccurate player ratings/attributes.

      Delete
  6. I was thinking some system that includes testing performance at the combine, position, size, then having the attributes also tied to performance. So for example a 40 time that converts to 0-100 attribute have the 10 yard, 20 yard, and shuttle break down into acceleration and then 3 cone drill equate to agility. But obviously we can't just count on those because then we would have the busts who are workout warriors dominating... You would want to be cautious with rookies as you would need more subjectivity or perhaps college statistics could be used?? But for the rest of the players have have a good sample size statistics would be a great way to adjust those ratings.. so a guy who runs a fast 40 time but never is able to gain separation (and does not have a high amount of big plays) should loose some of that speed because it has not translated.. Also I feel yards after contact should be a direct correlation to break tackle.. etc.. I think you can find a pretty good way to translate these attributes more realistically.. I guess most important is understanding how each rating works.. you have identified the importance of Block Shed.. How do you typically set your depth chart? What attributes do you look for? speed? CTH? hit power? tkcl?

    How far away do you think we are from a computer tracking players on the field and generating attributes? Similar to the one they use on tv to show a player acceleration and actual speed on replays?

    People always say EA does not have the money or resources. They sold millions of these things. Especially when they have so many willing and interested people willing to work not only for money but because of there interest in the game.

    ReplyDelete
    Replies
    1. Let me put it a different way. EA chooses not to put the resources ($$$) into ratings and attributes, and it shows.

      I agree, 40 time, splits and cone drills can be used for rookies. I have discussed such things. I have talked about drop% and the CAT attribute. I have discussed DAC and deep passing statistics. I have talked about tackle efficiency (missed tackles per tackle made) and many others. There are several statistics that can directly correlate to the objective attributes. Then you have the more subjective attributes, such has SPC. That is were a team would really help.

      I don't think we are that far away from a CPU generating attributes. It would beat the bias and inconsistency of one person.

      I don't do this for money. I do it because I love the NFL and Madden and want the game to be represented as realistically as possible. One of my main goals it to raise fan awareness. If more fans put pressure on EA to address this issue, then maybe it will be corrected sooner rather than later.

      BTW - Yards after contact can be used for Elusiveness (ELU) and maybe trucking (TRK). Trucking is difficult because that is more of a power move. Breaking a tackle isn't always about running over a player. Maybe a Break Tackle attribute makes more sense than Trucking (TRK). A powerful player will run them over (or through), while a weaker player with more elusiveness slips through their grasp.

      These are the type of things a team could discuss. You bring up some very good topics, it's too bad this discussion doesn't take place at EA.

      Thanks for all of your feedback.

      Delete