Tuesday, February 21, 2012

Attribute Spotlight - Running Back Carry (CAR) attribute (NFL Fumble %)

    When it comes to Madden football, there may be no position were the Carry (CAR) attribute is more important than at Running Back.  The very nature of the position requires accuracy and consistency for the CAR attribute.  In this blog, I will show you the fumble percentage of 40 different NFL running backs.  This is my most in depth blog yet when it comes to fumble percentage, and by the end of this blog I will propose a new way of rating the carry attribute in Madden football.  It's not my intention to just bash the current ratings system.  My goal was to find a better way to rate the CAR attribute so all players are rated more accurately and consistently in future Madden titles.

     I still haven't found a website that provides a fumble percentage statistic.  So all of this data is the result of NFL statistics I gathered from the web.  (You can find these stats on several websites, like:  NFL.com, ESPN.com, Yahoo.com, FoxSports.com, etc...) I entered the statistics into a spreadsheet to calculate a fumble percentage per touch.   I defined a touch as a reception, rush attempt, or kick/punt return.  As I said before, my focus will be on rushing attempts and receptions.  

 NFL Rulebook:  definition of FUMBLE: 
Article 4 A Fumble is any act, other than a pass or kick, which results in loss of player possession. The term Fumble always implies possession. (8-7-3).

    First I had to educate myself on the average fumble percentage for NFL running backs.  Please be aware that I only included players that were defined by NFL.com as a RB,  I intentionally excluded Fullbacks.  From 2009 to 2011, Full Backs carried the ball 1052 times and had 14 fumbles which results in a fumble percentage of 1.33%.  The average FB clearly fumbles at a different rate than the average RB, so I will treat them as completely separate positions.  This is what I found when looking at the last three regular seasons:

- In general the percentages are very small, with some variance from year to year.  For the most part, NFL running backs are great at protecting the football.  I believe the 3 year fumble percentage is a reliable number to set as a bench mark when looking at the last 3 years of the selected RB's.  There needs to be bench marks.  I plan on going back 10 years for all three of these fumble percentages.  So far I have only completed the last 10 years for RB's (rushing touches only).  This is what I found:

    From 2002-2011 NFL RB's had 119,683 carries and 1,274 fumbles for a 1.064% fumble percentage.  Compare that percentage to the 2011-2009 rushing fumble percentage of 1.049% and we see a slight change, but nothing drastic.  I will later use the 10 year rushing fumble percentage as a benchmark to compare with a career rushing fumble percentages.    

    You will see the category of "All Touches" in a upcoming spreadsheets, that includes kick and punt return touches as well as rushing and receiving.  Since the stats show that Kick and Punt Returners are at a higher risk of fumbling, I decided to disregard those numbers when I rated the CAR attribute. 

    The above table includes all kick return/punt return touches and fumbles from the last three seasons.  You can clearly see that being a returnman in the NFL increases the likelihood that you will fumble the football.  The kick return fumble percentage is 2X has high as the other RB fumble percent figures. The Punt return fumble percentage is over 5X as high as the other RB fumble percentages. 

    A Running back like Darren Sproles is at a disadvantage because he return punts and kicks which automatically puts him at a higher risk of fumbling the football.  I think the kick/punt return fumble percentage should factor into the KR/PR rating in Madden, but not in the CAR attribute when it comes to the player's primary position, such as RB. 

    In this spreadsheet you will see the fumble % broken down into four categories.  I believe that the "Rushing & Receiving Fumble %" will be the most reliable to use when rating the Carry (CAR) attribute.  These percentages are based on Regular Season statistics only.  Not that the playoffs are not important, but this allowed me to compare all RB's equally even if they didn't make the playoffs.  How I think the playoffs should impact Madden attributes is a totally different topic that I plan on addressing in future blogs.    

    Notice how Michael Turner and Marshawn Lynch both have a 99 CAR (M12 CAR) attribute, but neither really deserves it based off these numbers.  Brandon Jacobs and Willis McGahee are clearly overrated as well in this attribute.  I chalk up these poor ratings to the general perception of these players.  That is why perception should not factor in the attributes that can be objectively rated by using statistics.  Some attributes can't be rated based on statistics alone, but the CAR attribute is a great example of one that should be rated according to statistics.

   Green-Ellis has clearly earned a rating of 99 CAR by having never fumbled the football in his career.  In some cases such Green-Ellis, the attributes look pretty good.  Now look at Darren Sproles, he is 6th from the bottom on this table due to having a 78 CAR attribute in Madden 12.  Sproles has one of the best rushing and receiving fumble percentages in this group no matter what time frame you look at. 

    Right now in Madden 12, the four most important attributes for a RB's overall (OVR) rating are SPD, AWR, CAR, and BCV.  All four of these attributes are weighted the same when it comes to OVR rating.  I believe in some cases the CAR attribute is used to hold down a player's OVR regardless of their ability to hang onto the football.  If the CAR attribute is a player's ability to protect the football, than it should be based on reality and not used to lower or raise the OVR.

Blue = Good, better than the average.

Red = Bad, worse than the average.

* Please note that the NFL RB Rushing Fumble % used for career comparison (1.064%), was based on the last 10 years of NFL regular season statistics for ALL Running Backs.

    I included the spreadsheet below to show that a higher amount of touches will not result in a higher fumble percentage.  Look at some of the players who have the most touches the last 3 years;  running backs like Steven Jackson and Ray Rice have done very well despite a heavy work load.  I think this spreadsheet clearly shows that some players are just better at securing the football, regardless of the touches the receive.


   In this table, you can see the incosistency and inaccuracy of the current CAR attributes when compared to the actual Rushing and Receiving Fumble %.  The accuracy is roughly 50% when looking at these players.

Blue = Good, better than the average
Red = Bad, worse than the average


    My final spreadsheet below is my recommended CAR ratings for the 40 running backs.

    The hardest part was coming up with a procedure that I could follow for each RB while removing bias.  First, I decided to only consider the 3 year fumble percentage and the career fumble percentage.  By doing this, I believe it gives the appropriate weight to the most current season while also not overlooking a players entire career.  Basically, I didn't include the 2011 season because it is already represented in the other two figures.

    There are no rookies on this list (although rookies were included in total NFL numbers), I only included players with at least 2 seasons in the NFL.  I will go over Rookie ratings at the end of this blog.

    I would subtract 1 point from 100 for every tenth of a pecent (0.1%) that a player has in the fumble percentage.  To do this I rounded all fumbles percentages to the nearest tenth of a percent.  I did this separately for the 3 year and the career figures, than I averaged them out to come up with the new CAR attribute.  While a 2 or 3 year player has the same number in each column, it still reflects what they have done in the NFL so far.  Veterans don't get underrated for one bad season, or overrated because of one good season.

    Here are my suggested carry (CAR) ratings:

    *I normally don't like anyone receiving a 100 rating, but for Green-Ellis I made an exception.

    You can see based on my proposed rating system the average RB would have an 89 CAR attribute rating.  Currently the range for the CAR attribute in Madden 12 is 57 to 99 in regards to RB's.  I don't think you would see many RB's below a 70 CAR using this procedure.  A player should be that low unless the numbers support it.  This attribute is too important in regards to gameplay to be inaccurate.  It can't just be a tool to manipulate OVR.  Many players are not used in Madden 12 due to having a low CAR rating, by having accurate and consistent CAR attributes more players will be used in the game (online or offline).  

    Based on preliminary numbers, this procedure would work well for RB, FB, WR, and TE.  Due to QB's fumbling at such a high rate when running the ball, the procedures would have to be modified to some extent.  My earlier findings show that QB's fumble the ball about 6.45% of the time when rushing (I excluded sack fumbles).  That is 6X more than that of RB, WR, and TE this season.  QB's fumbled 5X as often as Full Backs (FB).  Some QB's fumble the ball when rushing over 10% of the time; for that reason, the QB carry (CAR) attribute rating will be a little more challenging.  I plan on doing that very soon in an upcoming blog.  

I also plan on redoing my Tight End and Wide Receiver blogs to reflect this new procedure.  There is no reason why the highest WR carry (CAR) in Madden 12 is only 85 and only 80 for TE's.  My early findings don't indicate that much difference in fumble % when each player is utilized at there primary position.  WR's do show a slightly higher fumble percentage when rushing the ball, but I won't be certain until I collect more data.       


How to handle rookies and limited role players -

   For rookies that answer is pretty easy, use their college statistics.  Now finding a website that actually tracks fumbles for individual college players is not so easy.  Many sites exclude the fumble statistic from individual college player statistics.  I did finally find a site that included fumbles in there stat profiles.  That website is http://www.teamrankings.com/college-football.

   By applying the same carry (CAR) rating procedure to incoming rookie RB's, this is what I came up with:

    Last, but not least.  How do you deal with a 1 year player with very few NFL touches.  I suggest still using his college stats until that player has been in the NFL for 3 years.  After every additional year in the NFL, you can drop off the oldest college season.  Here is an example:

    So you can see here that I dropped the 2007 out when assigning the carry rating.  Since the college website I found did  not specify wither the fumbles were from rushing, receiving, or kick return, I had to use all of the touches.  This is a slight change from the recommended procedure, but one that is needed to be to come up with a more accurate carry rating for Da'Rel Scott.  Since he only has 21 NFL touches, it doesn't seem accurate to rate him solely on his NFL experience.  While Scott did have a higher fumble percentage in college than the NFL average, his current 68 CAR rating is way to low.  He would have been an 87 CAR coming into the NFL based on his college stats.  The one fumble from his only NFL season, has essentially cost him one point to his CAR attribute which went down to 86.  With this formula he would been rated an  87 carry coming out of college.

    Players with limited playing time will always be a challenge to the player ratings person, this is just one example on how to make that rating easier to obtain by including college statistics.  I understand college is not the same as the NFL.  That said, if Da'Rel Scott continues to fumble at a high rate in the NFL while receiving limited touches, his CAR will continue to go down in future seasons.

    That's it guys.  I learned a lot from studying NFL fumbles.  I don't see any reason why this type of analysis can't be applied to the Madden Carry (CAR) attribute.  The main problem is the OVR formula.  By changing these carry attributes, the OVR will go and and down and not everyone will agree with those changes.  That is why I will continue to suggest that OVR no longer be connect to the player attributes, it should be a stand alone ranking.  I intentionally left out OVR from this data, it should not play a role in how we rate player attributes.

   Have a great week.

 **PLEASE READ ** I don't normally like 100 rating for any attribute, but it's hard to reduce a players rating who has never fumbled the football.  To avoid having a bunch of 100 rated players who have never fumbled the ball, but have a very low number of touches, you could set a minimum number of NFL touches that a player needs to achieve before a 100 CAR could be given.  Of course, 100 would only be given if the player maintains a 0.0 fumble %.

   I would probably set a minimum of 150 touches for the TE position, 200 touches for the WR position, and 300 touches for the RB position, before a player could earn the 100 CAR.  Until the player reaches the minimum number of touches at each position the highest CAR they could achieve a 95.  Remember the rookie CAR attribute and college statistics would also be considered for the first 3 seasons or until the player has reaches the position minimum.

    In general it would be very rare for a player to have a 0.0 fumble percentage after reaching the position minimum.  You can see in this blog that only Green-Ellis has a 0.0 fumble percentage.

    Stay tuned for more suggestions and projections on Madden 13 player ratings and attributes.  Follow me on twitter @mannmicj to receive updates on my newest blog posts.

Saturday, February 18, 2012

Attribute Spotlight - TE Carry (CAR)

    Today I decided to expand on the WR carry (CAR) attribute blog, by looking at the TE position.  I will use the same formula to calculate the fumble percentage of TE's, to see who protects the football the best and who is the worst.  I made this list much larger and took most of the starting TE's in the NFL and a few backups who get solid playing time.

    After looking at all of the data, I developed my own suggested CAR ratings for these TE's.  Please note, that out of the 23 attributes that effect a TE's overall (OVR) in Madden 12, the CAR attribute only impacts the OVR approximately 2%.  (How much due attributes effect OVR can be found here:  http://maddenmanniac.blogspot.com/2012/01/madden-12-what-attributes-effect-ovr.html)

    Since the CAR attribute has very little impact on a TE's OVR, it's possible the accuracy and consistency of this attribute is not a high priority.  I believe the CAR attribute is very important due to it's impact on Gameplay.  Although TE's do not fumble much in Madden, it's still part of the real NFL and should be represented as such in the video game.  There should be a clear difference between the best and the worst.

    I hope you enjoy this blog.  In the future, I plan on doing this same type of blog on running backs.


CAR = Current carry (CAR) rating in Madden 12 as of the final Roster Update for the season. 

MY CAR = This is my suggested carry (CAR) rating.

FUM% = The fumble percentage (fumbles/touches) of the player. This is calculated three ways, using career totals, 3-year totals, and the current season. This is based on Regular Season Stats.

CRFUM = Total career regular season fumbles

CRTCHs = Total career touches, which includes: receptions, rushing attempts, punt returns, and kickoff returns.

FUM% = The fumble percentage (fumbles/touches) of the player. This is calculated three ways, using career, 3-year totals, and current season.

11'FUM = The current number of fumbles for the 2011 NFL regular season. 

11'TCHs = The total number of touches for 2011, which includes which includes: receptions, rushing attempts, punt returns, and kickoff returns.

3yrFUM = The total number of regulars season fumbles from 2009 to 2011. 

3yrTCHs = The total number of regular season touches from 2009 to 2011, which includes: receptions, rushing attempts, punt returns, and kickoff returns.

Red Fill Color = (BAD) worse than the average TE in FUM%.

Blue Fill Color = (GOOD) better than the average TE in FUM%

(*) = indicates the player has been in the league two years or less.

*** Update:  For the 2011 NFL Season, 110 Tight Ends (TE) caught 2341 passes for 27,080 yards, 198 TD's, 23 Fumbles.  So the average NFL fumble percentage for 2011 is 0.98% (based only on receiving touches).  

-- All of these statistics were gathered from NFL.com.  The number of touches includes any rushing attempt or kick/punt returns as well as receptions.  To see all fumbles and touches for a player, you need to view the career statistics page for each player.


- The current range for the CAR attribute at the TE position in Madden 12 is 52 to 80.  For comparison, the range for the WR position is 46 to 85, FB is 60 to 88, RB 57 to 99, for a total range 46 to 99 excluding QB's.  

- Like the WR's, these percentages are low in general and it just shows how well NFL TE's protect the football.  Even with the small percentages it is easy identify which players are better at protecting the ball.  

- Be sure to look at all three columns.  Some player have more touches than other due to years in the league and how much they factor into their teams offense.  

- You can see that in general the TE's that are good are protecting the ball have be pretty consistent, in turn the TE's who are not as good at protecting have been pretty consistent as well.  That said, there are a few player trending in the opposite direction.  Daniels and Watson have really improved over the last years, while   Keller has gotten worse.

- Vernon Davis is vastly overrated in the CAR attribute based on the current range for TE's.  This goes back to perception not being the best way to rate a player.  While Davis is a dynamic TE, he does have flaws and those should be reflected in Madden when compared to other TE's.

- The other Davis on this list, Fred is vastly underrated.  He has the lowest CAR attribute in Madden 12 for a TE that is currently on an NFL roster, and he is only 3 points from the being the lowest among all the TE's.  Basically, if you switched the CAR attribute for Vernon and Fred it would result in a more accurate representation of their fumble history.  

- Currently the highest CAR for a TE in Madden 12 belongs to Nate Bynum of the 49ers with an 80.  He is a second year player who missed the 2011 season and was on the IR.  He only has 5 career receptions with zero fumbles, so I excluded him from this study.  I did look into his college history to see why he may have been rated so high.  In college he caught 47 passes and returned 1 kick for a total of 48 college touches with 1 fumble.  His college fumble percentage would have been 2.08% which should not have resulted in a 80 CAR based on the current range for TE's.

- Currently the lowest CAR for a TE in Madden 12 belongs to  Schuylar Oordt at 52.  He was an undrafted FA out of Northern Iowa who is not currently on an NFL Roster.  I could not find any college stats for Oordt that included fumbles.  He was also left out of these spreadsheets.    

Below is my spreadsheet with CAR recommendation for the selected TE's:

****See an updated spreadsheet at the bottom of the page with different suggested CAR attributes based on my new formula of removing one point for every 0.1% in fumble percentage.  I will start at 100 and go from there.  (UPDATED 2/21/12)

*** Update:  For the 2011 NFL Season, 110 Tight Ends (TE) caught 2341 passes for 27,080 yards, 198 TD's, 23 Fumbles.  So the average NFL fumble percentage for TE's in 2011was 0.98% (based only on receiving touches).  


- To determine my the CAR attributes, I first looked at the career numbers, then the three year numbers, and finished by looking at the 2011.  I would give more weight to the the career and three year numbers than just the 2011 number.  I also looked for trends, such as wither or not the player was improving and declining in their CAR ability.  If the career numbers were close between two players I would use the 3 years numbers as a tie breaker.  If both the career and three years numbers were close I would then use the 2011 as a tie breaker.  When I say tie breaker, I mean when comparing to very similar players with a similar number of touches.  I also looked at the number at career touches and factored that into my final ratings.  The procedures I used for rating the CAR attribute could also be applied to many other attributes and statistics, such as:  CAT, TAK, SAC, MAC, DAC, and KAC.  The key is to use all available statistics and analysis.  CAT would not be based on just the number of catches, but also the drop percentage.  TAK would not be based on just the number of tackles, but also the missed tackles.  You can see were I am going with this.  People might disagree with some of the ratings, but having clear defined procedures gives the rating more credibility.      

- My range for the TE carry (CAR) attribute was from 55 to 85.  I selected this range after looking the other offensive positions and their CAR range.  It could be argued that the range should go up to 99 and that 55 is to low, but based on the WR and FB positions this seemed far.  RB is a totally different issue because of the impact CAR has on their OVR (Approx 13% for RB and just 2% for TE).  Currently, the CAR is used to manipulate the OVR rating of a RB and not really based on a fumble percentage.  Again, I will do a RB blog in the coming weeks.  

-  It's strange how close my average and the current Madden 12 average is for these TE's, even though I changed some attributes significantly.  Since my range covers 30 points, I considered 70 to be the average for the CAR attribute when looking at the TE position and as it turned out, it actually was the mathematical average when I finished.

- I gave Tony Gonzalez the top CAR rating at 85. He is the only one at 85.  He has a very low career fumble percentage and no one compares to his number of touches.  He also he hasn't fumbled the ball since 2006 and has only one fumble since the beginning of the 2000 season.  For someone like Gonzalez you could easily expand the range and give him a 98 or 97 CAR rating based on his last 11 years.  I would not argue with anyone who wanted to give him a higher CAR rating, I was just working within my assigned range and he is clearly the best.

- Antonio Gates was close to Gonzalez for his career fumble percentage, but his 3 year percentage was higher than Gonzo.  Gates received a very good CAR rating of 82.

- Brent Celek received the 2nd highest CAR at 83, because he has yet to fumble in his career (5 years).  You could argue that he should be up there with Gonzalez, but the number of career touches made the difference.  Again, if you expand the range, Celek would easily be in the 90's for CAR rating.  

- 49er fans will disagree with my CAR rating for Vernon Davis, but I stand by it.  His career fumble percentage is almost twice as high as the career average for these TE's.  This year his fumble percentage was almost four times as much as the 2011 average.  While his 3 year average showed improvement, his career  numbers and his poor 2011 fumble percentage can not be ignored.    

- While Gresham has only played two seasons, I still gave him the lowest CAR attribute.  His first two seasons have been below average and he is current working on the worst career fumble percentage (I acknowledge on few opportunities than others).  He did show some improvement this year, but he needs to build on that and get into the blue before I would consider raising it.  

- Due to his low number of opportunities, I gave Chandler a 58 instead of a 55.  Chandler has less than half the opportunities of Gresham.  With more opportunities he could easily earn a better rating.

- Since Pitta, Ballard, and Dickson do not have very many touches at this point in their careers, I decided not to give them ratings of 80+.  They still received a very good ratings and can improve on that in the upcoming seasons.

- Ben Watson has been great the last three years with no fumbles.  Because the trend is show great improvement, he received a good CAR rating of 74.  In this case, a player can over come his past by performing much better in that last 3 years.  I treated Owen Daniels in a similar way.  He has been much better the last three seasons and therefore was rewarded for that with a 70, right at the average.

- Like Waston, Keller's career numbers are not the only thing I looked at.  Keller has been getting worse over the last 3 years, for that reason he received a 69 CAR rating which is just below average.

- Some of the TE's we not rated that bad when compared to the actually numbers.  I think this was mainly coincidence, since other CAR attributes were clearly inaccurate.  In the future, 70 could be the default CAR rating for TE's within this attribute range.  Rookies and other players with very little NFL experience could receive a 70 until they prove otherwise.  This can be adjusted during the season if a player has a serious fumbling issue, but otherwise it shouldn't fluctuate much during the NFL season and it should be reviewed prior to the next Madden edition.  

- We saw several players lose substantial points to their CAR ratings this year because of one bad game.  Even though the number of fumbles for the average NFL player during a season is relatively low, I am not totally against decreasing a CAR rating during the season.  I would ask that when the CAR rating is adjusted that not only this year, but the last three years and the career totals be considered.  By looking at the entire picture when rating the CAR attribute, I believe the result will be more accuracy and consistency.

Thanks again for following my blog.  The RB carry blog will take much more time due to the sheer volume of RB's and the statistics.  Have a great rest of your weekend. 

(2/21/12) - Updated TE CAR spreadsheet using my new formula of subtracting 1 carry (CAR) point for every 0.1% in fumble percentage.  See RB CAR blog for more informaton, http://maddenmanniac.blogspot.com/2012/02/attribute-spotlight-running-back-carry.html

 Please note:  A red star (*) next to the suggested CAR rating indicates that the player has less than a 150 total NFL touches and therefore I will not give them a 100 CAR rating until that minimum is reached with  zero fumbles.

 **PLEASE READ ** I don't normally like 100 rating for any attribute, but it's hard to reduce a players rating who has never fumbled the football.  In 2011 all TE's in the NFL fumbled only 0.98% per touch.  That is excellent, based on this rating procedure the average TE would have a 90 CAR.  To avoid having a bunch of 100 rated players who have never fumbled the ball, but have a very low number of touches, you could set a minimum number of NFL touches that a player needs to achieve before a 100 CAR will be given.

   I would probably set a minimum of 150 touches for the TE position, 200 touches for the WR position, and 300 touches for the RB position, before a player could earn the 100 CAR.  Until the player reaches the minimum number of touches at each position the highest CAR they could achieve is a 95.  Remember the rookie CAR attribute and college statistics would also be considered for the first 3 season or until the player has reaches the position minimum.  By doing this, Pitta, Ballard, and Tamme would all have to wait until the minimum of 150 was reached.  Until the minimum number of touches is reached, I decided that a 95 CAR (up for debate) is the max a player can receive.

    In general it would be very rare for a player to have a 0.0 fumble percentage after reaching the position minimum.  You can see in this blog that only Celek has met the minimum number of touches and still has as  0.0 fumble percentage.

Tuesday, February 14, 2012

Suggestions for EA Madden NFL 13 regarding player ratings/attributes

ATTENTION - As of now Madden 13 will not give gamers the ability to hold fantasy drafts, online or offline.  If you would like to let your voice be heard by voting and commenting on the "fantasy draft" topic, please go to this link  http://www.easports.com/gamechangers/ideas?campaignId=27782

* These are the suggestions I forwarded to the EA Games Changers for Madden 13 community day.

      1.  Please consider using a player ratings team instead of just one person.  To have accurate, consistent, and objective player ratings/attributes a team is needed in my opinion.  The job of player ratings is too big for one person, plus a team will reduce the subjective (bias) opinion of just one person.  Plenty of people in the Madden community would contribute to this type of program for free; each could be vetted to make sure that they are not rating their favorite team/division.  A ratings team would also reduce the chances of a team/players being neglected due to having a poor season or for being from a small market (unpopular).  All teams deserve to have accurate and consistent player ratings/attributes.

Rating Team Option:
    The lead ratings person would supervise eight individuals who are each assigned to one division in the NFL.   These eight individuals would watch every game in for their divisions each week and would do all gathering of stats/analysis for those teams.  They would follow each team in the division from the NFL draft/Free Agency to Super Bowl each year.  These eight individuals would come together every week to discuss their suggestions for increases and decreases to player attributes.  This ratings round table discussion would take place prior to the official roster update for the upcoming week.  Although, each of the individuals are not experts on all of the divisions in the NFL, they will be able to hold each other accountable.  They will also help each other decide on ratings that are more subjective like CIT, SPC and HPW.  If you wanted to take this a step further, you could have volunteers assigned to every NFL team that give feedback to the Division representatives.  ** Please note, each Division representative should be assigned to an opposite conference than that of their favorite NFL team, as well as any volunteers.  To have a more objective ratings system, no one should be rating their favorite team/s or division rivals. 

2.     Please consider establishing guidelines/procedures for Madden player ratings.  Guidelines would indicate what type of things should be considered when changing a players attributes.  Things to consider when rating a player include: NFL draft combine numbers, college career (for rookies coming in), NFL statistics, expert analysis, and  in-depth NFL statistics like those provided by ProFootballFocus.com.  By having guidelines/procedures to rate players, it will reduce bias and give the player ratings person documentation to support why changes were made.  These procedures for player ratings should be made public, so that the Madden community knows how players are being rated and why.  Part of the frustration for fans when dealing with the Madden Ratings Debate is that they have no idea what is being considered and universally used to rate each player.

3.  If the ratings person/team is going to utilize a website like ProFootballFocus.com for more in-depth statistics, then those ratings and analysis need to applied to all players on all teams equally.  Credibility is lost when one player’s rating/attributes are justified by the ratings staff based on ProFootballFocus.com, but another player with a similar grade from PFF is being ignored or possibly decreased.  No matter what the procedures are for rating Madden players, there must be consistency or you lose all credibility. 

 4.  In no way should the popularity of a team effect Madden ratings, such as the clear bias toward the Dallas Cowboys http://maddenmanniac.blogspot.com/2011/12/ea-madden-ratings-dallas-cowboy-bias.html . This type of bias must stop to have credibility within your ratings system.  Consumers from all markets (big and small) purchase Madden NFL and they deserve accurate and consistent ratings. 

5.  Utilize more statistics for the objective attributes in the game like: CAT, CAR, MCV, ZCV, and TAK.  The key is to utilize all available statistics to come up with a more accurate rating. 

For example:

CAT (catch) should not be based on just the number of catches made; you must also look at drop percentage.  Brandon Marshall is a good example of this.  While he makes a lot of catches every year, he is also one of the leaders in drop %.  If CAT is the ability of a player to catch a pass, than drop % must factor in.  Marshall currently has a 94 CAT in Madden 12; his drop % has not been factored in.  This works the other way for player who catches fewer passes.  Just because a player only caught 40 passes in a season does not mean he should be a 79 CAT.  What if he didn’t drop a pass all year?  I am not saying he should be a 99 CAT, but he is clearly more sure handed than a 79 CAT attribute.  Here is a link to a blog of mine for more examples:  http://maddenmanniac.blogspot.com/2011/12/inconsistent-catch-cat-attribute-in.html

CAR (carry) should be based on the number of fumbles a player has, not just used to increase/decrease an OVR rating.  By using statistics provided by NFL.com, you can create a number for each player based on fumbles per touch.  CAR should be a very objective attribute, the player either fumbles or he doesn’t.  Someone like Josh Cribbs who has 28 career fumbles should not have an 82 CAR which is the highest for all WRs in Madden 12.  Here is my in depth blog about NFL running back fumble per touch percentages:

TAK (tackle) should not only be based the number of tackles made, but also the number of tackles missed.  If TAK is the player’s ability to tackle, you can’t ignore the missed tackles.  The missed tackle stat has been available for quite some time, but ProFootballFocus has now made easier with their new tackle efficiency stat which is based on opportunities versus tackles made.      Also notice that they exclude a tackle efficiency rating for defensive tackles (DT).  I contacted them to ask why, they stated because DT/NTs in the NFL rarely miss tackles, but they do track the missed tackle stat for them.  The tackle attribute should not be used to decrease a DTs overall (OVR) rating.  The OVR formula is broken.  You have DTs that didn’t miss a tackle all season with TAK attributes rated in the 70s, this is just unacceptable.  It does not reflect their actual ability to tackle.  Here are some more examples in these blogs:

MCV/ZCV – There are many statistics available to help rate these attributes.  Some of the best statistics available are from ProFootballFocus.com, who the lead ratings person touts as his most reliable and trusted source for NFL player and team analysis.  Again, it appears the PFF grades/ratings are used for some players and ignored for others.  The rookie CB Patrick Peterson is a good example of that.  He still has elite MCV and very ZCV despite the fact that he struggled in coverage per PFF.  Also, please look into what type of defense a team actually runs when adjusting MCV/ZCV.  It makes no sense to give a player a big boost in ZCV when their team primarily plays MCV.  The OVR formula for CBs is broken and gives too much weight to MCV; this is why we see increases in ZCV, but little change in MCV after updates.  Many statistics are available to help make this a more objective attribute, please use those stats and use them consistently for all players.

KAC/KPW – This is also another set of attributes that can be very objective.  Statistics are available to help accurately rate both of these.  NFL.com even has average kickoff distance which can be an additional tool for determining KPW.  I propose you create one more kicker attribute and call it something like LKA (long kick accuracy) for FGs over 50 yards.  This would make the kickers more realistic and reduce the impact of KPW on long FGs.  Right KPW is the most important factor for making long FGs in Madden 12.  LKA would factor in each kickers history of making long FGs, which would result in a more accurate representation of all NFL kickers.  Here is a blog with more examples:  http://maddenmanniac.blogspot.com/2012/01/position-spotlight-kickers-kackpw.html

6.  Last, but not least.  Please fix the OVR formula or make it a stand-alone number.  Right now the current OVR formula is based on much more than 99 OVR points.  The current OVR formula is forcing the ratings person to decrease (inaccurately) player attributes to hold OVR at a predetermined number.  Since accurate attributes would result in a 99 OVR for many players, the current formula is preventing consumers from having an accurate representation of the NFL player’s true ability.  I have attempted to figure out what the formula for OVR is and what attributes effect OVR for each position.  I might not be completely right, but I think it’s close.  That said, the OVR formula for Madden player ratings should be released to the public so they understand the system better.  

    Right now, the OVR is being determined first in the mind of the player ratings person and attributes are manipulated from there to reach that OVR.  It should be the exact opposite.  Attributes are what matter to gameplay and they should be rated accurately and consistently regardless of trying to reach a predetermined OVR.   So, unless the OVR formula can be fixed to allow for accurate/consistent player attributes, it should be a stand-alone number that the player ratings person can just enter.  OVR would basically become a ranking not a direct result of attributes.  Here is another blog addressing this topic: 

7.  Thank you very much for listening and considering any of these suggestions.  I know I didn’t enter the #MaddenRatingsDebate the correct way.  I was a jerk and took personal shots at people.  I would like to apologize to everyone for that and let you know that I am now trying my best to keep all suggestions and observations focused on the game itself.  My goal is and has always been to make the game better by having more accurate, consistent, and un-biased player ratings.  The bottom line is we all love this game, which is all the common ground we need to make a difference.  

*** View and vote on all of my Game Changer Ideas here:  http://mgi-gc.easports.com/a/ideafactory.do?id=18503&mode=author&discussionFilter=active

Sunday, February 12, 2012

Attribute spotlight: Awareness and Play Recognition.

    I have been thinking about how to determine the effect of the awareness (AWR) trait on gameplay for some time now.  This is just the beginning of my in-depth look into AWR, but I wanted to get some information out there along the way.

    I picked a few players from the following positions:  QB, RB, WR, MLB, and CB.  Due to the large amount of time this takes, I couldn't do very many players.  I do think the information does show valid trends when it comes to AWR.

    After reviewing the statistics for the defensive players, it was clear that AWR didn't impact the statistics.  That is when I decided to look at play recognition (PRC) for the defensive players selected.  The results were surprising to say the least.  I will note my observations in between each position covered.  Again, this is just the beginning of this "study".

The following is how I did this:

1.  I went into player edit mode for the selected players and set the AWR to 99 for all of them, while leaving all other attributes alone.

2.  I saved and used this same roster in a offline franchise.

3.   I turned off injuries and preexisting injuries in the offline franchise.

4.  I simulated the first season of the offline franchise and recorded the statistics for each player after the season.  I verified all depth charts to ensure that all of these players remained starters for the simulated season.

5.  I than repeated this process with 50 AWR and 0 AWR for the selected players.  I started a new franchise each time and simmed the first season only.

6.  I did one additional simulated season for the defensive players, where I set AWR to 99 and PRC to 0.

P. Manning99625367584477341489.6263

QB Observations:

-  AWR had a clear effect on the QB's completion %, TD passes, and the number of times sacked.

-  Passing yards went down due to reduction in completion %.

-  With 99 AWR, Tony Romo was below his career QB rating (96.9) and career completion % (64.5).

-  With 99 AWR, Payton Manning was below his career QB rating (94.9) and career completion % (64.9)

-  In a simulated season, QB's completion percentage seems to be unrealistic, even at 99 AWR.

-   For Madden Fans who like to use the offline franchise mode, realistic stats are important.  Since awareness has a big impact on the simulated stats for QB, it's extremely important that this attribute be rated accurately and consistently.  

-  These numbers are only based on simulated games, I plan on playing games and collecting data to see how much AWR effects accuracy when a human is in control.  My guess is that AWR will play a role in QB accuracy even with a human playing (BTW - I was wrong here per 2/14/12 update).  If that is the case having accurate AWR attributes for QBs in Madden is even more important.

**** 2/14/12 UPDATE FOR QB'S:

    I have played some games against the CPU using Drew Brees with a 0 AWR and had the CPU use him as well with 0 AWR. I can now be certain that AWR does not affect the completion percentage of a human controlled QB. That is exactly how it should be.  Although the my 99 AWR QB had a slightly higher completion percentage, I don't think it's enough to say AWR had any impact.  It did seem like the CPU was more aggressive when it came to bringing pressure on my 99 AWR QB.

Sack % = Total sacks/(Attempts + total sacks)

Human  0 AWR
1 45 28 62% 1 2%
2 34 21 62% 0 0%
3 37 27 73% 1 3%
4 35 29 83% 2 5%
AVG 151 105 70% 4 2.6%
Human 99 AWR
1 30 21 70% 3 9%
2 44 30 68% 1 2%
3 40 32 80% 0 0%
4 37 29 78% 3 8%
AVG 151 112 74% 7 4.4%

    Against a human, the CPU QB with 0 AWR actually holds onto the ball longer than the 99 AWR QB, which does result in more sacks. The CPU QB with 99 AWR actually forced the ball more to get rid of it before the pressure got to him.

1 15 8 53% 8 35%
2 23 17 74% 8 26%
3 15 8 53% 9 38%
4 18 13 72% 7 28%
AVG 71 46 65% 32 31.1%
1 22 13 59% 9 29%
2 28 11 39% 6 18%
3 19 13 68% 4 17%
4 27 19 70% 7 21%
AVG 96 56 58% 26 21.3%

    Early indications are that the CPU QB with 99 AWR actually had a lower completion percentage than the CPU QB with 0 AWR. Of course, there was the very bad game #2 for the 99 AWR CPU quarterback. Game #2 throws of the average and it would be a lot closer if they game wasn't included.  I didn't ignore that game because it was the worst game for a CPU QB regardless of awareness.  
At times it seemed like the QB with 0 AWR had a slower release, but again the completion percentage was better because the QB took the sacks instead of forcing passes. 

    It's hard to be sure what caused the decrease in completion percentage in the simulated seasons. Could it be the response to pressure?  Maybe.  When simming a season; maybe the pressure does decrease the completion percentage. When I actually played the CPU with 0 AWR this was not the case, the QB with a 99 AWR had the lower completion percentage with more attempts. 

    Also, once there is less than two minutes left in the half or game, both the 0 AWR and 99 AWR quarterback became SUPER accurate and better at avoiding sacks. I believe this is due to traits and the fact that Brees has the "Clutch" trait.  Unfortunately, outside of the clutch trait, traits have been neglected in Madden 12 and are very inaccurate. 

    Anyway, I am still working on it. There is definitely a difference between playing the CPU and Simming.


RB Observations:

- Even with 99 AWR neither RB came close to their career average per carry (McCoy 4.8 and Foster 4.7).

- AWR has a clear impact on yards per catch and it wasn't until AWR was set to 50 that either player was close to their actual career averages per catch (McCoy 7.3 career YPC, Foster 10.3 career YPC).

- In general the yards per carry is to low for RB's when simulating a season in Madden, even with 99 AWR.  Also, in general the yards per catch is too high for the higher awareness RB's in Madden and unrealistically low for the RB's with a lower AWR attribute.

- Fumbles for RB's also seemed a little high when compared to the actual fumble numbers for these two.  Foster has 9 career regular season fumbles and McCoy has 5 career regular season fumbles.

- Again, I plan on playing games with these AWR settings and seeing what the difference is when it comes to gameplay.

A.J. Green9982128715.68212

WR observations:

- AWR did not seem to have any effect on the number of catches, both players were very consistent all three seasons.  It also didn't have any effect on drops.

- AWR had a huge effect on YPC for both players, similar to that of the two RBs.  Notice how close the averages are for the RBs and WRs with the corresponding AWR attributes.  This should not be the case.  During the 2011 regular season.  ALL of the top 34 WR's in receptions averaged over 10 yards per reception.  During the 2011 regular season, only 6 of top 34 RB's in receptions averaged over 10 yards per reception.

- For simulation fans, realistic statistics are very important.  Due to the connection YPC and TD's have with awareness, realistic statistics will be hard to achieve when simulating games.  Rookies and role players tend to have very low AWR attributes in Madden, this will result in poor YPC statistics and therefore lower receiving yards.  If you look at the top rookie WR's in the NFL over the last several season, you will see that their average per reception is very close to that of veteran WRs.  In my opinion, AWR should not dictate the yards per catch when simulating a season in Madden.  With the current system, AWR clearly does dictate YPC.

-  Again, I will play some games with these same attribute settings and record the results.  To do that accurately, I will set the attributes the same for all WRs, TEs, and RBs.

99 AWR/0 PRC5923200010
99 AWR/0 PRC6817401110
99 AWR/0 PRC7627300020

MLB Observations:

- This is where things got really interesting.  This is actually the biggest thing that came out of this research.  AWR had no effect on the simulated stats for the MLBs.  There was little fluctuation in the stats when you compare the average against each of the three seasons.

- When I realized that AWR had NO effect on these defensive players,  I decided to take a quick look at the play recognition (PRC) attribute instead.  I did one additional season with 0 PRC and 99 AWR for each of the three defensive players.  This not only reduced their number of tackles and assists when looking at the average, it drastically reduced their interceptions and pass deflections.

-  With 0 PRC these three MLB's totaled 0 INTs and only 3 Pdefs.  Prior to adjusting the PRC to 0, the average season for these three players would have produced  3 Int's and 18.5 Pdefs.

-  So with AWR clearly not having an impact for defensive players in a simulated season, why even have that attribute for defensive players?  Right now AWR is the 2nd highest contributing attribute for a MLB's overall (OVR) rating, tied with PRC.  Why is AWR needed in this case and why does it contribute to the OVR when it doesn't appear to effect performance?

- As with the offensive players, I will be testing these attributes in real games that I play and blog about my findings.

99 AWR/0 PRC146000000
99 AWR/0 PRC255001000
99 AWR/0 PRC97000000

CB observations:

- As with the MLB's, AWR did not effect the simulated stats of the CBs when compared to the 3 season average.

- Even with 99 AWR, the number of pass deflections for CBs is WAY to low and unrealistic.  According to NFL.com, 75 defensive backs (safeties and corner backs) had 10 or more PDEF'S this season.  During the 2011 NFL regular season, Revis had 25, Webb had 25, and Haden had 19 PDEF's.

- PDEF is clearly one of the worst simulated stats in Madden 12 and is absolutely unrealistic for CBs.

- Once PRC was reduced to 0 for the CBs and awareness raised to 99, none of them intercepted a pass or had a PDEF.  The PRC had a huge impact on the number of tackles the CBs made as well.  I also  confirmed that each one of these CBs was still CB1 on their team's depth chart, just to be sure.  They all played from the CB1 position for all 16 games.

- So when you add up the 3 CB's and the 3 MLB's with 0 PRC, they had a grand total of ZERO interceptions and 3 Pdef's.

- While I drastically moved the PRC to zero, I think it's clear the PRC matters to defensive players (especially in pass defense) and AWR has no impact.

- I am extremely excited to play some games with my defense having all 99 AWR and 0 PRC, to see the effect on human gameplay.  I expect to find out that 0 AWR will not reduce the performance of my players, but that 0 PRC will.

- When I did my breakdown of attributes that effect a CB's overall (OVR) rating, I found that AWR accounted for 13% of the possible OVR points and PRC only accounted for 9%.  If AWR doesn't matter in gameplay for defensive players, why should it matter in OVR rating.


- Remember, I only changed AWR and PRC, all other attributes where untouched.

- I know these findings are only based on simulated seasons, but I think it's pointing me in the right direction for how much impact AWR and PRC have on player performance in Madden.

- I also understand that until I play several games with these AWR and PRC attribute settings, I am only speculating with many of my observations.

- In future roster updates for defensive players, I think it's important for all of us to monitor PRC and AWR.  I have noticed this season, that in many cases the AWR and PRC are rated the same or very close.  It's quite possible that PRC is the most important attribute for defensive players when a human is not in control of that specific player, therefore the accuracy of PRC is extremely important to game play.

- It may take me some time, but I will have a followup blog to address the gameplay impacts of these attribute settings.

- I do believe that EA is aware of this already, but as consumers (fans) of the game we do not have easy access to this type of information.

Thanks for following the blog and a great week.