Madden NFL 11 News Post

FBGRatings Seeks to Recalibrate Madden Ratings, Change Game

02:46 PM - January 24, 2011 by RaychelSnr

I recentely sat down for a talk with FBGRatings.com's Dan Berens to discuss his site's vision and what's going on over there today. The site is currently working on getting accurate ratings for every player using real hard data converted into the Madden ratings universe. Dan claims that when these numbers are plugged into the game, it plays much better and much closer to real life. Check out the interview below and also check out Dan's website to see what he's got going on!

Interview with Berens on the OS Radio Show on BlogTalkRadio

Game: Madden NFL 11Reader Score: 6/10 - Vote Now
Platform: PS3 / Xbox 360Votes for game: 96 - View All

Madden NFL 11 Videos

Member Comments

# 941 K33057 @ 12/06/12 08:40 PM

Great Stuff Great Info..
Tho for those who like classic teams would be nice if some could get together and come up with a system for old teams like the ratings FBG has come up with..

# 942 coleguyton @ 12/09/12 06:48 PM

Quote:

Originally Posted by K33057


	Great Stuff Great Info.. Tho for those who like classic teams would be nice if some could get together and come up with a system for old teams like the ratings FBG has come up with..

Exactly what I would like to find!!!

# 943 DCEBB2001 @ 12/22/12 12:22 PM

Just a little info from a response I filed in another thread:

-------------------------------------------------------------------------

Hey guys thanks for the feedback on the ratings. Just to clarify, there were a few reasons why there were not any rosters available for download.

The first was purely my error in that I never even knew there was an external editor. Now that I know, I should be able to pitch something to a roster editing team.

Second, the work for the site right now is on such a massive scale that I have not had personal time to do any roster creation. Thus, I have been attempting to get a team together that I can "outsource" the roster creation side to. That, however, takes time because of issues regarding non-disclosure agreements, licensing (we are run by TSX who has several copyrights), patents on how we rate players, and distribution. They have been ongoing for almost 3 years now. Basically if I can find a trustworthy team to create the rosters who are willing to sign the proper paperwork, this would have been in the bag months ago.

Third, there were constant complaints about the methodology regarding some attributes like SPD and ACC, which were the two big ones. All players in Madden accelerate until 45 yards in a sprint, then stop accelerating for a period of time based on their stamina. If the STA rating is at 99, they never tire out during a run, so a player will hold his maximum velocity until the end of the play. That maximum velocity is determined by the SPD rating while the acceleration to 45 yards is determined at first by the ACC rating and then a combination of the two attributes until 45 yards. Breaking down that combination and the effects of one another on each attributes takes a lot of time. Since every player has their own specific "velocity/acceleration curve" we can measure it out for real NFL players. However, this process takes a TON of time and I had to work with a local research university's physics department in order to get it right. It should be good now, however, for the new future updates coming out this offseason.

Fourth, EA changed their OVR formulas again. Breaking them takes a TON of time to get them correct. I didn't do that this year initially because editing wasn't even patched into the game right away and there was uncertainty if it even would be. I can do that this offseason as well since I will only be updating player ratings once per month. From there I just have to hope that EA doesn't change it again, which is like wishing for a new Benz under my xmas tree next week.

Fifth, people complained about the compatibility with EA's ratings system, especially for rookies you import into your franchise (CCM). The issue arises because the rookies are rated differently (and erroneously IMO) than that of players on the FBG site. So essentially you have a population of players rated with raw attributes being averaged out to 70 and all players being rated logically on the same scale (as in your position doesn't affect your SPD type of logic, which EA ignores). That either requires a guide to re-rate your rookies in franchise OR requires us to fit our square peg (FBG Ratings) into a round hole (Madden's ratings system and substructure).

Sixth, I needed more feedback from users. Offering suggestions was great but thus far nobody has offered a legitimate and publishable review of the ratings and how they work in-game. That would be very useful from those who said they liked the gameplay in this thread. It tells me what we are doing right, and what needs to be improved. I read EVERY email and PM I get. Most emails are from people saying that they like the site and want a roster for download or something to that effect, but rarely is valuable critique offered.

The best way for you to get FBG-rated rosters available for download is to help us best determine what users like and dislike about the system so we can have a more thorough methodology that gives the users what they want. We have cracked the SPD/ACC code for the game and are making ratings according to real-life velocty/acceleration curves. We are also updating our data distributions to better reflect the data we have.

WE ARE ENTERTAINING THE IDEA OF RATING PLAYERS MORE CLOSELY TO HOW THEY ARE RATED IN MADDEN...as to solve the fifth issue stated above, in order to make the ratings more compatible with that of the game. Once again, feedback can either deny or affirm this potential change, so please sound-off if you care one way or another.

Sorry for the novel but I want to dispel any rumors out there. Feel free to simply reply to this thread or the FBG thread. Emails or PMs are welcome as well.

Dan B.

# 944 slyfox @ 12/27/12 06:42 PM

Are these ratings available on a spreadsheet somewhere? That would save alot of time and paper just printing off each teams 'individual player ratings. As it stands now, you'd have to print each players's rating individually. I've tried making a cut and paste spreadsheet myself, but I just can't get it to match up correctly.....I'm not very talented with Excel. I've tried by hand but that takes way too long.

# 945 DCEBB2001 @ 12/27/12 11:24 PM

Quote:

Originally Posted by slyfox


	Are these ratings available on a spreadsheet somewhere? That would save alot of time and paper just printing off each teams 'individual player ratings. As it stands now, you'd have to print each players's rating individually. I've tried making a cut and paste spreadsheet myself, but I just can't get it to match up correctly.....I'm not very talented with Excel. I've tried by hand but that takes way too long.

Sorry, there are not any spreadsheets available.

# 946 jfb33 @ 07/22/13 08:49 PM

What's happening with this project?

# 947 DCEBB2001 @ 07/22/13 11:55 PM

Quote:

Originally Posted by jfb33


	What's happening with this project?

The project is still underway offline. I had to re-do a bunch of stuff this offseason and re-rating over 25000 players takes a lot of time. All players on current NFL rosters should be updated by the time all camps are under way in the next week or so...more likely by the first week of the preseason (excluding the HOF Game). You will notice many changes, include that of the STR rating. The data I have allowed me to better rate STR for players and the results are...well...shocking. There are a lot of OL who are not nearly as strong as you would think, and a lot of position players who are quite strong. Using legitimate Bench and Squat 1RMs has revealed a LOT about the college and NFL population of players. This will undoubtedly throw many of you off, but it is what the data dictates, and I am sticking with it.

The same goes for many other ratings. Using the real distribution given by the scouting data (instead of altering and fudging it to fit into Madden) has yielded some very interesting results. I won't go too far into details now because I want to wait for all the hate mail and "WTF?"s I am bound to get, but that being said I really think that this will be the most accurate batch of ratings to date.

I am also going to be adding more personal data like Hometown info. The goal is to one day use the site as a complete CAP site, where all info including build and equipment is included.

# 948 hanzsomehanz @ 07/23/13 06:02 PM

Quote:

Originally Posted by DCEBB2001


	Third, there were constant complaints about the methodology regarding some attributes like SPD and ACC, which were the two big ones. All players in Madden accelerate until 45 yards in a sprint, then stop accelerating for a period of time based on their stamina. If the STA rating is at 99, they never tire out during a run, so a player will hold his maximum velocity until the end of the play. That maximum velocity is determined by the SPD rating while the acceleration to 45 yards is determined at first by the ACC rating and then a combination of the two attributes until 45 yards. Breaking down that combination and the effects of one another on each attributes takes a lot of time. Since every player has their own specific "velocity/acceleration curve" we can measure it out for real NFL players. However, this process takes a TON of time and I had to work with a local research university's physics department in order to get it right. It should be good now, however, for the new future updates coming out this offseason.

Regarding the info extracted in italics: did you test the effects on various min. Speed threshold settings and fatigue settings?

I can relate to the theory of your conclusions if you are using the base enviornment standard of '50'. If so, the roster data would be molded to fit the '50' enviornment but of course not everyone prefere these settings.

Considering we can alter the speed and accel enviornments universally and globally, as well as stamina and fatigue: I am beginning to wonder if their is room for alteration available within your data pool and what significance that data will hold when people choose to edit sliders and game speeds to their preference.

On a footnote, no roster is perfect in how they rank - people can always fundamentally argue attributes.

I applaud your endearing efforts albeit I see no saving grace in the cause.

If you study the madden 13 opening release roster file you will see EA has actually done a fairly equitable job of balancing the spread of overalls.

My findings tell me that 75% of the 60 - 99ovr pool is represented by 60 - 79 rated players. The other 25% is represented by 80 - 99 rated players with only four rated 99 (Calvin, Ware, Rodgers, and Revis).

You can see from the data hierarchy that they or Moore, are using a pyramid structure:

41% rated 60-69
34% rated 70-79
18% rated 80-89
7% rated 90-99

You can see from the above that the population shrinks significantly as you exceed the average overall bubble of 70-72. You can also see the bottom 41% base is wide enough to support the top three levels.

The pyramid structure is a better representation of rank than a cylinder with an even spread of volume (60-99 / 4 = 25% on each level which is even distribution.)

*I really do not see the answer to be a wider spread in ratings but a wider spread in animation programming.

Sent from my SGH-I727R using Tapatalk 4 Beta

# 949 DCEBB2001 @ 07/23/13 08:14 PM

Quote:

Originally Posted by hanzsomehanz


	Regarding the info extracted in italics: did you test the effects on various min. Speed threshold settings and fatigue settings? I can relate to the theory of your conclusions if you are using the base enviornment standard of '50'. If so, the roster data would be molded to fit the '50' enviornment but of course not everyone prefere these settings. Considering we can alter the speed and accel enviornments universally and globally, as well as stamina and fatigue: I am beginning to wonder if their is room for alteration available within your data pool and what significance that data will hold when people choose to edit sliders and game speeds to their preference. On a footnote, no roster is perfect in how they rank - people can always fundamentally argue attributes. I applaud your endearing efforts albeit I see no saving grace in the cause. If you study the madden 13 opening release roster file you will see EA has actually done a fairly equitable job of balancing the spread of overalls. My findings tell me that 75% of the 60 - 99ovr pool is represented by 60 - 79 rated players. The other 25% is represented by 80 - 99 rated players with only four rated 99 (Calvin, Ware, Rodgers, and Revis). You can see from the data hierarchy that they or Moore, are using a pyramid structure: 41% rated 60-69 34% rated 70-79 18% rated 80-89 7% rated 90-99 You can see from the above that the population shrinks significantly as you exceed the average overall bubble of 70-72. You can also see the bottom 41% base is wide enough to support the top three levels. The pyramid structure is a better representation of rank than a cylinder with an even spread of volume (60-99 / 4 = 25% on each level which is even distribution.) *I really do not see the answer to be a wider spread in ratings but a wider spread in animation programming. Sent from my SGH-I727R using Tapatalk 4 Beta

The speed threshold has no affect on the fact that all players stop accelerating at the same point. You can measure this in a practice mode by recording the results and counting the frames given the frame rate you are using. This project is not meant to alter gameplay aside from using real NFL scouting data to posit the attributes and OVR ratings.

The data I have shows a very different scale for NFL talent. EA's opening roster for M13 shows a pyramid scale, correct, but that does not mean that the true population of NFL talent also fits into this scale. In fact, if I measure the data I have, you get the following breakdown based upon an equal interval (unbiased) where the maximum score is equal to the maximum rating possible (99) and the minimum score is equal to the lowest rating possible (12).

The number of players measured in the population is 57,851. The minimum score is -198 (due to injuries) and the maximum score is 1000. Here is the breakdown of the number of players in each group if this was interpolated into Madden OVR ratings:

90-99: 9
80-89: 50
70-79: 469
60-69: 1930
50-59: 10995
40-49: 26085
30-39: 18237
20-29: 73
12-19: 3

When you apply a rational function to these groups, the extremum lies where x=35. So this data would show that the average football player in this study including those of high school, college, and NFL calibers should be rated around 35 if you want to equally represent everyone in the data from the maximum to the minimum.

Now since Madden only includes NFL players currently on teams (as well as some top FAs), this is how the distribution looks for all players currently on NFL teams:

90-99: 7
80-89: 50
70-79: 426
60-69: 1001
50-59: 1261
40-49: 54
30-39: 2
20-29: 77
12-19: 0

The extremum lies where x=54 in this distribution with a rational function applied.

The data I have shows that the most NFL players would be rated circa 54 OVR, once again if you use an equal interval of distribution.

The scouting data shows that the players at the top are EXTREMELY rare. If the interval between every point is equal (the difference between 0 and 1 is worth the same as the difference between 999 and 1000) then that means that the scouting department that derived these numbers really feels that players rated a 90 and above in Madden are truly rare and special. Even all the players rated above 80 (57 of them) make up only 1.98% of the total players currently on teams (2878 players).

When you compare this to Madden 13's distribution of 2465 players, the differences are quite stunning:

90-99: 169
80-89: 413
70-79: 870
60-69: 792
50-59: 198
40-49: 19
30-39: 2
20-29: 2
12-19: 0

Plotting these groups yields an extremum where x=65 and 6.85% of the population is rated above 90 while 23.61% of all the players are rated over 80.

It really comes down to what you prefer...a roster with almost 1/4 of the players rated 80+ or a roster where 1/50 are rated 80+. Now, you all know what I prefer, but keep in mind that the scouting data also shows the large differences between elite players and average players, while Madden tends to bunch them all together rather closely.

# 950 Se7eN Purposes @ 07/24/13 06:03 AM

Quote:

Originally Posted by DCEBB2001


	The speed threshold has no affect on the fact that all players stop accelerating at the same point. You can measure this in a practice mode by recording the results and counting the frames given the frame rate you are using. This project is not meant to alter gameplay aside from using real NFL scouting data to posit the attributes and OVR ratings. The data I have shows a very different scale for NFL talent. EA's opening roster for M13 shows a pyramid scale, correct, but that does not mean that the true population of NFL talent also fits into this scale. In fact, if I measure the data I have, you get the following breakdown based upon an equal interval (unbiased) where the maximum score is equal to the maximum rating possible (99) and the minimum score is equal to the lowest rating possible (12). The number of players measured in the population is 57,851. The minimum score is -198 (due to injuries) and the maximum score is 1000. Here is the breakdown of the number of players in each group if this was interpolated into Madden OVR ratings: 90-99: 9 80-89: 50 70-79: 469 60-69: 1930 50-59: 10995 40-49: 26085 30-39: 18237 20-29: 73 12-19: 3 When you apply a rational function to these groups, the extremum lies where x=35. So this data would show that the average football player in this study including those of high school, college, and NFL calibers should be rated around 35 if you want to equally represent everyone in the data from the maximum to the minimum. Now since Madden only includes NFL players currently on teams (as well as some top FAs), this is how the distribution looks for all players currently on NFL teams: 90-99: 7 80-89: 50 70-79: 426 60-69: 1001 50-59: 1261 40-49: 54 30-39: 2 20-29: 77 12-19: 0 The extremum lies where x=54 in this distribution with a rational function applied. The data I have shows that the most NFL players would be rated circa 54 OVR, once again if you use an equal interval of distribution. The scouting data shows that the players at the top are EXTREMELY rare. If the interval between every point is equal (the difference between 0 and 1 is worth the same as the difference between 999 and 1000) then that means that the scouting department that derived these numbers really feels that players rated a 90 and above in Madden are truly rare and special. Even all the players rated above 80 (57 of them) make up only 1.98% of the total players currently on teams (2878 players). When you compare this to Madden 13's distribution of 2465 players, the differences are quite stunning: 90-99: 169 80-89: 413 70-79: 870 60-69: 792 50-59: 198 40-49: 19 30-39: 2 20-29: 2 12-19: 0 Plotting these groups yields an extremum where x=65 and 6.85% of the population is rated above 90 while 23.61% of all the players are rated over 80. It really comes down to what you prefer...a roster with almost 1/4 of the players rated 80+ or a roster where 1/50 are rated 80+. Now, you all know what I prefer, but keep in mind that the scouting data also shows the large differences between elite players and average players, while Madden tends to bunch them all together rather closely.

Is it possible for you to release a roster say, once every week or so which has all the updated ratings compiled within? Very interested and I couldn't find a roster on your site.

# 951 Robcards @ 07/24/13 08:24 AM

Didn't read through the 100 pages of this thread but I get the feeling these ratings will be drastically lowering current players ratings? If so, what would this roster be used for? Clearly it couldn't be used for franchise mode past year 1 as incoming rookies would be insanely better than current players.

# 952 DCEBB2001 @ 07/24/13 08:49 AM

Quote:

Originally Posted by Se7eN Purposes


	Is it possible for you to release a roster say, once every week or so which has all the updated ratings compiled within? Very interested and I couldn't find a roster on your site.

We currently do not produce rosters for download.

# 953 DCEBB2001 @ 07/24/13 08:50 AM

Quote:

Originally Posted by Robcards


	Didn't read through the 100 pages of this thread but I get the feeling these ratings will be drastically lowering current players ratings? If so, what would this roster be used for? Clearly it couldn't be used for franchise mode past year 1 as incoming rookies would be insanely better than current players.

That is correct. It wouldn't prove fruitful to use these ratings for franchise mode. This project is used to merely rate players in Madden using real NFL scouting data. What you do with them after that is up to you. If we were able to edit draft classes like we used to be able to do on the PC versions, then of course they would be more useful.

# 954 Find_the_Door @ 07/24/13 01:18 PM

Quote:

Originally Posted by DCEBB2001


	We currently do not produce rosters for download.

Will you be now that Madden will have a roster share feature?

# 955 DCEBB2001 @ 07/24/13 07:20 PM

Quote:

Originally Posted by Batum Shaka Laka


	Will you be now that Madden will have a roster share feature?

I have not expecting to this year.

# 956 Find_the_Door @ 07/24/13 07:30 PM

Quote:

Originally Posted by DCEBB2001


	I have not expecting to this year.

Was there any reason in particular as to why you won't?

# 957 DCEBB2001 @ 07/24/13 08:48 PM

Quote:

Originally Posted by Batum Shaka Laka


	Was there any reason in particular as to why you won't?

Running the website dedicates a lot of time whereas creating and editing rosters would simply be too much.

# 958 Dj_MyTime @ 07/25/13 02:30 PM

Just looking over the rosters/ratings etc...either I'm missing something or I just do not agree with how slow all the players seem to be...then Steve Smith on CAR 94 trk, 99stf, and a couple other 99's in the moves category....

He's my favorite player and I watch every game but even I wouldn't take it that far as a homer. 60"s spec catch/cit, 89 spd???

Again I'm not picking they just seem very unusual...

Edit-Ted Ginn Jr. w/ 84 trucking at 5'11 180...what exactly do you use to determine trucking, it seems really high. I could see it being higher for Smith as hes one of the more physical receivers, but Ginn has never been that kind of physical player...

# 959 Dj_MyTime @ 07/25/13 02:34 PM

Quote:

Originally Posted by DCEBB2001


	Running the website dedicates a lot of time whereas creating and editing rosters would simply be too much.

Your site seems popular enough, youd' wrather tinker around on a site instead of actually having the masses play with these adjusted ratinngs when we finally have roster share to see if they improve gameplay or not...?

Seems a little strange you wouldnt be excited and want to submit your work, but you'll fluff it online???

# 960 DCEBB2001 @ 07/25/13 06:57 PM

Quote:

Originally Posted by Dj_MyTime


	Just looking over the rosters/ratings etc...either I'm missing something or I just do not agree with how slow all the players seem to be...then Steve Smith on CAR 94 trk, 99stf, and a couple other 99's in the moves category.... He's my favorite player and I watch every game but even I wouldn't take it that far as a homer. 60"s spec catch/cit, 89 spd??? Again I'm not picking they just seem very unusual... Edit-Ted Ginn Jr. w/ 84 trucking at 5'11 180...what exactly do you use to determine trucking, it seems really high. I could see it being higher for Smith as hes one of the more physical receivers, but Ginn has never been that kind of physical player...

All of the ratings currently posted were last updated in December using an old system. The new ratings, once posted, will be even more foreign to you, I'm sure.

For Smith, take his split times, create a cubic function, differentiate, and then compare it to the rest of the population. His new speed will be around 86 with an ACC of 85. So yes, these are very different from what Madden has been feeding us for years (players in the 90s in just about every category).

Ted Ginn - his R-value (given by the scouts to determine the number of tackles "broken or slammed") was 2.3 out of a maximum 3.4 for any WR. His new TRK rating will be around 67 with an ELU of 80 and a SFA of 63. Once again, a bit different from what Madden publishes.

« Previous 1 2…45 46 474849 50 51…76 77 Next »

More Links

FBGRatings Seeks to Recalibrate Madden Ratings, Change Game