FBGRatings Seeks to Recalibrate Madden Ratings, Change Game

[DCEBB2001]

Quote: Originally Posted by buckey00

That is very cool. So Rice's strength will allow him to break more arm tackles while mccoy will be able to finish runs more effectively? I have noticed ea has so little variation in the ratings and they are very inaccurate. Such as Arian Foster runs a 4.78 but he has speed in the 90's while Trent Richardson runs a 4.45 and has a lower speed. And Collin Kaepernick feels allmost the same as using Aaron Rodgers. I cant wait to use these after the patch comes out.

There are certain ratings that do not work alone. When running, the SPD and ACC ratings do work symbiotically up until 45 yards. At 45 yards, the SPD rating takes over until the STA determines how long the maximum velocity is effective. For SPM/JKM moves, the AGI and ACC ratings play a very large role in getting into and out of the move effectively. For instance, a RB with a JKM of 90 with ACC of 90 will get out of the move quicker than a RB with a JKM of 90 and ACC of 80. So when you look at the ratings on the site, don't sinle out any one attribute. Instead, look at how they may affect one another. The only "true" way to see how each player performs is to actually try it out.

I do not have a timetable for how long it will take to do the OVR calculations for M13 presently. It really depends on how much we CAN edit for CCM. If EA allows us to edit everything in a roster before starting a CCM, then it would be worth all the time it takes to get the calculations correct. If we can't edit as much as we would like, then I may have to skip that for this year and hope that full editing is back next year. I would also like to see if we can edit in CCM, especially draft classes.

The one thing I am constantly fighting is the urge to simply rate players the same way EA does, but with the better info. For instance, all attributes have a designated "average" rating as well as a max and min. This average for all raw attributes is 70. So an "average" vertical jump of 34.5 inches (the average since 1998) would yield a JMP rating of 70. However, in Madden the average JMP is 68. The average SPD is 74, ACC is 81, and AGI is 74, and STR is 71. So as you can see, the average would change if I were to rate players the way EA does. However, I would still utilize a normal distribution based on the data.

I would love to hear from some of the guys who are regulars to this thread about this in detail. The upside to using the same way EA rates players is that you can use rookies produced in CCM. The downside is that you get more ratings inflation toward the upper bound. Thoughts?

[PGaither84]

I assume you plug these ratings into Madden and actually test the game play. After all, the title of this thread is "Change Game."

Well, I was wondering about things like Zone Coverage. I am still playing Madden 12 and I use custom sliders, and even with 90-100 pass reaction time for the computer, the zone coverage by line backers was still quite lacking for my tastes. For example, I could call a simple cover 3 blitz with the players in yellow hook zones across the middle, and a LB with 64 ZCV would roll over to the wrong zone leaving his area uncovered. I also find many times when I have both MLBs assigned to hook zones over the middle, they often like to split wide leaving the middle WIDE open, an area that was supposed to be on lock down. There are countless other examples on how they cover flats or buzz zones as well [or don't for that matter.

Well, I decided to take a look at the roster of Madden 12 and found that the highest ZCV rating of any line backer was Brian Urlacher with 85, and it all went down hill from there. There were only 9 line backers in the entire game with 80 or higher ZCV ratings, and of those only 3 or 4 were over the 80 mark, and one of them was in the free agent pool.

So, the first thing I decided to do was edit all of the 49ers linebackers to 99 ZCV [not save] and go into practice and see how well they did their job when maxed out. Well, while they didn't always do a perfect job [programming] they did a very good job of covering their assigned area and locking down routes/receiver in those zones. Part of what I wanted to test is if the programming of the game itself was even set up to do it's job correctly, which it should with a rating of 99. Anything lower than 99 and per-programed random mistakes creep in more and more. Players start to become less perfect.

So, i reloaded the roster and began an NFL wide roster edit. I chose to edit ever line backer who had a ZCV rating of 65 or higher and give them all a universal +10 to their ZCV rating. I have found over the years in Madden that 70-75 is about the cut off of effectiveness of Madden positional ratings before it doesn't matter how low your rating is. For example, the QB Cone from Madden 06 was the same size for about a 70-75 Awareness QB as it was for a 40 awareness QB. It wasn't until you started to go above 75 that the cone would actually get bigger. There are other examples of this, but the real point is that in Madden, if you want your rating to matter, it had better be over 75. Also, what this really means is that if you had below 65 ZCV [in my opinion] you didn't really need that 10 point boost. You were never going to be a coverage line backer anyway. That's how/why I set that cut off point so I wouldn't waste my time editing everyone.

Well, fast forward a bit and I have really enjoyed the results of my efforts. Playing against he computer is more difficult [and realistic] as players DO THEIR JOBS. I can't over stress how important that is to me. They aren't perfect out there, but they look a lot less like high school kids and look more like pros out there.

Now, there is a separate talk about how high a rating should be, and what a 90+ rating means, but that isn't what I want to get into just yet. I just wanted to address how the ratings on your site actually impact game play in Madden. I find the incredibly low ZCV ratings I see on your site to be detrimental to a simulation experience. Weather or not I agree with how low those ratings are is irrelevant.

[RogueHominid]

Quote: Originally Posted by PGaither84

Now, there is a separate talk about how high a rating should be, and what a 90+ rating means, but that isn't what I want to get into just yet. I just wanted to address how the ratings on your site actually impact game play in Madden. I find the incredibly low ZCV ratings I see on your site to be detrimental to a simulation experience. Weather or not I agree with how low those ratings are is irrelevant.

This is the basic philosophical and practical issue here. I think it's interesting to see what values the secret scouting data produces, but Madden is a game where the gameplay hinges on animation triggering, and the ratings values at which new animations trigger bears no relation to real life scouting data about various aspects of gameplay.

Playmakers and Silent Nature and maybe a guy named Purple Haze (I think--it's been a long time) had actually worked these values out quite specifically. For example, a RB Vision value of 90 + would generate smart cutbacks; a truck move over 95-ish would generate CPU trucking animations that you'd never see otherwise. The same applies to WR/TE CTCH, and so on down the line.

I would ultimately like to see a point where players' actual physical data governed the outcome, but in order for that to happen, we'd have to have a 100% physics-driven game, I think, rather than an animation-driven game.

Perhaps hiring a lobbyist to pester EA Tiburon--or at least bring them Krispy Cremes every morning--would help bring change.

[DCEBB2001]

Quote: Originally Posted by PGaither84

I assume you plug these ratings into Madden and actually test the game play. After all, the title of this thread is "Change Game." Well, I was wondering about things like Zone Coverage. I am still playing Madden 12 and I use custom sliders, and even with 90-100 pass reaction time for the computer, the zone coverage by line backers was still quite lacking for my tastes. For example, I could call a simple cover 3 blitz with the players in yellow hook zones across the middle, and a LB with 64 ZCV would roll over to the wrong zone leaving his area uncovered. I also find many times when I have both MLBs assigned to hook zones over the middle, they often like to split wide leaving the middle WIDE open, an area that was supposed to be on lock down. There are countless other examples on how they cover flats or buzz zones as well [or don't for that matter. Well, I decided to take a look at the roster of Madden 12 and found that the highest ZCV rating of any line backer was Brian Urlacher with 85, and it all went down hill from there. There were only 9 line backers in the entire game with 80 or higher ZCV ratings, and of those only 3 or 4 were over the 80 mark, and one of them was in the free agent pool. So, the first thing I decided to do was edit all of the 49ers linebackers to 99 ZCV [not save] and go into practice and see how well they did their job when maxed out. Well, while they didn't always do a perfect job [programming] they did a very good job of covering their assigned area and locking down routes/receiver in those zones. Part of what I wanted to test is if the programming of the game itself was even set up to do it's job correctly, which it should with a rating of 99. Anything lower than 99 and per-programed random mistakes creep in more and more. Players start to become less perfect. So, i reloaded the roster and began an NFL wide roster edit. I chose to edit ever line backer who had a ZCV rating of 65 or higher and give them all a universal +10 to their ZCV rating. I have found over the years in Madden that 70-75 is about the cut off of effectiveness of Madden positional ratings before it doesn't matter how low your rating is. For example, the QB Cone from Madden 06 was the same size for about a 70-75 Awareness QB as it was for a 40 awareness QB. It wasn't until you started to go above 75 that the cone would actually get bigger. There are other examples of this, but the real point is that in Madden, if you want your rating to matter, it had better be over 75. Also, what this really means is that if you had below 65 ZCV [in my opinion] you didn't really need that 10 point boost. You were never going to be a coverage line backer anyway. That's how/why I set that cut off point so I wouldn't waste my time editing everyone. Well, fast forward a bit and I have really enjoyed the results of my efforts. Playing against he computer is more difficult [and realistic] as players DO THEIR JOBS. I can't over stress how important that is to me. They aren't perfect out there, but they look a lot less like high school kids and look more like pros out there. Now, there is a separate talk about how high a rating should be, and what a 90+ rating means, but that isn't what I want to get into just yet. I just wanted to address how the ratings on your site actually impact game play in Madden. I find the incredibly low ZCV ratings I see on your site to be detrimental to a simulation experience. Weather or not I agree with how low those ratings are is irrelevant.

Great post.

I have also found this to be rather troublesome with the OL/DL interactions, specifically with PMV/FMV rushes. Unfortunately, EA does have a weird way of making anything below a numeric average worthless. However, the data that I am privy to dictates that the variation in a skill like ZCV for a LB is far wider than the effective range that EA may provide us with (99-75) as you pointed out in your post.

This is another issue I take with EA. They simply don't utilize the full scale of ratings. Now I can stretch them all out as far as I like from a positional average toward the upper and lower bounds, but if they don't work in the game they are meaningless. However, I decided to stick to my guns and the data distribution in the hope that EA would change the effectiveness of their ratings range for M13. I don't have any evidence at this point to show that it has been improved but it is worth looking into.

The more we find out about how the ratings interact with the game engine the more disappointed I get with EA. The interaction is so broken that it seems almost pointless at times to even try to get around it. However, I still strongly believe that providing a new way to rate players with a wider distribution and showing EA that it is important to do so may accomplish something. Idealy, if I could find a way to get into the lion's den and push for a better process that can better marry the in-game engine to the underlying data, the game could make a change for the better. But until that happens we are stuck dreaming of a way to better represent the abilities of NFL players while stuck in a box that is far too small.

[DCEBB2001]

Quote: Originally Posted by JMP

This is the basic philosophical and practical issue here. I think it's interesting to see what values the secret scouting data produces, but Madden is a game where the gameplay hinges on animation triggering, and the ratings values at which new animations trigger bears no relation to real life scouting data about various aspects of gameplay. Playmakers and Silent Nature and maybe a guy named Purple Haze (I think--it's been a long time) had actually worked these values out quite specifically. For example, a RB Vision value of 90 + would generate smart cutbacks; a truck move over 95-ish would generate CPU trucking animations that you'd never see otherwise. The same applies to WR/TE CTCH, and so on down the line. I would ultimately like to see a point where players' actual physical data governed the outcome, but in order for that to happen, we'd have to have a 100% physics-driven game, I think, rather than an animation-driven game. Perhaps hiring a lobbyist to pester EA Tiburon--or at least bring them Krispy Cremes every morning--would help bring change.

Another good post. It is so sad to me that things are so driven by animations that are only triggered at certain intervals. If we did have a truly physics-induced game then the ratings would only enhance the given physical ability instead of triggering a differnt animation like we have at present. At this point all I can offer is one perspective and a voice for how Madden ratings would look if given a real-world/real-data approach.

This has it's limits as pointed out previously, however. Aside from some attributes that are definetly determining factors in in-game performance (like SPD, which seems to be the ONLY one at times) we are left with animation ranges for the time being. If I had a larger team that could analyze every single animation every single year perhaps we could break it down and effectively marry the data to the animations. But that seems pointless with as many "changes" EA brings to the game every year.

[PGaither84]

Quote: Originally Posted by JMP

This is the basic philosophical and practical issue here. I think it's interesting to see what values the secret scouting data produces, but Madden is a game where the gameplay hinges on animation triggering, and the ratings values at which new animations trigger bears no relation to real life scouting data about various aspects of gameplay. Playmakers and Silent Nature and maybe a guy named Purple Haze (I think--it's been a long time) had actually worked these values out quite specifically. For example, a RB Vision value of 90 + would generate smart cutbacks; a truck move over 95-ish would generate CPU trucking animations that you'd never see otherwise. The same applies to WR/TE CTCH, and so on down the line. I would ultimately like to see a point where players' actual physical data governed the outcome, but in order for that to happen, we'd have to have a 100% physics-driven game, I think, rather than an animation-driven game. Perhaps hiring a lobbyist to pester EA Tiburon--or at least bring them Krispy Cremes every morning--would help bring change.

I love this reply!

There is a difference between what we all think players "should be rated," and how those ratings work in the game.

Sliders are often times just as important in adjusting when those animations trigger. For example, I can have 20 user break tackle and 30 computer tackle, and even with a +10 advantage to the computer, the user can break a lot of tackles without even trying. I had a game where Alex Smith broke 3 tackle animations late in a game with those early slider settings [that I no longer use] by simply pressing up on the left stick. I didn't press stiff arm or anything. However, with much different settings, I was unable to break a single tackle, not even with a back like Adrian Peterson against CBs with low tackle/strength ratings.

We can sit here and say "With this data, this player should be rated X." However, if you plug that into the game, I often doubt that you will get realistic results that will mirror that player you spent time editing.

[DCEBB2001]

Quote: Originally Posted by PGaither84

We can sit here and say "With this data, this player should be rated X." However, if you plug that into the game, I often doubt that you will get realistic results that will mirror that player you spent time editing.

Well it depends on the animation. Some are far more effective to trigger than others, which is another odd thing...the fact that they are not triggered at the same rate is troublesome and tedious to determine.

SPD for instance is far easier to manipulate and determine the animations by increasing/decreasing the attribute value as opposed to something like BSH which sometimes seems random no matter what you do to it.

I still enjoy seeing how the players should be rated with the data I have at present, but I have fully understood from day one that the game engine will limit the effectiveness in the game.

[RogueHominid]

Yeah, there's an open source build your own football game site that's working with collaborative designers to customize a full 3D, RTF experience, but it takes a lot of money to do that independently. That kind of scenario would be the ideal place to input data like this since the premise is to build a customized experience where physics govern outcomes.

With Madden you have to hope that there's some kind of reasonable correspondence between external data and the game's internal logic.

I did some messing around last year and could tell that, whatever the specific rating for a WR CATCH was, a slider value of 75 or higher unlocked the animation set necessary to get WR to compete for the ball at the point. It's a very strange thing.

To the point of the project, though, I think it would be great to see the difference between players like a Richardson, who runs with power, often stiff arming and using massive leg drive, vs. someone like Foster, whose runs are defined by that 30 yard burst ability and great vision. I like the idea of long striders who will pass you at 15 yards and dust you vs. those who will get on top of you right at 15 yards but will have maxed out their speed and so won't win the long race. Lots of other examples, but Madden does need to differentiate players better, no doubt.