CFM Rookies Not Adjusted To Stock Player Scale (EA Is Investigating)

[Hooe]

Re: CFM rookies not adjusted to Stock Player Scale

Great post above from Hooe! Now in the time you have done this and spotted this issue why oh why have EA not. I am very positive about there product this year very. But really they could of spotted this by doing what you have as a simple test baffling.

Personally, I'm inclined to give the developer the benefit of the doubt based on my first-hand experience making games. Video game development projects have a ton of moving pieces, and in a game as huge as Madden NFL it's easy for a creative team, QA team, and everyone to miss something like a minor balancing issue.

And let's be honest, that's what this is, a minor issue; a major issue would be something profoundly egregious like every generated QB having THP 0, or offensive linemen routinely entering the league with SPD 80 or greater. Personally I would call the extreme SPD regression behavior from last year's game far worse balancing issue than what we're seeing here.

It's also possible (albeit unlikely, in my outside opinion) that this piece of the game is functioning as Tiburon intended when they built it.

[msdm27]

Re: CFM rookies not adjusted to Stock Player Scale

So it turns out Madden NFL 16 simulates seasons in cloud CFM really quickly. As such I've already got some data for you all, only took about an hour total. It's taken me longer to type this post than it took to gather the data, lol.

I focused on two ratings that have come up a lot at the start of this thread - WR SPD and QB DAC. I also added in WR ACC for good measure.

For these three position-rating combinations, here's some useful statistical info for them (these numbers computed using the ratings from the Madden NFL 16 base roster spreadsheet EA released on their web site):

RATING    μ        σ      μ-σ    μ+σ
WR SPD    88.2     3.5    84.7   91.7
WR ACC    89.9     2.4    87.5   92.3
QB DAC    63.6     5.9    57.7   69.4

A quick brush-up of terms: μ is the average (also the mean), σ is the standard deviation. μ-σ and μ+σ are numbers each one standard deviation away from the mean; it's expected that about two thirds of the ratings used to calculate μ and σ will fall between these two extremes, assuming that the ratings fall over a normal distribution.

Having these numbers in hand, I created a cloud Connected Franchise, playing as a coach, turned off coach firings, and simulated ten seasons twice in a row. In year 2035 of this CFM, there are the same computations for the same ratings:

RATING    μ        σ      μ-σ    μ+σ
WR SPD    91.3     4.6    86.7   95.9
WR ACC    86.7     5.8    80.9   92.5
QB DAC    75.6     5.8    69.8   81.4

And finally, I subtract the original 2015 numbers from the 2035 numbers to visualize the change:

RATING    Δμ      Δσ     Δ(μ-σ)    Δ(μ+σ)
WR SPD    3.1     1.1    2.6       4.2
WR ACC   -3.2     3.4   -6.6       0.2
QB DAC    12.0   -0.1    12.1      12

Plain-text English explanation of what's happening / what these numbers mean:

After a twenty-year period of CFM simulation in this game save, the average _top speed_ of all wide receivers in the game increased marginally. The standard deviation also rose, indicating a wider variety of speed ratings amongst all the players. This is pretty evident at the top of the scale - there are 29 WRs with 97+ SPD.

Simultaneously, in the same 20-year simulation, the average ACC rating for receivers fell marginally, and virtually the same amount as SPD rose. The game still hands out a good number of high ratings here - there are fourteen WRs with 97+ ACC - but also more low ratings, relatively speaking - there are 71 WRs with ACC at 83 or less, while there are exactly three such players on the default Madden NFL 16 roster. It also appears that 80 is the lowest rating the game will hand out to a generated rookie at WR, and there are no CPU-rostered players with stupidly low SPD or ACC ratings, as would happen in Madden NFL 15.

In effect, what you have going on at the wide receiver position over time is that the population of available receivers indeed has a higher top speed and straight-line speed, but less short-area athleticism. I probably need to do this analysis for WR AGI in order to capture the full picture here, but honestly I don't think this incredibly offensive. I imagine that people are going to be upset over the greater frequency of higher speed ratings, but I see it as an increase in variance with respect to the caliber of available athletes at the position.

QB DAC is a different story, this is something I think Tiburon should consider adjusting if consistency with the base roster is something they feel is important to CFM. That rating is indeed rising dramatically over time. 29 passers have DAC 80 or greater (there is one QB who tops 90 DAC). Compared to the default roster which has 19 players with a DAC of 70 or greater and only two players who crack 80 (Aaron Rodgers and Matt Ryan). The floor for the players with the worst DAC ratings also rises an alarming level; while there are 30 players on the default roster with a DAC rating of 59 or less and 101 with a DAC rating of 69 or less, there is not a single rostered QB in the 2035 league with a DAC rating less than 64 and only 16 with a DAC rating of 69 or less.

One caveat - this doesn't identify where the problem is. I paid no attention to rookie classes, I literally hit the "sim ten years" option twice. It's possible - though I'd say unlikely, given how much the minimum rating is rising - that the rise in QB DAC over time is caused by an imbalance in upgrade package XP cost for that rating.

Anyway, if anyone wants to see any more ratings analyses like these, I'll keep this CFM file around and can probably take a small handful of requests for additional ratings to look at if y'all would like. I imagine that the CFM servers will get wiped at official launch on Tuesday, so bear that in mind.

All this said, I'm personally still looking forward to playing CFM, I always have fun playing it and I don't really find anything in here particularly offensive under Madden NFL's current player ratings philosophy. I do hope this post finds its way to someone at Tiburon and provides them with one data point which could be helpful in their making any decisions with respect to CFM rookie generation / player progression balance adjustments.

Thanks a lot, CM!
Would you mind, if you have time, to delve a little bit into SPD for DT's? Lots of players in M15 had 85+ SPD at this position and were generally lightweight players, around 280lbs (granted that is more cosmetic)

Also, if you really have a lot of time to test, maybe you could choose 5 current players (skill positions) between 28 and 31 years old and see how their physical attributes decrease after simming 3-5 seasons?

[Hooe]

Re: CFM rookies not adjusted to Stock Player Scale

Thanks a lot, CM!
Would you mind, if you have time, to delve a little bit into SPD for DT's? Lots of players in M15 had 85+ SPD at this position and were generally lightweight players, around 280lbs (granted that is more cosmetic)

Also, if you really have a lot of time to test, maybe you could choose 5 current players (skill positions) between 28 and 31 years old and see how their physical attributes decrease after simming 3-5 seasons?

I'll look at DT SPD ratings tomorrow, sure thing. Not sure I'll be able to get to the second request, I'll try but I make no promises.

[AE Spits Fiya]

Re: CFM rookies not adjusted to Stock Player Scale

We have to remember that there was no Randy Moss before he got drafted either. His physical tools were unheard off at the time.
What about Calvin Johnson, Adrian Peterson, Clowney etc.


In Madden 13 or 25 (can't remember) we saw JJ Smoke with 99 SPD, 99 ACC, 99 AGI - I'll never forget him. We saw Jason Merian with 96 SPD as a QB. Unique players that are once in a lifetime. They HAVE to happen. BUT the problem last year was that you could find them in round 3-7. THAT shouldn't happen. They should automatically be a top 5 (or first round) prospect based on the physical tools.


I agree that there shouldn't be 10 per draft. Maybe not even 1 per draft. But they should be there.

[lowndsy]

Re: CFM rookies not adjusted to Stock Player Scale

Thanks CM Hooe I am giving them the benefit of doubt the game looks awesome and as a CFM only player it is an issue for sure. I am just saying I think any sim player had been invited to test with Madden early it may have noticed rather than just testing button mashers who show us aggressive catch every pass.
I hope I don't sound like a downer i am not trying to be can't wait for British release date. Just think maybe could of been avoided my 2pence worth.

[DeuceDouglas]

Re: CFM rookies not adjusted to Stock Player Scale

I think being that the initial worry was about CFM rookies being adjusted to the stock players, we should be looking the rookies when they come in versus after they've had time to progress. This very well could turn out to be an XP or progression issue more so than a draft class issue. I'd be interested to see something like the top 10 rookies in DAC as they're drafted in like 5-10 draft classes just to see if they come in too high or if they progress at a rate that allows for this. There is definitely some kind of issue there though.

[AE Spits Fiya]

Re: CFM rookies not adjusted to Stock Player Scale

AND some of you who are frustrated. I completely feel you.
Each year they announce some HUGE FEATURE to blindfold people from all the small balancing fixes they don't do.


Anyone know Football Manager? Their franchise is based solely on balancing a great idea to begin with. They don't implement a lot of new stuff. They just make it BETTER, BALANCED, GREATER in every way every year.


I remember when Vontaze Burfict won DPotY and in his Awards tab it said "Offensive player of the year" - still does. How can that not be fixed yet? That is so insanely easy/stupid.


The drafted players from real life are 70-72 ovr already in round 2. In Madden you find 85+ guys still in round 4. All backup players on stock rosters are useless after the first season. That sucks too. Been that for years.


Relocation, popcorn prizes etc. I could care less. Stop implementing new stuff, when old stuff doesn't Work right. Focus on the Things you already got in the game. Make them work. Some people will complain that they buy a "roster update". Let them. Use a year for balancing out issues.

[EricFreakingBerry]

Re: CFM rookies not adjusted to Stock Player Scale

It also appears that 80 is the lowest rating the game will hand out to a generated rookie at WR, and there are no CPU-rostered players with stupidly low SPD or ACC ratings, as would happen in Madden NFL 15.

Thanks for your testing Hooe, I found this aspect enlightening and informing based on their claims to fixing regression. I think the numbers are definitely higher (WR's not as alarming as the QB's) but it's nice to see that there may be a different regression system in place.

I would enjoy seeing some testing done on the ages/retirements in the coming few days if you have time. If not, no big deal - I'll have the game Tuesday, I'm sure.

[Stridah]

Re: CFM rookies not adjusted to Stock Player Scale

I am a little worried that even with hope's data, you have 40 year old wr's with 75 speed driving the averages down.

15+ Rookies a year with 95 speed is alot, i think the rosters need to be looked at a little more in depth after the sim.


Also hope, it might be a good idea to look at every position, because i have the feeling athletic ratings as a whole may be broken.

[howboutdat]

Re: CFM rookies not adjusted to Stock Player Scale

I just did a test run of a CFM draft.. In total, 16 WR's had 95 speed or higher... Of those 16, 1 had 99, 5 had 98, and 6 had 97


The current updated stock roster has 10 total WR's 95 speed or higher for the entire NFL

www.techraptor.net

Ok so right there , it seems pretty evident there are 6 more wrs with as much or higher speed than the tops stock wrs. In 1 draft.

In effect, what you have going on at the wide receiver position over time is that the population of available receivers indeed has a higher top speed and straight-line speed, but less short-area athleticism. I probably need to do this analysis for WR AGI in order to capture the full picture here, but honestly I don't think this incredibly offensive. I imagine that people are going to be upset over the greater frequency of higher speed ratings, but I see it as an increase in variance with respect to the caliber of available athletes at the position.

First up , thanks alot for running this test! Great job! I quoted the above to point out ,your correct. People are still going to be upset by the part in bold. Because at the end of the day , that was the very issue people was noticing and raising. I noticed you stated there were 27 WRs with 97 or greater speed after 10 drafts. There are only 2 WRS in the stock rosters even at 97 speed.So what to me still looks wrong here is those numbers themselves. In the current NFL there are only 2 wrs that fast. This number is due to the current past 10 drafts of the real NFL. 2 WRS that fast. Fast forward 10 years and there is a total of 27 WRs that fast. I mean, to me, that sounds like a pretty large jump if one looks at it in comparison to real life , which last i checked is what we are trying to get them to mock here.If it had only been say 5 -7 total , i might would agree, its a marginal difference. But 27 , compared to 2 .....seems rather high to me .

Here is my question on this and if someone who has the game would like to take the time to check this , i think it would be interesting:

Stock rosters has :

WRs-2- 97 SPD
WRs- 7 -96 SPD
WRs- 7-95 SPD
WRs- 10 -94 SPD

CBs- 4 - 95 SPD
CBs-6 - 94 SPD

After ten years , i wonder how much these numbers change . I may be able to get a run on this by tomorrow, not sure, i dont have the game but may have a way to get this info. If so ill post back when i do. If someone else wants to check it ,feel free. Im just curious .

As for the QB deep acc, there seems to be not even room to argue its not "that bad" . Thankful to the community for catching it and bringing it up.As well we should be thankful to Rex and Kolbel for communicating with us they have seen our info and are looking into it. Hate EA or not, thats also a step in the right direction, hard for anyone to deny that.

[bad_philanthropy]

Re: CFM rookies not adjusted to Stock Player Scale

Thanks CM. Posts like your are vintage OS, and what has made this place great over the years.

Next time I expect you to plug every rookie class into SPSS and provide a complete report. Only joking, of course.

[DerkontheOS]

Re: CFM rookies not adjusted to Stock Player Scale

So it turns out Madden NFL 16 simulates seasons in cloud CFM really quickly. As such I've already got some data for you all, only took about an hour total. It's taken me longer to type this post than it took to gather the data, lol.

I focused on two ratings that have come up a lot at the start of this thread - WR SPD and QB DAC. I also added in WR ACC for good measure.

For these three position-rating combinations, here's some useful statistical info for them (these numbers computed using the ratings from the Madden NFL 16 base roster spreadsheet EA released on their web site):

RATING    μ        σ      μ-σ    μ+σ
WR SPD    88.2     3.5    84.7   91.7
WR ACC    89.9     2.4    87.5   92.3
QB DAC    63.6     5.9    57.7   69.4

A quick brush-up of terms: μ is the average (also the mean), σ is the standard deviation. μ-σ and μ+σ are numbers each one standard deviation away from the mean; it's expected that about two thirds of the ratings used to calculate μ and σ will fall between these two extremes, assuming that the ratings fall over a normal distribution.

Having these numbers in hand, I created a cloud Connected Franchise, playing as a coach, turned off coach firings, and simulated ten seasons twice in a row. In year 2035 of this CFM, there are the same computations for the same ratings:

RATING    μ        σ      μ-σ    μ+σ
WR SPD    91.3     4.6    86.7   95.9
WR ACC    86.7     5.8    80.9   92.5
QB DAC    75.6     5.8    69.8   81.4

And finally, I subtract the original 2015 numbers from the 2035 numbers to visualize the change:

RATING    Δμ      Δσ     Δ(μ-σ)    Δ(μ+σ)
WR SPD    3.1     1.1    2.6       4.2
WR ACC   -3.2     3.4   -6.6       0.2
QB DAC    12.0   -0.1    12.1      12

Plain-text English explanation of what's happening / what these numbers mean:

After a twenty-year period of CFM simulation in this game save, the average _top speed_ of all wide receivers in the game increased marginally. The standard deviation also rose, indicating a wider variety of speed ratings amongst all the players. This is pretty evident at the top of the scale - there are 29 WRs with 97+ SPD.

Simultaneously, in the same 20-year simulation, the average ACC rating for receivers fell marginally, and virtually the same amount as SPD rose. The game still hands out a good number of high ratings here - there are fourteen WRs with 97+ ACC - but also more low ratings, relatively speaking - there are 71 WRs with ACC at 83 or less, while there are exactly three such players on the default Madden NFL 16 roster. It also appears that 80 is the lowest rating the game will hand out to a generated rookie at WR, and there are no CPU-rostered players with stupidly low SPD or ACC ratings, as would happen in Madden NFL 15.

In effect, what you have going on at the wide receiver position over time is that the population of available receivers indeed has a higher top speed and straight-line speed, but less short-area athleticism. I probably need to do this analysis for WR AGI in order to capture the full picture here, but honestly I don't think this incredibly offensive. I imagine that people are going to be upset over the greater frequency of higher speed ratings, but I see it as an increase in variance with respect to the caliber of available athletes at the position.

QB DAC is a different story, this is something I think Tiburon should consider adjusting if consistency with the base roster is something they feel is important to CFM. That rating is indeed rising dramatically over time. 29 passers have DAC 80 or greater (there is one QB who tops 90 DAC). Compared to the default roster which has 19 players with a DAC of 70 or greater and only two players who crack 80 (Aaron Rodgers and Matt Ryan). The floor for the players with the worst DAC ratings also rises an alarming level; while there are 30 players on the default roster with a DAC rating of 59 or less and 101 with a DAC rating of 69 or less, there is not a single rostered QB in the 2035 league with a DAC rating less than 64 and only 16 with a DAC rating of 69 or less.

One caveat - this doesn't identify where the problem is. I paid no attention to rookie classes, I literally hit the "sim ten years" option twice. It's possible - though I'd say unlikely, given how much the minimum rating is rising - that the rise in QB DAC over time is caused by an imbalance in upgrade package XP cost for that rating.

Anyway, if anyone wants to see any more ratings analyses like these, I'll keep this CFM file around and can probably take a small handful of requests for additional ratings to look at if y'all would like. I imagine that the CFM servers will get wiped at official launch on Tuesday, so bear that in mind.

All this said, I'm personally still looking forward to playing CFM, I always have fun playing it and I don't really find anything in here particularly offensive under Madden NFL's current player ratings philosophy. I do hope this post finds its way to someone at Tiburon and provides them with one data point which could be helpful in their making any decisions with respect to CFM rookie generation / player progression balance adjustments.

EA should just hire you to test CFM draft classes, this can be your resume.

[Hooe]

Re: CFM rookies not adjusted to Stock Player Scale

EA should just hire you to test CFM draft classes, this can be your resume.

I'd rather personally rebuild how draft classes are generated.

I wrote a small app for personal use in the lead-up to Madden NFL 15 to do exactly this, actually; my motivation was to be able to build draft classes which match the FBGratings roster set. The idea behind it was that, if EA enabled draft class editing in CFM (which they didn't in M15 and haven't in M16), I would feed the program a roster spreadsheet and it would generate players using the existing player ratings as the valid ranges for each rating. For example, in Madden NFL 15, the mean SPD for WRs was 90 and the standard deviation was 4. Using those values I can spit out a rating which fits within the range, given the ratings provided. Here's an example of SPD ratings my app spits out for a generated population of NFL wide receivers in a hypothetical custom draft class:

97 
96 
95 
94 94 94 94 94
93 93 93 93 93 93 93 93 93 93 93 93 93 93
92 92 92 92 92 92 92 92 92 92 92 92 92 92 92 92 92
91 91 91 91 91 91 91 91 91 91 91 91 91 91 91 91 91 91 91 91 91 91 91 91 91 91
90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90
89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89 89
88 88 88 88 88 88 88 88 88 88 88 88 88 88 88 88 88 88 88 88 88
87 87 87 87 87 87 87 87 87 87 87 87
86 86 86 86 86 86 86 86 86 86
85 85 85 85 85 85
83
81
75 75 75 75 75

Notice that the SPD ratings bunch up around 90, just like the base Madden NFL 16 roster, and ratings on the extremes are pretty rare. For comparison's sake, the SPD ratings for the WR population of my 2035 season CFM:

99
98 98 98 98 98 98 98 98 98 98 98 98
97 97 97 97 97 97 97 97 97 97 97 97 97 97
96 96 96 96 96 96 96 96 96 96 96 96
95 95 95 95 95 95 95 95 95 95 95 95 95 95 95
94 94 94 94 94 94 94 94 94 94 94 94
93 93 93 93 93 93 93 93 93 93
92 92 92 92 92 92 92 92 92 92 92 92 92
91 91 91 91 91 91 91 91 91 91 91
90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90
89 89 89 89 89 89 89 89 89 89 89 89 89 89
88 88 88 88 88 88 88 88 88 88 88 88
87 87 87 87 87 87 87 87 87 87 87 87 87 87 87
86 86 86 86 86 86 86 86 86 86 86
85 85
84
83
82
81 81 81
80
77 77 77

This distribution is far more linear, resulting in a more diverse player set but also one with more values on either extreme of the range.

This app is not quite robust enough to be used publicly, and I won't be releasing it for now - it doesn't consider player type and scheme, for example, and I'd want that information to better build more distinct player archetypes. In addition, there's no point to me releasing it publicly, since Madden NFL draft classes can't be edited anyway. I'd be happy to talk about what I did with Tiburon, though, if they are interested in hearing about it.

[griffy20]

Re: CFM rookies not adjusted to Stock Player Scale

The simple fact is that EA will continue to overlook these types of things in CFM due to the fact that CFM is particularly hard to monetize. Their main focus is MUT because that provides a steady revenue stream even after players have already spent $60 on the game.

I'm not saying this is wrong, or that I blame them, it's just a fact. The only way things will change are if CFM becomes a bigger part of the game, which I don't foresee happening because they currently can't monetize it the way they can with MUT.

I would actually argue that it's in EA's interest to NOT make CFM playable for long periods of time due to the fact that if people were satisfied with their draft classes and overall gameplay many seasons into their CFM's, the incentive to purchase next year's edition of the game is greatly diminished when one could presumably just keep playing the old game.

[Mauer4MVP]

Re: CFM rookies not adjusted to Stock Player Scale

The simple fact is that EA will continue to overlook these types of things in CFM due to the fact that CFM is particularly hard to monetize. Their main focus is MUT because that provides a steady revenue stream even after players have already spent $60 on the game.

I'm not saying this is wrong, or that I blame them, it's just a fact. The only way things will change are if CFM becomes a bigger part of the game, which I don't foresee happening because they currently can't monetize it the way they can with MUT.

I would actually argue that it's in EA's interest to NOT make CFM playable for long periods of time due to the fact that if people were satisfied with their draft classes and overall gameplay many seasons into their CFM's, the incentive to purchase next year's edition of the game is greatly diminished when one could presumably just keep playing the old game.

I tried that kool-aid too but had to spit it out.