Madden 19 Ratings Spreadsheet

[adembroski]

Quote: Originally Posted by SyncereBlackout

I love this forum Sent from my iPhone using Tapatalk

I don't think I'd call that post "science" so much as an excessively wordy way to say, "it's relative" followed by a fairly in-depth commentary on the idea. My next post gets a big more mathematical if you're into that though

[mtmetcalfe]

Quote: Originally Posted by adembroski

Completely different discussion. Do the ratings matter enough in gameplay? is there a great enough difference? I could answer either way depending on when you ask. I could say I can't *feel* the difference one day, the next day I'm watching a game and reminded of just how little difference there really is between NFL players. These are the top 1% of top college players, remember. This is why we get slideritis. See my sig

So would you say what some of us like myself are looking for would be more traits and tendencies to differentiate players more rather than just ratings? Maybe more so for the upper echelon of players?

[canes21]

Quote: Originally Posted by adembroski

It's not for the roster at large, it's for draft classes. That's for elite prospects, not elite NFL veterans.

I think you must have skipped my further replies on this. I knew it was from the draft class, 80+ being elite gave that away. my issue with it was that these ranges are 4-5 points only. I'd rather see something where a poor player is under 60. Average is 60-72, Good is 73-84, Elite is 85+ for draft prospects. That's just an example, but the ranges are all 2x bigger than they are by default.



My biggest issues with Madden's roster is that the ratings scale is hardly used and you can take QB's for example. The strongest armed QB and the weakest armed QB are within 20 points of one another. That should be stretched out further to 30 or more in my opinion. Same for all position specific ratings. Having such small ranges for the generated prospects will only further the lack of separation between players making everyone in your franchise mode feeling so similar as already is a huge issue with the base rosters.

[adembroski]

Quote: Originally Posted by mtmetcalfe

So would you say what some of us like myself are looking for would be more traits and tendencies to differentiate players more rather than just ratings? Maybe more so for the upper echelon of players?

I like the idea of a system of ratings and traits. I've always said EA's biggest weakness is they never follow up on a good idea. They put it out half-baked and never come back to it. The trait system Madden has was a good idea, but it was never iterated on.

[Other Guy]

Quote: Originally Posted by adembroski

Let me ask you this; by what model would you distribute "talent" through the league, and how strict would you be with its imposition? How many statistical anomalies do you allow? I'm not sure the average Madden fan would accept that. I think they'd say the number of 99s is much better, but they'd go nuts over 54 players ovr 90. You could go Pareto distribution, which is probably a more realistic model, but then you'd be making the "everyone feels the same" problem even more pronounced.

Thanks for all the feedback, it's much appreciated! I have never really drilled down into ratings data before, but the last few days I just sort of had a random urge to look at it. I didn't envision a particular curve fitting to ratings distribution; just wondered if anything had surfaced throughout the years of rating players.

I can't add much more to what you posted above; you have all the bases covered. Normal curve definitely would produce too many highly-rated players.

I did find one journal article on the subject, and it found that athlete performance fits the Paretian distribution. Which, again, is exactly what you pointed out. Here are a few tidbits.

Quote:

Results summarized indicate that the distribution of individual performance follows a Paretian distribution more closely than a Gaussian distribution.

Quote:

Across a variety of sports, nationalities, and levels of play, the Paretian distribution yielded a significantly better fit than the Gaussian distribution. In all but one sample, the performance distribution favored a power law more strongly than a normal distribution.

Thanks again for your insight!

[briz1046]

Quote: Originally Posted by Other Guy

Thanks for all the feedback, it's much appreciated! I have never really drilled down into ratings data before, but the last few days I just sort of had a random urge to look at it. I didn't envision a particular curve fitting to ratings distribution; just wondered if anything had surfaced throughout the years of rating players. I can't add much more to what you posted above; you have all the bases covered. Normal curve definitely would produce too many highly-rated players. I did find one journal article on the subject, and it found that athlete performance fits the Paretian distribution. Which, again, is exactly what you pointed out. Here are a few tidbits. Thanks again for your insight!

I don't think you can accurately describe the Madden ratings using population distributions , because what Madden shows us is only a part of the true population
At the to end of the spectrum it should work pretty well , but at the bottom end what you're shown is artificially cut off , as they rather arbitrarily include or exclude players in order to achieve 74 man rosters with acceptable numbers at each position
Throw out the40 or so LSs with silly low ratings , what I believe you're actually seeing is one half of a bell type curve with a very jagged tear somewhere near the middle of the bell

In adembroskis post you see the numbers rising steadily as the OVR decrease until you reach 65 or so , then it drops markedly at the lower end
Excuse me if my maths isn't very ' mathematical it's been too long since I used any of this stuff to remember anything but a broad outline

[adembroski]

Quote: Originally Posted by briz1046

I don't think you can accurately describe the Madden ratings using population distributions , because what Madden shows us is only a part of the true population

Pareto distribution is fractal, so it doesn't matter how much of the population is sampled.

[briz1046]

Quote: Originally Posted by adembroski

Pareto distribution is fractal, so it doesn't matter how much of the population is sampled.

Ok .got you , warned you my maths is basically a foggy memory