Donny Moore, the 'Madden Ratings Czar', Leaving EA

[charter04]

Quote: Originally Posted by roadman

Dan, isn't scouting circumstantial as well though. It was already stated by Charter that scouting isn't always 100%. You bring up WR's and something close and dear to our hearts. Look at the Packer WR's, no separation, Cobb, Jones and Adams. Adams played well as a rookie, McCarthy was praising him in TR camp, and kerflewy. Cobb strikes it rich in the off-season based on past performance and he isn't playing up to his normal standards, injuries or not. Jones, we all know he has regressed. I would love to see Adams and Cobb take ratings hits for the lack of production they've had during the year. I am the type of person that doesn't care what Jerry Rice, Randy Moss and Sterling Sharpe did 20 years earlier and compare them with Cobb and Adams.

Yes I said it's not 100%. Nothing is. But, I would still take a trained eye that's misses to be much closer to 100% than a non trained eye.

Non trained eyes are reactionary like most of us. A trained eye see's it as potential. They see a player as he his. They mess up too but, some of their mess ups were not the player ability but, all the other things. His own laziness or the wrong offense or any number of things.

I just like the idea of attempting something that could be revolutionary instead of just settling for how it's always been.

Sorry for the long reply. Lol

[DCEBB2001]

Quote: Originally Posted by roadman

Dan, isn't scouting circumstantial as well though. It was already stated by Charter that scouting isn't always 100%. You bring up WR's and something close and dear to our hearts. Look at the Packer WR's, no separation, Cobb, Jones and Adams. Adams played well as a rookie, McCarthy was praising him in TR camp, and kerflewy. Cobb strikes it rich in the off-season based on past performance and he isn't playing up to his normal standards, injuries or not. Jones, we all know he has regressed. I would love to see Adams and Cobb take ratings hits for the lack of production they've had during the year. I am the type of person that doesn't care what Jerry Rice, Randy Moss and Sterling Sharpe did 20 years earlier and compare them with Cobb and Adams.

The set of circumstances are different. As a WR, your production value is largely based not only on your own skills, but the skills of other players (QB, OL, DB, etc). Scouting data takes only into account the individual abilities of the player. Can he run good routes? Can he catch while under duress? Does he have solid, mundane, use of his hands? In this way, the player himself dictates his grades. The success of other players does not matter.

Ask yourself which is more likely for the WR scenario: will the number of catches a player makes likely result in him being a better route runner OR will the quality of the player's route running more likely affect his ability to catch passes?

[Ueauvan]

Dan, i understand what you mean, a part of me immediately thought of Pavlov, i.e. wr runs good routes gets catches, repeats action for more positive outcomes

[khaliib]

Quote: Originally Posted by charter04

Here is why I don't like the big focus of production driving ratings. It becomes a little like chasing your own tail. I just don't like the overreaction feel of weekly updates based on stats personally. It reminds me of one of my favorite games of all time, Tecmo Super Bowl. The best CB in the game was a guy on Tampa Bay named Wayne Haddix. If you know who he is then you either remember him from the game or you were a Bucs fan in 1990. Lol He had a lot of ints that's year. Was he the most talented CB in the NFL in 1990? No. He was rated in Tecmo SB the equivalent to a 99 overall. All 7 of his ints came that year. If we want a true sim football game I want sim ratings. IMHO ratings largely based on stats are more arcade than sim. Just my opinion though. From my experience using FBG ratings in madden 25, 15, and 16 they do a much better job replicating real life even stats wise then some assume. People complain about how crazy accurate the CPU QB's are on default sliders or how the defense catches ints too easy or how any WR seems to be able to spec catch too much. FBG rosters gives the desired result that is closer to real life than EA's stock rosters. I've played with them a lot. I've played with EA's a lot. FBG gives more realistic results. And that's in an engine that should be 100% geared toward its own rosters. But, most QB's are deadly accurate because their ratings are too high. What ever the case something needs to change. Either the engine or rosters do no replicate real NFL football as it should. The devs say they want a sim FB game so everything must be considered. Again just my opinions

But in doing this, you just took away the ability for each gamer to experience a "dynamic" CFM and removed the ability for each to have a life on it's own.

Because the focus of the player ratings is utilizing exact data (per say), there can be "no" deviation, otherwise, we're back to what we have now.

If ratings can't deviate away from their source being so hard data driven, then CFM becomes nothing more than "Play Now" games tethered together.

Because player "Back Stories" cause deviation to the ratings, this element must be stripped away due them being of a fantasy nature that's not in line to the real life happenings of players.

How can players be dynamic through out the course of a game?
- Hot or Cold moments

The Draft becomes useless as gamers will know who the best players are for that draft. (remember no Back Stories because they cause deviation)
- this just killed online leagues also with the data revealing who to take.

How are coach's rated, which affects AI gameplay?
- the only thing that seems logical is their records, which is performance driven.

How is player Stamina determined, as well as, how much weight is applied towards altering a players ratings?

Coach's decide what plays will be called, what players will play, yet all the focus seems to be just on the players.

Just curious how some aspects not talked about will fit into the grand scheme of things?

[ggsimmonds]

Quote: Originally Posted by DCEBB2001

I would like to say a few things based upon your points above: You are correct in saying that Madden is not a scouting tool. However, Madden uses attributes and a scale for them. Scouts do the same thing. They attempt to quantify qualifiable information. Madden does the same thing. They attempt to give value to what is seen. If the goal is to replicate real life results, then why do they have ratings for attributes at all? Why not just have a YDS rating or a REC rating?

Because a yards rating or a reception rating is far too basic. By using the component attributes it allows for more depth in the game. I don't argue in favor of 100% production.

Quote: Originally Posted by DCEBB2001

What production stat tells us how fast a player is or how tough he is? Name one that measures a player's stamina. Madden already has a framework built (albeit not perfect) that can harness scouting information. All that needs to be done is a proper application of it based on quality data. Sure, that would require matching animations (a complete overhaul, IMO), but it can be done. Will it? Who knows. All I know is that I don't expect anything soon...likely too much old code stacked up.

You are correct that no production stat tells us how fast a player is. Indeed you could go further and say using only production would make it impossible to rate rookies! But as I said, I'm not in favor of 100% production. Madden currently uses 40 times for speed (to varying levels it would seem). However the FBG advocates are not pushing for the status quo, they want more reliance on scouting than is already present.

Quote: Originally Posted by DCEBB2001

That being said, production is the product of a multitude of things happening on any given play, seemingly infinite at that. I use the "Open Receiver" example whereas a WR gets open every play because he is so fast, agile, and technically sound that he is merely uncoverable. However, if the QB never gets to the top of his drop before being planted on the turf, that WR will never "produce". According to the production line, a WR in that offense would be TERRIBLE. The fact of the matter is, that WR is not a worse player because of his lack of production. Take that same player and place him on a competent team with a decent line and maybe his production rises as he is finally getting the ball. Is he now better because of it, or is he the same player he was, just under different circumstances? Production is very circumstantial. How many of you had Brandon Marshall in your top 5 WRs in the league before this season? I can tell you that I caught a bit of heat on these very forums for people mocking the fact that he was rated as a top 5 WR according to the FBG Ratings. The guy set a career high in TDs and may set one in receptions; placing him likely among the top WRs in the game right now, I will add. The point is that few expected him to produce up to the level at which the scouts knew he could play at. Did he magically get better because of having Fitzpatrick throw it to him, or have the OL that the Jets have? I would posit that he is the same player that the scouts slated him to be when the year started.

Quite right about the multitude of variables and it actually points to why I am weary about going full on scouting data. I don't have confidence in Madden to be able to adequately factor in all those variables. Take for example something that scouts grade QBs on: composure. I don't think Madden can properly model that in the game.

However at least in this context production =/= stats. You disapprove of the use of PFF in rating Madden players so this is where we will disagree, but if a WR gets open and the QB fails to get him the ball it will still look upon the WR favorably.

Quote: Originally Posted by DCEBB2001

The bottom line is that we need to rate the players based upon their POTENTIAL attribute values in EQUAL situations to see how they really differentiate from one to another. Production follows talent, not the other way around.

I'm not opposed to using scouting data, I just think the argument is all wrong. There should not be a scouting vs production debate. Use them both, but when there is divergence (e.g. scouts say a young player should be performing well but his production lacks) more often than that production should trump what the scouts say. To throw this out there I think we absolutely need something of a "ball locating" attribute for CBs, so that would be something that falls under the scout umbrella.
What I don't want is to use scouts and then cover my ears to everything else.

What the conversation should really be about is the system. Currently there does not seem to be much of one. Right now how EA rates players it seems like he can watch a game on Sunday and write down notes in his tablet based on what he sees, but if he were to watch the same game on Tuesday when he is in a different mood the notes could be very different.

In contrast take your FBG ratings. You will never publicly reveal your methodology because with that, I could take the same input source (scouts in your case), and create an identical roster set. This is what EA needs.

We only differ in what should be the source of data. Your position is scouting data.
Mine is a combination of scouting data and advanced metrics like PFF.

[ggsimmonds]

Quote: Originally Posted by charter04

Here is why I don't like the big focus of production driving ratings. It becomes a little like chasing your own tail. I just don't like the overreaction feel of weekly updates based on stats personally. It reminds me of one of my favorite games of all time, Tecmo Super Bowl. The best CB in the game was a guy on Tampa Bay named Wayne Haddix. If you know who he is then you either remember him from the game or you were a Bucs fan in 1990. Lol He had a lot of ints that's year. Was he the most talented CB in the NFL in 1990? No. He was rated in Tecmo SB the equivalent to a 99 overall. All 7 of his ints came that year. If we want a true sim football game I want sim ratings. IMHO ratings largely based on stats are more arcade than sim. Just my opinion though. From my experience using FBG ratings in madden 25, 15, and 16 they do a much better job replicating real life even stats wise then some assume. People complain about how crazy accurate the CPU QB's are on default sliders or how the defense catches ints too easy or how any WR seems to be able to spec catch too much. FBG rosters gives the desired result that is closer to real life than EA's stock rosters. I've played with them a lot. I've played with EA's a lot. FBG gives more realistic results. And that's in an engine that should be 100% geared toward its own rosters. But, most QB's are deadly accurate because their ratings are too high. What ever the case something needs to change. Either the engine or rosters do no replicate real NFL football as it should. The devs say they want a sim FB game so everything must be considered. Again just my opinions

I think you are off the mark here. Avoid the temptation to phrase everything in a sim vs arcade argument. You remind me of politicians during campaign season!

I also dislike the overreactions we see with the weekly roster updates, but that is a separate issue.

Quote: Originally Posted by DCEBB2001

The set of circumstances are different. As a WR, your production value is largely based not only on your own skills, but the skills of other players (QB, OL, DB, etc). Scouting data takes only into account the individual abilities of the player. Can he run good routes? Can he catch while under duress? Does he have solid, mundane, use of his hands? In this way, the player himself dictates his grades. The success of other players does not matter. Ask yourself which is more likely for the WR scenario: will the number of catches a player makes likely result in him being a better route runner OR will the quality of the player's route running more likely affect his ability to catch passes?

I cannot speak for everyone, but I am definitely not supporting player ratings derived from box scores.
An illustrative scenario from an NFL game:
WR A: Constantly beats his man to get open. Gets off the line cleanly with a proper release. Makes crisp cuts in his routes and attacks the ball to make the catch. Finishes the game with 11 targets. Makes 8 catches for 122 yards.

WR B: Constantly beats his man to get open. Gets off the line cleanly with a proper release. Makes crisp cuts in his routes and attacks the ball to make the catch. Finishes the game with 5 targets. Makes 3 catches for 44 yards.

Both players get open the same number of times, but for whatever reason WR B doesn't get as many targets. From this game both players should be rated identically.

At this point Dan you may be thinking, "but gg that sounds an awfully lot like what I am arguing for with FBG ratings."
Maybe so, but there are subtle differences.

Scouts: Grades things like hips, crisp cuts in running routes, attacks the ball at its highest point,etc. Scouts ask "does he show these characteristics? From each of these component grades scouts conclude he is a skilled WR who should be able to produce in the NFL (production follows talent). In this sense it is predictive in nature.

alternatively: Does not look at technique so much as the result. Did he get open? Did he run a clean route? Did he get off the line cleanly? Thus it is not predictive but rather it is results driven.

Dan I think you know exactly what I am talking about. When a trained scout watches film they ask things like did he get off the line quickly and immediately attack the corner or did he waste time with a useless shimmy? How many steps did he take in his breakdown? In this case even if the Wr got open the scout may see that his technique was sloppy so would not expect this to be a long term trend.

Then watch film with a guy who grades players for PFF. What does this guy ask? "Did he get open." Much simpler in comparison.

In terms of overall knowledge and credibility the trained scout will be superior. But Madden does not have things like a "fluid hips" rating. It has a "route running" rating that determines whether or not a WR gets open. So the simplistic "did he get open?" works for Madden.

[roadman]

Quote: Originally Posted by khaliib

But in doing this, you just took away the ability for each gamer to experience a "dynamic" CFM and removed the ability for each to have a life on it's own. Because the focus of the player ratings is utilizing exact data (per say), there can be "no" deviation, otherwise, we're back to what we have now. If ratings can't deviate away from their source being so hard data driven, then CFM becomes nothing more than "Play Now" games tethered together. Because player "Back Stories" cause deviation to the ratings, this element must be stripped away due them being of a fantasy nature that's not in line to the real life happenings of players. How can players be dynamic through out the course of a game? - Hot or Cold moments The Draft becomes useless as gamers will know who the best players are for that draft. (remember no Back Stories because they cause deviation) - this just killed online leagues also with the data revealing who to take. How are coach's rated, which affects AI gameplay? - the only thing that seems logical is their records, which is performance driven. How is player Stamina determined, as well as, how much weight is applied towards altering a players ratings? Coach's decide what plays will be called, what players will play, yet all the focus seems to be just on the players. Just curious how some aspects not talked about will fit into the grand scheme of things?

This is my thought process, exactly.

Thanks Khaliib, you stated better and more thoroughly than I did.

Also agree with GG's last post, post was too long to quote with Khaliib's, but both on point where my train of thought is.

And here is the rub....... I don't think either side is right or wrong. I say to each their own and think both systems could work at EA.

[DCEBB2001]

Quote: Originally Posted by khaliib

But in doing this, you just took away the ability for each gamer to experience a "dynamic" CFM and removed the ability for each to have a life on it's own. You can build variation into the model with scouting data just as well as you can with production. Players drop and fall every week according to the scouting data. Within a CFM, you can build in the proper distribution, both positive and negative, to affect not only the overall grades, but also the individual attributes. Since we have 20 years of data to draw from, we have a very large sample to build that model upon. In essence, the scouting data serves as the "base" for the ratings when you start CFM, but we can tie things like production into the model for how those ratings change. We can also use "potential" grades, fixed, and either available to the user or hidden under the hood, to control the progress, lack thereof, or rate thereof. Because the focus of the player ratings is utilizing exact data (per say), there can be "no" deviation, otherwise, we're back to what we have now. A historical analysis of the data shows that there is always variation. Players are always moving up and down. Guys who move up tend to have higher potential. Guys who stay the same from their rookie year, or advance very little, have lower potential grades. That grade, either revealed to the user or hidden, in itself can control progression without utilizing production. In this way, the user actually has less control over how well his players advance. Instead of simply racking up endless stats to make your player a 90, you are limited to his potential. This makes the user have to make some tough decisions. If you have a player who you think is "maxed out" at a 70 OVR (solid starter), who also produces well for you, but come across another player in the draft who you think may have an opportunity to be better (having better potential), you could be risking what you know (in the player you currently have) for a player that may not turn out to be better. Maybe contracts come into play as well. The point is that a model can be built to control progression/regression without relying on production. Production, after all, is all relevant to the user. What qualifies good production for one user may be entirely different for another user. If ratings can't deviate away from their source being so hard data driven, then CFM becomes nothing more than "Play Now" games tethered together. See my comments above as to how they can deviate within CFM. Because player "Back Stories" cause deviation to the ratings, this element must be stripped away due them being of a fantasy nature that's not in line to the real life happenings of players. Not necessarily. We just change the model for WHAT determines the deviation in the ratings. Instead of using stats/production, we use an already predetermined potential grade with variation built into their ability to progress. For instance, if you draft a 52 OVR player with A+ potential, you know that he has the ability to be a HOF player despite his low OVR grade when drafted. Using a normal distribution model, you know that he has a 95% chance of reaching an OVR grade within 2 standard deviations of his draft grade. However, there is also a 5% change that he doesn't meet that potential and becomes a bust. The opposite also holds true. You could draft a player with a D- potential, knowing he only has a 5% chance of becoming a 90+ OVR, but he does. He would be an obvious draft gem. In all reality, this actually plays out in real life. Teams take risks on players based on the potential odds that the player will turn out. Why not add that into the game? How can players be dynamic through out the course of a game? - Hot or Cold moments The Draft becomes useless as gamers will know who the best players are for that draft. (remember no Back Stories because they cause deviation) If the potential for the player is in itself based on a normal distribution that controls the odds of which a player will succeed or not (a probability model), then you are taking a risk with every player. The higher the grade of the player, the higher the risk, obviously. Combine that with his potential grade, also based upon his probability to progress, and you just squared the number of possibilities for that player. There are two factors you now need to consider: the player's draft grade (and likelihood on where he will be drafted) affecting the risk, and the player's potential affecting the likelihood that he will progress. When you combine both, the progression of the player could be quite dynamic. If you then throw in a third part where progression is based upon production, you now cubed the possibilities. That is a LOT of variation built into the model. - this just killed online leagues also with the data revealing who to take. How are coach's rated, which affects AI gameplay? Let's throw in a 4th variable that controls progression: coaching abilities. Some are good at developing talent, others are not. - the only thing that seems logical is their records, which is performance driven. How is player Stamina determined, as well as, how much weight is applied towards altering a players ratings? Every attribute has a mean and standard deviation as well as a model for how it can or won't progress tied into the potential of the player. Coach's decide what plays will be called, what players will play, yet all the focus seems to be just on the players. Then, add coaching ability into the progression "soup" to progress players. Just curious how some aspects not talked about will fit into the grand scheme of things?

See my responses in bold above.