Just stumbled across a great find (re: pass rush)

[Ian_Cummings]

This might be a long one, so bear with me...


After all the worry over the pass rush (and the heat the NCAA 09 took last year as well as Madden for the "robo-qb"), it's obviously been at the top of my list all year to make sure it was effective. Typically we use QA as our gauge on how effective things are, and throughout the year they have been providing reports, stats, etc that show how much pressure they are getting and stuff like that. The pocket never quite felt "hectic" to me (as many of you have mentioned) but it was always pretty damn close. So after actually watching the whole thing again (rather than playing it), and reading this post from LBz...

Quote:

“But pass rush. Front four pressure of both teams showed nothing. KVB gave us the only front 4 pressure of the first half on a half line jet stunt (even though it was not run correctly, I'll take it) at the 4:17 mark of video 3. I don't even know who Terrell Suggs, or Jevon Kearse are? And barring that one play KVB was absent as well. No pressures/hurries by the front 4. QB's chillin in the pocket. Suggs is completely absent in the first half. No pressures, swallowed up by the run ect., Ngata and Pryce did some good against the run. Not against the pass.”

...I figured the best I could do would be to actually snoop around the code and formulas with one of the gameplay guys and see if we saw anything weird. Well...after about 30 minutes of stepping through code and breakpoints, I found a wonderful little gem:
weaponcurve.JPG

"Ian, what is that strange looking graph?" you may ask...

Well my friends, you are feasting your eyes on a "WEAPON CURVE".

"Ian, what is a weapon curve?"

Firstly, when I worked on the gameplay team for the 08 products, we implemented what we thought was some really neat tech that allowed us to really fine tune the game - and that is a "ratings curve". Since Madden's inception, every single formula that has existed in the game to determine success or failure has been linear. So to keep it simple, let's say a guy with 100 catch rating catches the ball 100% of the time, and a guy with a 0 catch rating catches 0% of the time. With a "linear" formula, that means a 40 rating = 40%, 75 rating = 75%, etc.

The "curve" was added to allow us to send in a rating and grab a different result based on how we wanted to tune each specific area. It meant that once we were happy with the ratings of the players, we could still tune the game accordingly without having to go re-rate everyone. So let's say on the catching example, we wanted a 40 rated guy to catch 75% of the time and scale up from there, we could make a graph like this:
catchcurve.JPG

Over the years, as the ratings started to become closer and closer together (and not use the whole scale), we also used the ratings curves to separate out players in gameplay. NCAA fans may remember the oft-discussed "speed curve", which was in essence a way to make a 99 guy drastically different from a 97 guy, whereas if a linear scale had been used before he wouldn't really have been noticeably different.

The above examples don't really look like "curves"...but we got the name because we could build curves on how we wanted players to perform. Here's an example curve that allows us to make a flat line for the common players and then a drastic increase from the the 70-100 ratings.
curve2.JPG




Hopefully I did a good job explaining what the ratings curves are, and what they are used for. Enter Madden NFL 08, and "Weapons". The weapon feature was a huge hook for the team that year, so everyone really wanted to make sure that the weapons stood out differently from everyone else. Since it would have seemed weird to just have the weapons be 98 and above and everyone else be sub-90, the curves were used to make those guys more powerful. It meant that the curve would drop SUBSTANTIALLY at the number that was the rating cutoff for the specific rating. For example, let's say you were a 98 in Finesse Moves...this curve made sure that you acted like a 98. But if you were a 90, this curve dropped you down to perform like an 81 rated guy would. By using these curves, all of the players were still able to be rated in the high 80's and 90's, but anyone with a "weapon" tag was actually a good 10% better than someone that may have been one rating point lower than him (since he wasn't marked as a weapon). Here's what I mean:
bigdropoff.JPG

It was tough for me to understand at first, but imagine I am playing war, but in order to use my card, I have to handicap myself with the dealer. To check to see if I should win against my opponent, I hand in an Ace to the dealer, and he gives me back an Ace to play. Now imagine I hand him a King...he gives me back a 10 to play. That's kinda how the 'weapon curve' works...except I'm passing in a 96 finesse move rating and getting an 88 in return.

Well, you may have already guessed how this story ends.

The curves for the weapon guys were still being used in the calculations for the pass rush successes (even though weapons have been removed). As you all know, Donny stretched the ratings out in a big way, which is in essence doing what these ratings curves used to do...make sure the elite guys are well ahead of the rest of the pack. With these curves still in place, it means the effectiveness of the pass rushers was being reduced on top of their already lowered ratings. I think there are only like one or two LE's in the league that are over 90 finesse moves, yet anyone that was under 94 was effectively getting 10% or so shaved off that rating when the calculations were made going up against an offensive lineman.


The gameplay guys and myself have flattened everything out in the curve and will be testing the hell out of it over the next few days. I am confident this will help the pass rush quite substantially without being too risky of a change (meaning the good guys will just get better...it's not a global change where all of a sudden the pocket will be collapsing all the time).


I said this quite a few times last year, and I'm not trying to be cheesy or anything, but this is really just f***ing awesome. Being able to openly communicate with you guys - the folks that are extremely passionate and knowledgable like LBz and Adembroski and countless others - it's just an amazing thing. Being on the other side of this for so long where we'd work as hard as we could and just get ripped apart in August every year, it is such a great feeling to be able to find issues and make changes STRICTLY because of the passion and helpfulness that you guys provide.

Next up, run styles and defenders cheating inside too much on outside runs. I don't think fixing those will be this easy...but we'll do our best

[Von Dozier]

Very nice.

Quite refreshing to see someone putting their heart and soul into a game like this. It improves more and more every day.

[NoTiCe_O]

Thanks Ian, you guys are always coming through alot this year

Reading Now

[T9R]

Wow!

[sportzbro]

you continue to push the bar when it comes to interacting with the community compared to other well known sports titles. i continue to dream about a day when they show this type of dedication.

crazy information btw.. and the titans vid looked pretty awesome.

[Paytonsplace]

Great Job Ian. Here is a wonderful example of taking input from the community and correcting the performance curve to yield (hopefully) a better rush. Great work on a Friday night!!

Now about those easy Field Goals..........

[Purple28Pedestrian]

damn Ian, well worth the read, good to hear that this should help sort the pass rush out. Thanks for being so candid, let us know how it works.

[Jackdog]

Quote: Originally Posted by BORN4CORN

you continue to put the NCAA to shame. i continue to dream about a day when they show this type of dedication. crazy information btw.. and the titans vid looked pretty awesome.

Please leave NCAA out of this. Stick to Madden talk please.