SUPERSIM Modding thread

[adembroski]

Reposting my explanations from the larger modding thread.

Quote: Originally Posted by adembroski

OK, and here it is, folks, Supersim's big scary file: Compound ability. There's a couple of dozen files in this directory and there's a ton of entries in each. Thankfully I can tell Supersim has been updated extensively since I worked with it, but that also means I don't really necessarily know what everything does. There are 6 fields I pretty clearly remember. min_player1, max_player1, min_player2, max_player2 These convert attributes to ranges of numbers that can be used to calculate success. So if you set min to 1 and max to 99, the attribute would remain the same. If you set min to -10 and max to 0, a 90 would become a -1. The game is going to generate a number between min and whatever the attribute actually is. The basic mechanic is that this number is compared against a number representing difficulty, and if it's higher, it succeeds. At the bottom of each compound ability file are two other fields; does_player2_oppose and needs_two_players. needs_two_players does what it says on the tin. Break TD chance, for example, doesn't have an opposing player. Catch chance requires a QB to throw and a WR to catch. A tackle requires a tackler and a tacklee. does_player2_oppose depends on the action. If it's a cooperative action, this should be unchecked. If it's an opposed action, it should be checked. If this is unchecked and needs_two_players is checked, the two derived numbers will be summed. If it's checked and needs_two_players is checked, then the higher derived number wins. I'm not sure how much I'm going to remember about all this. I will say this; this **** is detailed as heck. EA could make an exceptional text sim with this thing. it's probably the highest quality thing in the game. Tuning it, however, is very, very hard.

Quote: Originally Posted by adembroski

I REMEMBERED!!! So the other fields are attributes, "Player1_Acceleration," etc.. Those are weights. So, for instance, supersim_breaktdrun_chance has the following... player1_acceleration = 10 player1_ballcarriervision = 1 player1_clutch = 1 player1_speed = 5 player1_breaktackle = 10 So the number used by min_player1 and max_player1 is going to be the weighted average of those listed. So... I need to come up with a language to explain all this. Any of you familiar with RPGs? It works a lot like D&D 2nd Edition's skill system. I'll come back to that... First, it's going to go ACC *10 + BCV *1 + SPD *5 + BKT *10 + Clutch * 1 (we check supersim_trait_conversion to find that Clutch is 50 if false, 99 if true). So lets do LeVeon Bell. ACC = 93, BCV = 96, SPD = 90, BKT = 92, Clutch = 99 ( ( ( 93 * 10 ) + ( 96 * 1 ) + ( 90 * 5 ) + ( 92 * 10 ) + ( 99 * 1 ) ) / 27) = 92.4 So, min_player1 = -99 max_player1 = 25 That's a 124 pt difference, so 124 * .92 = 114, which means our derived number (or LeVeon's break long run rating) is 15. Whenever LeVeon carries the rock, the game is going to produce a random number between -99 and 15 to determine whether or not he just flat takes it to the house. NOW begins the process of remembering how the opposed number is produced. There is a number his skill check will be compared too, and I don't remember what it is. I mean, obviously there's a player2 outlined in the xml, but "needs_two_players" is false, so I don't know how it works if there is no opposition, nor do I know who the opposition is. MLB I suppose? I'm not sure. FS? either would make sense. So in the old Advance D&D 2nd Edition rules, the an opposed skill check worked like this- each combatant would roll a 20 sided die, and whoever got higher but still under their skill would win. So lets say you're in a joust, and you want to see who rides better to give them a bonus to their attack... one guy has a 10 riding skill, the other one has a 15. The first guy rolls a 9, and the 2nd guy rolls a 17. The 2nd guy has the better skill AND the higher roll, but he didn't keep it under 15, so he loses. This works kind of like that. The difference is you CANNOT go over your skill, because the skill determines the size of your die. If you've never played tabletop RPGs, ignore this, including it because it'll make it a bit easier for those who have to wrap their minds around it. Also, you might be wondering why we're stretching the rating all the way to -99... it's because the ranges were measuring are more like 80-99, but Supersim just looks at the whole rating. You stretch the low end down to take those 75-ish points every HB has in BTK and the like. Also, it helps to normalize the weighted average. There were probably easier ways to go about this, but it works. The conversion system is to get the chance of something happening to where we want it. We know that a guy with an 90 BTK vs. a 90 TAK shouldn't be a 50/50 proposition, just like a pitcher is going to beat an equally matched batter 70-75% of the time. The tackler wins most of the time, so we create this derived chance that we can manipulate to be the correct win/loss ratio to produce accurate stats.

Quote: Originally Posted by tdawg3782

I know that QB rushing sim stats has always been an issue in this game. Especially this year. I was messing with the sim engine and was able to get Marcus Mariota to rush for over 700 yards. Not to bad. Some potential here. Down side is it took a lot of carries. Cam did good as well. EDIT: I looked it up and Cam did have 139 carries last year (for 754 yards). So maybe not as crazy as I thought. Next is to try to find a way to up the YPC to try and find that balance.

[adembroski]

One thing I've discovered- raising the rushing rate also increases completion percentage. I've had Newton in the right range of carries compared to pocket quarterbacks, but his completion percentage jumps into the 80s.

[adembroski]

Quick FYI: These do NOT work on Cloud. I figured this, the data all lives on the server, but I wanted to be sure.

[adembroski]

I'm remembering what I'm missing here.

When I was doing this for '13, I had a game log... like, every dice roll, when it happened, the results, all that. I could play a season, go back and look at any game, and see exactly what happened with these numbers on a play-by-play basis, and it's a lot more complex than you think.

There's a mistake I've found though. player2 is the receiver, but it treats him like a DB. Not opposed, but things like ZCV and PRC are weighted, but not Route Running.

I think priority number one for me is to cut down on the number of attributes being used in all these calculations. Too many weighted attributes just clutters things up and makes things too even.

Right now the completion percentages are a tad high, the yards per attempt (a VERY tough one to tune) is about a half yard high. I think I'll leave the latter and try to bring the completion percentage down a couple of percentage points, THEN I'll start working on scramble chance and then come back to completion percentage to correct.

Anyways, I think I'm done for the night. I don't know how soon I'll get back to this but I think I'm gonna let Tdawgs progression numbers stand and abandon my progression project. I'm a lot more useful here.

[tdawg3782]

Quote: Originally Posted by adembroski

I'm remembering what I'm missing here. When I was doing this for '13, I had a game log... like, every dice roll, when it happened, the results, all that. I could play a season, go back and look at any game, and see exactly what happened with these numbers on a play-by-play basis, and it's a lot more complex than you think. There's a mistake I've found though. player2 is the receiver, but it treats him like a DB. Not opposed, but things like ZCV and PRC are weighted, but not Route Running. I think priority number one for me is to cut down on the number of attributes being used in all these calculations. Too many weighted attributes just clutters things up and makes things too even. Right now the completion percentages are a tad high, the yards per attempt (a VERY tough one to tune) is about a half yard high. I think I'll leave the latter and try to bring the completion percentage down a couple of percentage points, THEN I'll start working on scramble chance and then come back to completion percentage to correct. Anyways, I think I'm done for the night. I don't know how soon I'll get back to this but I think I'm gonna let Tdawgs progression numbers stand and abandon my progression project. I'm a lot more useful here.

Let me know if there is anything you need me to test. You obviously know this stuff better then I would think everybody here at OS. This type of project is right up my alley as far as numbers and data collecting. Excited to see what tweaks you can uncover.

[adembroski]

Quote: Originally Posted by tdawg3782

Let me know if there is anything you need me to test. You obviously know this stuff better then I would think everybody here at OS. This type of project is right up my alley as far as numbers and data collecting. Excited to see what tweaks you can uncover.

Dude, just do what you do. You're smart enough, you're gonna find **** I didn't even find in '13 when I was paid to figure this **** out (EA's documentation is ****).

I will give this advice. Make your low and high the same at first. This way you can find the average and work outward from there. Every time you raise max, you bring the whole scale up, so working out from a central point is most effective. I'm still working on a decent way to predict what numbers WILL do, and I think I'm just gonna have like a thousand randbetweens in excel to do it, which is inefficient, but it's more accurate than guessing.

[adembroski]

Also, getting this right could be a massive project. I'd love to have an army on this, some working on their own thing, some backing up the dudes working on their own thing. The tools we have are powerful, we can make this great if we get enough people wanting to take it on.

I wanna finish what I started.

[tdawg3782]

There is some funny stuff in this sim engine. I'm taking a look at supersim_fumble_forcefumble_chance. Player1 is the defensive player and Player2 is the offensive player as far as I can tell. But under Player2 it barley factors in Carry (2), doesn't factor in the coverball trait at all, and for god knows what reason puts a 25 weight on zonecoverage lol. Weird.


EDIT: Holy crap. I can see why they zeroed out the coverball trait. It's way to volatile. With the trait conversion being the way it is a guy who has never cover ball (Alex Collins I'm looking at you lol) he had 28 fumbles lol. But a guy who had cover all the time had 0. Back to more tweaking.