XIM Community

XIM APEX => Game Support => Topic started by: GoodAimButATankMain on 08:03 AM - 05/16/20

Title: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 08:03 AM - 05/16/20
Hey, not sure if mist knows this but i suspect he just thought linear was a strait line in the curve editor so when he made the ST he just used a line. Well, the linear mode is actually a couple S curves stacked on top of one another so the single line will not work the best for 1:1 input's. This isnt too noticable since the s curves are pretty flat at a value of 0 aim ease in, but if you go into the practice range and slowly spin your character from 1 side of your mousepad to the other, you will go further than if you spin at a faster speed. This messes with your flicks a lot, and luckily i have a way of making my own ST's and made one for everyone to use until mist fixes this issue. I created it using 33 aim ease in so it offers better micro movements than strait linear as well as keeping good tracking. Here is the link to the config:

 https://community.xim.tech/index.php?topic=87413.msg810915#msg810915

And here is to the link to my guide on how i made the ST.

https://community.xim.tech/index.php?topic=87372.msg810631#msg810631

Also, here is the link to imgur of me using the data i got from my test's to see that it is in fact an S curve.

https://m.imgur.com/gallery/IviMVTv
Title: Re: Overwatch Linear ST Needs A Retrain
Post by: GoodAimButATankMain on 09:53 AM - 05/16/20
Just tested exponential ramp as well, seems like they both probably use S curves instead of the traditional exponential curve. I think they both need a new ST
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: Boomboks on 03:50 AM - 05/17/20
I was skeptical about these settings because there's a lot of "xim scientists" on here sometimes with bad math/theory. But these actually feel really great.

Using it with an extremely fast sensitivity, like 15cm/360. But playing some Tracer on xbox, the tracking is better than any settings I've used before.

I'm inclined to believe you that we need a new ST for the game. I wonder how exponential would feel with this kind of fix too
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 04:34 AM - 05/17/20
Yeah, the curve isnt "perfect" by any means, but i plan on redoing this ST in the near future because when i created this config, i was pretty tired at the time and got lazy with my tests. Also, had the same thought's lol, so i am planning on creating multiple STs with different aim ease in values so people can choose which they like the best :)

I dont think i will ever be able to create "Perfect" ST's with my method, but spending enough time on each game can get u pretty close to it lol and this curve can definitely still be improved on. Especially with these S curve games, they require a lot more spin tests than normal to get good curves working lol, n im still trying to come up with a way to graph the curve, but for right now my best method is to split all the individual s curves up, and then find each one separately and try to link them up together. This isnt the most accurate way of doing it, but it gets the job done good enough to create STs for games that use them lol
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 06:48 AM - 05/17/20
I was skeptical about these settings because there's a lot of "xim scientists" on here sometimes with bad math/theory. But these actually feel really great.

Using it with an extremely fast sensitivity, like 15cm/360. But playing some Tracer on xbox, the tracking is better than any settings I've used before.

I'm inclined to believe you that we need a new ST for the game. I wonder how exponential would feel with this kind of fix too

Because of this post.... It got me thinking "Hmmm, what is the s curve shape n why do games use it???" And heres an update everyone! It occured to me that u can create a s curve using 2 parabolas, 1 thats flipped upside down, and one thats right side up :) So, that got me thinking that maybe games arent using s curves stacked on top of eachother, maybe their using parabolas stacked on top of eachother, and then a lightbulb went off in my head haha. The data shown in my guide on how to deal with S curves (which is the data from my ow spins), is actually just 2 parabolas stacked on one another. For some reason this didnt occur to me and i thought i needed to keep using s curves to piece together the data. After this discovery, i recreated the ST using the same data points, but with 2 parabolas instead of 3 s curves. I just finnished the curve n it already feels better than before! I know a lot of people like the old curve though so i dont want to just strait up delete it, so ill post the updated curve in my config :)
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: kaiju on 02:04 PM - 05/17/20
No curve for PS4? :(
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 02:19 PM - 05/17/20
Sorry, but im afraid not atm :( Deadzones are different on the xbox from the ps4, so the curve wont work on ps4. BUT, assuming its only the deadzone that changes from console to console and the aiming properties stay the same, I might have a solution to this but its not 100% guarinteed to work so dont get ur hopes up. Itll still take a few days to get the info i need lol
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: strandomring on 08:41 AM - 05/23/20
How can we make retraining the ST a reality? I'm on PS4 and I've spoke to a few t500 players and they all concur that something is definitely "off" now about Overwatch and its aiming. They say something changed; I wouldn't know because I took a break for the game but am now back.

Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 09:24 AM - 05/23/20
mist4fun can check if Overwatch has changed, but, was there a recent update to the game that came out and did they specify that a change to their aiming was made?
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 12:18 PM - 05/23/20
No, the problem lies within the ST creation itself I think. If you own an xbox you can see this yourself with trying my overwatch curve compared to either of the custom ST's and doing accuracy tests with them. My curve is much more consistent even though i do it all manually because I think the way mist create's the ST's is flawed now that games are changing the way the aiming works. (It used to be that in just about every game, it was a strait exponential curve, but now it is an S curve shape in a lot of games) My theory is that mists translation method probably takes all the data his AI collect's, and puts them all together to make 1 exponential curve.... but the problem occures when the  game isnt a strait parabola because now the data is inaccurate. That is why with my method, I find the parabola between 3 data points at a time, and stitch the curve together that way. This gives you a curve that more closely resembles the game your playing and feel's more 1:1. Another game this is obvious in is apex legends, you can test the accuracy of each game very easily. All you have to do is line ur crosshair up with a target, move your mouse to the side of your keyboard, flick to the side, and slowly move your mouse back to the point. The closer u are to ur target, the more accurate the ST is. You do this multiple times with multiple flick speed's and soon u can see the curve isnt accurate in apex once u start flicking arround 75%-100%. Same thing in overwatch, but in modern warfare for example ur aim stays 1:1 no matter the flick :) Now, using my method, it showed me that the curve for modern warfare was strait parabola (Which is why i think this game works well with xim), but overwatch (A game everyone is having troubles with no matter the new st created or not) use a S curve instead. Im 99% sure apex also uses an S curve which would explain why Apex and overwatch are 2 games everyones having trouble with. Sure, doing my method all by hand isnt 100% accurate but it sure feels way more accurate in games like overwatch than the normal ST. If you dont believe me go ahead and read my guide i wrote on the topic, or look at my overwatch post's comments. Everyone loves my curves MORE than the normal ST's because i manually made this change lol :) So i think this is more of the ST creation at fault than the overwatch game specifically
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 12:22 PM - 05/23/20
This is not how STs are created.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 12:33 PM - 05/23/20
I dont want to give any name's incase this could get them introuble, but i have a VERY reliable source that told me the ST's are not created manually, and instead are created with AI. So all im saying is that i think the actually coding that takes the data the AI collects, and turns it into a curve needs rewritten. Because if this is not the case, then theres no way my custom created curves should be more accurate during these flick tests than the normal ST's, and its not just me who feels this since i have many players backing my curves up. Im not trying to argue here, im just trying to point out something isnt right and needs to be fixed :)
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 12:34 PM - 05/23/20
And until this is fixed, i wrote a guide on how to manually do this on any game so people can fix their games. I even have a config list that i am helping fix these broken games on lol. Overwatch is the game i took the most time on to create, so this one is the best example of what 1:1 should feel like compared to what the standard ST's feel like :)
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 03:28 PM - 05/23/20
Training does run via an automated computer-vision based system. However, this statement: "My theory is that mists translation method probably takes all the data his AI collect's, and puts them all together to make 1 exponential curve.... but the problem occures when the  game isnt a strait parabola because now the data is inaccurate", is not correct. We don't fit curves to the data -- that would be too limiting.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 08:00 AM - 05/24/20
But when doing accuracy test's on overwatch (flicking your mouse away from the keyboard and slowly bringing it back to see if the crosshair land's on target), the crosshair lands far away from target on the standard ST's, but when using my custom created curve, the crosshair land's very close to target (Also, this game has an S curve). But in a game like Modern warfare (a straight exponential curve), doing the exact same test's show's this game is completely 1:1. Not to mention, overwatch has 2 official ST's that both suffer from the SAME 1:1 problem's, so the chance's of this 1 game being just a fluke is very unlikely. Also, if this is not a problem with ST creation, why is it that my manually created curve is more 1:1 than 2 software created curve's??? That just doesnt add up in my eye's.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 08:11 AM - 05/24/20
Also, im starting my apex config today, and that game also is not 1:1 according to these test's. Id almost bet you this is because the game probably uses an S curve instead of a parabola, even though i have not gotten the data yet. I will however update this thread with my info on apex legend's if it is an S curve for even more proof that this is just a problem with the ST creation process. And again, im not trying to argue or come off as being rude :) I just think if people are going to pay for the xim they deserve to get as best 1:1 experience as possible.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 09:52 AM - 05/24/20
But when doing accuracy test's on overwatch (flicking your mouse away from the keyboard and slowly bringing it back to see if the crosshair land's on target), the crosshair lands far away from target on the standard ST's, but when using my custom created curve, the crosshair land's very close to target (Also, this game has an S curve). But in a game like Modern warfare (a straight exponential curve), doing the exact same test's show's this game is completely 1:1.

There can be 2 reasons for this. If you are flicking and you are hitting the maximum turn speed of the game, then, then you will not land on the same spot when you return. When doing tests like this, you need to be sure not to hit that maximum speed. The other issue is Overwatch has temporal acceleration. Normally we mark games like this with a 'U' in our support list, but, it has less than some of the worst games (like R6S). But temporal acceleration will definitely cause tests like these to fail. If you are able to compensate with a curve, it doesn't mean it'll work for all others (who have different setups and play style).
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 10:00 AM - 05/24/20
(1.) I am NOT flicking faster than the max turn speed. I am also flicking at the same exact speed's to test this on both so that's not it either (like slow flicks, medium flicks, and fast flicks on both moving my arm the same speed on each ST). (2.) If that is the case, then my curve would not be 1:1 and landing on the target no matter my flick speed (as long as its under the turn cap), because the aim acelleration would mess with my curve JUST like the official ST. But my curve land's on the target almost every single time no matter what I do, and the origional ST does not even come close to the target on faster flick's. If u have an xbox and overwatch, load the game and test it yourself.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 10:02 AM - 05/24/20
What's your curve look like?
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 10:06 AM - 05/24/20
>>> XIM APEX [Translate Ballistics] START COPY >>>
X5BC:hm6I/4qki/WNHY4njx6QBJDeka2SdJMyk+qUnJVIleeWeJcDl4aYCJiFmP+Zdpnzmm+a7Jtom+ScYZzdnVmd1Z5Sns6fSp/IoEagxaFEob+iOqK0oy2jqaQlpKGlH6WbphemlKcQ:X5BC
<<< XIM APEX END COPY <<<

If u take a look at the curve... About half way through the curve is where the exponential parabola changes to more resemble an S curve, which explains why flicking below 50% works relatively the same on both, but over 50% speed my curve is way more accurate than the standard. Also, this is on BOTH origional st's not just 1, so im very convinced it has to do with the S curve.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 10:06 AM - 05/24/20
1/3rd not half*
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 10:09 AM - 05/24/20
http://imgur.com/gallery/S5Z5Etp

Here is the desmos curve. It shows it much more better than the curve editor. As you can see, the curve is not a strait parabola, it is an S curve. In the beginning it goes above the exponential curve, then it goes below the curve arround 40% through the curve, then curves back up arround 80%
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 11:33 AM - 05/24/20
I can ask mist4fun to take a look.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 11:36 AM - 05/24/20
Thank you! :)
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: Addytoostrong on 04:39 PM - 06/08/20
Just curious if mist looked into this or if we are working on any updates?? If not it's cool but always curious.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: End on 07:53 AM - 06/09/20
the STs are flawed in some manner, as you can clearly see from the evidence provided. i have also tested this method, as well as the curves that have been created via this method, and can absolutely vouch that everything stated is legitimate. both OW STs need a proper retrain
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 08:54 AM - 06/09/20
Yeah, i wonder why mist hasnt touched on any of this yet? Many people are using my curve now, and almost every single person says its way better than the origional ST, so id think mist wouldve checked into it by now. Something is definitely messed up with both of the main ST's.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 11:13 AM - 06/09/20
Just curious if mist looked into this or if we are working on any updates?? If not it's cool but always curious.

He did and the STs are all still valid with OW's aiming system. In fact, we also have the same type of asymmetric swipe tests as mentioned here. However, ours is computer controlled. Meaning, they run exactly at specified angles and speeds for exact times (down the the microsecond). In all cases and tests, the reticle lands exactly where the test started.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: TheCityWok on 01:24 PM - 06/09/20
Just curious if mist looked into this or if we are working on any updates?? If not it's cool but always curious.

He did and the STs are all still valid with OW's aiming system. In fact, we also have the same type of asymmetric swipe tests as mentioned here. However, ours is computer controlled. Meaning, they run exactly at specified angles and speeds for exact times (down the the microsecond). In all cases and tests, the reticle lands exactly where the test started.

Off topic but is the updated Titanfall 2 translator near release?
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: Boomboks on 02:03 PM - 06/09/20
Off topic but is the updated Titanfall 2 translator near release? [/quote]

The Overwatch 60AEI is next buddy, wait your turn. But also, he's taken an indefinate break from making the STs. Might be essentially cancelled if nobody else does them 
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 02:28 PM - 06/09/20
Titanfall 2 is currently being worked on, I don't know what the ETA is right now: https://community.xim.tech/index.php?topic=30363.0
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 03:49 PM - 06/09/20
Just curious if mist looked into this or if we are working on any updates?? If not it's cool but always curious.

He did and the STs are all still valid with OW's aiming system. In fact, we also have the same type of asymmetric swipe tests as mentioned here. However, ours is computer controlled. Meaning, they run exactly at specified angles and speeds for exact times (down the the microsecond). In all cases and tests, the reticle lands exactly where the test started.

If that's true, then why is it that when I do these swipe test's I make sure I never exceed the maximum xim speed limit (no flashing), meaning im always somewhere in the curve, and I do this for like 5-10 minutes with different speed's so I can see if its 1:1 with my curve. With my curve, im very close to target every time (if not exactly on target), but when doing the same exact test on the origional ST's, its off target (For both the origional linear and exponential ST's)? Maybe the data from the computer isnt wide enough away so its not noticable? (Like maybe it uses 25% speed flick, then 20% speed back so its roughly the same spot, even if its not 1:1 its close enough. When I do the flick test's I flick at high speeds, and slowly come back. Medium speeds, and slowly come back, slow speeds, and slowly come back, so I get a big range of potential movement's and they all come back to be 1:1, but its not the case with the origional ST's.)
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 04:00 PM - 06/09/20
Also, anyone can do this to see what I mean. Just load up the training range, turn off your aim assist, and use the origional ST's with their setting's. Now, find a target, and aim ur crosshair so your directly on target and move your mouse so its touching your keyboard. Now, flick away from ur keyboard making sure u dont go above the xim speed rate (so it doesnt start flashing), and go back to the same spot on the keyboard in a slow motion keeping the mouse as strait as possible. The closer the curve is to 1:1, the closer your reticle will be on the target. Do this for like 5 minutes with different flick speed's and u can see the overwatch origional curve is NOT fully 1:1. Now, copy my xbox linear 33aei curve i made (ps4 was made from someone else so i cannot guarintee its 1:1), and do the same exact tests. You can see my curve lands very close to the target on any flick you do. So I have trouble believing the origional ST's are "100%" 1:1, when my manually created curve is literally doing better than either of them in these tests (made by hand, no AI)
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 04:09 PM - 06/09/20
I'm not sure. All I can say is that the systems we use are automated. Meaning, we call tell it so swipe out at 35-degrees at a certain speed for 1.5s and swipe back at 215-degrees at 1/3 of that speed for exactly 4.5s (and, of course, any combination). Everything lands in the exact place it started. Timing, angles, speed and stability are exact.

The main difference is, when you do it by hand, you are constantly changing directions (it's impossible not to do that). Perhaps the curve is helping compensate for that since there is still some temporal acceleration in Overwatch that may be kicking due to reticle instability.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 05:34 PM - 06/09/20
Maybe, but I think its unlikely since i spend 3+minutes spinning to create my overwatch curve (at every 5% increased stick tilt percentage). Any aim acelleration would have a very, very minimal impact on the curve since I also start the spin before my target, meaning there is no more acelleration happening by the time I actually reach the target for my first spin. And I have smoothing at 0, so if there is any acelleration it is a tiny ammount and wouldnt be causing this.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 06:16 PM - 06/09/20
I'm not sure what to say. If we change the STs, they will start failing our asymmetric swipe tests.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 07:56 PM - 06/09/20
Maybe the test's are flawed in some way from the program then? Im curious what would happen if mist tried running my curve through the test? That might show more insight on the problem. My curve definitely isnt perfect since its manually drawn, but I cant think of any other reason my configuration should be more accurate in my 1:1 test's than the origional ST's. I wonder if this is a problem in other game's as well, because when doing these tests in modern warfare's origional ST, it lands in the same spot everytime? The only game ive found that doesnt really work well is overwatch, and judging by the curve, this is the only game ive done with the wierd curve (Its not a strait exponential curve, its kinda wiggly). I wonder if maybe the tests go off the data point's themselves? Because that would explain why its 1:1 in the program. What i mean is if mist does 25% stick tilt, and 50% stick tilt as part of the data to make the curve, then flicks at 50% speed to a set angle, and then goes back at 25% speed, the program will show that its 1:1, but when flicking at something like 32% (some value thats not a data point collected), then maybe it wouldnt be 1:1 anymore??? I know modern warfare uses a strait parabola curve, and that game has very close to 1:1 movements, but a game like overwatch where the curve is wiggly, the 1:1 test fails?
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 08:33 PM - 06/09/20
We can use any mouse value, speed, and time we want. I think he did try your curve but didn't see the same results you are getting. I can ask him.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 08:35 PM - 06/09/20
Would it maybe be possible for mist to post the curve he created using the program so we can see them side by side and try to find out whats happening easier???
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 10:16 AM - 06/10/20
Would it maybe be possible for mist to post the curve he created using the program so we can see them side by side and try to find out whats happening easier???

Using what program?
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 11:42 AM - 06/10/20
Training does run via an automated computer-vision based system. However, this statement: "My theory is that mists translation method probably takes all the data his AI collect's, and puts them all together to make 1 exponential curve.... but the problem occures when the  game isnt a strait parabola because now the data is inaccurate", is not correct. We don't fit curves to the data -- that would be too limiting.

This program, the program that creates the curve. Would there be any way for us to see the curve it generated so we can compare my curve to the origional ST curve? I wouldnt even mind making a new exponential curve 1 last time with my method so we have the exact same settings so the curves are from the same data. That way, we can clearly see where the curve changes between the 2 and it could help us find out whats going on?
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 01:06 PM - 06/10/20
The ST trainer? It doesn't create discreet curves like you are thinking as it's not expressive enough to represent the actual math that goes into a game's aiming system. It would be too limited (like XIM2's "standard" translators).
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 02:51 PM - 06/10/20
But it has to have some sort of data it collects right, like a spreadsheet or something? Could I get that data to convert into a graph and get a curve to compare it to? I dont need a xim curve if thats what you thought, I just need the data so i can input it myself into desmos or something to compare the 2.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 11:57 PM - 06/24/20
The ST trainer? It doesn't create discreet curves like you are thinking as it's not expressive enough to represent the actual math that goes into a game's aiming system. It would be too limited (like XIM2's "standard" translators).

Okay, so i made an exponential curve using my method with the same exact ingame settings as the original Exponential ST. I didnt collect enough data so it still doesnt feel as 1:1 as i could make it (im going to add more data soon to get it as good as i can), but it still feels more 1:1 than the original ST. Is there any data u can send me that the ST trainer got for the exponential curve so we can see exactly where the 2 curves are different and find whats causing the issue??? Again, I dont need a xim ballistic curve to see this, all i need is the data collected by the ST trainer and i can convert it to desmos to see the 2 curves (mine and mists exponential curves) side by side to see how they are different from eachother and get more insight to the problem. You can literally see my curve is more accurate than the ST yourself by doing the accuracy test i described earlier, even if u have a ps4 since someone in my discord made a ps4 equivalent curve as well, and says its more 1:1 than the normal curve too. You cant just say the st trainer is 100% accurate when in real world examples u can clearly see its not. U dont even need to do accuracy tests to see this lol, literally drag ur mouse across ur mousepad fast to see how far u spin on expo ramp (without going over the max speed cap), and then do it going slow. I guarintee u the crosshair will not end in the same spot on the real ST.... not even close
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 12:03 AM - 06/25/20
And if i am right about the system being flawed in a game like overwatch, if the trainer is showing the original ST as 1:1, then if you throw my curve in there of course it will show mine is not 1:1, because its flawed. Im quite certain it is flawed though because doing the accuracy tests i described by hand SHOULD be landing on the same spot if its 1:1.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 11:24 AM - 06/25/20
I can tell you, for certain, that if we used a curve with our asymmetric swipe test, it wouldn't land in the same spot. Maybe there is a difference in in-game settings or character used? I can't think of anything else that would explain it.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 01:05 PM - 06/25/20
No, the ingame settings are completely the same and characters aiming stays completely the same between all heros. Also theres no way the ST trainer swipes are accurate if their landing in the same spot during the ST trainer test's, because just doing the simple flick and pull back test i came up with shows its nowhere near 1:1 lol, if it was accurate, you would be landing in the same spot when u flick and pull the mouse back to the same location but with the normal ST curve, its not even close to the same spot. Thats why i think the ST trainer is messed up in this game, and it ultimately means the trainer has a flaw and needs to be checked on a coding level or something.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: Jeff87356 on 01:27 PM - 06/25/20
OBsIV Iíve tested it myself and made a PS4 curve using his guide and I can assure you that when you use the curve itís more 1:1 than the actual ST. Feels much better in game too
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 02:05 PM - 06/25/20
Thats why i think the ST trainer is messed up in this game, and it ultimately means the trainer has a flaw and needs to be checked on a coding level or something.

The asymmetric swipe tests aren't done by the trainer, those are done using a separate tool that we run to ensure the results of the trainer are looking good. It's the same thing you are doing, except, the input is driven by a machine with perfect timing and speeds and also along all aiming angles (not just horizontal).
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 02:50 PM - 06/25/20
Okay, so what your saying is with the tests, the curve is 1:1 across ALL values??? Then why is it that doing a swipe across your whole mousepad slowly does not land the crosshair anywhere near the same spot as when you do a fast swipe across the whole mousepad? If it was 1:1, it would land in the same spot no matter how fast or slow you move your mouse across the pad. So why is the cursor not landing in the same spot or anywhere close when moving the mouse the same exact distance as when moving the mouse fast??? U cant say the test is accurate when theres proof showing it is not by just doing some 360s. Idk, maybe the test is accurate and mist just didnt check enough values to see where that it isnt 1:1. What im thinking is when you do the test by hand, you actually fluctuate between all the different speed values, and when u do the test your going a set speed the whole time. This means during mists test, hes only testing 2 values at a single time, but my test is fluctuating between the whole curve so maybe this is why mists test says its 1:1 but doing it by hand shows its clearly not??? Like lets say we have 2 exponential curves. 1 curve is (y=x^2), and another is (y=x^3). Both curves will pass through the points (0,0) and (0,1), but the data between the points is completely different. So, if you do mists test using a speed of 0, and a speed of 1, it will say that the curve is 1:1, but doing it by hand you will fluctuate between the whole thing and thats why its not 1:1 when done by hand?
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 02:54 PM - 06/25/20
I mean i know theres more tested then just 2 values, but the concept stays the same. If you test only a few points its not enough testing, and another thing that could corrupt the data is if you do the tests with the same data points the translator got, because if it got points 0,0 and 0,1, and you test a speed of 0 and 1 then of course it will be 1:1, but u have to test things like .33 speed and .66 speed to see if it stays 1:1 if that makes sense?
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: antithesis on 03:36 PM - 06/25/20
The biggest issue with your methodology is human error. There's no way you can replicate the precision of a computer that is programmed specifically to map and test exact vectors and velocities.

The most important factor is console look mechanics are based on velocity, not distance. Unless you can exactly replicate the vector and velocity, you will see variance.

You're a human being, it's impossible to have anything other than variance because no test can be exactly replicated by the nervous system. OBsIV's tools on the other hand precisely replicate the vector and velocity, the results of which are both on-target and replicable.

If the OW ST feels off, it feels off, but manually creating a curve using nothing more scientific than guesswork and claiming the people who have been doing this for over a decade and have created hundreds of STs as having a flawed system is a bit rich when your own methodology is riddled with flaws (it's manual, it can't be anything but flawed).

By all means keep doing what you're doing - XIM allows us the freedom to tool around with these settings ourselves, but user bias is an amazing thing. I know for a fact that XIM has published firmware updates that have zero effect on the look mechanics of Smart Translators, yet XIM users will claim that Game A or B feels better or worse, when it's exactly the same.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 04:00 PM - 06/25/20
I never said im perfect, in fact i said my curves arent perfect and im human so doing this by hand will never be 100% precise. But saying just because i dont move 100% perfectly, my whole method is flawed is completely wrong. You dont have to move 100% in a strait line to tell if something is off or not, were not talking about a few pixels here. With my sens, moving slowly across my mouse pad i do arround a 340 degree turn, when moving from my keyboard to the edge of my desk (in a strait line). Then, when i move the mouse fast i do over a 360 from the same exact spot (in a strait line), i do arround a 380. Thats not a few pixels that just proves somethings off, literally check it yourself. It is 100% obvious something is flawed because doing this in a game like modern warfare, itands 100% in the same spot, doing the same test on the same sens and settings. Overwatch is the 1 game where this is very off, literally do this yourself and u will agree with me just like everyone else who uses me and jeffs curve.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 04:01 PM - 06/25/20
And i also never said my curves are perfectly 1:1, i just said its more 1:1 than the ztandard curves.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 04:09 PM - 06/25/20
Also are curves arent guesswork btw, we gather a lot of data to make the curve, and then use math to make a curve through the data. Its not 100% accurate but it is close (like 95%+).
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: antithesis on 04:16 PM - 06/25/20
Everyone...who's everyone? You've had a few people agree with you, which is natural when they feel things are off themselves and are willing to try anything to fix their issue. We've seen sheep fawn over a few oddball concepts over the years, simply because they feel different.

Eventually, they all go back to vanilla Smart Translators because it's the most accurate mapping of the game's look mechanics. The one exception to that rule is mjfame's config - his reasoning and technobabble was all over the place, but his results worked in the games he tested. They're not better than the default ST, just different and somewhat effective.

As a human being, you cannot, ever, replicate the exact same vector and speed. That's a huge flaw in your methodology. Until you can plug that gap, your results can easily be dismissed due to factors such as mouse sensor, mouse surface, general cleanliness, display lag, input lag, network lag and simply being human.

We know that some console games have asymmetrical deadzones, Fortnite being the prime example. The issue of under or overshooting can be caused by lopsided deadzones and it shows up in the trainer. The games are also marked U for Unpredictable Aim in the games support list.

Again, console look mechanics are designed around velocity not distance. You will have two different results in every single test, unless vector and velocity can be matched precisely. Only one of the two parties in this discussion can make that claim and the results are consistently on the mark.

Claiming that XIM's system is flawed because the results don't match your own flawed methodology is just plain silly. We've seen this sort of thing several times before where a XIM end-user knows better than those who designed, built and coded the device. Unless you can create a scientifically accurate and replicable test to demonstrate your hypothesis, your results can't be anything but invalid.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: antithesis on 04:17 PM - 06/25/20
Also are curves arent guesswork btw, we gather a lot of data to make the curve, and then use math to make a curve through the data. Its not 100% accurate but it is close (like 95%+).
You're doing it by hand and by feel. That's guesswork.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 04:25 PM - 06/25/20
No, we actually arent??? U even quoted what i said?????
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 04:29 PM - 06/25/20
What we do is map the right stick to the arrow keys, then bind walk button to the right arrow key so when you hold it down you spin at that exact stick tilt the entire time.... Then, we record this using recording software on the phone with frametimes and count spins for about 3 to 5 minutes for each data point. We then find the exact frame that it is closest to a full spin and divide that time by the ammount of spins to get a accurate 360 time for each data point. We then enter this data into desmos, and use a line of best fit finder through each data point to stitch together are curve. This process takes hours to do, and is not just "feel and guesswork". We do 30+ spins to get are data... sometimes 50+ and we do this for many many values, so saying we are guessing couldnt be more false
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 04:30 PM - 06/25/20
I wrote a whole guide on this topic, i explained this many times, this is a accurate way of creating curves. Ask anyone that uses my curves and they will tell u they are more accurate than the official ST curves on overwatch
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: antithesis on 04:51 PM - 06/25/20
No, we actually arent??? U even quoted what i said? ??? ?
Not 100% accurate. Say no more.

What we do is map the right stick to the arrow keys, then bind walk button to the right arrow key so when you hold it down you spin at that exact stick tilt the entire time.... Then, we record this using recording software on the phone with frametimes and count spins for about 3 to 5 minutes for each data point. We then find the exact frame that it is closest to a full spin and divide that time by the ammount of spins to get a accurate 360 time for each data point. We then enter this data into desmos, and use a line of best fit finder through each data point to stitch together are curve. This process takes hours to do, and is not just "feel and guesswork". We do 30+ spins to get are data... sometimes 50+ and we do this for many many values, so saying we are guessing couldnt be more false
That's a clever approach. It's less error-prone than using a mouse, but it only works along at most 8 vectors. I'm assuming you're only doing it for 2, which takes a few hours. I'm also assuming you're mapping each data point based on Walk values of between 1 and 100 to generate the curve. It's as scientific as you can make it, but you have gaping holes in your data due to the limited number of vectors and values you're testing.

XIM's look mechanics are mapped along all vectors by computer vision, artificial intelligence and machine learning. Let's call it 360 vectors, minimum. It takes 24 hours per view, i.e HIP & ADS (depending on game) and per platform i.e PS4 and XB1. XIM's Ballistic Curve fidelity is far higher than 0 to 100, it contains 10,000 data points.

Subjectively, your curve feels great, to you and to like-minded individuals on Xbox One. On PS4, you're not getting the same feedback. Objectively, you can't map the entire look mechanic, whereas XIM can. If mist used your methodology for the official STs, XIM would be laughed out of business.

Thankfully, XIM provides the ability to do what you're doing, so continue doing it if it yields good results. But calling their system flawed based on your very limited testing and methodology is still a very long bow to draw.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 05:10 PM - 06/25/20
Okay, but still after using my method to get a curve from like 15-20 data points, it still is more 1:1 when doing my accuracy tests. You cannot say that when moving my mouse across the mousepad slowly on the normal ST, and the game does like a 340, and then when moving it fast it goes 380 that that is 1:1. If it was 1:1, it will land in the same spot if you move the mouse the same distance, no matter the speed. When doing this with my curve, it lands like 360-380. Its not perfect, which is why im going to add more data points to get my curve more accurate (jeff already did this with his ps4 curve), but the fact that my manual curve is more 1:1 than the original ST with only 15-20 points means theres something wrong with the original ST.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: antithesis on 05:29 PM - 06/25/20
If it was 1:1, it will land in the same spot if you move the mouse the same distance, no matter the speed.
Your mistake is thinking of your mouse as a mouse. It's not, it's a giant thumbstick. XIM translates mouse vector and velocity into thumbstick coordinates, not physical distance.

Distance is irrelevant for console, velocity and vector is 100% of the ballgame. You are accounting for that by using WASD and Walk as vector and velocity modifiers, so kudos for thinking it through to eliminate other variables, but your testing is still very rudimentary.

My only objection here is the use of the word "flawed" when describing the Smart Translator system. It's far, far more accurate than you can hope to be and the results of similar tests for vector and velocity are 100% accurate and predictable. Unfortunately, as human beings, we're not.

I think what you're doing is neat, but it's very labour-intensive. You've revised your methodology a few times, which means you're getting closer to a more reliable result. But, it's nowhere near close to what the Smart Translator system is doing to map the look mechanics. Your curves are a good supplement to the generic XIM STs and it's the best methodology I've seen to date, but it's not "better" than the XIM ST at mapping the entire look mechanic.

RML has made a tonne of Ballistics Curves over the years and in many ways, they do feel "better" than the default ST. He adjusts the velocity at points along the curve, which adds acceleration or deceleration, but it also diverges from XIM's 1:1 mapping of the entire look mechanic, which is what you're doing too.

As I said, keep doing what you're doing and let the results speak for themselves. I appreciate that your objective here is altruistic and that you may have something to contribute to improving the XIM. But as OBsIV implied in a previous response, you're doing it via circa XIM 2 methods from over a decade ago that pre-date the Smart Translator system.

They've checked the Overwatch look mechanic in the trainer and have found no changes. They've tested their data using precision vector and velocity tools and have found it be 100% accurate and predictable along all vectors. If I read it right, they also reviewed your data and found it to be inaccurate. But if it feels good, use it.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 05:38 PM - 06/25/20
I should note that a translator will model all reticle movement angles but also changes to behavior based on speed too (which is the norm). That said, horizontal is very important a d if there is an issue with any of it was want to know about and improve if possible. The issue here is that there is no area to improve and we aren't getting the same results as being reported using the curve noted using the Generic Alpha ST.

mist4fun tried it and this was his findings today:

"Ok, I read all of it and found their shared config. They are playing on Xbox but had someone else on PS4 share some of the PS4 settings for the curve.

I tested on Xbox with:

Their config settings
Sensitivity 1
Two keyboard wrist wrests 20Ē part
Slow forward swipe and fast return swipe
Within the turn speed limit

The person plays on Linear so I tested it first. Their config is not linear. I did three swipes using a slow forward swipe and a fast return swipe among many others but hereís that data:

800 DPI = The return swiped moved about a quarter of the distance needed to reach the start position.
6000 DPI = The return swiped moved about half of the distance needed to reach the start position.
12,000 DPI = The return swiped went beyond the start point by about a half of an inch.

I tried using their config settings that they shared for Exponential but there is a deadzone so it doesnít work at all.

Ö Iíve now tried current support, both Exponential and Linear on X1 and in real world conditions 2/3 tests return to the start position and each of the return swipe that didnít was very close. If I didnít have tools and were just using these tests and feel I would say itís linear.

I canít explain why they think that itís easy felt in swipe tests, itís been the opposite for me. About how they are getting the data, maybe prolonged spinning is less reliable with some mechanics?"
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 06:18 PM - 06/25/20
Okay, 1:1 movenents mean if you move 10% mouse speed, then you move 10% turn speed in game. 5% mouse, 5% in game. With a perfect 1:1 ratio, the graph for turn speed and mouse speed will be x=y, a strait line. Therefore, as long as you dont go over the max turn cap speed for the xim, it will not matter how fast you move your mouse u will still end in the same exact spot if you move the mouse the same distance regardless of speed.  It makes 0 difference if it is emulating a analog or not if the movements are 1:1. Next, idk why when u do my accuracy test, you get worse results using my curve with my settings than the original STs because literally everyone besides you who i did my test says mine has better results. But its obvious this is turning into an argument rather than a actual discussion, this whole post is about spreading the info not bashing the mods or anything. But i guess we will just have to agree to disagree and let users decide because i have said everything i had to say and u can choose to do whatever u want with my findings.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 06:21 PM - 06/25/20
And 1 more thing, having more time w the xim actually means literally nothing. Its like trying to say "Im older than u so im right" which is not a way to win an argument lol
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: antithesis on 06:27 PM - 06/25/20
And 1 more thing, having more time w the xim actually means literally nothing. Its like trying to say "Im older than u so im right" which is not a way to win an argument lol
No-one made that argument.

OBsIV invented XIM and wrote all of the software. No-one knows what it does better than him. It's gone through 6 generations in just over a decade. Experience and product knowledge is the argument, not time.

If your curves work, use them. Many people use RML's too. But tossing the word "flawed" around without the level of product knowledge that XIM staff has is bush league and merely demonstrates your lack of experience.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 06:35 PM - 06/25/20
If the OW ST feels off, it feels off, but manually creating a curve using nothing more scientific than guesswork and claiming the people who have been doing this for over a decade and have created hundreds of STs as having a flawed system is a bit rich when your own methodology is riddled with flaws (it's manual, it can't be anything but flawed).

Right here u say because they did it longer their right. Even college professors have proof readers for their paper's so time means nothing.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 06:38 PM - 06/25/20
Also, you forget i came up with the method to literally use the xim to make a 1:1 curve for any game so i do have knowledge on how curves work, and the math behind them. Also, earlier u said ps4 players have different experience, thats because i only have an xbox and ps4 has different deadzones which means it needs its own curve made, and people like jeffs curve more than the original ST on ps4 too.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: antithesis on 06:40 PM - 06/25/20
Right here u say because they did it longer their right. Even college professors have proof readers for their paper's so time means nothing.
Experience matters. As does replicable results. Their results stand up under scrutiny. Your's do not. Armchair experts are less reliable than experienced, qualified professionals.

I don't want to devolve into an argument either. You've shared your opinion and XIM has looked into it. That's all you can expect them to do.

If something good comes of it, great. If nothing happens, just accept that your curves work well for you.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 06:43 PM - 06/25/20
Ive said what i needed to say, its not guesswork im doing, and other people who do my accuracy test can tell you the same thing im saying. I have nothing else to say, i can personally feel my curves are more 1:1 and if u guys dont think the normal ST is flawed then i cant change your mind. Ill just let the users decide since my curves are public.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: antithesis on 06:54 PM - 06/25/20
In which case, thanks for sharing. I'm sure your efforts are appreciated.

If nothing else, you've taken a more scientific approach to curve creation for the X-axis. I'd be interested to see your Y-axis and 45 degree vector results.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 08:26 PM - 06/25/20
As I said, I always want to know how we can improve the system. Maybe what mist4fun said is right -- it could be the difference here is how the game behaves when observing this while spinning for multiple seconds. We don't do anything like that. But, if he says that he can't reproduce the behavior, I trust that and can't assume there is a general issue that is affecting everyone. We need to be really careful about updating STs since people are very sensitive to any changes. I'm not sure why there is a difference. In order to investigate it, we would need to be able to reproduce it. I do appreciate the effort to try to identify issues. Especially when it comes to the translation system.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 09:09 PM - 06/25/20
Yeah I agree. I think the xim is awesome and works great in a lot of games, and i didnt expect mist to just delete the normal ST and use mine or anything like that, all i wanted was to get the info out there that this game in particular doesnt feel too good compared to other xim ST's made, and that there could be a problem hidden somewhere that isnt obvious right away. Here is the info i have to share summed up:

1. Overwatch uses a wavey curve instead of a strait exponential curve or linear curve. (only game ive tested that has this, and is also the game i have the most issues with the normal ST's so maybe this is correlated in some way)
2. Since the curve is wavey, it takes me a lot more time to get enough data to make a 1:1 curve because 1 exponential curve doesnt accurately represent the data.
3. When i do my accuracy test (lining the mouse up with the keyboard, flicking away at various speeds and slowly coming back to see if crosshair lands in the same spot), my curve performs better for me than the normal ST's.
4. When doing 360s on the normal ST, my crosshair doesnt line up in the same spot depending on the speed i used. But when doing this on my curves, it lines up closer for me.

Here is what i think could be problems:

1. The ST creator maybe creates its curve using a single exponential ramp, which would not match the data perfectly.
2. The ST tester could have a fault in it, like using data points that the ST trainer used (would say it is 1:1 when in reality it could be off).
3. The ST trainer not testing enough data samples.
4. Tests/creation were done with aim assist enabled by accident (unlikely but possible. A 0 window size still has aim assist, maybe mist thought 0 would disable it or something)

Maybe the xim team would wanna investigate further after i add more data to my curve, but thats pretty much all i have for now. Again, im not bashing the xim team, or saying their wrong, i just want people to have good setups and im trying to think of whats causing people to have issues, and possible solution's. Users can choose what they want to use, and I understand people are scared of change. Thats why i keep even my old, inaccurate curves still up so people can choose what they want to use the best. If i find anything else I will post it in here.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 10:00 PM - 06/25/20
So, is it not possible that there is a #5 on your list of potential problems -- that is your curve actually doesn't do what you think it does for everyone's hardware setup?

I'm not going to disclose our IP,  but I feel you've concluded that the way that our trainer works is in the simplistic terms as you've described. It doesn't work like that and if I were to describe it in terms of numbers of curves (as you've defined a curve) then it would be in the millions (all in a single translator). You've made the assumption that the angle of the reticle always matches the angle of the stick and all that needs to be considered is the vectors magnitude when doing the transformation. This is why focusing on horizonal only works (for some games).
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 11:28 PM - 06/25/20
I dont see how hardware setup could affect the curve? Dpi change just affects how sensitive ur mouse is, sensitivity just affects how sensitive ur mouse is, and polling rate just affects how fast ur mouse gets the data. As long as you use the curve with generic alpha, and dont use any boost and the right ingame settings, then idk how hardware could change the curve at all. Again, im not saying my curve is exactly 1:1, im just saying with all the players in my discord, and me have all felt that my curve is more 1:1 than the original ST. My curve still needs to be worked on, its not a final product yet
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: mist4fun on 11:17 AM - 06/26/20
I'm not sure how our results are different in the swipe test but my results can be reproduced. If you go through the step list that was shared you should have the same results.

My best guess about the data you've taken is that the game's look mechanic might be behaving differently when spinning for a prolonged period of time. Some games have timed acceleration that activates only when the stick has been maxed for X amount of time so it could be this.

On why you and other people think that your config feels great it may be that it behaves better with aim assist. The Aim Ease in probably changes things a lot too.

About the mouse being another variable in testing, some mice use sensors that have positive or negative acceleration. You would think that sensor makers would want a linear sensor but it's more complex than that. Small corrections are also built into the sensor to make tracking better, sometimes they can get it wrong though. Your mice might not even have this problem but I can't find which models you're using.

edit: I went through the thread again yesterday and I just wanted to say that we appreciate the report and we did take the the original report seriously. All four STs were checked and and passed acceleration tests using automated tools. This time around I did manual swipe tests and logged acceleration results for both your config and current support.

In the end you should use what feels best. If this config feels and performs better than current support then go for it. One of the great things about XIM are it's advanced features that allow you to make changes that you prefer over default settings.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 04:16 PM - 06/26/20
Thank you for looking into this, but i do all my curve tests and swipe tests with no aim assist at all so that has no effect. Also, when i record my spins for 3-5 mins, i check the first spins rough time compared to the final time i get when i do the math, and its always basically the same time, just a few tenths/hundreths off from the final value which means there is no acelleration (or if there is its such a tiny ammount it would have such a minimal impact its not noticable). And since i do all my curve work from these spin tests without aim assist, there is no bias at all or going by feel to edit my curve. Im not sure how u can replicate 1:1 mouse movements with the standard ST because I cannot do it with my swipe tests no matter what i mess with, and with my curve its not perfectly 1:1 either because my curve isnt 100% finished yet (i still need more data), but it still lands noticably closer to the target when doing the same exact swipe tests on the same exact ingame settings and sensitivity. Idk whats happening since i dont know how they are made, im just trying to spread the info that at least for me, this game does not match up 1:1. Also, you said maybe there is acelleration added once u hold the stick down for X ammount of time, could there maybe be something like that but it happens in the very beginning and u guys arent holding the spin long enough during the test/creation???

And again, I dont want u guys getting the idea that im saying u guys are bad or anything like that, im just trying to spread the info that with my tests, this game is not 1:1, but doing the same exact test with the same exact sens in any other game ive played so far (like cod modern warfare for example), using no curves at all, my movements are 1:1 and line up perfectly. Its just this 1 game where both the standard linear and standard exponential configs dont line up for me. Maybe i can make a quick youtube video later demonstrating how i do my 1:1 test and show the result's from both w included xim settings and ingame settings so u can see im using the same settings and right settings in both.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 04:22 PM - 06/26/20
About the mouse being another variable in testing, some mice use sensors that have positive or negative acceleration. You would think that sensor makers would want a linear sensor but it's more complex than that. Small corrections are also built into the sensor to make tracking better, sometimes they can get it wrong though. Your mice might not even have this problem but I can't find which models you're using.

Also, I didnt think about how people could be using mice like this so thanks for pointing that out, but for me personally I use a g502 hero on 1k hz and xim set to 1k hz as well, and 3200dpi/12k dpi based on how im feeling. And have tested the 1k hz and my mouse can handle it. Also, doing my swipe tests in other games shows its 1:1 so it cant be my mouse thats having issues here either.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 04:32 PM - 06/26/20
In the end you should use what feels best. If this config feels and performs better than current support then go for it. One of the great things about XIM are it's advanced features that allow you to make changes that you prefer over default settings.

Also i really love how we have the ability to do this, but this post isnt about "what feels best" at all so im sorry if i made it sound like that, its about the aiming not being 1:1 in my swipe tests. Sure, we could say that my swipe tests arent 100% accurate because im human, but i do this same exact thing in every game and every single 1 lands in the same spot showing its 1:1 using the default settings with no curve, boost, or anything like that. Overwatch is the 1 config i found that doesnt, so im trying to figure out whats messed up with it and spread the info. My configs do work better for me and a lot of others, but its also not "plug and play" since u need to add the right curve and stuff, and its task is the complete same as the default STs, creating a 1:1 translation of mouse movements to ingame movements.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: antithesis on 04:44 PM - 06/26/20
Also, when i record my spins for 3-5 mins, i check the first spins rough time compared to the final time i get when i do the math, and its always basically the same time, just a few tenths/hundreths off from the final value which means there is no acelleration (or if there is its such a tiny ammount it would have such a minimal impact its not noticable).
It needs to be the first spin value as that's the only one with the initial acceleration. We don't spin our way around the map at 100%, we start from 0 and ramp through acceleration for 99% of mouse movements.

A tenth of a second, or 100ms, is a MASSIVE amount of acceleration, not tiny. Even 10-20ms will have a noticeable effect on turning. XIM accounts for this acceleration along all vectors and all velocities. It also accounts for velocity jumps as you ramp up from 0 to 100% turn speed, whereas your method is only measuring 100% turn speed for a given data point (there's still acceleration to get to that point).

The difference here may simply be that OW has poor acceleration modelling. This tends to happen in game engines that are built primarily for controllers. I recommend testing Destiny 2 if you have the game as another example of an acceleration-heavy look mechanic. Fortnite's also worth testing. Well established PC game engines in CoD, Battlefield, Titanfall etc are much more mouse-like in their behaviour and don't exhibit the same acceleration problems. These games always feel the best when using m/kb on console (though AA has poisoned most of them in recent years).

My configs do work better for me and a lot of others, but its also not "plug and play" since u need to add the right curve and stuff, and its task is the complete same as the default STs, creating a 1:1 translation of mouse movements to ingame movements.
The objective is the same, but the task is nothing like XIM mouse translation. You're accounting for one or two vectors (0 degrees and 180 degrees) at a fixed speed for each data point, whereas XIM apparently accounts for millions of vectors and velocities.

As said, if it works well for you, use it. But I'd caution against investing too much time into the process of creating a curve that assists just the X-axis as most mouse movements are along X&Y vectors. I like and support what you're doing as it's as scientific as the task can be given the tools on hand, but don't toss around words like "flawed" and "fault" when you (and I) have absolutely no idea of the complexity involved in creating a Smart Translator.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 05:41 PM - 06/26/20
Oh yeah, i understand that but the thing is with my method, im only able to record at 30fps, so i have to judge where about the exact time the crosshair touches my target, and then judge about where the exact time is when it crosses it again. This is why my final time is tenths or hundreths off, its just because i cannot perfectly see when it crosses... But if i do more spins, this number gets more and more accurate over time because of how many spins im doing. And, I do spins for each individual stick tilt value i chose from (99% out of 100% is not max tilt) so i doubt the acelleration is there. Also, i did think about this in the beginning, as u can read in my guide i start my spins a bit to the left of my target just to account for starting acelleration if there is any so by the time i reach the target im spinning at max speed for that stick tilt value already. Of course, if there is starting acelleration there would be no way for the xim to get rid of this at all, which is why i try to start the spin before my target so the acelleration is all gone by the time i reach my target. Also if there is acelleration, both of our configs would feel not 1:1, but in my tests my curve feels more 1:1, and i havent even finnished getting it as 1:1 as i could using my test. Also, i think i get what ur saying about the angles, does the ST take into account every single stick angle? (like instead of just horizontal movement it also does like diagonals and stuff?), because that could be the cause right there.... Maybe overwatch has different aiming curves depending on where the stick is actually positioned (whether your going up and to the left, or strait to the left) because my curves only account for atrait horizontal movement, and my flick tests. This could be why my horizontal movement feels more 1:1 than the official ST but maybe the official ST has better diagonal movement than mine and the loss in horizontal 1:1 is just a side effect of better diagonal movement??????
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 05:43 PM - 06/26/20
If that is the reason, would there be any way to update the ST to map out the curves on all stick directions (or like 8 of them), so horizontal movement stays 1:1 while still keeping the diagonals 1:1???
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 05:45 PM - 06/26/20
If that isnt a possibility, my input is focusing more on horizontal movement as that is what is most used in games.

Again, i might not be fully understanding how it works, but if it is the case that could explain why their different.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: antithesis on 06:42 PM - 06/26/20
I'd check the Y-axis and see if it produces similar results. Check one X+Y vector too (bind W & S to the same key and see if it travels at 45 degrees).

If they all match, you're onto something, at least in terms of curve creation with generic STs.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 06:55 PM - 06/26/20
Well the thing is, my curve creation requires to be spinning for like 3 minutes to get an accurate data point... And, since u can only bind up, down, left and right, the most i could do is 45 degrees, but i would need to hold the 45 degree spin for 3 minutes to get accurate data, and thats not possible since after a few seconds i would hit the vertical limit and stop looking upward, so unless im missing something this wouldnt be accurate enough to test that. Also another problem would be not being able to see the target to line it up correctly after doing a spin, which would make counting the spins really inaccurate.... so unfortunately i think its best to leave the non horizontal movement tests to the official ST trainer. I think ive come up with the best way to make a 1:1 curve given the tools i have, but i can definitely see how doing vertical tests as well and mashing all the data together could be causing the difference in horizontal movement and thats why it feels off. Maybe overwatch just has a wiere aiming system where the vertical curve is different from the horizontal one so when mashing the data together to make just 1 curve will produce a curve that doesnt 100% match either direction, but is a middleground between the 2 so 1 doesnt feel worse than the other??? I havent tested vertical movements with either curve, but i suppose if i move my keyboard to the bottom of my desk, touch my mouse to it and flick up, then come back down slowly you could test how 1:1 it is the same way i do with horizontal movement.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 07:15 PM - 06/26/20
I did just think about how the xy ratio could be found though... First, you find the 360 time like normal on the horizontal field, then, you shoot a bullet in the floor and in the cieling so u have a start and end point, and these will be 180 degrees feom eachother, so now you find the time it takes u to go from the start to finish at 100% stick tilt and then divide ur 360 time by 2... If they match, then theorietically you would have a xy ratio of 1.00, but if they dont u can divide the vertical 180 time by the horizontal 180 time to get the xy ratio more accurately than going by feel haha. However, once u find the xy ratio and use that, but you do micro circles and your actually doing ovals, this would mean the vertical curve is different from the horizontal curve and would need a separate vertical curve to have a true 1:1 experience.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 07:18 PM - 06/26/20
I think im going to try this out in the next few days and let you guys know what i find. I know finding the time of vertical 180s isnt totally 100% accurate, but it should be close enough to feel if the 2 directions need a different curve, and that would ultimately explain why the horizontal swipe tests dont match on the standard config for me.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 04:56 AM - 06/27/20
Okay, so i just finnished updating my config and noticed it was still not quite 1:1 yet (it should be very close with how much data i used), so i decided to test vertical movement using my method the same exact way and discovered on 100% stick tilt, the xy ratio is suppose to be "1.44" to be 1:1, which is close to 1.5 like i was using, but then i tried using 35% stick tilt (a data value i had in my normal curve, so after doing the math using this point it should still be 1.44) the xy ratio was suppose to actually be "1.20" to have 1:1 movements. This means that the game in fact has 2 separate curves for vertical and horizontal movements (at the very least 2 curves). Since i discovered this, it makes sense that my swipe tests werent exactly 1:1, so what i did was turned my xy ratio to 0 and used my updated curve to do only horizontal tests, and it was a lot closer to 1:1 than before. Not perfect with very slow movements, because i can only put a certain ammount of points in the curve editor, but with speeds of like 6% speed to 100% speed it lines up 1:1 (Taking into consideration that since im human, i cant move my arm in a perfectly strait horizontal line meaning my crosshair will be a little off target even with a pure 1:1 curve). Basically, from doing this i discovered the game uses 2 different curves and this might be whats causing the inconsistencies in the main ST creator if it doesnt create 2 separate curves for vertical and horizontal movements.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 06:27 AM - 06/27/20
Later, im going to test a few more stick tilt values just so there is enough data to 100% say for certain the vertical and horizontal directions have different curves.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 03:15 PM - 06/27/20
Okay, did more test's and the xy ratio seems to actually be close to 1.44 throughout the curve. Mustve misjudged the 35% stick tilt by like a frame because the rest of the curve seems to stay the same... So maybe this game just has unpredictable movement and isnt able to get perfect 1:1 movements???
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 03:23 PM - 06/27/20
I really think overwatch must have an unpredictable aim pattern now. If this is the case (which i think it is), then can we update the official ST settings page to say that the game has unpredictable movement like other games???
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 01:21 PM - 06/28/20
I really think overwatch must have an unpredictable aim pattern now. If this is the case (which i think it is), then can we update the official ST settings page to say that the game has unpredictable movement like other games???

Overwatch used to be marked with a "U" for this reason, but, we removed it because one of the updates in the past made in better in this area (i.e. less temporal acceleration). I'll speak to mist4fun about it.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 02:27 PM - 06/28/20
Okay thank you! This would explain everything haha, I always thought the game had nothing wrong with it so thats why I was so confused as to why the movements dont line up 1:1, but if there is unpredictable aim then that explains everything lol
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: OBsIV on 02:46 PM - 06/28/20
Blizzard isn't the worst on designing quality aiming systems, but, there are up there. I'm not sure why this is so hard for some game makers. People using a XIM tend to notice it more, but, whatever we notice, people using controllers will notice too.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 03:17 PM - 06/28/20
I personally never noticed it when i played controller before my xim, but when i got my xim, compared to other games, it feels really bad to me... Then, when i looked at the settings page to make sure it was set up right, it doesnt say anything about the game having unpredictable movements so i assumed it felt this bad because a problem with the creation itself. As for why my configs feel better than the original for me and others, i still have no clue... maybe its just the way the original ST creation differs from my creation? I spin for 3 to 5 minutes on each point and only on the horizontal axis trying to get rid of any acelleration at all, where it sounds like the original ST doesnt spin but instead flicks in each direction (could be wrong on this but its still different from mine in some way)? Either way, if the game has unpredictable aim then neither 1 will be perfect so that explains why they feel different. I can suggest that in games w temporal acelleration, spinning for longer to make sure it has virtually no effect would help i would think, if you guys want to take that into consideration when creating ur ST's. I realize this wont help in every game, but this might explain why my config feels better to me and others.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: antithesis on 05:33 PM - 06/29/20
The ST trainer doesn't flick, that was mist manually testing to replicate your result.

The trainers maps the entire look mechanic by recording movement along all vectors at all velocities and working its magic for deadzone size, shape, acceleration, velocity jumps etc.

As OBsIV said, that'd be akin to creating millions of curves in one ST, but it doesn't work that way (we don't know exactly how, that's XIM Tech IP).

You can see a little bit of that automation in this video (keep in mind this is from 9 years ago) -

Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 06:24 PM - 06/29/20
Yeah definitely not flicks, but its still similar (moving in 1 direction for a short ammount of time and then changing the angle). Maybe since the translation is changing angles so much, the acelleration would affect the ST more than my curve creation where i spin for 3 minutes at one angle? I realize the translation process is probably different but still very similar
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: antithesis on 06:29 PM - 06/29/20
The trainer moves constantly for 12-24 hours per viewport (HIP vs ADS) along all vectors and velocities to map the look mechanics. You're way oversimplifying it.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 06:37 PM - 06/29/20
Okay, but the amount of time u do circles doesnt matter if the acelleration kicks in when changing angles, and if the accelleration does change that could explain why my configs feel better for me and many others. Im not saying its a simple process or anything, im just saying if the accelleration keeps changing it will mess up the data collected
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: antithesis on 10:21 PM - 06/29/20
XIM accounts for changes in acceleration and velocity in OW, that's what OBsIV means by temporal velocity and velocity jumps. OW has had this issue from day dot, but it has improved significantly over time to the point where the game is no longer marked "U". That doesn't mean OW's look mechanics are perfect, but it's in the best state it's ever been.

OBsIV is a computer scientist who has built the most advanced console look mechanic modelling in the world over the course of 12 years. Please trust that he knows his job and there is nothing you've mentioned that hasn't already been considered.

mist has stated that testing showed that the default OW ST has more predictability than your curve. If you like it, use it, but it's futile to keep looking for fault where none lies. OBsIV has got this covered.

As a new player in the FPS market, Blizzard's look mechanics in OW feel nowhere near as crisp on console as the established Unreal, Frostbite and iD Tech engines powering most PC and crossover games like CoD and Battlefield (they've been ruined instead by AA). The same can be said for other controller-centric titles, like Destiny. Ubisoft does a consistent job of ruining controller look mechanics when m/kb works fine on PC. Even Epic screwed the pooch with Fortnite's controller look mechs and they invented Unreal Engine.
Title: Re: Overwatch Linear And Exponential ST Needs A Retrain
Post by: GoodAimButATankMain on 07:27 AM - 06/30/20
But when i test my curve compared to mists, mines more predictable. Idk what were doing differently, but with my setup I land noticably closer to the target with my curves than using the default ST during my accuracy tests. And i never said Obs1v isnt smart or anything like that, but since no one has ever come up with my way of creating a 1:1 curve (or at least have done it), then I thought giving my input on it is still good. Sure, maybe he has come up with a way to work arround the acelleration for his ST creation, but with my method the accelleration is 100% gone, and im going to post my thought's because if they have always used circles to make their curves or whatever, then maybe they never thought about trying it my way or incorperating it in. Im not saying my way is perfect, and im also not saying their way is perfect, but if its a new method never taken into consideration before then maybe a combination of the 2 could help with accelleration games, or even maybe just the info would be good. And again, the time you have doing something does not matter, other people sometimes think differently and can catch your mistakes or come up with new ideas.... Just like with medicine in the dark ages compared to now, you could be a doctor in the dark ages for 50 uears and still have worse medicine treatments than a doctor of our time, because some people came up with new ideas and discovered new things. Same thing can be applied here, im not saying hes not smart or anything, all im saying is if theres a new idea that could potentially help then i am not going tl assume they thought about it, otherwise no progress will ever be made.