Magic Data edit = Screw up performance data [Fixed]

Posted by XenonS 
Magic Data edit = Screw up performance data [Fixed]
Date: September 18, 2016 02:15PM
Posted by: XenonS
Hi,
I've been testing the edition of 2 magic data values these days, the AI GRIP which is said to increase Agressivness of AI cars by reducing the value, and AI ERRORS, which is usually from 40-96, higher values mean more errors from AI.
I must say that I encounter no effects like described in the FAQ by TomMK who has published the descriptions of these values. What I found out instead is that any slight changes of the Magic Data will definitly screw up any Performance File.

If I set a somewhat higher AI ERROR (65) and set the AI GRIP to 41000, I expect more overall car spins and more overtakings in places where these would otherwise not occur. But nothing like that happens: what it does instead is decreasing the overall Pole performance by adding 5-6 seconds to the best cars, meaning that if you play on Semi-Pro you are almost garanteed to get on pole position, without any effort or special car setup. Any performance settings become nil, at least in the Season 1991 mod where I have tested all this.

I think what GP4 is trying to do with the AI GRIP is increasing the relative speed of the cars on a track, and the game does this very artificially: on high degree corners you see cars almost come to a stop for some seconds allowing others to overtake them, this translates in 'Agressiveness', but it's artificial and actually makes the game very exploitable where you can easily overtake such cars.

These 2 magic data values should have nothing to do with the Performance Data (usually 17 files are set for each track that GPxPatch will load automatically), but for me this is a fact, and it maikes the edtion of Magic Data superfluous, or these descriptions are plain wrong.
I will have to test and see what the Standard Game (Season 2001) does instead of a mod, but who is still playing the vanilla game, I guess nobody...

What is your experience? Does the Magic Data what YOU want to achieve, and what is this?

Greets,
XenonS



Edited 3 time(s). Last edit at 09/19/2016 12:04AM by XenonS.
Re: Magic Data edit = Screw up performance data
Date: September 18, 2016 06:31PM
Posted by: Atticus.
AI aggressiveness works in a very unusual way indeed. I would try to explain it, but a graph, an admittedly rough one, is much more telling than that.



The main takeaway for you is that decreasing the AI grip value will actually decrease AI aggressiveness - right down to 32768. At 32767 it is high again and then it goes down again all they way to 0. (I don't actually know what happens if you set it to zero.)

As you've also noticed, it is an extremely sensitive value - in the original files, it is always around 64000-65000 and that has a nicely subdued effect. I think it is intended as a "master key" kind of value which should be used to finetune the AI corner behaviour when everything else is settled. Of course, you would then need to use increment steps and not trolling it with values like 0 or 40000...

As for AI errors, it is not at all significant, certainly not for the pace of the AI. It is what it is: set it high and the AI will make more driver errors, missing corners and spinning off and set it low for a nice professional, faultless race. (A personal advice: decrease the number below AI errors to about 16 instead of the default 32 or 64 for the AI to be able to actually carry on after a spin and avoid a DNF by simply waiting on the edge of the track for the field to pass. This value sets the spin recovery range for the AI, i. e. it sets how many sectors the AI checks for incoming cars while sitting on the edge of the racing line trying to rejoin the race.)



My workthread - [www.grandprixgames.org]
Full of classic F1/non-F1 track layouts

My blog about F1 performance analysis - [thef1formbook.wordpress.com]
Re: Magic Data edit = Screw up performance data
Date: September 18, 2016 11:52PM
Posted by: XenonS
Hi Atticus,
thanks to elaborate on this, such a non-linear function explains it all, that's why I never got results from any settings that I have set. On the other side, I never believed these values were nonsense, they are there for a reason ;-)
This is all very interesting, it means that the default value for AI GRIP (I generally see 64768) is actually almost the highest possible and shouldn't be changed if I want cars to be aggressive. Also reducing the value (in my case to 41000) will give horrible racing results that now turn to be fully logic: to obey for more passivity the cars must slowdown and mimic an overly safe guiding style like dodging ice in August...horrible...After this experience, and seeing your function, I would urge everyone not changing the AI GRIP, or only make very small changes around the default value.

I can confirm that the AI ERROR trigger works as intended, but my obeservations were severly hampered by the wrong settings of the AI GRIP, it's difficult to imagine how cars that run overly safe will do a lot of spins ;-), but keeping the default above they will do so.
Thanks for your advice to also change the Fine Tune value to see effects.

There is much more under the cover of all these values indeed!

XenonS



Edited 3 time(s). Last edit at 09/19/2016 12:05AM by XenonS.
Re: Magic Data edit = Screw up performance data
Date: September 19, 2016 11:31AM
Posted by: Atticus.
XenonS Wrote:
-------------------------------------------------------
> to obey for more passivity
> the cars must slowdown and mimic an overly safe
> guiding style like dodging ice in
> August...


Hahaha! :-)

You're welcome. :-)



My workthread - [www.grandprixgames.org]
Full of classic F1/non-F1 track layouts

My blog about F1 performance analysis - [thef1formbook.wordpress.com]
Sorry, only registered users may post in this forum.

Click here to login

Maintainer: mortal, stephan | Design: stephan, Lo2k | Moderatoren: mortal, TomMK, Noog, stephan | Downloads: Lo2k | Supported by: Atlassian Experts Berlin | Forum Rules | Policy