The definitive AI Settings Thread

Lgel said:
Instead of whining I looked for information, and learned by whatever mean available (video included).
Click to expand...

My "whining" comes after literally 10+ hours of searching and reading and experimentation in which I literally got nowhere. The game just kept rejecting my packages because my modified Joesville didn't match the checksum. The game did not do what the (4 years old) instructions for packaging said it would do. I'm a professional coder, these things should not be this obtuse when I've been playing with the ISI codebase since I was 12 years old.

If this is the attitude towards newcomers who are literally begging for help, it's really no wonder this game is not taking off like rF1 did. When I tried to get into rF2 modding in 2013 the packaging tutorial PDF on the page you linked was already outdated and wrong and I see it's still the same outdated and wrong 2012 information. How does anyone expect new people to get involved in the game when the only people who can understand how things work have been here since the beginning and aren't interested in helping newcomers understand, instead pointing them to incomplete and incorrect information?

Although it is kind of funny how people here and on other forums are enjoying my tweak and I can't even use it because I can't get the game to work. I guess I'll look at the humorous side.
 
Last edited by a moderator:
peterchen said:
Some are faster, some are slower, some are trolls and there is mattlamberston..... *facepalm*
Click to expand...

I have no idea what this means but if you're trying to harass me away from the forum then you win. I only came back here because I found the AI tweak I posted on the last page (from a 2013 thread) on another forum and realized that if people were still using it, it must have been a really good tweak and I might want to try harder to get back into modding. But apparently this community is not interested in helping newcomers. Oh well.
 
I wish people would stop cocking around with ISI cars before it is even finalised.

Like the modded F3 cars, some of the rF2Rank times are ridiculously fast, makes the whole exercise pointless. :rolleyes:
 
Some of us just want to play around with the AI and don't give a single darn about ranked times and stuff, like I've been doing with ISI games for 15 years. Which just goes to my point about how offline modding and online modding should work separately and offline should be easier to tweak without having to make new official packages. How is wanting to create a more realistic AI Indycar race at IMS "cocking around"?
 
I realize that !

I never said not to do in private did I. ?

Modded ISI cars are "cocking up" the Rank is all, makes it less relevant.
 
From OP...

Fixed AI Setups="0" // whether AI use the fixed setups, only applicable if "Fixed Setups" is also enabled (and can be used in single player to have the AIs use your favorite setup)
Worked great in rF1, need confirmation that it works the same in RF2.


Not sure if others were able to confirm but I am not able to unless results.xml file is inaccurate W/R/T fuel?

From player.JSON...

"Fixed AI Setups":true,
"Fixed AI Setups#":"whether AI use the fixed setups, only applicable if \"Fixed Setups\" is also enabled (and can be used in single player to have the AIs use your favorite setup)",
"Fixed Setups":true,
"Fixed Setups#":"use fixed setups specified in UserData\\<plr>\\FavoriteAndFixedSetups.gal (based on track and vehicle class)",

From FavoriteAndFixedSetups.gal...

AutoLoadEntry
{
Track="Southshore_GT"
Classes="Spark_F2"
Session=10 // not currently used
File="C:\Users\rf2sucks\Documents\rFactor2b946\UserData\player\Settings\Southshore_GT\sf2_5laps.svm"
}

Results.xml reports that my fuel percent after lap 1 = 7.1% while all AI = 97.6%, which appears to mean they are starting with a full tank.

Edit - Looks like AI times are ~1 second faster though compared to without forcing setupso may just be AI fuel reporting inaccurate? Now that I think of it I think I remember how to find AI default .svm so I will check that...

Edit #2 - Too convoluted. I give up. Any help appreciated...
 
Last edited by a moderator:
Is there a way to choose what's laps the AI pit on? This was possible with rf1 and even F1 Challenge back in the day. This was done in the tracks VEH file, though I can't find this in rf2.
 
Anything I change to stop Bobby running wide here EVERY lap at Mills? Kills immersion because simply would never happen in IRL driver making exactly same mistake every lap

5uinaen1duebuwbzg.jpg
 
mattlambertson said:
Sorry, I thought you were commenting on the recent discussion rather than the concept of the thread in general. I was confused.
Click to expand...

nah nah. :)



I just saying a "tweaked" car can get the best rF2Rank, that is fact not supposition. ;)


P.S. I like hot lapping, only "Pro" sims you can trust is IRacing and RRE and honourable mention to pCars. lol
 
Last edited by a moderator:
P.S.R. said:
Anything I change to stop Bobby running wide here EVERY lap at Mills? Kills immersion because simply would never happen in IRL driver making exactly same mistake every lap

5uinaen1duebuwbzg.jpg
Click to expand...


I hadn't seen this until today, but that is an odd bug or something with the AI at certain tracks. For example in the Celica GTO, there are certain drivers that push wide on certain corners and I cannot figure out for the life of me what's causing it. At one point I even deleted/recreated some of the .rcd and AI learning files to fix it, and it only worked on a couple of drivers. Other drivers, namely Bobby Rahal, will simply continue doing the same thing over and over. It's like something caused them to learn the track the wrong way.
 
Okay to check back in on that issue, I've found that I'm pretty certain the issue is caused by the Waypoint Step Speed. I haven't really found anything that definitively explains exactly what these settings do.

I have found that the smaller the StepSpeed number, the faster the AI tends to go, to a point. If the number is too low, they will pull a Rahal and just dumptruck off the track. If the number is too high, the AI seems confused and is slower. The top number, StepDistance, again, not exactly sure how this works... but it seems to have a sweet spot. I've found that at 5.0 they seem limited to how fast they can go. At 6.0, they have the capability of going faster but take longer to learn the track. At even higher numbers, like 6.6 the AI has the potential for really fast laps, but they are few and far between mistakes and spins and struggle to finish AI learning.

StepDistance = 5.8
StepSpeed1 = 18.0
StepSpeed2 = 25.5
StepSpeed3 = 35.0
StepSpeed4 = 45.0
StepSpeed5 = 55.0
StepSpeed6 = 65.0
StepSpeed7 = 75.0
StepSpeed8 = 80.0

This is just an example and I would love to know more about exactly what these values do and how they effect the AI. For example, is this point speed in m/s? Also, I have been treating StepSpeed1-8 sort of like rally pacenotes. I've been under the assumption Speed1 is for slowest corners, Speed8 is for slight bends in the road. Does anyone know if this is correct?
 
I'm wondering if anyone can help me configure, or at least understand, AI pit stop frequency settings. For example, in a 20-lap race at Malaysia using the F1 2015 mod, AI starts on supersofts. I personally wear out supersofts in about 11-12 laps. Based on the relative speed of the AI and watching their lap times drop off, I'm assuming their tires are wearing out at the same rate, however, the AI never elects to make a pit stop. I'm wondering if they're actually experiencing the same amount of tire wear, but also I want to know if there's a trigger in the AI logic that dictates when they should pit based on tire wear.
 
AI pits when tyres are worn out to a certain amount. How fast this appears is only known by the modder.
As far as I know, fuel settings have no influence on that.

It´s a little bit tricky to calibrate AI-wear to human standard.
Good modders do this. But that implies that the modder is a good driver too!
Maybe tyres last much longer than this 11-12 laps and you just driving too harsh. (no offence)
I don´t know this mod, however nothing you can do about his, other than be nicer to the tyres if possible.
 
First post updated with a WARNING as I'm "washing my hands" of any responsibility now, it was fun to tweak & fiddle but those days are behind me as RF2 has become simply good enough "out the box"....in fact it always was good enough but it's fun to play around with stuff isn't it!? :)
 
peterchen said:
AI pits when tyres are worn out to a certain amount. How fast this appears is only known by the modder.
As far as I know, fuel settings have no influence on that.

It´s a little bit tricky to calibrate AI-wear to human standard.
Good modders do this. But that implies that the modder is a good driver too!
Maybe tyres last much longer than this 11-12 laps and you just driving too harsh. (no offence)
I don´t know this mod, however nothing you can do about his, other than be nicer to the tyres if possible.
Click to expand...

Oh I'm doing just fine on my tires...it's the AI that probably needs to be pitting more frequently lol. Eventually they start dropping 2-3 seconds a lap to me.
 
Made a find yesterday:

The Player.JSON setting:
"Auto Line Smoothing":
Is definately broken! I´ve used to set this on "7" but I noticed now that AI with these setting is going funky.
They don´t follow the line, going wide totally destroying a good AIW-line. This behaviour change only minor with several laps driven.
I noticed that in actual build as in older build 1098.

Leave that at "0" !!!
 
You must log in or register to reply here.
Back
Top