Does the Adaptive AI keeps learning? So the more/longer you play the better they get adapted to your level?
Oh yes. Make no mistake, using the tool is a short-cut for giving the AAI a decent starting point. Once there though, the AAI will continue to adapt. So if you start getting faster, so will the AI (up to a point, naturally). Which is one of the reasons I'm so enamored with the whole concept. Yes, there are some quirks and the odd pitfall, but dear lord is the end result worth it.
Nice, thanks... will they also adapt to less bumping my rear? I made in the beginning some miscalculation on their braking points. They tend to brake a bit early so I bumped in them from time to time. Now I know this I don’t bump them anymore. Although I notice that they still hit my rear from time to time. Could by my error of course, wrong braking point or wrong race line. Just curious..
Not directly. Despite the name, the Adaptive AI doesn't adapt to you as such. It doesn't e.g. look at your braking-points and adjust to them. Nor does it try new lines based on your driving.. What it does is use the predetermined difficulty levels, and adjusts these to suit your pace. On the other hand, adjusting to your pace will, even if by accident, also mean less bumping since the braking points are more likely to be the same as the players, given that the pace is roughly the same. So there will still be the odd bump'n'run. In my case it's very rare by now, maybe once a month or so.
A fact that is very annoying for me. I use a manual clutch and a shifter. Maybe that is the cause that i'm brake a little bit too early. The pace of the AI is ok, but the braking point isn't it when you shift manual. In my opinion. ...or this is my fault as well.
@MattStone You can take a look into the following file: Steam\steamapps\common\raceroom racing experience\Game\GameData\General\r3e-data.json Open it with a text editor and search for WTCR 2019.
Updated, see OP. New assets.txt enabling Bilster Berg and Dubai. Added Cayman Trophy, KTM X-Bow, TCR19.
We are in year 2019, i just canot understand why the AI is still so bad, actually the AI should do all this you said: breaking point, driving line, speed in corners and in straight, now thats the way AI would be great to drive with, for my understanding, its up to the AI to adapt to the player and not the inverse. Just my 2 cent...
you're are gonna need a computer that is worth a hell of a lot more the 2 cents if you want AI that would require that amount of processing power.
Yup. You're asking for an AI that can adapt to all track/car combos for all possible users that might use that combo. We're talking machine-learning on a massive scale. And while it's theoretically possible, you'd need a bit more CPU oomph than in your standard PC, plus it's somewhat outside the scope of most software developers. Not to mention the time it would take. The current AAI 'only' has to adjust to the lap-times. Imagine how many times you'd have to run a combo for it to adapt to specific lines, braking-point etc. Then you'd have to find a way to get a spread within the grid, otherwise all cars would run exactly the same lap-times all the time. Then you'd have to also figure out how to teach it alternate lines if the primary one is occupied. Oh, and you'd also need to program it so it doesn't just ram into the player when he/she messes up a turn. At the end of which you'd have an AI that runs at exactly the same pace as the player.... Where's the fun in that? Then there's the question of whether the AI should adapt to the player or vice-versa. Well, whenever you do a multi-player race, you're constantly adapting to each players driving-style. I fail to see why racing the AI should be considerably different. But by all means, go ahead and program an algorithm that takes all of the above into account. I look forward to testing it out.
@CheerfullyInsane IMPORTANT change, the tool as of now tries to find the R3E and use its r3e-data.json rather than the assets.txt which no longer exists. If it cannot find the directory you must enter it manually in config.lua file.