Information for Customizing AI drivers in AMS2

Is there an AI setting in the XML file that deals with the magical skill boost an AI gets after they have been overtaken? I follow some AI for a couple of laps (not fighting for position), find them ridiculously slow in certain corners with no-one in front of them but the moment I overtake, they magically can take these exact same corners so much faster even then able to over take me on the outside.

Is there a setting in the XML file I can set to zero to completely eliminate this behaviour? Its frustrating enough that they all seem to be comparatively faster in straights and slow in corners.

 

Agreed. I think this behavior is hard coded. It's actually pretty cool that it does this IMO, and it's less exaggerated now than it was in earlier builds, but I feel it still needs fine tuning to make it a little more realistic. Also, it would be cool if the AI made the occasional mistake when pushing too hard to retake.

 

Some people (including me) used to complain about "rage turbo mode" where you'd pass an AI car, they'd drop 5 seconds back, and then in the space of a half a lap they'd make up that gap and be right behind you spoiling for a fight. I haven't seen that for quite a while.

The AI don't give up when they get passed--I don't either. But I'm not experiencing any magical boost, at least in the cars I've been playing the last month (GT 3/4, Stock Cars, F-Classic mostly).

 

What is the xml name for the Porsche Carrera Cup? I haved tried Porsche_991_GT3 but it doesn't work. It wasn't listed in the original xml names on the first post.

Thanks

Edit, I'm a fool, its Carrera Cup.xml!

 

If anybody is interested, I have uploaded the first video for the championchip with the custom AI mentioned earlier:

 

when the sim assigns a qualifying skill rating to an ai driver does that depend in any way on the driver's race skill rating or is it completely independent ?

 

When the sim does it, I'd guess the race skill and Q skill are closely correlated. With custom AI you can adjust it however you want.

 

yeah that's what i was guessing. i was making an ai xml and wasn't sure how to assign the qualifying skill. i think i'll just use an rng to produce a figure +/- a smallish amount. it's really great that we can do this especially for people like me who's main interest is in racing the ai.

 

Maybe I'm a little too obsessed with AI levels, but for each series and AI xml I create, I work out the average qualifying positions and average race finishing positions for each driver, then rank them and assign a level based on this. The consistency value is interesting and I've been playing with various statistical formulas to find a good way to produce one. Latest I've used is variation, which basically looks at how wide range a drivers finishing position may be. Using this, JV is most consistent with least position variation, whereas Barichello has highest variation, so therefore give him lowest consistency.

 

That's pretty cool!

 

For the consistency/stamina values, has it been discussed what the ranges are that are used by the default AI? I've done some AI files where I've just used random numbers between 0 and 1 for these values, and I can really mess someone's results up with low values in those fields even though they are otherwise higher-skill. It's obvious to me that I should be usinig a narrower range--didn't know what others or the default suggest would be a good ones to use.

 

Well this makes me feel that maybe I am a normal person . I did not get to this point, but certainly wanted to do it, just had not the time. Some result seems strange though. IDK howyou exactly you calculated this, but one thing I would consider is that car problems may lead to bad results that are not related to the driver consistency. Jos Verstapen ahead of Rubens and Hakkinen, for instance, looks a bit strange.

 

I didn't include any DNFs, just cars classified. Remember this is just for the consistency value. Hakkinen still has the 4th best race skill, whereas Verstappen has 20th best.
In my experience, this consistency value certainly helps have a less predictable finishing order over a season.

I do feel sorry for Rubens though. After strong qualifying, he nearly always plummets down the field during the race!

 

just kidding.... I think the world of you Rubens!

 

The opening post defines consistency as:

consistency: Low consistency value means the driver skill can be randomly reduced a bit.

I've realized I'm not clear if this refers to lap-by-lap consistency or race-by-race?

 

Good point. I always assumed it was on a lap by lap basis, but if it is race by race, could be why Rubens has so many shockers! Laptimes in a race may confirm.

 

I adjusted my AI file a bit, the gap of 0.04 between each driver was a bit to large, I was soon lapping drivers. Put it down to increments of 0.02

 

Did anybody experience exceptionally slow AI in the wet? I was racing at 82 AI difficulty with the file from above in Goiânia. In the dry, I had a hard time keeping up, but once the track became wet, I was building a gap of 10 seconds and more. It was especially noticable on the straights. The AI was slowing down, whenever they hit a puddle. Any thoughts?

 

Hi there Blaine7. I think this should go in the thread about AI performance so Reiza can consider it. Overall I haven't tried yet but heard that AI is quite inconsistent on the wet. Sometimes too fast sometimes too slow

 

Cheers, I have just done so.