PC2 Weird qualifying speeds in rain at The Glen

[Jurld]

I just picked this game up a few days ago when it went on sale, finally. Started a career in Formula Rookie. First race seemed pretty standard. Got to race 2 at Watkins Glen and found myself in the light rain. All I can say is, I'm really impressed by these environmental effects. I think the cars are a little forgiving, even with rain tires, but you do have to adjust and it looks really cool, for sure.

In practice I was a full second faster than the next car @ 95%. Then I got to qualifying and the best I could do, with barely any fuel and plugging the radiator vents was about six full seconds behind the pole. I noticed the rain was a lot heavier during qualifying. I didn't notice this messing me up much. My car acted like it just had zero power, would not rev to limit, etc.

In the race it starts out dry and starts to rain about lap 3. I started in 12th and had managed to pass the leader before the end of the first lap.

So... what am I missing about the qualifying?

 

[Dmitriy Martynov]

AI is rediculisly fast at qual's. At race their pace is much slower.

 

[Jurld]

All the other races in that series so far (1/2 way through the second stage) I’ve been within a second and managed pole at Texas. So there was something weird going on there.

 

[2112]

I was qualifying for a race at Imola last PM. It was in very light rain with a wet, but dry line developing, track. I was four or five seconds faster than the next guy to qualify.

Race was on dry track with dry tires and I could not match my drying track time.

Seems on a drying track with wets the cars are very quick. Could be due to the nature of a wet tire on a drying track for a few minutes.

 

[Jurld]

I dunno. I hate to sound like a bitch, but now I think the AI is just messed up when it's wet. I got invited to the historic Lotus race at Monza, where it is raining at the start and the track becomes progressively more lake-like. The AI at the start seemed even slower than usual; I easily made it to P1 by Lesmos and was well ahead by the end of lap 1. As the track became more wet the car became increasingly "fun" to drive, which is what I'd expect. But as far as I can tell the AI never slowed down much from their pace on lap 2. By the last (4th) lap, Parabolica was like ice, no traction at all, but the AI continued on their merry way, barely slowing down. Eventually, after many attemps, I did manage to get far enough ahead on lap 1 and 2 to maintain the lead in spite of the conditions at Parabolica on lap 4.

Or maybe I just suck at driving in the wet.

 

[Msportdan]

yep had the Ai quicker in wets on a drying track. would love these little ai issues I be fixed in the wet.

 

[Jurld]

I've decided the AI in general is just... hit or miss. Continuing on in the same Formula career track, I came up to Road America. I managed to beat the AI but had to turn their rating down into the 70s to do it. They just carry crazy speed at certain points of the track. I tried downloading several of the top setups from time trials and was able to keep within a second of their ghosts without much sweat. Their record times were seconds off the race pace with AI at 95. The track conditions for the race were actually quite a bit slower than the TT, to boot.

Fast forward to the next race at Sonoma. I set the AI back to 95% and rolled the dice. I think I won that race by over 10 seconds... Similar results at Long Beach and most recently Portugal. So... I need to up them on most tracks but then there are some where they are just magical creatures.

As my boss at work likes to say, "I think they have some opportunities here."

 

[Andy_J]

Yes despite SMS working on the AI with every update you can't rely on them to be consistent. Just what can they do to fix it though? To me they will give up soon and work on producing PCars3 complete with the dodgy AI.

 

[Jurld]

You're probably right. Like everything else, it's an economic problem of ROI. Pretty much no game in the history of PC racing games has had what I'd call brilliant AI. The bar is pretty low. It eats a lot of CPU cycles to do it "right" (whatever that is) and that comes at a huge premium when you are also crunching physics. Between the two priorities, I'd pick physics every time. I get the impression most racers who are into "more realism" would rather race (non-wrecker-type) humans anyway so it's clearly not a huge priority for them to spend much time here, vs. trying to create some kind of online ladder system to facilitate those desires.

I don't want to sound like I'm pooping too much. I like the game a lot, especially that there are so many of the tracks I enjoy and they are all really nice looking and I don't need to go hunting down a bunch of mods to race on them.

The track to track inconsistency to me sounds like it's more of a configuration issue with various tracks than a problem with the core AI programming. Most tracks I've hit so far seem "good enough" as far as being predictable. The rain/dry thing sounds like the core code is still just not right at all, either the AI or the weather physics, or both.

Solving the first one problem could be low hanging fruit, solving the second one would probably deliver important value to the customers if the problem is with the physics and not the AI code.

This is just totally out of my rear guessing though.

 

[Kurupt CDN]

http://forum.projectcarsgame.com/showthread.php?60908-Unofficial-issue-list-Post-Patch-4-0

 

[Andy_J]

You're probably right. Like everything else, it's an economic problem of ROI. Pretty much no game in the history of PC racing games has had what I'd call brilliant AI. The bar is pretty low.

Microprose GP series especially GP4 by the mighty Geoff Crammond did have have what I would call almost human brilliant AI. In fact I was doing a race at Silverstone on it yesterday and the AI not only seem to be aware of you they actively race you. It's still brilliant if aged.