Hello again! Can someone tell me exactly when (if they actually do) do AI cars pit for new tyres? Is there a % threshold? Scenario: I am racing the GT3 championship againts the AI. 25% race length and 4x tyre wear. Fuel consumption off. In Nordschleif race, which was 45 min. Most AI cars went to the pits around half the race. But then, next race in Spa, again 45 minutes, I pitted pretty early because a braking lock destroyed my fron tyres. But none of the AI pitted at all... I did notice they were way slower at the end, so I ended up in the same position as before pitting. How does it work? Are those different strategies for different tracks? Do they prefer to keep racing slower rather than loosing time in the pits? Thanks!
Can't answer the question I'm afraid but I can share my experience. After having some long and frustrating races with GT3 cars I decided to do some testing. I am using custom AI files with tyre wear for all AI set to 1 which, as I understand, is minimum tyre wear. Stamina, consistency and aggression are all set to 1 to remove these variables. A spread of AI race skill from 0.5 to 1 for each car model and max (31) opponents. All tracks mentioned are the most modern varient on default date, light cloud in single race. AI at 108, (matching my fastest times to the AI within 3 tenths), max aggression. I drove a Mercedes. With normal tyre wear enabled (also fuel consumption normal), all AI lose up to 4.5 seconds per lap over a 1hr race compared to their fastest lap. No AI made pit stops, my tyres were approx 70% worn according to the hud and had no drop off in performance. Typically my lap times remain pretty consistent across the whole race, the car being perhaps slightly faster towards the end. The result is a pointless race as I end the race being ridiculously fast in comparison to the AI. I raced at Spa and Interlagos having the same result - starting last and easily finishing first! Then I went to Hockenheim with tyre wear disabled and fuel consumption normal. I again started last but finished 7th after a really entertaining and competitive race. The AI I was directly racing against in the last 4 laps set their fastest lap after I passed them. The cars in front were 0.5 to 1 second slower at this point but their fastest laps were comparable to mine and the two AI I was racing in those last laps. Again my lap times were pretty consistent across the hour race gaining perhaps 0.5 seconds over the hour race, not enough to be significant. I plan on doing another race (Laguna Seca) again with tyre wear disabled and fuel consumption enabled to confirm this when I get chance. The conclusion I draw from this is that the AI suffer much greater tyre wear than the player, rendering tyre wear utterly pointless at the moment versus the AI in longer races. Hopefully the devs will take notice and rectify the problem.
Having so many different tire compunds, must make getting the wear rates right a pretty big effort, so I imagine it will be a while before they are all as good as they can be. Something that I have seen though is the fuel weight doesnt affect both the AI or the player much, I can do very similar laptimes with 5 laps of fuel or with a full tank, maybe 2 or 6 tenths slower, should be a higher difference. And thats something that should be easier to balance, imho.
Thanks for the testing and the reply! I used to drive without tyre wear and fuel consumption, but I wanted to make the races more varied. However, I wasn't able to make the AI pit consistently unless they have mandatory stops... It probably works fine in other clases, and worse in others I'll let you know if I find something else interesting.
The consistency and stamina parameters make a big difference to the AI after the 60-75% range no matter the length of the race in my testing... Which will compound their tyre wear if the AI for that class have tyre wear, which I'm convinced that a few do not... The problem here is there's no enough CPU resources on the average users PC to run both the player and AI on the same tyre model... So there's going to be large discrepancies between the AI and player until the physics for the player side is locked down...
No worries. I want to set up a GT3 championship, so was testing for myself and ran into the same issues you mentioned. Always happy to share my experiences and hope the feedback can help Reiza get the AI tyre wear sorted too! The race at Laguna Seca went very much the same way as the Hockenheim race, AI holding their pace to the end, all lapping within 1 second of their fastest lap. Slower when in lapped traffic - it's a difficult track to pass on! Couple of AI crashes, one of which ruined my race at the corkscrew, but that's how it goes sometimes. Again an enjoyable race. The only real difference was I had to up the AI to 113 fo them to be competitive. I think that if you want a competitive race(s) with the AI in the GT3 class, it's best to disable tyre wear for now. Perhaps if you want pit stops, enable fuel consumption and increase the multiplier. A 1hr race uses practically a full tank so I guess with accelerated (2x?) you would get a pit stop. 3x will be tight for a 1 stop race at some circuts, but probably good if you do 45mins. More things to test...
Which is why I set the custom AI to have max aggression, consistency and stamina. It negates the effect! When my testing is complete I'll tweak them to taste and set up a championship. I fully understand that the AI uses a simplified physics model. It is what it is, all I can do is hope the great minds at Reiza can program convincing AI. No doubt the feedback here will be of use in the optimisation of the AI physics/scripting. The AI is, in my opinion, already way better than in the other sims I own, certainly better than the bore fest that is ACC's AI. However as Reiza freely admit, the game is work in progress. I think in time they'll get it right, AMS2 is already exceeding what SMS achieved with the Madness Engine and I am glad they took on the challenge! Maybe by the time CPU power is good enough to have neural nets running the bot cars with real artificial general intelligence, my rural internet will be fast enough to enjoy the MP sim racing others seem to take for granted... Alas until such times, I can only enjoy the single player experience!
Thanks again for the data and experience shared! I am already in the last two races, so maybe I'll just race without pit stops and then try another class. As for the fuel usage, one thing really keeping me away from playing with it is the waysetups work... It would be much better if we had a "full tank" or "minimum necessary" option so we don't have save a new setup every time we change fuel consumption. Though as some of you said, fuel weigth doesn't seem to mae much of a difference? Maybe I can just use full tank in every race regardless of length? PS: I completeley agree the AI is better than in most racing games. Sure, frustrating at times but...humans can be frustrating at times XD.
Even at 1.0 there's still effects from both stamina and consistency which warp the tyre figures.. Whilst this sim is a work in progress, it's not like Reiza is new to what AI needs to be realistic... They've had several games where things like fuel weights and other important parts of the AI code existed that are just absent from the Madness engine... And unlike the MP side Reiza have put considerable effort in getting the AI to the level they are at now... So they may have hit the ceiling on how good the AI can be... I hope not, but at the same time I'm realistic about these limitations that effect every sim that strives for sophisticated physics... Given a rural internet upgrade maybe 5 years away, or more, you may well be right that you'll be able to enjoy both offline and online properly at the same time... As it's going to be a good few years before the AI isn't running off limited resources and is forced to have physics that are oversimplified just to get numbers on the grid with AI... To get the best experience vs the AI turn the fuel usage off and run the fuel levels at whatever the default set up gives you... Otherwise you'll always be gaming the AI... Qualifying with a full tank to have a decent start to the race and dominate from the middle of the stint onwards, or qualify with an empty tank and get swamped by the AI at the start in order to have a race from the middle of the tank to the end of the tank... Because fuel weights only matter to the player... You can gain 3-6 seconds over a stint depending on the track... Up to 30 seconds at the Nordschleife... There's a bunch of Simhub overlays and dashboards that will give you information on fuel, Crew Chief can even give you audio calculations on the fly...
I started exploring AMS2 with the formula trainer/advanced, then the f3 cars but found the AI frustrating (better than PC1/2 though) in the single seater classes with the somewhat wonky collisions. I moved on to the Ginetta Junior, GT5,4 and now 3. Only now have I come across the tyre wear problems which I put down to the soft tyres of the GT3, the other GT's use hard if I remember correctly. Custom AI became a thing when I was at GT4 and started experimenting and found ways to get a much better racing experience. That has consumed about the last 4 months of my sim racing time! I find if I use AI quali skill around 0.1 below race skill, the AI pace is the same in practice/quali as it is in a race. Hence have never bothered with fuel as a variable of pace against the AI. Fuel doesn't affect players pace, either that or I am the most consistent driver ever! Others will argue it does, but not in my experience with the GT cars. I just fill the tank and go qualifying and racing! With a 1hr race I've never run out of fuel on any track I've raced at. Sure it'd be good for fuel weight to effect the pace of the car but in AMS2 it doesn't seem to. I'd always assumed that tyre wear offset fuel weight but my recent tests without tyre wear seem to disprove that. I raced at Silverstone tonight with the GT3 cars 1hr, no tyre wear, fuel consuption accelerated, same custom AI, (fuel management at 1, but I think I read it only effects F-USA cars). Started full of fuel as ever, 121 litres in the Merc. Didn't qualify, started from 20th of the 30 car grid by choice. 31mins into the race 11th place, and I had two laps of fuel left. I use Z1 dashboard/analyser for telemetry and it tells me how much fuel I need to finish the race. The AI had started to pit and I did too, put in 45 litres, which got me to the end with a lap to spare. During the pit stops I gained 8 places, finished 2nd battling with the 3rd placed car through lapped traffic. Good race but how did I gain 8 places in the pit stops? Looking at the replay of the winning Merc AI, it's dashboard read 40 litres as it crossed the finish line. So it appears the AI over fuelled by about 40 litres costing them the time. AI lap times were consistent throughout the race with no significant drop off apart from in lapped traffic. I'm going to do another race like this to see if it's repeated. Looks like another problem Reiza needs to sort out... Makes me wonder if it's worth the effort to be honest, perhaps best to forget pit stops and just enjoy races without them!
They've been talking of upgrading the Internet for the 6 years I've been here, doubt it's going to happen in the next 5 let alone 10. They don't care, we pay more now for 1.5mB down/0.5mB up than I did in the city for 10mB/2.5mB! No incentive to upgrade us at all. Which car/class? I've not experienced this.
Yeah, definitely openwheelers AI needs a different logic. They drive too close to the player and the "get back to the ideal racing line when in a corner, no matter what's on the way" thing makes those classes frustrating, forcing you to just avoid them, letting them pass most of the time. I'll probably just try with pit stops in other classes and if I encounter any problems, get rid of tyre wear and fuel. It's a shame because it's the first time I play a sim with pit stops since F1 2010 I think. And it almost works...
After another couple of races, testing fuel and pit stops in 1hr races, the way to do it is to have fuel consumption accelerated, fill your car with fuel at the race start, set a pit stop strategy to max fuel your car, then pit when you choose. The AI fill their tanks at the pit stops regardless of the fuel reqired to finish, so you should be stationary in the pits for the same time as them. I tend to be a second or two faster on pit entry/exit so there is no need to push the pit lane speed limit lines. For 45min races you could probably go to 3x fuel consumption. Hope that helps...
In any formula car it's very noticeable if you aren't on softs where the tyre deg keeps up with the fuel weights until it's time to pit for new ones... Example the CART gen 1 at Elkhart lake... I'm at least 3 seconds faster at the end of a stint whilst the AI in all cars only get slower over a stint due to the consistency and stamina effects coming in... If it's a 3 stint race than the 3rd stint is where they start losing time... They never get faster over a stint...
