New source. Steam. 2015. Player NU13. I'd go so far as to say that not only is the AI terrible but so is the whole cutting track warning thing. Here are two examples that have happened recently. I was on the charge in 92 DTM Merc, got through the first corner but in the second corner misjudged it badly and went off the track and through the sand trap, got back on the track and find i have been warned for cutting the track....cutting the track?? Since when has pushing to hard and having an off ever been cutting a track? Second example, in a ADAC race in a Camaro, been off the track three times through my fault in pushing to hard but when i catch up to the other back markers i am trying to pass them and they appear to not know i'm there because they keep nudging me off the track. This happened again and i get knocked off the track by Stevie Wonder in a Merc and i end up getting a drive through penalty. Dumb AI and dumb programming by the develpers. The last sentence is lapidary, since 2015 the developers could never find the solution Matt Stone by the way this is a good train wreck. Crazy train (sorry Ozzy)
@Nano 10 Have you ever considered that it could be your fault by driving against the AI? What makes you so sure that you are on the right path? Laptimes? Sectortimes? Brakingzones? I think that the AI in R3E is not perfect - but it is closer than any AI comparing to other titles IMO. Anyway, concentrate on your driving and take the AI as a filler - and at least try to have fun
Speaking of the 92 DTM, I have been having a blast racing the AI in them. I've been rotating between the merc, the audi and the mustang. I have to say the racing has been very intense and exciting, I will be in a group of three or four cars fighting for position. And the AI are great. Sometimes the ones behind me will line up in back of me then jump to the side essentially stealing the draft so they can pass me on the inside. And just like in real life there isn't much I can do to stop them and I will loose 2 or 3 position in just a few seconds. Sometimes I will out break myself in the corner trying to gain a position, and run wide coming out. And it will cut behind me to retake the position. And if I'm not careful I will lose even more positions. Other times I get it right and get a clean pass on the inside for the position. But the AI will get a good run on the outside. SO it's a fender to fender drag race to the next corner. It's like a game of chicken to see who will brake first. Sometimes I win sometimes I loose. Other times I will be in a battle for position with one car for several laps before I finally get the position. Even then it will follow me throughout the next several laps. And if I make any mistakes at all it will be on my ass again. And in true DTM fashion there is always a some bump and grind with the other cars throughout the entire race. And it used to send me spinning off the track. But I've spent a lot of time learning how to drive. So now it's all part of the race. Knowing how to keep the car on the track took some time to learn. But it was well worth the effort. And there is a fail safe move in the single player mode. Assigning a key to the re position car option. So if I do get bumped off. I just get car stopped as quickly as possible hit the key and only loose a couple of position. I could go on and on, but I'm already rambling a bit. Anyway, I always think the AI couldn't be better. And yet S3E always seems to prove me wrong. Now this isn't just with the 92's either. Since I have learned MY personal AI levels, almost all of my races are like this. The only times I'm disappointed, is when I get the level wrong. Thanks S3E, you guys rock the single player mode WAY better than all the rest. Have a great race y’all Gregoryleo.
I am liking the adaptive ai right now. I spent a few weeks with other Sims racing online, this helped me learn some about racing in packs. The ai in Raceroom, if you are able to hold a line, and keep with the race pace really do a nice job now racing with you. I tried training them with 10 lap races. I stayed as clean as I could on the first race. The second was challenging, but very fun. Much better than the previous build. Tracks I've raced on, Sonoma, a1 ring, Bathurst. All done in gt3 cars, I think you are getting close to perfect sector3
I agree fellas; the update that Sector3 did late last year has made the AI very good save those 4 points I mentioned earlier. But where the AI is a little slow is balanced out by where they seem a little fast so it makes for some interesting tactical planning of your passing moves. But the quality of Sector3 AI leaves the competition behind...keep up the good work guys
I would know if i am the only one to have this feeling : when i set the AI on Adaptative, i start every race in pole position because my qualification lap is the best and i win every race. But if i set AI on 102, opponents are very competitive, it's quite too hard for me Is it the same for you ?
^ Maybe you did not correctly understand how the Adaptive AI works. It needs to be trained for every car-track combo. It does not give you optimal outcome already in race one, adapting and altering itself within that one race. It alters between races only. You need at least 2-3 races with one full lap of yours to have it setting itself within a performance range close to yours - for that car, for that track. Its like a pendulum swinging back and forth several times, every time the amplitude shrinking until it just vibrates around the centre. Thus, your first use of a given track-car combo will have the AI always performing at the minimum value of 80. To train it, the AI takes your first full lap in a race (not training) session (means: lap 2). After you completed that lap, the rest of the race can be skipped and you could just launch the next race. Doing so 2-3 or 4 times, should see the AI being closer to your level. You will see an easy ride in race 1, a much more dffficult one in race 2, a mildy reduced performance again in race 3 (if race 2 saw you failing), and then maybe from race 4 on something in your range. The corrections done by the AI in the next race become smaller and smaller, and go up and down according to your last performance. You need to do it with every car and every track. Thats what makes this system tiresome if you jump a lot between cars. Drivers sticking with just one or two cars, find it more useful. I have used adaptive AI a lot for many months, since it was revamped, but in the imminent past have gone back to fixed AI levels of around 97, for the above mentioned reasons. The need to train the AI was what finally had me loosing interest in the adaptive AI. P.S. To train the aI, set yourself on the last place in the starting grid, and let the others chase away. This is to make sure you have enough space for one clean full lap in lap 2 with no car getting in your way. This is what counts, the rest of the race is not relevant. After that lap just quit, and start new. Do this 3 or 4 times, and you should be fine then.
Now you do. self-adapting AIs within one race are usually so-called "rubber-band AI." Nascar 2003 was a very prominent example for that. It worked and acchieved what it should, but was quite obvious in the artifical nature of the resulting effect. Gran Turismo Sports is another exmaple: it tries to distrubute the cars's perfomance in a way that the player will face overtakings and duels until short of the end of the race and will usually win in the last seconds. The result is that the race is entertaining, but predictable, and that the AI cars do not really drive competitvely and as best as they can, but as they need in order to provide excitement until the race ends. Its no bad method to secure entertainment, both titles work very well. Its just something different than simulating race events by individual perfomance characteristics.
hm, not sure if I remember right, but isn't the Adaptive AI even learning when the AI-Level is fixed? If yes, it could be a good solution to train the AI first with a fixed AI-Level matching your own performance. So the training-period would be a bit more entertaining. Maybe @Christian Göpfert can clear that up a bit.
To the best of my knowledge that is the case, the performance index will also be written when using a fixed AI level, provided you're doing valid laps in a race session. Also, adaptive AI works on a class-track base, so no need to train each and every car in, lets say the GTR3 class. And it might be worth giving this a look: https://forum.sector3studios.com/index.php?threads/adaptive-ai-primer-database-tool.5632/ (Which is basically the result of this very thread.)
Is that truly certain and fully confirmed? I asked for this long time ago, and got exactly the opposite reply back then. Also, from having used GT3 almost exclusively when doing adaptive AI races a lot, I did not get the impression that the whole class adapted when I switched cars. In a new GT3 car the AI cars again drove like noobs - or so it seemed to me. I am no superfast driver, so when I easily can roll up the field from behind within just one or two laps, that gets my attention. At fixed 97, I cannot do that. Not that I care for adaptive aI that much anymore, I am quite happy with now using fixed 97, it does for me what i expect from the game.
Yep, it is. The index-file doesn't store information on individual cars in the first place. If you've experienced the AI driving like noobs, I can think of two possible reasons. 1 ) The GT3 class has a lot of drivers, and not all of them are equally fast. There is a fair bit of variation within each class. 2 ) More likely, it's on a new track. Bear in mind that if the AAI has no info on that particular track, it'll use the average of all the other tracks it has data on for that class of cars. So if you have data in there from i.e. the beginning of a training session, or something as simple as a track where you're slower than usual, that'll affect the average. Also, you don't have to start from scratch with every class. The AAI also uses the average for all classes on a given track, when introducing a new class on it. The AAI does have its own set of quirks though, and the averaging, while fast, does lead to erroneous data unless you treat it methodically. Short version, it is far better if you have a general idea of what level you're at before you start training it.
This sounds interesting ! Ive always used fixed at 106 but recently had my interest sparked into this adaptive Ai , should i delete the Ai file and start again teaching the Ai or just switch to adaptive and race and they will continue to learn and evolve ?
If you've been using fixed AI up to this point, there's no sense in deleting your index-file. It has already collected data from the fixed AI races, so it's got a decent starting point. One issue you may run into is getting the damn thing to adapt in the first place. It's a little loath to move on its own accord unless it has some data further up or down the line. Easy fix though; Run a few laps with an AI set 5 levels higher than normal (111 in your case), then switch to adaptive. What should happen (should, mind you. It can be a bit fickle at times) is the AI will then start to adapt to somewhere between 106 and 111 depending on your lap-times. And for those combos where 106 might be a little too fast, run a few laps against 101 AI, and switch to adaptive. The main problem with the AAI is that the algorithm that governs whether it tries a new level or not is a little wonky. So once you have a good idea of your actual level, it pays to bracket that level by +/- 5-7 levels, just to give it some data to work from. Either by running against fixed AI, or simply editing the index-file if you're so inclined. But once you've got the AAI set up, it delivers some of the best racing you're going to find.
Just finished up one of the best races I have had in a while. Audi TT VLN at Knotstorp Ring, 30 laps. AI was running low 102's. I started 15. It took till the last 5 laps to be with the leaders. Excellent racing, lots of fighting with the ai, and the ai with each other in front of me and behind me. If I let up and coasted for a lap, the guy behind me challenged me, small slip up, and I had to fight for my spot. I really enjoy that track, usually use the frjr there, or even the tt nsu. Decided to give the tt vln a try there... OUTSTANDING!!! The racing was alot of bumping, and poking , squeeze into a spot , time the corner to take advantage of the car ahead, stay close to the car ahead and wait for him to make a mistake, or keep the pressure on him so he doeas make a mistake... Unfortunately the next time i race there, I wll likely stay 15 or only advance a position or 2, which is just as much fun too Excellent job s3
I'll have a go at the AAI , thanks for the advice . With the Ai at 106 i dont usually race and win but rather fight for the podium or even a top 10 and I enjoy that . I dont like not having cars infront of me so having them range up to 111 sounds great