Actually, that's exactly what it does, maybe I should've been more clear. The Adaptive AI adapts according to the last ten laps for that particular car/track combo. If you switch class or track, it adapts to your pace with that combo. There's a bit more to it than that, since it also 'borrows' info from other combos, but no need to get technical about the whole thing. Short version is that the more you race with adaptive AI, the more info it has, and thus the more likely it is to hit the sweet spot for any particular combo. In your case, if your 'standard' level is 105, try running a short race against 110 on the tracks where you're strong, and a short race against 100 on your weaker tracks. Then switch to adaptive. The point being to give the Adaptive AI a bracket, so to speak, of five levels which you fit into somewhere. It's much more likely to adapt if it has an idea of what levels to try out, as it were.
I'm trying to figure out the best way to approach AAI training as a new user. I've barely done around 10 hours in the game and only own the DTM 2015 series but plan on purchasing probably all content very soon. My intention is to use RRE mostly for offline championships. I deleted the xml file since it probably barely has any data in it and I want a fresh start, so the way I want to approach it is: Let's say I want to do a DTM championship in 5 different tracks..... Before starting the championship, pick the car I will be using and do practice sessions with just one AAI (as someone else mentioned in a previous post in order to avoid running into other cars and keep my laptimes consistent) and just hotlap for a while so the game can gather enough information on my laptimes and do that for all 5 tracks. For single races and once I own more series, just hotlap for a while with 1 AAI in any given car/track combo, so by time I decide to run a single race using that same car/track combo, the game will already have data to go by. Is this a good approach or am I high?
You're high. The basis of the approach is sound enough, but the problem is that the AAI doesn't adapt during a race, only between races. So 5 laps of hotlapping will only give you one entry. So when you're just starting out, short races are your best friends. If you're trying to do a 5 track championship with a single class, start doing single races on one track. Standing starts, full grid and starting last. After 2-3 laps, hit Esc and restart race. Rinse and repeat until the AAI has gotten up to your speed. Next track, same thing, only difference is that it'll look at your previous AI levels for that class, so it'll start closer to your level. And so on, and so forth. Every combo you do makes the next one start at the average of the AI levels it already knows about, so the more tracks you do with one class, the faster it'll start to adapt. EDIT: Btw, the reason for using full grids is that if you only race with one AI, you admittedly get less chance of a collision. But your opponent is picked at random, and not all AI drivers are equally fast. So in order to get a nice average of the grid, you need at least 12-15 cars on the grid, preferably all available cars in the pack. Just don't go over 29 opponents.
Thanks for the clarification, especially that last paragraph was very insightful. I wouldn't have guessed that racing one AI would only modify the behavior for that specific driver. Will do as you suggest. Cheers
@CheerfullyInsane Hey man, one more question....If I'm gonna be doing short 3-4 lap races with 20 AI and starting last, what do you think is better? a) Race them normally and in the process probably gain a few places?......In this scenario my lap times would be lower than my potential hotlaping pace as I'll be fighting for places and sometimes get stuck behind a car b) Give them a 3-4 sec advantage at the start of the race, so i'll probably still be last after 3-4 laps but my laptimes would be closer to my hotlaping potential and more consistent as I'll be racing "alone". Does AAI looks strictly at laptimes or also your position in the race?
It only looks at lap-times. It couldn't care less which position you're in. As to whether to race it or not during the 'practice' races, there are pros and cons to each possibility. If you don't race them, you'll get times close to your hotlapping ability, and thus the fastest AI that you can still race. But when you start to race them 'seriously', you occasionally end up having the top guys leave you behind because you're stuck in traffic. If you race them, you get an AI adapted to your actual race-pace. But it also means a very high chance of spins and crashes which will affect either yours or the AIs laptimes (or both). Personally, I start out with the hot-lapping option. Then, when I think I have the right pace set, I put myself in P6 and try to actually race them. Usually works out OK. Besides, it's not the end of the world if the setting is slightly off. One of the reasons that I like the AAI option is that it's dynamic, so it'll continue to adapt as you progress through whatever series of races you have planned.
Ok thanks, I think I'll go for the hotlapping option as well. "But when you start to race them 'seriously', you occasionally end up having the top guys leave you behind because you're stuck in traffic." ^^ I don't think that's too bad as that kinda thing also happens in real life racing, unless they slow each other down fighting for positions which I guess can/will happen as well. I'm thinking another pro of this approach is, if their bar is set closer to my hotlaping pace, that should have more realistic and challenging effects during quali sessions. But yeah, the fact that it's dynamic and will continue to adapt is great. Thanks!!
I have been racing the AI at 107% for quite some time, before I recently decided to give the adaptive AI a chance. And I was blown away - this is spot on and eternally better than a fixed level! For my own part, I immidiately noticed a couple of things: 1. After a couple of races, it matches my skill level perfectly, gives a good challenge yet not frustrating. 2. It also adapts quickly to new tracks where I'm not as fast as on other tracks that I know well. This was my initial reason for not using adaptive AI, I was afraid it would take too long for it to adapt, having to adapt for each track/car etc. I don't spend hours and hours of racing each week, so I want quality time when I first have the time to sit down. But this proved to be a completely irrelevant problem - on most tracks it's already adapted better than the fixed setting I've used, in such a way that I often end up fighting for a podium finish. Perfect challenge for me! 3. What I also noticed, is that the field is much more spread out with adaptive AI. When using a fixed setting, it seems to me that all the cars are almost equally fast, so if you spin or make a mistake, chances are that you will immidiately end up last, unable to catch up. That sucks. At least for me, who actually makes the occasional mistake... With adaptive AI, the field seems to spread out much more, and will usually be spread by more than a minute even in a short 15 minutes race. This gives me more room for small (and big) mistakes, without taking away the challenge of fighting for positions. The podium will most likely be out of reach after a spin, but it's far better fun to re-enter the field in 10th position and maybe climb back up to 8th or 6th, than knowing that a spinout will most likely put you last without any chance of catching up. And if I ever stop making these occasional mistakes, the AI will (hopefully) adapt. All this might just be me, but an already brilliant game got even better after switching to adaptive AI. Highly recommended.
Two little questions on that subject. Is there a list of an index of the tracks ? Can someone provide me one maybe devs ? What about fuel load regarding collected data ? Is the AI keeping up the level figured out and connected lap times to that no matter of the load of fuel ? I like to rework the aiadaption file by myself.
You mean a list of the numerical codes corresponding to each track? If so, here: Spoiler: Track listing Raceroom raceway Grand Prix: 263 Bridge: 266 Classic: 264 Classic Sprint: 265 National: 267 Monza Grand Prix: 1671 Junior: 1672 Mid Ohio: Full: 1674 Chicane: 1676 Short: 1675 Zandvoort Grand Prix: 1678 National: 1680 Club: 1679 Lakeview Hillclimb Full Run: 1682 Reverse: 2181 Zolder Grand Prix: 1684 Nürburgring Grand Prix: 1691 Short: 2011 Sprint: 3377 Hockenheim Grand Prix: 1693 National: 1763 Short: 1764 Raceroom Hillclimb Full Run: 1705 Reverse: 2214 Portimao Grand Prix: 1778 National: 1783 Club: 1784 Club Chicane: 1785 Suzuka Grand Prix: 1841 East Course: 2012 West Course: 2013 Bathurst: 1846 Indianapolis Grand Prix: 1852 Moto: 2014 Sonoma Raceway WTCC: 1854 Sprint: 2016 Long: 3912 IRL: 3913 Mazda Laguna Seca: 1856 Hungaroring: 1866 Salzburgring: 2026 Shanghai Grand Prix: 2027 Intermediate (WTCC): 4041 West Long: 4042 Slovakia Ring: 2064 Macau: 2123 Motorsport Arena Oschersleben: 2384 Eurospeedway Lausitz Grand Prix: 2468 Automobilsport: 3291 Moscow Raceway Sprint: 2473 Full: 3383 FIM: 3683 Norisring: 2518 Brands Hatch: 2520 Red Bull Ring Speilberg Grand Prix: 2556 National: 5794 Paul Ricard Solution 3C: 2867 Solution 1A: 4246 Solution 1C-V2: 4247 Solution 2A Short: 4248 Sachsenring: 3538 Spa-Francorchamps Grand Prix: 3870 Classic: 4542 Combined: 4543 Chang International Circuit Full Curcuit: 4253 D Curcuit: 4252 Nordschleife Nordschleife: 2813 VLN: 4975 Tourist: 5093 24 Hours: 5095 Silverstone Grand Prix: 4039 International: 5816 National: 5817 Historic GP: 5862 Stowe Long: 6055 Short: 6056 Gelleråsen Grand Prix: 5925 Short: 6138 Mantorp Park Long: 6010 Short: 6167 Anderstorp Grand Prix: 5301 South: 6164 Knutstorp Ring GrandPrix: 6137 Falkenberg Grand Prix: 6140 Imola Grand Prix: 1850 Sepang Grand Prix: 6341 North: 6578 South: 6579 Spoiler: Car Classes FRJ (20): 253 Aquila (24): 255 HillClimb Icons: 1685 GTR1: 1687 GTR3: 1703 GTR2: 1704 GTRX: 1710 Touring Classics (39): 1712 GTO CLassics (17): 1713 Group 5 (44): 1708 P1 (15): 1714 Silhouettes (24): 1717 DTM 13: 1921 WTCC 13: 1922 P2 (46): 1923 BL Cup: 2322 ProCar: 2378 ADAC 13: 2922 DTM 14: 3086 ADAC 14: 3375 DTM 92 (25): 3499 WTCC 14: 3905 DTM 15 (24): 4260 ADAC 15 (19): 4516 WTCC 15 (23): 4517 FR2 (22): 4597 Audi TT 15: 4680 Audi TT 16 (20): 5726 NSU (21): 4813 F4 (20): 4867 F3 (26): 5652 Audi TT RS (20): 5234 DTM 16 (24): 5262 KTM (29): 5385 FR US (20): 5383 WTCC 17 (16): 6309 BMW 235i (23): 6344 FRX: 5824 Porsche GT4: 6648 Porsche GT3 Cup: 6345 (Numbers in parentheses is the number of liveries in the class) As for fuel-load and its effect on the AI, I couldn't tell you. It's very rare that I run races exceeding 30 minutes.
The index data can be found here as well: https://github.com/sector3studios/r3e-spectator-overlay/blob/master/r3e-data.json
No you are not right. He wanted to explain that the drivers have differnet skills. As you just wanted to race one Ai oppenet you could randomly get a opponent not that fast as another one. So if you match up the pace of your AI other drivers could be faster so you will be not fast enough. Or the other way around. As you drive for example GT3 against K.v.d. Linde and you can match his pace then you will kill other drivers on same level easily.
Oh btw, if you're going to muck around with the AI file manually, here's a few tips. 1) Be absolutely sure of your syntax. Any mistake will result in the file being discarded, and a new one generated. So unless you're used to working with XML syntax, I suggest you use cut'n'paste from other similar entries and just modify the values themselves. And of course, back-ups are mandatory. 2) No more than 10 AI driver entries per combo (index 0 through 9), Any more will get the file discarded and....well, see point 1. Usually 5-6 entries are more than sufficient to do the job. 3) Don't bother editing the file while R3E is running. R3E loads the whole thing into memory when it starts, and rewrites it upon shut-down. So any manual corrections that you make while R3E is running won't affect the AI as they weren't loaded, and they'll also get overwritten upon exit.
Are there any news about the "adaptive ai primer" tool? I edited the assets file by myself, so I can still use the tool. I really like the possibility, to see which settings I have to do to "traine" the AI faster.
The AI is so aggressive. Watch this The game needs to incorporate "%AI Aggressioin" like GTR, AC, PC2, RF2 and problem solved.
I’m sorry I didn’t see anything aggressive there. I saw late braking, following their race line and you hesitating.
At least Video Nr. 2 has nothing to do with braking That AI Guy eventually carries the name of Mark (Marquez)
Project Cars 2 has an aggresion meter for AI from 0 to 100. It works perfectly , i think that it should be the way to go in order to improve the AI behaviour in RR .
i find the Ai ok , sometimes a bit timid even but very raceable , i run at 110-115% with most combos , and honestly its a blast Andi