With this post I would like to highlight some of my reflections on the usefulness of having rain, sun, variable track conditions accordingly. The conclusion of my reflections is that, with the current conditions of the A.I., live track is absolutely useless from the point of view of the single player who loves to compete with A.I .. In fact, there is no consistency in the performance of the A. I. In the different conditions of the circuits, either too fast on a dry track or too fast on the same wet track or vice versa, with all possible combinations of inconsistencies. Therefore, it makes no sense to have a championship with variable weather conditions. I think that the maturity reached by Ams2 in many other aspects of the simulation has reached excellent levels, as well as the amount of content (cars and tracks) absolutely satisfactory. So, I think Reiza should dedicate the next DLC to optimizing the shortcomings I mentioned above that destroy the playability of a title with enormous potential. Therefore, I pray that Reiza would like to accept my observations and devote himself to equalizing the performances of the A.I. making them homogeneous both in the various meteorological situations and between circuit and circuit and between classes of cars. I fully understand that the work to be done is truly huge and time-consuming. Personally I would be absolutely willing to pay for these optimizations like any other DLC, indeed, even more. I would be happy to hear the opinion of other users. Happy new year to everyone.
I don't think the ai is affected by track temp at all, they do affect the track tho, so lots of cars will heat up the track or dry it up if wet I am generally not a fan of this extra physic for the ai, would be much smarter to apply the normal physics to them and use time trial splines for the ai to follow. Assetto corsa does it basicly like that, u can even look at it if u enable the ai debugging
In several AI runs i did, AI laptimes also increased with hotter track temps appropriately to mine. Very cold track temps seem not to hit AI times as hard like my player times though. I agree with many points in the OP btw. and it's a slow process, that wasn't helped by the recent updates to physics and some AI behaviour of course. This part of the game certainly needs more attention, but it was mostly coming down to just not enough time in the last months due to obviously important things to sort in advance. It's on the scope to do better weather related calibration, though.
whats really important imo is if they just generally get faster/slower as the overall tracktemp increases/decreases, or if they are affected by each tile like the human is. as a bit of an insight, livetrack is essentially a grid of tiles that have some properties to them, such as temp, dirt, water etc. the human vehicles grabs the current tile its driving on and calculate its effect on the tire (temp change, grip etc) and the tire effects the tile aswell increasing its temp/wetness etc when debugging the game i couldnt see any of the AI accessing those tiles in a meaningful way (other than changing the properties of the tile). since this game is pretty overcomplicated when atleast from a reversing perspective, theres always a chance its happening somewhere else and since i wasnt looking for the ai being affected by livetrack i could have simply missed it, but yea
I just hat two occurrences that also play into this kind of Issue with the AI, namely that the AI has a hard time stragegizing which tire to use in a given situation involving changing weather conditions at race start. Example 1: G40 Cup at Silverstone, default date, real historic weather. damp track, few puddles, no rain. AI start on slicks and pit in before the end of lap to switch to wets, then pit in again 2-3 laps later to switch back to slicks as the track dries up. Would have been nice for the AI to recognize the trend of the track to become dryer over time. Example 2: F3 at Curitiba, random weather. Light rain at the start if the race, but still dry track. AI start on slicks and pit within the first three laps to switch to wets. This was the 2nd race in a new Championship, so it's extra meh because I can either easily win by default or go into pits as well taking all the risks that come with that. Would have been nice for the AI to start on wets because they know that rain tends to make the track wetter over time. TBF, I don't know if the "automatic tire by waether" setting would have started my car on slicks as well in this scenario. My examples basically only show that the AI has a hard time with transitional weather conditions at race start. I also had races that started dry but rain came a few laps in and the AI had no issue pitting for the correct tire, and they even pitted on different laps, so that already seems to work quite well.
I agree with you guys for the meaningfull reasons that seem very complex, i mean weather, physics and such, but i think all could be solved by just one thing. the physic of AI is a good point and i think Reiza is working on that but the big question, at least for me, is the fact that we struggle to race with AI is simply theyr speed in corners and straight. Thats why i, since AMS1, tought about it and found the best way to achieve a good race is to have a AI EDITOR that would use the best lap from player as a basis for theyr speed in corners and straight, so we woulnd be frustrated every now and then behind a AI at some corners and so the pace would be optimised. Why a AI editor? because we want the AI to be as good as the player and even better, even at a new track, after certain amont of laps i know exactely where the AI are too slow or too fast and obsiously i will take the advantage to pass them at those corners but does this make it a better race? the answer is no. A good race is for the player to feel immersed in believing he is racing against real people, and not by winning just because the AI have so a discrepancy in speed (corners and straight ) versus the player. I dont know i make myself clear, being not english nativ, all i can hope is that Reiza, who is doing a great sim, will think about AI in another way and will go that route instead of patching here and there small adjustments that just dont give a big smile on our faces. Thank you for reading
Eventhough I dont care to "race" AI bots in any of the sims I use then the most efficient way of optimising the bots to human players would be the same strategy as good old NR2003 had integrated almost 20 years ago. In NR2003 it was possible to set the sim in a special Testing Session mode where you should run about 5-8 laps at realistic speed and lines. Afterwards the sim saved some special lp-files for the used track. These files would then be used to set the "boundaries" for the AI bots in different situations when you raced them. The names of some of the lp files does reveal what they are intended for. race.lp raceline.lp minrace.lp maxrace.lp minpanic.lp maxpanic.lp limp.lp pit0.lp minpit0.lp maxpit0.lp pit1.lp minpit1.lp maxpit1.lp .lp files tutorial ByTheWay: There was some bugs/difficulties creating the files. But the intention of creating and use of the files was/is lightyears ahead of most socalled racing sims even today.
Totaly agreed Bruno, i remember it, and yes it wasent without bugs but never the less that was lot of years ago when the Dev's team were passionate about theyr work, i think it got lost somehow afterward sadly. EDIT maybe Reiza Dev's team that seem passionate too, could go this route, who knows...
After posting above I just wonder if iRacing who basically is build upon NR2003 are still using a similiar live AI generating mode. But now as a house intern generator. Because without knowing anything about it then I hear a lot of people describing the AI bots in iRacing as the best since the invention of sliced bread
Yes Bruno, My dream is AMS2 whit Iracing A.I. In Iracing A.I. is very competitive and " smart" , you can make great door to door battle whit them, in all the available circuit. The core of A.I. racing is battles whit smart opponents, whitout this racing is boring.... The current situation of the A.I. makes AMS2 a racing game without the pathos of racing, useless ... and this, given the progress achieved in vehicle physics, FFB, graphics (especially in VR) is a real shame. I repeat the primary concept of my opening post: REIZA, if Madness Engine allows it, make A.I credible ,competitive and homogeneous in various situations, even with paid DLC. Content can wait. We are basically talking about the essence of car racing. The amount of content is secondary, it probably attracts sim racers who like to do 3/4 laps on the new car or track. We must make the competition exciting, that's the important thing ...
