After doing many races at silverstone gp with the GT4's I noticed that my qualifying position was always last around 3 seconds slower than the leaders and my race pace was also around 2-3 seconds slower than the leaders ; the adaptive ai never slowed the ai down. So I did some testing to see what the aiadaption file was doing after each race. The first column in the picture below shows how the file entry looked before my first test race. Note how the AI adaption index is 120 and AI time 2:13.47 (I came last in the previous race and I clearly remember the ai doing 2:11's and 12's whilst I was at best high 2:13's So I thought that in theory the AI adaption process should slow the AI down in the next race. So after my first test race I noted the results in the aiadaption file (second column below). Again I qualified last with the ai around 3 seconds faster then me. Again the ai pace was around 2-3 seconds faster then me. Looking at the AI Adaption Index record I noted that it didn't write a new index but it did change the AI time. I'm not exactly sure what that time represents but the leaders in that race were certainly lapping in the 2:11's and low 2:12's After the second race I recorded the results in column 3 and as you can see the results were pretty much the same as the first race and again no new index was added to the ai adaption file! I also tested this with a new adaption file. Initially everything looked like it was working while the ai was getting up to speed but it kept going all the way to an ai index of 120 and that's where it stayed! I have a limited knowledge of how this all works but I suspect the way the AI Index time is calculated is wrong for silverstone UPDATE: I just looked through my aiadaption file for all tracks raced with the GT4 and I can see that all tracks have only 1 AI Index recorded and it's set at 120 There was only one that had two entries; the first was 80 and then it jumped to 120 and has stayed there over several races. I've attached my aiadaption file for you to analyse
It's not particular to Silverstone, it's a known (well, to me at least) quirk with the AAI, and one of the reasons I never use AAI alone. What I suspect is happening is that the AAI finds a time that is 'close enough' and then stops looking. The AAI works far better when it has known lap-times to fit in between. So there's an easy fix; run against fixed AI at 110, then switch back to adaptive. What should happen is that the AAI now knows that 110 is too low and 120 is too high, so it'll start to explore the levels in between.
@CheerfullyInsane You were spot on thank you I did 3 races with fixed AI at 100 and watch the AI index value. I was surprised the 100 index didn't move much (only +/- 250 ms) as I thought it may have had to work out it's 100 strength speed or something. So I then switch back to Adaptive AI and it took the usual 3 races to get close and then by the 6th race it had it dialed in exactly to where I had figured out the best strength was when I did it manually and the racing was close and fun. It was quite interesting watching the aiadaption file do it's thing after each race (see image below or attachement) Do you think I will have to go through this process with every track? So now I guess the developers will have to look at why when I started racing the GT4's every one of my tracks had the initial AI index set to 120 and why it needs me to do this manual kick up the arse to get it working again. It will be a painfully long process if I have to do this with each track
In theory, yes. What I do is run a few laps against Adaptive, nip out and check what level it chose, then run a few more laps against fixed AI adjusting the level up or down as needed. After a while you get a feeling for how many levels you need to adjust. I do have an idea as to how to utilize the Primer tool in order to speed up the process significantly, but I'll have to test it first. Unfortunately, work is absolutely nuts at the moment so I don't have the energy for it on work-days. So it won't be until at least this coming week-end.
@CheerfullyInsane All good mate; I've now found a method that works pretty good for me I do a 4 lap race at my goto strength of 118 (seems like a lot of tracks I need to be between 118 and 120 as the ai are slow for the gt4's) Then I set to adaptive AI and do 2 or 3 (just depends on how far out the 118 was) 4 lap races with a 10 minute qualifying to see where I start in the pack and it's finds my speed perfectly. At the end of each race I check and record the results of aiadaption file to see what it's doing and when I can see it's settled on one of the indexed strengths I move on to the next track Whilst it's a bit of a pain now that I've set my starting point at 118 the racing is always fun for those 4 laps LOL What is a pitty tho; and I hope the devs fix this soon; is that I'm maxing out the 120 and I can feel the aiadaption file would love to put me somewhere between 120 and 125 so I hope the devs can let the AI go up to 130 or even 140 for the aliens out there. My Zolder results below are pretty typical of the 6 tracks I trained this evening. Started at 118 (column 2) and won that race Then the next race set to adaptive it selected 120 and I won that race too after qualifying second. I feel that an AI strength of around 122 would be great for me. These below results were pretty similar to Red Bull, Brands Hatch and Zandvort. The only ones that challenged me tonight were Chang and Silverstone where the aiadaption process appropriately settled on 117 but I suspect with a bit of setup tweeking I'll soon max them out to 120 as well