I am an online racer and discovered RR a while ago because I felt it had the best AI of all the racing sims I have played. I am going to give adaptive ai a shot and after doing some research on the forum I still have a few questions. 1- I plan to run a series of 4-5 5 lap races on a number of tracks and wonder what position I should place my car. 2- After I get enough data for a track does this process need to be repeated for a different car running on the same track ? 3 - When racing on a track that has accumulated sufficient Adaptive ai data what happens during qualifying ? It's my understanding that data is only gathered for races. 4 - I believe the adaptive ai takes the average of my last 10 races. If I have a very slow lap like a crash will that be enough to skew the data so that I may get some unrealistic results ?
I am not an expert regarding AAI, and I wonder why there is no (at least short) documentation answering those questions, how to use AAI and what to expect. 1. As the final goal of AAI seems to be finding a pace for you to run somewehere top-ten to mid-pack you may want to start there. But if the AI are much too slow you might also start in pole and show some fast clean laps? 2. As far as I understand the AAI first checks for the car/track combo. If there is no data it looks for other cars on this track, and so on. It tries to guess your pace. 3. Probably nothing happens. It would be somewhat strange if the AI adapted between qualy and race, wouldn't it? 4. I have read that consistent lap times are needed for a good AI adaption. I expect that there is no filter at all. Probably this won't be a problem unless it happened regularly.
Thanks for the info. I guess another alternative might be to pick a percentage, run several short races and take a look at overall performance against the ai. In the past I have been quick to change the ai %. This did not allow for some natural variation in ai performance. I'll give adaptive a try and see what happens. It might not be a bad idea for AAI to not use laps that have a very large variation to avoid skewing problems. I would like to get more info on how it actually works.