I have had this one tumbling in my mind for some time and I think you guys are the only ones out there that could actually implement it if you and the community found it intriguing. The problem: Most cars have a qualification time and ability too close to each other. This is far from unique to Automobilista or Reiza games. The Result is that if you take your average single player race most drivers will start often behind the bulk of the AI driver by a couple of seconds, then as one relearns the track and gets more familiar with it, the human driver times improve and he/she catches up with the AI. All too frequently as the lap times improve, I go from last place in both practice or qualification, to blowing by the entire field in first place. When that happens, I have 3 choices: 1) Race like that, starting in P1. Boring as hell and likely turning into a rout where I barely see any other car 2) Exit the session, increase the AI strength and go at it again. This can get boring and hit and miss. 3) If I raced past the AI in qualification, at the start I pull up and let half the field pass and then go chase the car. OK, but not ideal. The solution: As you undoubtedly know, a lot of guys edit the AI files to best reflect Real Life driving abilities of past and present drivers. It's not hard. It's just a bit of a pain to do on all series and all cars, but it's doable. But it's not for everybody and I am actually not sure it is even possible right now as Reiza has been hiding more and more files for reasons I can certainly guess and agree with. I am a programmer, so for me it's not difficult to manually make these changes to the AI and use some of my text editing tools to make things easier, but I got thinking that if reflecting actual drivers abilities is possible, then why not change the way the AI drive to create a bigger spread among the cars. That way regardless of the difficulty level, a SP driver should always have some cars to race against just like in RL. There would be more chances to lap the AI and so forth. Alternate solution: Of course, some sort of on board interface to let us adjust individual drivers abilities would be great, even if it used some sort of third party tool. The way I envision it would be to simply select the car and race as one usually does Set up the average AI ability Then have a screen showing a list of AI each with a slider next to it and their ability registered with the average ability. It would be easy for everybody to use and I don;t think it would be a terribly complex thing to implement. When I took the time to change individual driver abilities in the files, I could have easily written a simple script for those drivers that could have let me change those values in a more effective way. In fact, while we are at it, why not let us edit the AI files even more, like let us choose our own names. From my understanding of the Gmotor system, it should not be a monumental task, but I'll bet you you'll end up with the most fun AI in the business. AND an AI that more closely reflects real life races in both the rubber and gasoline racing world and the simracing world. I never entered a race where the entire field qualified inside 1 or 2 seconds from each other. I understand there could be issues I am not considering. Even though I know enough to know that if you have a value, you can usually out a variable in there and have the value change based on some condition. The AI driver abilities are based on a number of values. Why not let us either affect them individually or maybe give us something a bit more rough, like a tool to separate the AI field in 3 or 4 groups? Something like The Heroes, The Journeymen and the Bums. If each group falls within a second of each driver, now we have a qualification board with a nice 6 second spread from P1 to last. Much more difficult to improve so much that one will go from last place to first. Anyway, I hope you find this intriguing. The few races where I took the time to edit the AI abilities this way were much more interesting and less boring. I think a system like this would be much better than the R3E AI, which is usually a pain to train and often gets it wrong. Thanks for reading this.