Well, there are some reasons for the mainly poor AI, but as mentioned, it`s absolutely possible to have quite good raced with the AI. @Vince: i don't think that those "magical-variables" you have posted are the solution for a proper working AI. The AI-problem is much more profound. It`s hard to get a start to describe those problems, but i will have a try (please don`t bash me for my english):
There are many variables in some track-files, the car`s-physic files, the plr-file and so on, that have to interact with the ai-files. As a result we would need more Mods with Trackpacks! This is the most important thing in my eyes, but it`s not all, what's needed. Getting the ai working is a task, wich imho most modding-teams don`t want to encounter, because:
- the documentation about how to get the AI working is highly unregarded
- it`s quite a lot work needed.
Well, luckily there a handsome mods, wich try to get used of the capabilities of the GMotor-AI, with varying success. If you want have a try what is possible with the AI try our VLN 2005 Mod.
> Here <
you can find some download links
All 180 included cars have their own realistic laptimes and scriptet pitstops. Because the VLN (german long-distance-championship) has only races at the nordschleife, the trackpack is reduced to 10 versions of the Nordschleife. There are learning files for every car included, wich a hardly needed, because the Nordschleife with it`s unique track-sections is a very hard task for the AI.
Be aware, that the mod's AI will only
work with the included Nordschleife-tracks, wich are optimized for the mod`s AI-variables. There is also a batch-file included in the mod, with edits the plr-file to get rF ready for the AI-variables in the talent-files. We've documented all changes to the plr-file in a .doc-file with comes with the mod, for those who are interested in those changes. The steps to get the mod-ai startet are documented in the readme! Sadly, there are no good out-of-the-box AI-variables released with vanilla-rF, in a result of this things get a little bit inconvenient.
Sadly, most of ISI`s documentations about the AI are lost due to the RSC-breakdown. Luckily i had saved some of the explanations at my hard-drive. Maybe someone who is interessted in AI-Modding will have a use for those informations, about how the learning files work:
Ok here's the current situation with RCD files:
Aggression 0 - 100 (%)
As this number increases the AI will give other cars less room, both while passing and while following. It also increases the frequency at which they try to pass and the increases the threshold they are willing to endure before giving up a pass (or crashing, as the case may be...)
The aggression slider in the UI also affects this directly. RCD Aggression * UI Aggression = final....and should always be between 0-100%.
All but worthless or no longer used.
The lower the number, the higher the frequency of "intentional" mistakes. (if AI Mistakes is > 0 in the playerfile) If AI Mistakes in the playerfile is set to 0.0, then the AI drivers never make mistakes. *cough* Well, mistakes that I planned. "intentional" mistakes are things like, taking a turn too wide, or missing a breaking point.
Increasing composure also decreases the time between bad driving zones. (also affected by the playerfile variable AI Limiter, see MinRacingSkill below)
How close to "optimal" this driver drives a track. This number directly interacts with a variable in the AIW file called "AIRange" AIRange tells us how wide the gap, from 0-100% (expressed as 0.0 to 1.0 in the AIW) The default is 0.1 meaning that the full range of Speed RCD values from 0 to 100, are directly mapped to the top 10% of the max speed = 90-100% For example a driver with 100% speed will drive at (0.9 + 0.1 * 100%) = (0.9 + 0.1 * 1.0) = 100% of what he thinks the track can handle, while a driver with speed = 0 will drive at (0.9 + 0.1 * 0%) = (0.9 + 0.1 * 0) = 90% what he thinks the track can handle. There is a random fluctuation in this value, but it's much less than in previous version where the random fluctuation was so great that it made the original value meaningless.
Crash: 0 - 100 %
CompletedLaps: 0 - 100%
Only used in auto completing laps (skipping practice/qualifying)
MinRacingSkill=0 - 100%
When the AI Limiter variable(from the playerfile) value is > 0.0, the AI drivers go through cycles of optimal driving and sub-optimal driving where their driving skill falls to MinRacingSkill * Speed. (Set the AI Limiter variable to 0.0, and the AIs will always drive to the best of their ability...every lap) So for example if you have a driver with Speed = 50, a AIRange in the AIWfile = 0.2), and MinRacingSkill = 90 This AI guy will usually drive around at (0.8 + 0.2 * 0.5) = 90% speed (really very very slow....but this is just an example), but will sometimes dip as low as (this speed * MinRacingSkill = (0.9* 0.9) = 81% while having a bad lap.....that is extremely low, but this is just an example with easy numbers.
Actually, this is just a part of many informations, that are needed to get a proper start with this kind of modding. But i think these basic-informations explain it very well, why it is so difficult to mod a good working AI, and why we have almost no GMotor mods/games with good working AI.