- 3,102
- Finland
- thematic604
I find that the AI pace potential of GT6 is good, it's just dumbed down in career mode. If this hindrance to their ability was diminished there could be subtle variation instead. Since some tracks are quite long I don't think AI practice/warm-up session with black screen makes sense, but perhaps the first lap of the race could determine "individually" the strong and weak spots. I'm not going into adaptation, just memory.
Suppose in the very first turn of the race, the lead driver "remembers" the corner well and if left to their own devices, makes a good entry and exit of it. The rest of the drivers would typically be unique, but the car's performance would matter much more. In subsequent laps, each driver would perform in a given corner as well as determined on lap 1, regardless of whether they had a chance to do it right then. I'm assuming here that the AI can read the "track map" - even for RNG tracks - enough to be able to have an array of bends and jumps to follow.
The difference to just adding some noise to the AI inputs is especially for one make races. On many tracks the individuals that got the best exits where it counted, would eventually be at the top. A race could still be restarted in the same conditions by the player, so long as the randomizing was done before loading the event.
Downside is that the range "I have an idea about this corner, it's on this paper slip" to "I know this one better than I know myself", would be not fun to tweak.
Suppose in the very first turn of the race, the lead driver "remembers" the corner well and if left to their own devices, makes a good entry and exit of it. The rest of the drivers would typically be unique, but the car's performance would matter much more. In subsequent laps, each driver would perform in a given corner as well as determined on lap 1, regardless of whether they had a chance to do it right then. I'm assuming here that the AI can read the "track map" - even for RNG tracks - enough to be able to have an array of bends and jumps to follow.
The difference to just adding some noise to the AI inputs is especially for one make races. On many tracks the individuals that got the best exits where it counted, would eventually be at the top. A race could still be restarted in the same conditions by the player, so long as the randomizing was done before loading the event.
Downside is that the range "I have an idea about this corner, it's on this paper slip" to "I know this one better than I know myself", would be not fun to tweak.
