Locked in Qualifying / Finishing Order Based on Ratings?

[Slipster17]

Something I've been thinking about over the last couple days and I'm wondering if there is a way to implement this in game. If I'm creating a carset for a single race - i.e. the 1997 Suzuka expo race - I know the qualifying and finishing order of all of the drivers present. Is the ratings system for the game precise enough to allow me to set it in a way that would, for all intents, force the AI to qualify and finish drivers in the real orders if left entirely to their own devices?

I played around with a scale of 100 points that decreased by 3 each position. Marking them so that the top QUALIFIER had their qualifying rating setting at 100/100, the second place qualifier had their rating set to 97/97 and so on down all 30 competitors. Then did the same thing with the finishing ratings so 1st = 100/100, 2nd = 97/97.... etc... So, taking Mike Skinner as an example. He finished 1st and started 4th therefore his two avg stats were entered in as 100/100 and 88/88. At least to my thinking - this should have guaranteed him to start and finish in that order. But after running a test race with just the AI - it didn't work out that way with regards to either. Geoff Bodine - set at avg finish 70/70 and avg start 82/82 - sat on the pole and won the race.

So, I'm guessing the avg start and finish rating doesn't weigh quite as heavily on the outcomes as I had thought - or at the very least the other driver ratings can override those 2 averages to a degree.

Thoughts? Am I asking too much of the ratings system?

 

[Highbank]

There is variables built in that make that not happen

 

[Slipster17]

There is variables built in that make that not happen

Had a feeling that was probably the case - but wanted to defer to someone more knowledgeable than me in regards to this. Thank you!

 

[fortine_oo]

There is variables built in that make that not happen

To extrapolate the “variable” aspect.

The exe decides what ratings values to apply to each ratings category.
The exe assesses a value from within the min/max range for each category.

The exe will always have a “worst” car, no matter the ratings.
To exaggerate the premise, consider a 42 car roster that has 4 ai of mostly equal ratings and exhibit laptimes comparable to the Player. If only those 4 ai comprise a roster, the Player would easily dominate at least one ai, be substantially better than the third rated ai, and probably slightly better than the second rated ai, leaving only the number one rated ai as the stiff competition.

So, even if you have equal min/max qualifying values, the exe will use the other categories to differentiate the pecking order, where the Player will always be better than someone, assuming you’re using an appropriate Opponent Strength value.
If you made all the categories equal min/max values and then gave each ai a lesser qualifying value, you might get your premise to work, although, knowing NR2003’s penchant for being random when you want conformity (and vice-versa), I doubt it.
I’d try it with a 6 car roster, 100/100 rating in all categories but pitting*** and qualifying. For qualifying, I’d start with 100/100 and decrease by 2 (98/98, 96/96, etc) for a 42 car field. For a smaller field you could increase the difference to 3, ??, whatever the largest difference value is that the size of the field allows.
*** pitting - at 100 ai sometimes push it and run out of fuel, 95 is normally high enough to get maximum fuel and tire wear before pitting.

 

[DaleTona]

To extrapolate the “variable” aspect.

The exe decides what ratings values to apply to each ratings category.
The exe assesses a value from within the min/max range for each category.

The exe will always have a “worst” car, no matter the ratings.
To exaggerate the premise, consider a 42 car roster that has 4 ai of mostly equal ratings and exhibit laptimes comparable to the Player. If only those 4 ai comprise a roster, the Player would easily dominate at least one ai, be substantially better than the third rated ai, and probably slightly better than the second rated ai, leaving only the number one rated ai as the stiff competition.

So, even if you have equal min/max qualifying values, the exe will use the other categories to differentiate the pecking order, where the Player will always be better than someone, assuming you’re using an appropriate Opponent Strength value.
If you made all the categories equal min/max values and then gave each ai a lesser qualifying value, you might get your premise to work, although, knowing NR2003’s penchant for being random when you want conformity (and vice-versa), I doubt it.
I’d try it with a 6 car roster, 100/100 rating in all categories but pitting*** and qualifying. For qualifying, I’d start with 100/100 and decrease by 2 (98/98, 96/96, etc) for a 42 car field. For a smaller field you could increase the difference to 3, ??, whatever the largest difference value is that the size of the field allows.
*** pitting - at 100 ai sometimes push it and run out of fuel, 95 is normally high enough to get maximum fuel and tire wear before pitting.

I'm so glad to have someone like you in this community with such a wealth of knowledge about AI behavior. I've been trying to dial in AI performance myself, starting with a few "fast" cars and adding slower cars in to see how the pack shakes out, but have yet to find out something that works for me.

I have the issue where either it is too random or not random enough, and I can't find that sweet spot where the field generally runs in a petty predictable pattern yet two or 3 cars outperform what they usually do. Ideally I'd have a few surprises in qualifying who grid further up/down the pack then they usually should, but everything mostly normalizes out during a 25% race (the longest I usually run my races for).