New season ratings update + standings script
- Thread starter ai_line_mod
- Start date
Sorry, last post here, but looking CTS used a 65 degree spoiler since they started racing at SS in 2000, so the 65 makes sense for them. I never thought the CTS physics were good for cup, and this is yet another reason why.
Finally tried out Talladega. works very well. Very similar to Daytona. The ai starts spreading out after a few laps. The default fast setup with 45 degree spoiler however seems to be a lot looser than at Daytona, especially the first few laps. I didn't have time to try longer runs to see if it's better with warmer tires.
Thank you guys for sharing this information! I have only watched some a.i. races so far at Daytona and Talladega but using your ratings and exe and the nr4 lap files, the racing is much improved. I've never liked the plate racing on NR2003 but I'm exited to race at these tracks now.
I should mention, after some testing, I've settled on using just the race.lp from NR4 and the min/max from NR2003. I couldn't take the AI checking up... Also, I reverted to the 55 degree spoiler for now, as it feels like this is what the SS physics were adjusted to, so without some other major reworks I stuck with that. But now the exe is in your hands so more testing/exploration can occur.
That's interesting to me. I actually transferred every lap file from NR4 and at least in watching the a.i. I don't see the check ups in the speeds. That is with cup physics, 45° spoilers, weather set to clear 85°, and a.i. set to 97%.
Now cts physics is a whole different story. Weather set to clear 60°. The pack doesn't spread out as well and the runs are much bigger causing major check ups.
I wonder if increasing drafting distance a little and playing with panic decel might help with the checking up?
Now cts physics is a whole different story. Weather set to clear 60°. The pack doesn't spread out as well and the runs are much bigger causing major check ups.
I wonder if increasing drafting distance a little and playing with panic decel might help with the checking up?
Panic decel is an idea. As you know, cloudy weather also provides an HP boost. I've found that cloudy provides a nicer spread in the field (maybe because more rating emphasis goes to Engine?). It seems that a good Aero/Engine spread results in the kind of SS we're looking for, as there is more "follow the leader" type racing, which I think you'll agree mimics the actual SS racing of the 80s/90s.
Been doing and watching a bunch of NR4 and NR2003 superspeedway racing, and I've had a realization about what sets apart NR4's AI racing on superspeedways.
Cars in NR4 in a single file train tend to go significantly faster than when cars are side-by-side or three-wide in a pack (especially through the corners). From memory, it's a difference of 5 mph or more (e.g. 190 vs. 185 mph).
This has a single huge benefit. It means that in NR4, by getting into single file, one group of cars can actually catch up to another! Whereas in NR2003, if a pack breaks away at the front, it runs away... a group behind it will never catch up under green flag conditions.
It makes longer races with AI much more engaging in NR4 if you start or slip towards the rear of the field. You might actually catch up. Even towards the back of the field, the cars never quite permanently lose touch. And, like in real life, there's actually a benefit to running behind a fast car and bumping them along the straight to try to catch up to a group ahead. It makes it not feel like a snooze-fest when cars are in single-file... you know you may be catching cars ahead.
I wonder if there's a parameter or two in the NR2003 EXE (perhaps related to side draft) that you could adjust to slow cars down when side-by-side to create the same effect as NR4. Given what I've realized about NR4, it might make a huge difference.
A few other things I've noticed about NR4 – some of which have already been noted in other posts above, but I'll put here for completeness' sake:
1. Even at Daytona, you can't flat-foot the car through the corner (especially on entry) which makes the driving more engaging, and is true to life even for the early-mid 2000s superspeedway packages;
2. The speed gains from the draft seem stronger and to extend further back than in NR2003, which makes the racing more engaging and may (or may not) also affect the extra speed in single-file trains;
3. There is more grip on the bottom than the middle, and even less on top, which is an issue that is lesser but still present into NR2003 (Brian Ring tracks aside) but artificially helps create variety by inducing a line of faster cars on the bottom (which can then break away in single-file and catch up to groups ahead);
4. I'm less confident about this, but it seems speed in the draft is sometimes affected by some kind of 'adaptive speed control' where your speed is partially dictated by that of the car in front... I've had my speed drop significantly (e.g. 5 mph) with my foot flat to the floor when approaching a pack of cars up ahead that are running slower.
Finally, I'll just share this video from GPLaps, in case anyone hasn't seen it. A fantastic showcase of what NR4 can do.
Cars in NR4 in a single file train tend to go significantly faster than when cars are side-by-side or three-wide in a pack (especially through the corners). From memory, it's a difference of 5 mph or more (e.g. 190 vs. 185 mph).
This has a single huge benefit. It means that in NR4, by getting into single file, one group of cars can actually catch up to another! Whereas in NR2003, if a pack breaks away at the front, it runs away... a group behind it will never catch up under green flag conditions.
It makes longer races with AI much more engaging in NR4 if you start or slip towards the rear of the field. You might actually catch up. Even towards the back of the field, the cars never quite permanently lose touch. And, like in real life, there's actually a benefit to running behind a fast car and bumping them along the straight to try to catch up to a group ahead. It makes it not feel like a snooze-fest when cars are in single-file... you know you may be catching cars ahead.
I wonder if there's a parameter or two in the NR2003 EXE (perhaps related to side draft) that you could adjust to slow cars down when side-by-side to create the same effect as NR4. Given what I've realized about NR4, it might make a huge difference.
A few other things I've noticed about NR4 – some of which have already been noted in other posts above, but I'll put here for completeness' sake:
1. Even at Daytona, you can't flat-foot the car through the corner (especially on entry) which makes the driving more engaging, and is true to life even for the early-mid 2000s superspeedway packages;
2. The speed gains from the draft seem stronger and to extend further back than in NR2003, which makes the racing more engaging and may (or may not) also affect the extra speed in single-file trains;
3. There is more grip on the bottom than the middle, and even less on top, which is an issue that is lesser but still present into NR2003 (Brian Ring tracks aside) but artificially helps create variety by inducing a line of faster cars on the bottom (which can then break away in single-file and catch up to groups ahead);
4. I'm less confident about this, but it seems speed in the draft is sometimes affected by some kind of 'adaptive speed control' where your speed is partially dictated by that of the car in front... I've had my speed drop significantly (e.g. 5 mph) with my foot flat to the floor when approaching a pack of cars up ahead that are running slower.
Finally, I'll just share this video from GPLaps, in case anyone hasn't seen it. A fantastic showcase of what NR4 can do.
Last edited:
Thanks, these are all good points. I have thoughts on some of them:
Of course, it's usually never one factor. I think another is that the effect of the draft was diminished in the patch (it's in the readme: "Reduced draft effects.") which you touch on in point #2.
Regarding point #1, I agree, and found the best way to simulate this is what I have done as described in previous posts, namely to borrow the *race.lp files from NR4. The AI frequently check up going into turns with these lines and as a result so does the human player, unless they are up front of course (if so, time to up the AI difficulty!).
#3: I'm not sure about the grip you mention, but I fixed this well enough for myself by using the *race.lp files as above.
#4: That might be the physics itself, a combination of drag/downforce/dirty air, etc.
I've seen the video; it is what inspired me to take the LP lines from NR4 and place them into NR2003. Honestly, the effect of just moving the LP lines has greatly increased my enjoyment of SS racing in 2003. It's not quite the same, but due to the AI checking up the field spreads out, yet remains "together" at the same time. There's something about the information of the physics in the creation of the LP lines in NR4 that translates well in NR2003 for SS racing. Interestingly, if you take the LP lines from NR2003 and place them in NR4, the race becomes a bore: once the field somewhat sorts itself out the AI just drive in a line. This may be due to some effects of the track.ini. I wish there was a way to make LP lines in NR4 but alas I do not believe there is a way.
By the way I should note that most of the testing I've done is with my ratings. I'm not sure how others work with the NR4 LP files, but mine were specifically designed with SS racing in mind (and all the other chassis/track types of course).
This *may* be due to ai_squeeze_pcnt. If you look between NR4->02->03, squeeze percent diminished to an effectively non used value (only default track to use it in NR03 is Kansas at 0.1, and even then I wonder if it was just missed). Some function in the exe must have been tweaked so that ai_squeeze_pcnt became ineffective, or rather counter effective, hence the 0 on all tracks. I've tried playing with it, but the results I get when using it are subpar. The only maybe workable value at SS is 0.1. It seems to cause AI to back off, or give more room, when latitudinally "squeezed" in the racing lines (why they sometimes just drive into the wall if minrace.lp is sufficiently close).
Of course, it's usually never one factor. I think another is that the effect of the draft was diminished in the patch (it's in the readme: "Reduced draft effects.") which you touch on in point #2.
Regarding point #1, I agree, and found the best way to simulate this is what I have done as described in previous posts, namely to borrow the *race.lp files from NR4. The AI frequently check up going into turns with these lines and as a result so does the human player, unless they are up front of course (if so, time to up the AI difficulty!).
#3: I'm not sure about the grip you mention, but I fixed this well enough for myself by using the *race.lp files as above.
#4: That might be the physics itself, a combination of drag/downforce/dirty air, etc.
I've seen the video; it is what inspired me to take the LP lines from NR4 and place them into NR2003. Honestly, the effect of just moving the LP lines has greatly increased my enjoyment of SS racing in 2003. It's not quite the same, but due to the AI checking up the field spreads out, yet remains "together" at the same time. There's something about the information of the physics in the creation of the LP lines in NR4 that translates well in NR2003 for SS racing. Interestingly, if you take the LP lines from NR2003 and place them in NR4, the race becomes a bore: once the field somewhat sorts itself out the AI just drive in a line. This may be due to some effects of the track.ini. I wish there was a way to make LP lines in NR4 but alas I do not believe there is a way.
By the way I should note that most of the testing I've done is with my ratings. I'm not sure how others work with the NR4 LP files, but mine were specifically designed with SS racing in mind (and all the other chassis/track types of course).
Addendum: Since you brought this thread back up, I've been doing more research. A few things:
I'm very please with how the AI race, and in turn racing the AI with these settings. You cannot just take the bottom line to the front, and the field spread makes it feel like you're racing the track just as much as the field. With my ratings, PWF expert setups, and difficulty between 100-105% depending, I can get to about P20 in a 10% race at either track. I have yet to drive a longer race, but have simulated several, also with promising results.
- 1 ) I brought over both the Dayton and Talladega race, maxrace, minrace, maxpanic and minpanic LPs from NR4.
- 2 ) For Daytona, I changed the ai_line_modifier from 1.00 to 1.01. This relieves the bottlenecking that occurs into turn 1. Note it still happens, but just slightly.. enough to keep the field spread. Bumping to 1.02 prevents the checking up at all, but you also lose the field spread.
- 3 ) For Talladega, leaving the ai_line_modifier at 1.02 seems to work best. Taking it down to 1.01 causes the field to check up too much into turn 1. Any higher and you just end up with a wad of cars.
I'm very please with how the AI race, and in turn racing the AI with these settings. You cannot just take the bottom line to the front, and the field spread makes it feel like you're racing the track just as much as the field. With my ratings, PWF expert setups, and difficulty between 100-105% depending, I can get to about P20 in a 10% race at either track. I have yet to drive a longer race, but have simulated several, also with promising results.
FYI, I uploaded the first version of my changes here: https://stunodracing.net/index.php?resources/90s-superspeedway-ai.17185/
Hot Links
