Creating LP lines

[ai_line_mod]

What is the best method you have for making LP lines? I generally follow fortine's advice from nrtracks, however I felt like I couldn't make a line like the original papy team. Messing around, I've found that using chassis type 0 for the track I'm trying to edit/create gets me the closest. Here's my general method:

1. 1. dial in the grip level. (most of the tracks I'm working on are community made tracks where the grip level is too high. it's really an unfortunate trend and results in bad racing both in terms of ai and player experience.)
2. 2. dial in a setup. part of this includes some back and forth with the track_tire_heat, as well as with the grip level. i look at lap times for around the era i'm working in and try to get my fastest lap close to or below the record
3. 3. change the chassis type to 0
4. 4. make your race lp AS IF you were driving with the normal chassis (e.g. lift where you would normally lift, gas where you normally gas, just like how you had to when you ran good laps with your setup)
5. 5. make all other lines as necessary
6. 6. change chassis type back to original value
7. 7. test with the AI (set ai_accel_modifier, ai_decel_modifier, ai_grip_modifier, ai_drag_modifier, ai_line_modifier to 1). if you don't like the race lp, go back to step 3 and repeat
8. 8. once you found a good race lp, start driving with the AI and adjust the values listed in 7 so that they become competitive around 97-100% difficulty. if i find myself having to raise the ai_grip_modifier much above 1.05, i'll leave it be and just race the AI at a higher percent. NOTE there should generally be an inverse relationship between track surface grip and ai_grip_modifier. if both are high (> 1.05), then something's wrong with your race lp and go back to step 3.
9. 9. when you get the AI to race well with you for the first 10 or so laps, dial in track_tire_wear, track_tire_wear_loss, and track_tire_wear_exp along with the ai_tire_wear and ai_fuel_use so that they have parity and are appropriate for a full pit cycle. this is where some research/judgement comes in and i'll watch old races on youtube from the era i'm working with to hear how the commenters are describing track conditions, pit windows, etc.
10. 10. tweak remaining ai_track values as necessary

Has anyone else successfully used chassis_type = 0 to make their lp lines? Why can't I find anything about this anywhere, is it a secret??

 

[ai_line_mod]

Scratch that. chassis type 0 is at best as good as using the standard chassis for race.lp, likely worse, just as finicky. It still has utility for other lp lines, but race.lp proves to be just as elusive as previously thought.

 

[fortine_oo]

... however I felt like I couldn't make a line like the original papy team.

Huh?

Having made a "Controlled speed" race.lp that out performed my very good "full speed" race.lp, I've made the "Controlled speed" method my normal "go-to" method.

Chassis: Using this "Controlled speed" method, which chassis you use to create the race.lp is less critical than the the full speed method.
Just because I'm 'old school', I tend to use the series for which I'm creating the ai. That would be cup used to cover all nascar series, unless it's only a gns or cts track, in which case I would use that series.
For pta or gtp tracks, I use the series for which I'm creating the ai.

NOTE: nr2003.exe records/saves the series type as well as the physics of the race.lp.

----- ----- ----- ----- -----

Controlled speed___better than acceptable for racing, and if done well, it rivals (even beats) the results of the "full speed" method.
a) If you absolutely can't get a good speed lap, use this method.
b) If you have a track that requires you to drive a line (for a race.lp) that is not the best racing line in order to force the ai to drive the best racing line when the track construction is pulling the ai offline, the "Controlled speed" method seems to net you a better race.lp.
c) If you want to create an excellent race.lp in the shortest amount of time, use this method.

NOTE:
This method is particularly suited for a RC, where smooth driving and car control is a premium for good race.lp creation and slowing and acceleration occur regularly. Real race speed is not required, TECHNIQUE*** IS CRITICAL.
I have successfully used this method to create a race.lp for ovals.
(I may have created a race.lp for a high racing line daytona using this method, but if I did, I don't recall the throttle technique I used, since current superspeedways are run without lifting the throttle in the turns. Unless you're having trouble controlling the car while flat out, it may be just as easy to create a race.lp for a superspeedway driving at full speed.)

*** Technique:
1) When acceleration is required, apply the throttle gingerly and keep applying the throttle and gaining speed until it's time to slow for a corner. It's easier to stay in control if you never reach the "actual" top speed for the straight. You may have to reduce the rate of throttle application to keep the speed down. The rate of throttle application isn't paricularly important. What's critical is to continually increase the throttle and continually increase the (car's) speed until it's time to slow down. AVOID on/off/on throttle application, this causes the ai to slow (dramatically) when it occurs in an acceleration zone.
2) Let off the throttle to reduce to corner speed; you can lift early and coast into the turn. Only apply the brake if necessary (and apply it as gingerly as the throttle). Release the brakes soon as possible and begin to apply the throttle prior to the corner apex, again apply the throttle gingerly and keep applying the throttle until it's time to slow for the next corner. NEVER apply the throttle and then get out the throttle because you've misjudged your speed or lost control of the car, this will only cause the ai to checkup dramatically and lose momentum at race speed.

 

[ai_line_mod]

Thanks, fortine. How is it known that physics is saved in the lp? I haven't been able to open the file in any meaningful way to see what's contained, or how it's used.

I think I'm starting to understand your controlled speed method. You drive at 90-95% of normal full race aggression, which helps you keep your inputs very smooth.

By the way, you might find this article interesting, if you haven't seen it yet: https://racesimcentral.net/richard-...gends-papyrus-and-the-final-days-of-a-studio/

 

[J.R. Franklin]

Loved reading the interview with Rich Yasi. Good stuff.

 

[fortine_oo]

Thanks, fortine. How is it known that physics is saved in the lp? I haven't been able to open the file in any meaningful way to see what's contained, or how it's used.

I think I'm starting to understand your controlled speed method. You drive at 90-95% of normal full race aggression, which helps you keep your inputs very smooth.

Re: physics
Sorry, I chose the wrong word. I meant the reflap process saves the Player's "inputs" while driving the lap. Those inputs are then interpreted by the exe to create the physics that are imparted to the ai, hopefully creating a human-like opponent to race against.

Re: technique
I never get anywhere near real speeds or real braking. The exe is marvelous at creating the proper actions from the least restrictive inputs, i.e. drive the path you want the ai to take, show where the high speed is by increasing the speed and show where speed reduction is appropriate by releasing the throttle.
This easy drive lap method is infinitely the easiest, hence quickest, method of creating the best ai lap. It also makes track.ini adjustments easier because you're not trying to correct for the Player-driver-created anomalies.

Re: ai_line_mod said " ... however I felt like I couldn't make a line like the original papy team. "
You didn't answer.
I'm curious, what does that mean? The path? The ai's driving ability? Something else?
It's been a long time since I've driven an original track with original lp's, but I'm pretty sure some Papy tracks require some tweaking at the very least, be it lp's or ini values. I can surmise the Papy team was under the gun to get the product "out there" and they often had to settle for "good enough".

Also loved the interview, thanks for the link.

 

[ai_line_mod]

Let me try to explain. Before I was interested in making LP lines, I was working on a ratings formula which would be dynamic yet accurate over the course of a season. Through the process of making this ratings formula, I would simulate seasons in order to compare the game results with real results. I worked with cup physics the most, but also with gns and cts, as I wanted the ratings to be applicable to all series/physics types. In order to establish some "baseline" in the ratings research, I used the game "as designed" and used the original exe with original tracks, etc. Since e.g. gns and cts schedules were different, aka shorter, I wanted to test their (2003 ) schedules and meticulously searched for the community made tracks which were the "best" i.e. raced like the papy tracks.

This is where I encountered problems. Finding a track where the AI ran like the papy tracks was near impossible! Main issues I noted:

• - AI weren't able to pass each other/bottom lines check up in the turns (unsure what causes this, may be ini, I suspect LP lines, likely a combo)
• - AI raced in a "pack" aka did not spread out as in the original papy tracks (what I will call longitudinal spread; pretty sure this is due to high grip settings)
• - AI drive in "lines"/don't have the same sort of latitudinal spread that papy tracks have/vise versa and drive too wide (suspecting LP lines)

These of course interfered with my ratings "data," affecting things like average finishes, lead lap finishes, etc. They also affected what it was like to just watch a simulated race, e.g. in my ratings the qualifying positions are variable, and drivers move through the field, up or down, over the course of the race, at least on the papy tracks. Not so much on many community made tracks. Likewise they affect what it's like to race the AI as a human player.

The ratings I finally developed can be found here: https://stunodracing.net/index.php?resources/full-season-ratings.14511/. I am not trying to brag but they work quite well for what I was aiming for. I used the ratings from the default cup and pwf bgn/cts mods as I suspected they knew a bit more than the rest of us which values/value relationships worked best, along with community researched and made ratings (e.g. the ones that come with NRatings).

Now, I intend to rework the tracks that were at least on the 2003 bgn/cts schedules so that they race like the papy tracks, without the issues noted above. PWF released several of these tracks, and they race nicely (except for PWF Homestead 2003 of course) and I want to fill in the gaps. Hence my inquiries. If you would like examples and how they deviate from the papy tracks I can happily provide several.

Which papy tracks do you recall needing tweaking? The one that stands out the most is the bristols, which were butchered with the conversion to the single pit road. This can be fixed simply by reverting to the two pit road system, which I did here: https://stunodracing.net/index.php?threads/bristol-ai-pit-issues.10783/. Martinsville has issues with half the field going a lap down when pitting under caution, I'm not sure how to address this one. Talladega/Daytona the pitting could be smoother so the AI don't check up. Darlington is difficult because the spread between min/max LPs are so small in the turns. But these are small issues when compared to just how the AI race, imo.

NOTE: I know as a rule of thumb that flatter tracks tend to have higher grip, thus less field spread. This correlates roughly with real life results however: tracks like new hampshire or pre2003 homestead generally had a higher amount of cars finish on the lead lap than more banked tracks, on average, considering cautions etc.

 

[ai_line_mod]

I've done some testing according to your methods. In this case, I'm trying to "fix" Zager's Atlanta 97 as the original LP lines are atrocious. See:

View: https://youtu.be/pept4Kxr0x8

There are three laps: 1) at 90% of race speed (full speed), 2) saving reflap (throwaway lap), 3 ) at controlled speed (or my interpretation of controlled speed).

What I found is full speed is finicky to create, but works like the original papy tracks. Controlled speed AI is too slow. Furthermore, it's hard to get the "line" right, as in, drift down into and out of corners as you would at full speed, therefore affecting the AI lines. Maybe controlled speed is superior at road courses (I can barely get a full speed lap in NOT under pressure).

How do I know this? I try both methods in an original papy track and see how the AI race. If I need to dramatically alter ai_track parameters, I know something's amiss. Of course, some changes are necessary because it's a different LP file, but they shouldn't be drastic to e.g. get the AI at the speeds they were with the original LPs.

In the below post, I pasted a table from my excel which has the papy tracks with the relevant track and ai_track parameters. If the LP line requires substantial deviation from these figures, at ANY track, then that is a good indicator that the LP line is not like the papy LP lines. e.g. there should be an inverse relationship between track grip and ai_track grip. (I can't upload a csv in the post, but I can upload it to downloads if there's interest.)

 

[ai_line_mod]

track

track_type

chassis_type

track_tire_heat

track_tire_wear

track_tire_wear_loss

track_tire_wear_exp

track_asphalt_grip

track_concrete_grip

ai_accel_modifier

ai_decel_modifier

ai_fuel_use

ai_grip_modifier

ai_drag_modifier

ai_line_modifier

ai_qual_modifier

ai_wall_offset

ai_tire_wear_left

ai_tire_wear_right

ai_pacing_distance

ai_bunching_distance

ai_drafting_distance

ai_panic_decel

ai_dlongpad_scale

ai_dlat_pad

ai_squeeze_pcnt

ai_inverse_slipcurve_k

atlanta

2​

2​

0.89​

1.02​

0.38​

0.94​

0.94​

1​

1​

0.95​

0.96​

1.08​

1.03​

1​

1.038​

100​

0.4​

0.48​

1.8​

1.65​

1.25​

7​

5​

0.85​

0​

0.63​

bristol

0​

2​

1.11​

0.53​

0.22​

1.03​

1.12​

1.1​

1.1​

0.95​

0.89​

1​

1.05​

1​

1.056​

100​

0.98​

1.1​

1.9​

1.6​

1.15​

6​

5​

0.8​

0​

0.28​

california

3​

2​

0.95​

0.84​

0.27​

1.17​

0.99​

1​

1​

0.9​

0.98​

1.05​

1.065​

1​

1.038​

100​

0.47​

0.57​

1.9​

1.6​

1.3​

6​

5​

1.3​

0​

0.66​

chicagoland

2​

2​

0.94​

1​

0.31​

0.92​

1.02​

1​

1.2​

0.9​

1.01​

1.03​

1.05​

1.01​

1.05​

100​

0.41​

0.5​

2.2​

1.7​

1.3​

6​

6​

0.9​

0​

0.6​

darlington

1​

2​

0.86​

1.08​

0.43​

0.76​

1​

1​

1.23​

0.95​

1​

1.05​

1.07​

1​

0.986​

100​

0.48​

0.56​

2​

1.7​

1.55​

6​

5​

0.7​

0​

0.35​

daytona

4​

1​

0.95​

1.29​

0.2​

1.2​

1​

1​

1​

1​

0.98​

1.1​

1.07​

1​

1.02​

100​

0.3​

0.39​

2​

1.7​

1.5​

6​

10​

1​

0​

0.79​

dover

1​

2​

1.08​

0.78​

0.4​

1.05​

1.05​

1​

1.05​

1.02​

1.05​

1.03​

1.05​

1​

1.06​

100​

0.58​

0.69​

2​

1.7​

1.24​

6​

5​

0.85​

0​

0.45​

homestead

1​

3​

0.92​

0.81​

0.3​

0.8​

1.07​

1.07​

0.38​

0.92​

1​

0.98​

1.08​

1​

1.015​

100​

0.73​

0.82​

2.1​

1.8​

1.36​

5.4​

6​

0.9​

0​

0.65​

indianapolis

3​

3​

0.84​

1.25​

0.2​

1.2​

1.02​

1​

1​

0.98​

0.99​

1​

1.07​

0.99​

1.022​

100​

0.43​

0.56​

2.4​

2.1​

1.5​

6​

5​

1​

0​

0.58​

infineon

5​

4​

1.13​

0.87​

0.2​

1.07​

1.05​

1.05​

0.4​

0.8​

1​

0.98​

1.1​

1​

1.039​

100​

0.9​

0.88​

2​

1.55​

1.15​

4.9​

1.5​

0.8​

0​

0.65​

kansas

2​

2​

1.07​

1.03​

0.28​

0.93​

1.01​

1​

1​

0.85​

0.98​

1.06​

1.04​

1​

1.055​

100​

0.455​

0.517​

2.2​

1.7​

1.2​

6​

5​

0.8​

0.1​

0.65​

las vegas

2​

2​

1.11​

0.89​

0.26​

0.92​

1.03​

1.01​

1​

1​

1​

1.02​

1.056​

1​

1.053​

100​

0.48​

0.58​

2​

1.7​

1.3​

5.8​

7​

1​

0​

0.58​

lowes

2​

2​

0.83​

0.98​

0.25​

0.9​

1​

1​

1​

1​

0.99​

1.07​

1.08​

1​

1.015​

100​

0.39​

0.5​

2​

1.7​

1.2​

6​

5​

0.8​

0​

0.68​

martinsville

0​

3​

1.87​

0.35​

0.19​

0.96​

1.11​

1.16​

0.38​

0.99​

0.925​

0.99​

1.05​

1​

1.045​

100​

1.58​

1.95​

2.1​

1.8​

1.08​

6​

5​

0.9​

0​

0.28​

michigan

3​

2​

0.86​

0.88​

0.24​

1.15​

0.93​

1​

1​

0.88​

0.96​

1.09​

1.05​

1​

1.025​

100​

0.39​

0.48​

1.9​

1.75​

1.3​

6​

5​

1.1​

0​

0.88​

new hampshire

1​

3​

1.3​

0.56​

0.18​

1.07​

1.12​

1.1​

0.38​

0.9​

0.98​

0.98​

1.1​

1​

1.041​

100​

0.95​

1.08​

2.1​

1.8​

1.15​

6​

5​

0.8​

0​

0.32​

north carolina

1​

2​

1.08​

0.98​

0.43​

0.8​

1.12​

1.08​

1​

0.96​

1.05​

0.99​

1.08​

1​

1.051​

100​

0.66​

0.76​

2.1​

1.9​

1.3​

6​

5​

1​

0​

0.46​

phoenix

1​

3​

0.81​

0.71​

0.25​

0.93​

1.08​

1.08​

0.44​

0.9​

0.96​

0.99​

1.11​

1​

1.024​

100​

0.46​

0.55​

2​

1.7​

1.14​

5.5​

5​

0.9​

0​

0.26​

pocono

3​

3​

0.91​

0.85​

0.26​

1.13​

1​

1​

1​

0.95​

1.03​

1​

1.04​

1​

1.022​

100​

0.71​

0.83​

2.4​

1.9​

1.3​

6​

5​

1.1​

0​

0.32​

richmond

0​

3​

1.08​

0.51​

0.29​

0.95​

1.1​

1.08​

0.41​

1​

0.99​

0.99​

1.09​

1​

1.061​

100​

0.88​

1.02​

2​

1.8​

1.1​

6​

2​

0.82​

0​

0.38​

talladega

4​

1​

0.95​

1.29​

0.2​

1.09​

1.06​

1.06​

1​

1​

1​

1.15​

1.07​

1.02​

1.018​

100​

0.33​

0.4​

2​

1.7​

1.5​

6​

10​

1​

0​

0.8​

texas

2​

2​

0.8​

1.04​

0.29​

0.93​

0.95​

1​

1​

0.98​

0.96​

1.07​

1.06​

1​

1.042​

100​

0.31​

0.41​

2.1​

1.7​

1.3​

6​

5​

1​

0​

0.82​

watkins glen

5​

4​

1.12​

1​

0.2​

1.11​

1.05​

1.09​

0.39​

0.85​

0.95​

0.982​

1.11​

1​

1.043​

100​

1.08​

1.07​

2​

1.55​

1.15​

5.9​

1.8​

0.8​

0​

0.65​