AI behaviour at the start of a race

[thunderbolt]

hi all. I recently did an edit of all the tracks I am using for my 2 seasons I'm currently running. Cup and Xfinity.
I edited so I can run AI @ 100% and still be competitive. I spent about a month on this project and the results so far show my edits were successful in so far as it is very competitive and our lap times / speeds are (for the most part) close to "real life" My car can run within the top 10 competitors in both series and neither I or the AI are dominant. So good close competition. And highly satisfying if / when I actually win. After 5 races in each series, I have 1 win in each.
The only issue I have; and this has been from my perspective an issue since the start of Nascar Racing, is the start of a race.
My car has "cold tires" and typically runs best after 5-10 laps. ok. But the AI do not seem to have the same issue. I have so far had some decent starting positions based on qualifying efforts but when the race begins, the AI most definitely have an advantage and I often lose about 10 spots. I then have to work my way back to where I started before I can hope to move towards the front. In the race I ran yesterday (84 laps at Texas) I qualified 4th; fell back to 13th and it took me almost 40 laps to get back to 4th; and then had to pit a few laps later for the 2nd half of the race.
Wondering if anyone has any suggestions to put us more on an even base here. Or is it something I just have to live with?
maybe @J.R. Franklin might have a suggestion for me?

Thanks in advance.

If anyone is interested, I could post my track edits / test results which I have saved in an excel format.


Points being tracked using the online points system found here

SCORING

NR2003 has many different mods, featuring different Nascar series, IZOD Indycar series, V8 Supercars. And scoring systems change over time. You might want to calculate your championship scores with...

nr2k3.weebly.com

 

[J.R. Franklin]

hi all. I recently did an edit of all the tracks I am using for my 2 seasons I'm currently running. Cup and Xfinity.
I edited so I can run AI @ 100% and still be competitive. I spent about a month on this project and the results so far show my edits were successful in so far as it is very competitive and our lap times / speeds are (for the most part) close to "real life" My car can run within the top 10 competitors in both series and neither I or the AI are dominant. So good close competition. And highly satisfying if / when I actually win. After 5 races in each series, I have 1 win in each.
The only issue I have; and this has been from my perspective an issue since the start of Nascar Racing, is the start of a race.
My car has "cold tires" and typically runs best after 5-10 laps. ok. But the AI do not seem to have the same issue. I have so far had some decent starting positions based on qualifying efforts but when the race begins, the AI most definitely have an advantage and I often lose about 10 spots. I then have to work my way back to where I started before I can hope to move towards the front. In the race I ran yesterday (84 laps at Texas) I qualified 4th; fell back to 13th and it took me almost 40 laps to get back to 4th; and then had to pit a few laps later for the 2nd half of the race.
Wondering if anyone has any suggestions to put us more on an even base here. Or is it something I just have to live with?
maybe @J.R. Franklin might have a suggestion for me?

Thanks in advance.

If anyone is interested, I could post my track edits / test results which I have saved in an excel format.


Points being tracked using the online points system found here

SCORING

NR2003 has many different mods, featuring different Nascar series, IZOD Indycar series, V8 Supercars. And scoring systems change over time. You might want to calculate your championship scores with...

nr2k3.weebly.com

You've identified one of the major flaws (imho) of NR2003... the HUGE advantage the A.I. cars have over The Player on the start or restart of a race. It seems that the only one affected by cold tires is The Human Player, not the A.I. (Nothing is more frustrating on NR2003 than working your way up to the top 5 only to have a caution come out with just a few laps left. You know in your heart-of-hearts that you stand virtually NO chance of winning the race due to the "cold tire effect" on the upcoming restart.)

Many have tried to remedy this, but frankly no one has ever found anything that solves the problem. What I do to "level the playing field" on this issue, is get a very stable and "tight" setup where I can drive the wheels off of it at the drop of a green flag without feeling like I'm on ice. Something where I can "hammer down" and stay with the A.I. cars and have a "chance" to maintain my position or even gain a spot.

Now.... this approach comes with a cost. First of all, many people prefer to drive a car that's loose. Of course... it's way faster, but only once your tires heat up. However, by that time, as you pointed out, you have lost a bunch of positions or worse yet, got in a wreck trying to hang onto a loose car in a bunched up restart field of cars and ended up getting into a crash. So to compensate, I have to match the A.I. to my "tight" setup and driving ability. "Loose is fast", but in NR2003... not on starts and restarts.

To achieve this requires a bunch of tweaking and testing by running races against the A.I. Tweaking track_tire_heat, track_tire_wear, track_asphalt_grip, ai_accel_modifier, ai_decel_modifier, ai_grip_modifier, ai_drag_modifier, ai_inverse_slipcurve_k, etc., etc. until I find that "sweet spot" of balance between myself and the A.I. Sounds like you are already doing much of this already. This is a long and time consuming exercise to get it just right.

One other potential "downside" is that speeds may not be totally realistic to real life. Personally, I focus my entire work on A.I. with the goal of making the racing against the A.I. enjoyable and somewhat realistic. I put very little focus on making sure that me and the A.I. turn laps at real-life speeds/times. As long as I have them close to the ballpark, I'm satisfied. In the end, it's all about having the sim be FUN! And quite honestly, it's really very far from a true simulation anyway. It's the best there is out there in my opinion, but it's not really close to being a true sim.

Many people don't like the fact that the speeds may not be perfectly accurate, but I counter with the fact that it is highly unrealistic to have the A.I. enjoy such a huge advantage on starts and restarts. so the goal is to find a compromise... a balance between realistic and fun.

You will find the tracks that I have done the A.I. on are like this providing that you use the "Fast" setups that I have included with those tracks and use the same ratings formulas that I use. You can find most of the tracks I've worked on at Smiffsden (SDT tracks), Project Short Track (PST tracks), some of the FSE tracks and many of the tracks we host at NR2k3Tracks.com have the qualities I outlined above. In most of the tracks I have done the A.I. on, there is a section in the README where I discuss the many factors that come into play that ultimately determine how enjoyable your racing experience against the A.I. will be.

 

[thunderbolt]

As always sir, your knowledge of this subject is very insightful. Thanks for your response..
Yes I have had wrecks in the closing laps following a caution because of my cold tire restarts. I've lost races I really should have likely won or at least had better finishes than what I did as a result. frustrating as hell. Overall, I am pleased with the results of my edits and as stated, it is truly satisfying when I can actually pull off a win.
Regarding real life speeds; I too put more emphasis on the competitive side but if I can also accomplish realistic speeds then I try to.
Same is true with track graphics. I use some of the older tracks because their AI behave more appropriately; at least from my perspective.
I also consider frame rates; which with some of the newer tracks are not great and can cause issues. The tracks I currently use are a combination from various track makers as a result.
Here's an example of a 20 lap practice run I just finished at the Day version of the BBMC Homestead track.
My car is the 127 and the 128 & 129 are fictional teammates rated equal to the #4

 

[DaleTona]

I love looking at peoples ideas on editing tracks to create a better racing experiance. I would do so myself, but I'm still on the process of figuring out ideal ratings, which is a complex puzzle in and of itself. If only I could nail that, then I could move forth and work on tailoring specific tracks.

Also seeing your spreadsheet would be nice to get some ideas on what to do myself. Always can use more knowledge

 

[thunderbolt]

@DaleTona - haven't forgotten your request. I am updating the excel sheet as I recently had some issues with a couple of tracks causing me to do more testing - stay tuned

I have more to say about track AI in general so here goes...

First, let me say I have the utmost respect, admiration and gratitude to ALL who create tracks and share with all of us. Your contributions are a large part of what keeps this sim alive. Along with the car painters, mod makers etc etc. to ALL of you I say THANK YOU

ok, back to track AI comments.
During my testing / editing project my goal was to achieve a competitive scenario where I can compete on a more or less even level with the best AI.
Years ago, I did a similar project and spent countless hours adjusting setups so I am comfortable with 98% of what I have in that regards.
I recently got the remaining 2% sorted out as well with some setups I found for my worst tracks.
So currently my speed / laptimes are inline with the top 10 performers running at 100% strength at all tracks - exception being the roadies which I stay away from - for now anyway. I use alternate tracks for those races.

Real life speeds are taken into consideration, but are secondary to the competitive aspect. Other factors in play are frame rates, carset ratings etc.
Carsets I use are rated using NRatings and I pull a min of 3 years worth of data. (2019 carset I am currently running based on rating from 2017,18,19)
I apply the MasGrafx formula for consistency as they are also available for the GNS CTS mods.

Now to the issue: I ran my Cup 2019 Martinsville race this week. Practice sessions showed I am in the top tier speedwise. I qualified on the pole by a whisker.
However, that is pretty much the end of the good news.
The race was a total cluster F**k to be blunt about it.
cold tires played against me and I struggled to stay in the top group right off the get go. Ok. I can live with falling back, and working my way back to the front after a few laps...right? In this case, not a chance in hell. The AI hit my car and basically just pushed me around so bad it was terrible. We had 5 cautions inside of 31 laps and after the 5th round my car was totalled to the point of a DNF 41st place finish.

Ok so how can I fix this. In looking back a few years, there was a day I used the "stock" tracks from Papyrus and although Martinsville (as example) has never been a strong race for me, I could at least run around without getting wrecked every other lap. So, I added back into my track folder the "original" track; made my edits and went out for a practice run. Speeds are where I would expect and again I'm competitive. Great. I was also able to make a 30 laps before pulling into the pits and no wrecks.
Next I'm going to run a "quick race" outside of my season and see how the AI behave; but so far it seems the AI have far improved "race etiquette" than the version I was using in my season. In other words; they are much more inclined to avoid wrecks while still being aggressive enough to give me a "bump" if I'm holding them up.
Now, my challenge is to see if I can "blend" this AI and pair it with the more current track graphics offered by the other version. I plan to take the track.dat file, add in my edited version of the track.ini, and then add the track.lp files from Papyrus. With a pinch of salt for flavour and luck, we'll have a modern looking copy of Martinsville that is actually something I can race on and have a shot of maybe a top 20? I guess we'll see. Stay tuned for actual results of this concoction.

What's the bottom line here? I think track makers need to be "proceed with caution" if they meddle too much with the ini / lp files. Some of these lines can be quite sensitive and cause havoc to an otherwise great piece of work. Get too aggressive with ai bunching etc can result in a race with more caution laps than actual racing. I am aware that most likely do not run AI @ 100% for every (if any) tracks. And truthfully, it probably doesn't matter if its 97, 98 or 100%. Each trackmaker has their own thoughts and their work may be designed to perform at (example) 97%. And that's fine. That said, it then obligates the users to adjust their AI settings for every race they run; assuming that like most of us, they are using tracks from a variety of creators. Hence my desire to edit to a consistent level so if I'm at it anyway, I might as well go for 100%.
Now I have not stated which version of Martinsville caused my issue; nor do I intend to. Party because I still appreciate the work and talent of those who made it. And because my comments / thoughts can be applied to virtually any track out there. Certainly the short tracks will see the most chaos if the AI are too aggressive.

Comments / thoughts welcome.