Hi,
I am trying to find
the best performing cars controlled by AI. Such cars can then be used for competitive racing offline and fill up empty slots in online races.
My aim is to find best cars in all ranks: rookie, amateur, advanced, semi-pro, pro and super-pro.
According to rough calculation there are approximately
3000 pro cars alone on RVZ but the problem could be balancing of the cars.
I started testing
pro cars. At first I wanted to single out the
best 3-5 cars from available packs diligently selected by community members which are believed to be well balanced.
I tested all pro cars from these packs
as of 31.01.2021:
► Show Spoiler
- abandoned_designs_project
I grouped them into
tiers. Each tier consisted of
10 pro cars. By coincidence there were exactly 8 tiers which equals 80 cars.
Tier1
► Show Spoiler
Acclaim F1
Aeroflash
AMW
Artair
Cougar
Fulon X
SNW 35
Purp XL
Humma
Toyeca
Tier2
► Show Spoiler
Ciagnik
Cintach
Dukhar
Eaglet
Electric Sheep
evasive
Superior
RVRC 20
Vitesse
SR-8000
Tier3
► Show Spoiler
Asahi
Oval Runner
Prime Target
Gear GT
G3X
EXE TC
Sir Gleam
Gust
Indy B
Mini Champ
Tier4
► Show Spoiler
Patriot
Pix Skull
Shinobi
S13 Alltune
Mark Jenkins
Tasco
Sally Taylor
Splat Buggy
Rally XL
TT Raider
Tier5
► Show Spoiler
After Image
BanKing
RC Bulldog
Black Widow
Chimera TC
Ayrton SP
Keyakizaka
The Knight
Quinx
Dev Panga
Tier6
► Show Spoiler
Bradyon
Cerebus
Stadium Star
Zoondiac
Metalzoo
GFX 00
Mugello
Sling Master
Hydro Flame
Maverick
Tier7
► Show Spoiler
Cherencov
Shark Bite
Jet Spike
Spectrum
Mid-Musc
RVXXL 5
Outlaw
Power Loader
Sunrise
Ryu
Tier8
► Show Spoiler
F-Zero SX
Formula 17
Stunt-Bowl Special
Team Machine
Steezy
SuperBoss
Twilight
Velter Ultron
Varanus K
Wing King
I chose
11 tracks to race against those cars. I was interested how they perform in
hardcore AI environment created by Whitedoom.
Number of laps was adjusted such that each race on a given track had to last long enough to avoid bad start or stupid mistake influence on the final result. While adjusting number of laps, approximately
240 sec race on each track was my reference. 11 tracks 4 minutes each gave about
50 minutes test for each car.
The tracks selected for the first championship were:
► Show Spoiler
[*]Toys in the Hood 1 - 6 laps
[*]Supermarket 2 - 14 laps
[*]Botanical Garden - 11 laps
[*]Toy World 1 - 11 laps
[*]Museum 2 - 7 laps
[*]Toy World 2 - 8 laps
[*]Wonderful Skylands 1 - 8 laps
[*]Sakura - 5 laps
[*]Donut Plains 3 - 8 laps
[*]Museum 1 - 5 laps
[*]Toys in the Hood 2 - 7 laps
I used
custom cup for the purpose of this championship.
This is the cup each car was put through:
► Show Spoiler
Code: Select all
;---------------------------------------
; Hardcore AI CUP
;---------------------------------------
Name "Hardcore AI CUP"
Difficulty 4
Obtain 0
NumCars 11
NumTries 100
Classes 0,0,0,0,10,0
Points 10,9,8,7,6,5,4,3,2,1,0
STAGE 0 nhood1 6 false false
STAGE 1 market2 14 false false
STAGE 2 garden1 11 false false
STAGE 3 toylite 11 false false
STAGE 4 muse2 7 false false
STAGE 5 toy2 8 false false
STAGE 6 fdt14skylands1 8 false false
STAGE 7 Sakura 5 false false
STAGE 8 Donut3 8 false false
STAGE 9 muse1 5 false false
STAGE 10 nhood2 7 false false
I usually waited long enough for the first
5 cars to arrive at least
before skipping the rest of the race.
I got the following results:
tier1
tier2
tier3
tier4
tier5
tier6
► Show Spoiler
![Image](https://i.postimg.cc/k430F3q2/pt6-1-s.png)
Because
there was draw between Sling Master and Hydro Flame playoff round was required:
Code: Select all
;---------------------------------------
; Hardcore AI CUP
;---------------------------------------
Name "Hardcore AI CUP play-off"
Difficulty 4
Obtain 0
NumCars 4
NumTries 100
Classes 1,0,0,0,2,0
Points 10,9,8,0
STAGE 0 medieval-redux 8 false false
STAGE 1 pr_harbour 6 false false
STAGE 2 ship1 6 false false
![Image](https://i.postimg.cc/GmjKgWFB/pt6-playoff-1.png)
Session log including playoff
pt6 - Pro Tier6 session log
tier7
tier8
Winners from each tier could then compete in the championship final which comprised of 8 best cars from each tier.
Final tier:
► Show Spoiler
SNW 35
Superior
Prime Target
TT Raider
Keyakizaka
Hydro Flame
Outlaw
Velter Ultron
The most exiting race was with these cars. I was really eager which car takes the 1st place and had my own expectations:
► Show Spoiler
1. Outlaw
2. Keyakizaka
3. Prime Target
4. Superior
8. Velter Ultron
In this way
I got the best performing car in this specific championship. Rusults could have been different if cars had been tested on a different set of tracks.
Final results:
Best performing car showcase:
Most interesting part of the competition was over. Now I had to single out the
2nd best performing car. Logically it could only be a car from 1st and 2nd place in each tier. 2 tiers minus the best car in this championship equals 15 cars left for 2nd place competition.
I put all 15 cars through the same cup:
► Show Spoiler
SNW 35
Superior
Prime Target
TT Raider
Hydro Flame
Outlaw
Velter Ultron
Humma
Dukhar
Mini Champ
Sally Taylor
Chimera TC
Sling Master
Ryu
Twilight
with the following results:
2nd best performing car showcase:
Fight for the
3rd place had to consist of cars left over with 1st, 2nd and 3rd place in each tier, altogether 24 cars.
I put 12 cars in tier1 and 12 for final round with the winner from tier1, all through exactly the same cup.
Results of final round for 3rd place tier1:
Results of the final for the 3rd place:
► Show Spoiler
![Image](https://i.postimg.cc/SxjJ25S0/screenshot-2021-02-25-11-30-51s.png)
Session log not available - forgot to save. Replays still available for the interested.
3rd place showcase:
► Show Spoiler
I decided that I will add to the pack other 2 best performing cars: Superior and Minichamp.
I would like to put other cars to the test
on various tracks and winners would be added each time to the
best 15 team but under a certain condition: from the next championship
only these which beat any of the 5 already selected cars would enter the best 15.
I enjoyed a lot this championship. The winners really
don't allow you to rest on a track and there is no much room for errors to win the race, especially on
Museum 2, Toy World 1, Supermarket 2, Wonderful Skylands 1, or Donut Plains 3.
Please
find replaysof the final for the 1st place:
Final for the 1st place replays
xxxxxxxxxxxxxxxxxxxxxxx
Continuation of testing.
When I look at possible cars I would like to put through the test
I see problem with top speed which is way too high on some of them or acceleration and so on. Then I don't know what I should do to test them fairly?
Please give me suggestions and if you like the idea
join this effort to find 15 best performing cars in each rank.
Below is the list of cars in tier1 from 2nd championship I was planning to start. They are not balanced properly and I don't know how and if I should continue testing all pro cars at all.
► Show Spoiler
Army: 45 mph,
McLaren F1 Elite GT: 50,
Ferrari 456: 39,
Harvester Toxin: 550 (yes 550!),
Marvin's TVR: 39,
Red Dragon: 40,
Toyxtpack: 50,
Ultimate Icey: 45,
VW Transporter: 46,
XJR Sovereign: 43