Logo
Current time: 05-27-2019, 01:57 AM Hello There, Guest! (LoginRegister)


Post Reply 
 
Thread Rating:
  • 0 Votes - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
AI testing results
12-10-2018, 08:18 PM (This post was last modified: 12-10-2018 08:32 PM by JR82.)
Post: #1
AI testing results
Have just completed a thorough test to see what each AI car performance parameter does, i used 9 cars 1st with the 899 900 parameters for power with gaps of 100 all the way down to the value 099 100, also did the grip/drag/aggression the same way, each time all 3 other parameters were 599 600 for the sake of them being the same.
the power & grip gaps are rather as expected (higher = quicker) but with the drag setting the default quali times (the grid when u 1st start quali) had the lower drag fastest but the higher value seemed quicker in the race....
The aggression setting, the higher value is just a more aggressive driver in terms of overtaking quicker than the less aggressive driver. I also noticed the more aggressive driver blocked/chopped the other car at Michigan.

I used 8 tracks for a season, 2 street (Surfers & Toronto) 2 Super speedways (Michigan & Fontana) 2 short ovals (Chicago & Milwaukee) & 2 road courses (Elkhart & Laguna), races were 50% in length. I didn't see any parameter that effected DNF's at all so it seems to be simply a random thing.

What i get from these is that it is possible to have oval & street/Road specialists tho i have not yet figured out the balance between them so they are competitive on both but one beats the other on thier favoured track (think Zanardi '98 or Brack '01).

Seems the ideal gap for each parameter is about 150-200 (low high value) (as is the case with most carsets i have seen anyway).

the 1st time gap is the gap between each car, the top/bot is gap from 1st to 9th (last). Aggression didn't seem to alter the grid in any way as all cars had the same time.

short oval
BHP - 1 MPH top/bot - 8-10 MPH race .1 per lap
Drip - 1.2 MPH top/bot - 9 MPH race .1/.4 per lap
Drag - low - faster 1.2 MPH top/bot - 9 MPH - race - high faster

street
BHP - .4/.6 sec top/bot - 4-5 sec
Grip - .4/.6 top/bot - 4/4.4 sec per lap race 1.3/1.5 per lap
Drag - low - faster - .4.6 sec top/bot - 4 sec

superspeedway
BHP - 1.5 MPH top/bot - 11.5MPH race .2 per lap
Grip - 1.5 MPH top/bot 11.5 MPH Race - .1 per lap
Drag - low - faster - 1.5 MPH top/bot - 11.5 MPH - race high faster

Road
BHP - .4/.6 sec top/bot - 3.3-5 sec
Grip - .4/.6 sec top/bot - 3.5/5 sec per lap Race 1.5/2 sec per lap
Drag - low - faster .6 sec top/bot - 3-5 sec
Find all posts by this user
Quote this message in a reply
12-11-2018, 05:30 PM
Post: #2
RE: AI testing results
Looking at building some Drivers2.txt files from scratch does anyone that has created carsets before etc have any tips?
Find all posts by this user
Quote this message in a reply
12-11-2018, 07:14 PM
Post: #3
RE: AI testing results
I just modify the original one.
Find all posts by this user
Quote this message in a reply
12-11-2018, 11:22 PM (This post was last modified: 12-11-2018 11:34 PM by JR82.)
Post: #4
RE: AI testing results
(12-11-2018 07:14 PM)samsepi0l Wrote:  I just modify the original one.

Do you have any parameters/guides that you use to do them? or just use the values from original carset as the base?
Find all posts by this user
Quote this message in a reply
12-11-2018, 11:50 PM
Post: #5
RE: AI testing results
When I did my '97 and '98 carsets, I began with the results of the yearly championships - engine, manufacturer, tire and driver, and used them to determine the DINFO line values. (Values here are from '97)

Power:
Honda 775 850
Coswth 750 825
Merced 725 800

Drag:
Reynard 500 600
Penske 600 700
Swift 550 750
Lola 700 800
Swift has a wider range because while it occasionally won (because Mikey), it was generally more mid-pack.

Traction and aggression values are a bit more complicated and I don't quite recall my decade-old thought processes, but I think I used some combination of tire and chassis for Traction and I think Aggression was just arbitrary numbers that I thought seemed about right for each driver. Basically, I tried to have overlapping ranges for everything while still weighting toward the top and bottom ends of the rankings. Here's the drivers2.txt that from the carset. Hopefully it will make sense what I mean.

VERSN 1
NUMDR 33
NUMCH 5
NUMEN 3
CHASS Lola_T9700 Penske_PC-26 Reynard_97I Swift_007.I
ENGIN Cosworth Mercedes-Benz Honda
NUMT 2
TIRES Goodyear Firestone
DINFO 00 2 0 0 375 562 375 562 387 562 375 562 Pace_Car Pacer Chevrolet_Camaro USA USA INDYCAR
DINFO 00 0 1 1 725 800 750 800 500 600 500 562 Champ_Driver Slick Reynard_97I/Mercedes USA USA ICR2/CART_Forever
DINFO 01 2 2 1 775 850 735 800 500 600 700 850 Jimmy_Vasser Jimmy Reynard_97I/Honda_971_HRR USA USA Target_Chip_Ganassi_Racing
DINFO 04 2 2 1 775 850 745 800 500 600 850 900 Alex_Zanardi Alex Reynard_97I/Honda_971_HRR Italy ITA Target_Chip_Ganassi_Racing
DINFO 05 2 2 0 775 850 740 800 500 600 750 900 Gil_de_Ferran Gil Reynard_97I/Honda_971_HRR Brazil BRA Walker_Racing
DINFO 17 2 1 1 725 800 730 800 500 600 700 820 Mauricio_Gugelmin Mauricio Reynard_97I/Mercedes_IC108C Brazil BRA PacWest_Racing_Group
DINFO 18 2 1 1 725 800 720 775 500 600 600 800 Mark_Blundell Mark Reynard_97I/Mercedes_IC108C Great_Brit GBR PacWest_Racing_Group
DINFO 02 1 1 0 725 800 685 750 600 700 800 900 Al_Unser_Jr. Little_Al Penske_PC-26/Mercedes_IC108C USA USA Marlboro_Team_Penske
DINFO 03 1 1 0 725 800 725 775 600 700 899 900 Paul_Tracy Paul Penske_PC-26/Mercedes_IC108C USA USA Marlboro_Team_Penske
DINFO 27 2 2 1 775 850 670 725 500 600 750 900 Parker_Johnstone Parker Reynard_97I/Honda_971_HRR USA USA Team_Kool_Green
DINFO 99 2 1 1 725 800 715 775 500 600 775 900 Greg_Moore Greg Reynard_97I/Mercedes_IC108C Canada CAN Forsythe_Racing
DINFO 06 3 0 0 750 825 710 775 550 750 800 900 Michael_Andretti Mikey Swift_007.I/Ford_Cosworth_XD USA USA Newman/Haas_Racing
DINFO 11 2 0 0 750 825 675 725 550 750 800 900 Christian_Fittipaldi Christian Swift_007.I/Ford_Cosworth_XD Brazil BRA Newman/Haas_Racing
DINFO 20 2 0 1 750 825 705 775 500 600 700 850 Scott_Pruett Scooter Reynard_97I/Ford_Cosworth_XD USA USA Brahma_Sports_Team
DINFO 40 2 0 1 750 825 700 750 500 600 550 700 Raul_Boesel Raul Reynard_97I/Ford_Cosworth_XD Brazil BRA Brahma_Sports_Team
DINFO 07 2 0 0 750 825 690 750 500 600 700 800 Bobby_Rahal Bobby Reynard_97I/Ford_Cosworth_XD USA USA Team_Rahal
DINFO 08 2 0 0 750 825 695 750 500 600 650 800 Bryan_Herta Bryan Reynard_97I/Ford_Cosworth_XD USA USA Team_Rahal
DINFO 31 3 2 1 775 850 680 750 700 800 650 800 Andre_Ribeiro Andre Lola_T9700/Honda_971_HRR Brazil BRA Tasman_Motorsport_Group
DINFO 32 0 2 1 775 850 660 725 700 800 550 700 Adrian_Fernandez Adrian Lola_T9700/Honda_971_HRR Mexico MEX Tasman_Motorsports_Group
DINFO 36 2 2 0 700 775 640 700 500 600 450 600 Juan_Manuel_Fangio_II Fangio Reynard_97I/Toyota_V8_Turbo Argentina ARG All_American_Racing
DINFO 98 2 2 0 700 775 620 675 500 600 450 600 P._J._Jones P.J. Reynard_97I/Toyota_V8_Turbo USA USA All_American_Racing
DINFO 24 2 2 1 700 775 625 675 500 600 450 600 Hiro_Matsushita Hiro Reynard_97I/Toyota_V8_Turbo Japan JAP Arciero-Wells/MCI_Racing
DINFO 25 2 2 1 700 775 635 700 500 600 600 800 Max_Papis Mad_Max Reynard_97I/Toyota_V8_Turbo Italy ITA Arciero-Wells/MCI_Racing
DINFO 19 0 0 1 750 825 615 675 700 800 450 600 Michel_Jourdain_Jr. Michel Lola_T9700/Ford_Cosworth_XD Mexico MEX Payton_Coyne_Racing
DINFO 35 0 0 1 725 800 600 675 700 800 350 500 Charles_Nearburg® Charlie Lola_T9700/Ford_Cosworth_XD USA USA Payton/Coyne_Racing
DINFO 34 0 0 0 750 825 600 675 700 800 600 800 Paul_Jasper® Paul Lola_T9700/Ford_Cosworth_XD USA USA Payton/Coyne_Racing
DINFO 09 2 1 1 725 800 645 700 500 600 600 800 Dario_Franchitti® Dario Reynard_97I/Mercedes_IC108C Scotland GBR Hogan_Racing_LLC
DINFO 16 2 1 0 725 800 665 725 500 600 500 700 Patrick_Carpentier® Patrick Reynard_97I/Mercedes Canada CAN Bettenhausen_Motorsports
DINFO 21 0 0 0 750 825 650 700 700 800 350 500 Richie_Hearn Richie Lola_T9700/Ford_Cosworth_XD USA USA Della_Penna_Motorsports
DINFO 77 2 0 0 750 825 655 725 500 600 600 800 Gualter_Salles® Gualter Reynard_97I/Ford_Cosworth_XD Brazil BRA Davis_Racing
DINFO 64 0 0 0 750 825 610 675 700 800 350 500 Arnd_Meier® Arnd Lola_T9700/Ford_Cosworth_XD USA USA Project_Indy
DINFO 100 2 0 1 725 850 600 800 500 800 850 900 Fred_Jones Fred Reynard_97I/Ford_Cosworth_XD USA USA ICR2/CART_Forever
DINFO 95 2 0 1 725 850 600 800 500 800 350 900 Bob_Murphey CowtownBob Reynard_97I/Ford_Cosworth_XD USA USA CowtownRacers
Find all posts by this user
Quote this message in a reply
12-12-2018, 06:24 PM
Post: #6
RE: AI testing results
Sweet, with those gaps do you find that cars that won (even the slower ones here) can still get up for a podium etc while the contenders are more consistently up the front?
P.S You have Fittipaldi in a Reynard not a Swift Wink2
Find all posts by this user
Quote this message in a reply
12-12-2018, 06:37 PM
Post: #7
RE: AI testing results
my 89 set




VERSN 1
NUMDR 40
NUMCH 3
NUMEN 3
CHASS _ _ _
ENGIN _ _ _
NUMT 2
TIRES Goodyear Firestone
DINFO 0 2 0 0 375 562 375 562 387 562 375 562 Pace_Car Pacer Denver,_CO USA USA INDYCAR
DINFO 0 2 2 0 500 562 500 562 100 150 998 999 Ignazio_Germi 11/14/1966 Lola_T8900/Judd-Honda Italy ITA Firenze_Motorsports
DINFO 20 2 2 0 365 550 365 550 395 570 300 800 Emerson_Fittipaldi 12/12/1946 Penske_PC-18/Chevrolet-Ilmor Brazil BRA Patrick_Racing
DINFO 4 2 2 0 364 549 364 549 395 570 300 800 Rick_Mears 12/3/1951 Penske_PC-18/Chevrolet-Ilmor USA USA Penske_Racing
DINFO 1 2 2 0 361 546 361 546 398 573 300 800 Danny_Sullivan 3/9/1950 Penske_PC-18/Chevrolet-Ilmor USA USA Penske_Racing
DINFO 25 2 2 0 327 512 327 512 425 600 300 800 Al_Unser-Sr 5/29/1939 Penske_PC-18/Chevrolet-Ilmor USA USA Penske_Racing
DINFO 6 2 2 0 364 549 364 549 395 570 300 800 Michael_Andretti 10/5/1962 Lola_T8900/Chevrolet-Ilmor USA USA Newman/Haas_Racing
DINFO 5 2 2 0 349 534 349 534 407 582 300 800 Mario_Andretti 2/28/1940 Lola_T8900/Chevrolet-Ilmor USA USA Newman/Haas_Racing
DINFO 8 2 2 0 364 549 364 549 396 571 300 800 Teo_Fabi 3/9/1955 March_89P/Porsche Italy ITA Porsche_Racing
DINFO 2 2 2 0 361 546 361 546 398 573 300 800 Al_Unser-Jr 4/19/1962 Lola_T8900/Chevrolet-Ilmor USA USA Galles_Racing
DINFO 3 2 2 0 343 528 343 528 411 586 300 800 Scott_Pruett 3/24/1960 Lola_T8900/Judd-Honda USA USA Truesports
DINFO 18 2 2 0 343 528 343 528 411 586 300 800 Bobby_Rahal 1/10/1953 Lola_T8900/Cosworth-DFS USA USA Kraco_Racing
DINFO 7 2 2 0 327 512 327 512 425 600 300 800 Arie_Luyendyk 9/21/1953 Lola_T8900/Cosworth-DFS Netherlands NET Dick_Simon_Racing
DINFO 22 2 2 0 310 495 310 495 439 614 300 800 Scott_Brayton 2/20/1959 Lola_T8900/Buick_V6 USA USA Dick_Simon_Racing
DINFO 28 2 2 0 239 424 239 424 513 688 300 800 Randy_Lewis 7/18/1945 Lola_T8900/Cosworth-DFX USA USA Dick_Simon_Racing
DINFO 30 2 2 0 321 506 321 506 429 604 300 800 Raul_Boesel 12/4/1957 Lola_T8900/Judd-Honda Brazil BRA Shierson_Racing
DINFO 10 2 2 0 321 506 321 506 429 604 300 800 Derek_Daly 3/11/1953 Lola_T8900/Judd-Honda Ireland IRE Raynor_Racing
DINFO 11 2 2 0 321 506 321 506 429 604 300 800 Kevin_Cogan 3/31/1956 March_88C/Cosworth-DFX USA USA Machinists_Union_Racing
DINFO 29 2 2 0 310 495 310 495 439 614 300 800 Pancho_Carter 6/11/1950 Lola_T8900/Cosworth-DFX USA USA Machinists_Union_Racing
DINFO 65 2 2 0 299 484 299 484 449 624 300 800 John_Jones 10/19/1965 Lola_T8900/Cosworth-DFX Canada CAN Protofab_Racing
DINFO 69 2 2 0 288 473 288 473 459 634 300 800 Bernard_Jourdain 8/16/1950 Lola_T8900/Cosworth-DFX Mexico MEX Andale_Racing
DINFO 14 2 2 0 283 468 283 468 465 640 300 800 AJ_Foyt 1/16/1935 Lola_T8900/Cosworth-DFX USA USA AJ_Foyt_Enterprises
DINFO 9 2 2 0 283 468 283 468 465 640 300 800 Didier_Theys 10/19/1956 Lola_T8800/Buick_V6 Belgium BEL Granatelli_Racing
DINFO 70 2 2 0 283 468 283 468 465 640 300 800 John_Andretti 3/12/1963 Lola_T8800/Buick_V6 USA USA Granatelli_Racing
DINFO 19 2 2 0 255 440 255 440 494 669 300 800 Guido_Dacco 9/10/1942 Lola_T8900/Cosworth-DFX Italy ITA Dale_Coyne_Racing
DINFO 21 2 2 0 294 479 294 479 454 629 300 800 Roberto_Guerrero 11/16/1958 March_89CE/Alfa-Romeo Columbia COL Alex_Morales_Motorsports
DINFO 86 2 2 0 288 473 288 473 459 634 300 800 Dominic_Dobson 9/14/1957 Lola_T8900/Cosworth-DFX USA USA Bayside_Motorsports
DINFO 12 2 2 0 277 462 277 462 470 645 300 800 Fabrizio_Barbazza 4/2/1963 Penske_PC-17/Cosworth-DFX Italy ITA Arciero_Racing
DINFO 71 2 2 0 277 462 277 462 470 645 300 800 Ludwig_Heimrath_Jr 9/9/1956 Lola_T8800/Judd-Honda Canada CAN Hemelgarn_Racing
DINFO 59 2 2 0 250 435 250 435 500 675 300 800 Steve_Saleen 4/2/1949 March_88C/Cosworth-DFX USA USA Saleen_Motorsports
DINFO 56 2 2 0 250 435 250 435 500 675 300 800 Jeff_Wood 1/20/1957 Lola_T8800/Cosworth-DFX USA USA Gohr_Racing
DINFO 16 2 2 0 233 418 233 418 520 695 300 800 Jon_Beekhuis 3/31/1960 Lola_T8800/Cosworth-DFX Canada CAN Bettenhausen_Motorsports
DINFO 50 2 2 0 233 418 233 418 520 695 300 800 Scott_Atchison 7/16/1962 Lola_T8800/Cosworth-DFX USA USA Euromotorsport
Find all posts by this user
Quote this message in a reply
12-12-2018, 06:54 PM (This post was last modified: 12-12-2018 07:02 PM by JR82.)
Post: #8
RE: AI testing results
What stands out there is bigger gaps in the values, which matches that era as there were bigger gaps in the field i had noticed when looking at all the results from memory.

I am looking at doing the seasons from '90-01 most of the carsets i have are pretty good tho in the '91 season i did Mi.Andretti lost the title in the end because Matsushita pipped him on the last lap to win at Nazareth which clearly isn't a realistic result. 1 or 2 others the order is right but only gaps of 25 with the parameters so the results are a lot less varied than that particular season was in real life. (01)
Find all posts by this user
Quote this message in a reply
12-12-2018, 09:17 PM
Post: #9
RE: AI testing results
I use an excel sheet I made (libre office calc actually on linux) to give me the numbers.

I score the drivers from 100 to 70, kinda like grades back in school. The best drivers are close to 100 and the slowest are closer to 70. The spreadsheet makes the numbers for me. (what you see). I have been racing a full season on my 89 carset and with the exception of Scott Pruett doing awesome (for some reason I don't know) the results seem pretty realistic.
Find all posts by this user
Quote this message in a reply
12-13-2018, 12:43 AM
Post: #10
RE: AI testing results
(12-12-2018 06:24 PM)JR82 Wrote:  Sweet, with those gaps do you find that cars that won (even the slower ones here) can still get up for a podium etc while the contenders are more consistently up the front?
P.S You have Fittipaldi in a Reynard not a Swift Wink2
Yes, it seemed to act fairly realistically, both in race results and in how the pack strung itself out over the course of a few laps after start or restart. It may take an unlikely dose of RNG for a Payton-Coyne car to get on the podium, but it is possible.
P.S. Hmm, it shows him in a Swift on my monitor. Try degaussing yours and see how it looks. Wink2
Find all posts by this user
Quote this message in a reply
Post Reply 


Forum Jump:


User(s) browsing this thread: 1 Guest(s)

Privacy policy