- 807
I've seen some mention of the AI cheating badly at Grand Valley before, but I thought some people might find this interesting.
I was just doing the Amateur MR Challenge in my Elise 190, and I noticed the AI cars were cheating more obviously than any other race I've seen. On the 1st turn of the 1st lap I took the lead by 1 sec, then the AI cars stayed about 1 sec. behind me for the rest of the race. All the cars were in a tight pack, but whenever one would pit in, it would catch up almost immediately.
For example, the Clio gets up to 122 MPH going down the right side of the map before the hairpin. He barely hits 117 MPH before he brakes for the 2nd tunnel after the bridge. He then gets up to 146 MPH for the main straight. He was in 3rd, 1.7 sec down at the end of the 1st lap. Lap 2 his speed was almost identical and he was staying with the pack, then he pitted in at the end of lap 2. Now get this. He leaves the pit over 20 sec. down, and his speedometer registers that in every section he is going almost the exact same speed as the first 2 laps, but one look at the car tells you the speedometer is wrong and he is rocketting down the right side of the map doing at least 200 MPH. He makes up most of the difference and is back in the pack in less than half a lap! He pitts again on lap 4 and catches most of it up again.
I wanted to be a bit more objective, so I wrote down his sector times and figured out his average speed.
As you can see, the 2nd lap he only speeds up 0.7 MPH because he is only down 2 sec. The 3rd and 5th laps are ridiculous though. After he pitts, he goes so fast that it brings his average speed up almost 20 MPH and he does the back half of the track 10 sec. faster! As all this is happening, his speedometer still shows the same speed as his first lap.
All the AI cars do this, but the Clio was the best example because he pitts twice. Apparently this is why you can only run the analyzer on cars you've actually driven, because analyzing an AI car would show a lot of numbers that don't add up.
To get a better look at how badly the AI cheats, let's compare my run to that in Der Alta's compendium:
I was running a 265 HP Elise with a stock transmission, but it probably doesn't matter what you use, the AI will speed up. As you can see, using the 787B and finishing the race 1 min 29 sec faster only granted Der Alta another 2 sec. lead on the AI Elise.
Here's my replay:
I was just doing the Amateur MR Challenge in my Elise 190, and I noticed the AI cars were cheating more obviously than any other race I've seen. On the 1st turn of the 1st lap I took the lead by 1 sec, then the AI cars stayed about 1 sec. behind me for the rest of the race. All the cars were in a tight pack, but whenever one would pit in, it would catch up almost immediately.
For example, the Clio gets up to 122 MPH going down the right side of the map before the hairpin. He barely hits 117 MPH before he brakes for the 2nd tunnel after the bridge. He then gets up to 146 MPH for the main straight. He was in 3rd, 1.7 sec down at the end of the 1st lap. Lap 2 his speed was almost identical and he was staying with the pack, then he pitted in at the end of lap 2. Now get this. He leaves the pit over 20 sec. down, and his speedometer registers that in every section he is going almost the exact same speed as the first 2 laps, but one look at the car tells you the speedometer is wrong and he is rocketting down the right side of the map doing at least 200 MPH. He makes up most of the difference and is back in the pack in less than half a lap! He pitts again on lap 4 and catches most of it up again.
I wanted to be a bit more objective, so I wrote down his sector times and figured out his average speed.
Code:
Clio Sport V6 24V - pitted on 2 and 4
Lap 1 Lap 2 Lap 3 Lap 4 Lap 5
-------- -------- -------- -------- --------
T1: 0'30.228 0'21.704 0'34.886 0'20.573 0'31.590
T2: 1'10.624 1'01.285 1'07.609 0'58.574 1'04.021
T3: 1'31.316 1'22.290 1'24.728 1'19.248 1'22.029
T4: 2'11.663 2'07.365 2'01.668 2'04.645
Time to cover back half (sector 2/3, 1.45mi):
Lap 1 Lap 2 Lap 3 Lap 4 Lap 5
-------- -------- -------- -------- --------
1'01.088 1'00.586 0'49.842 0'58.675 0'50.439
85.5mph 86.2mph 104.7mph 89.0mph 103.5mphAs you can see, the 2nd lap he only speeds up 0.7 MPH because he is only down 2 sec. The 3rd and 5th laps are ridiculous though. After he pitts, he goes so fast that it brings his average speed up almost 20 MPH and he does the back half of the track 10 sec. faster! As all this is happening, his speedometer still shows the same speed as his first lap.
All the AI cars do this, but the Clio was the best example because he pitts twice. Apparently this is why you can only run the analyzer on cars you've actually driven, because analyzing an AI car would show a lot of numbers that don't add up.
To get a better look at how badly the AI cheats, let's compare my run to that in Der Alta's compendium:
Code:
Der Alta driving the 787B
Amateur MR Challenge at Grand Valley
car Time down avg.speed
----------------- --------- ------- ---------
Mazda 787B 8'48.026 106.4 MPH
NSX Type S Zero 8'58.020 +9.994 104.4 MPH
Motor Sport Elise 9'00.346 +12.320 104.0 MPH
Elise 190 9'01.944 +13.918 103.7 MPH
Me driving the Elise 190
Amateur MR Challenge at Grand Valley
car Time down avg.speed
----------------- --------- ------- ---------
Elise 190 (me) 10'16.919 91.1 MPH
NSX Type S Zero 10'19.503 +2.584 90.7 MPH
Motor Sport Elise 10'24.909 +7.990 89.9 MPH
Elise 190 (AI) 10'28.771 +11.852 89.3 MPHI was running a 265 HP Elise with a stock transmission, but it probably doesn't matter what you use, the AI will speed up. As you can see, using the 787B and finishing the race 1 min 29 sec faster only granted Der Alta another 2 sec. lead on the AI Elise.
Here's my replay:
That 17mph speed up on laps 3 and 5 for that Clio is also suspicious.
