General AI Performance Discrepancy Reports (No Discussion)

I think the AI do not use the low downforce setup. It is the same on Monza.

 

Not just the vintages, also retros (not discussing, I just point this related post)

Track: Kyalami 76
F Retro Gen 1 (me in the lotus)
AI Difficulty: 110
Aggression: Medium
Laps: 39
Fuel 1x (About 80 liter at the start of the race), tires 2x

front row AI did the good laps at low 1.15s consistently all the race. I started doing low 1.16s/high 1.15s and at the end was 1.14.0 without even pushing. 2-3 sec difference on pace. Probably the AI improved a bit considering tires but just a bit. I am pretty consistent on this track with thoese cars, definitely did not gain the time due to driving.

 

LMDh tyre performance / fuel load effect needs to be looked at.

Did a 30 lap race at Spa 22 (starting with max fuel + pitstop for additional 6/7 laps of fuel to make it to the end). 110 AI difficulty - was constantly fighting the AI throughout the first few laps, AI was impossible to keep up with through fast corners like Pouhon and "No name" but at other corners they are around the same as me and in the last chicane and first corner they are just stupid slow so I could still take the lead, and at the at with low fuel load my fastest lap was 3 seconds faster then any AI.

tl;dr

• LMDh AI with high fuel load -> too fast compared to player in fast corners, I could barely keep up
• by the end of the race with low fuel I was lapping 3 seconds faster

AFAIK AI always starts on max fuel load if the full tank is not enough to complete the race so it's a fair comparison but correct me if I'm wrong

 

Did the AI pit as well towards the end? Generally if you are way faster towards the end of the race it could be either the AI tire wear because they didn't pit, or the fuel weight not affecting them too much on the beginning, or if they did pit, they overfueled. What was the AI pace in qualy? Was it competitive? If it was and the AI did pit, it could be that they filled their tanks instead of taking fuel for only 7 laps, or they didn't change tires, though in my testing they always do as they are instantly faster after they pit. But I haven't really paid attention to how much fuel (if any) they put when they pit,because most of my testing has been with old F1 cars that didn't really pit for fuel.

In the classes that I use and can monitor the AI fuel load in game, they will fuel for the required amount to complete the race, if they need half a tank, they will do half, if it needs a full tank, they will fill it up. So they are smart enough to do that, but dont really know how they refuel when they need it.

 

I pitted a lap after the AI behind me and required pitstop was on with min. 4 tires. Anyway, I just tried the same scenario again:

LMDH, 115 AI, Spa 2022, 30 lap race so full tank is not enough

Qualy

• got pole by 0.8 but the fastest AI was faster through the 2nd sectory by 0.3 still

Race

• I decided to start with only 10l of fuel instead of 110l to see if it helps
• AI is massively faster through Pouhon and "No name"
• also interesting how the AI is stupid fast through Eau Rouge/Raidillon on the opening lap but then take it slower on every other lap

I also tried Stock car 2023 around Spa 22 and Spa 20 - since it's a slower car it is more apparent that the AI gets a speed boost out of "no name" and through Pouhon. On Spa 20 the AI is slower and more on par with the player through Pouhon.

 

I tested the same scenario with F-Reiza, DPI, GT3Gen2, GT3Gen1 and GT1. From these classes I have only experienced the same performance difference vs the player in the aforementioned corners with GT3Gen2, the other classes have provided a much more consistent experience across the whole lap so I firmly believe this to be caused by OP AI tyres for some cars/classes

 

On the setting light rain, the AI is in another world compared to the player.

Tested with Hypercars, F-ClassicG4 and F3. With the F-ClassicG4, they were around 15-20 seconds quicker per lap (!) on dry tires, with the Hypercar it was around 15 seconds, especially as the track got wetter. With the F3, the discrepancy was around 7-10 seconds.

Something needs to be done there - this weather setting isn't really suited to dry tires but the AI exclusively use them. It's also one of the most fun weather settings it's sort of in between completely wet and dry, but sadly it's not usable offline.

 

Can you let us know what track, date, and AI strength ?
I had similar experience with F Classic G4 but not like 15-20 sec per lap.

 

Jacarepagua, 112 AI strength, date was something in August.

It's such a huge discrepancy that I didn't think it required me to name the settings, as it's in the completely wrong ballpark and needs exceptionally big adjustments (or the AI switching tires sooner).

It should be noted that the game puts the player on wet tires in these conditions, but the AI is collectively on dries. Player performance will be better relative to AI performance if player is on wets.

 

I had a similar experience but less than 15-20 seconds.
I tried the Formula Classic Gen 4 (Benetton), around Monaco. Track was supposed to be wet, with light rain. I set the date in July. I was surprised to see the AI with slick tyres, as the recommanded tyres for player was wets... i was so slow on wets, lot of understeer (car was default setup but still...).
Then i tried to put the same tyres as AI : slicks. The car was much better, at least i could defend my 4th place from Prost, but i was still lapping 3 to 4 seconds per lap slower than Patrese, Mansell and Senna.

One thing i noticed during this test : no rainspray. So i guess the high temperature had a big effect on this.

But you see, we had the same troubles with similar settings

 

Hah, you're right. I just did the same experiment in August in Laguna Seca and the track barely got wet by the light rain, so I was able to keep up with the (110) AI pretty decently, being around half a second slower for the first couple of laps (if even that) and even able to make some moves against midfield running cars.

So the setting is a bit deceptive because the amounts of water on the track can wildly vary depending on temperature, which in itself is pretty cool. However even with a lot of water the AI keep to their dry tires and they always seem to be exactly as quick as they are in Laguna Seca in much dryer conditions.

So maybe the issue is that the calibration only accounts for very little water on the track, but not for the wetter conditions. This could also be a result of testing everything at Barcelona, where there's probably not that much water.

On a side note, track temp of 33°C when there's 19°C air temp and rain seems a bit high...

 

It's been a long time I haven't tested that scenario, but if you start a race whose first weather slow is 'Light rain' the AI starts on slicks (while you start on wets if 'auto by weather') and pit at the end of the first lap to put wet tires.
For the AI to start on wets you have to set the initial Live Track state to 'Damp' or 'Wet'.

 

Is there a way to see what strategy the AI used for pitstops?

 

"Second monitor" will show pit stop time and some basic telemetry for the AI. However the AI have no strategic planning whatsoever, iirc Renato said it was an improvement planned for the future...

 

Thanks,
Can you also see setup of AI?

Sorry, what is second monitor?
Edit: is it Second Monitor - Timing/Telemetry App

That looks amazing, never knew this existed!

 

There is no way of seeing AI setups afaik.

And yes that's the app, very useful and well named, it's always running on my second monitor!

 

FUG2. Silverstone, 110 AI Med aggro.
AI braking too slow in the braking zones of turn 6 and turn 16. Seems like it depends on where you are, but it's very easy to catch you offguard.

Turn 6 you should brake at around 65m, they brake there sometimes, sometimes they brake at 80m
Turn 16 you shoulb brake at the 100m, sometimes they do it, sometimes they brake at 120.

Also their performance is very inconsistent, the fastest laps from P1 are around the 1.29.9, but they drop their speed to the 1.31.5 after a couple of laps. May have something to do with their battery going from Attack to Balanced.

AI seems too quick on cold tyres, first lap is very difficult to follow, it gets better in the second lap.

 

FUG2. Barcelona no chicane, AI 110 Med aggro.

AI are generally too fast on the last corner, carrying an extra 10-15kmh on the exit.

 

Baseline Strength: 110/High Aggression (AI was at 100 for the race report, 110 for the qualy one)
Car Used:Stock Car 2024
Track: Brazilia
Session: Quali/Race

The AI is way too slow on the wet in this class, I gained 5 seconds from dry to the wet on the AI. I was 2 seconds slower than the AI on qualy in the dry and in the wet I was 2.8 seconds faster after doing only 1 timed lap, photos below.




As you can see I only lost 11 seconds from the dry to the wet which is too low, even more so in a longish track like Brasilia, the AI lost a more respectable 16 seconds, dunno if the player wet tires haven't been updated yet but its a big difference.

Also did a 50 minute race and the AI lose too much performance by the time they pit, I lost about 5 tenths from the start to lap 13, the AI were losing 3 seconds, rendering the race a cakewalk. So either the player or AI tire wear needs adjustment. They were behaving rather well, but never did cover the inside either against me or themselves and were having a lot of trouble overtaking, backing out of clear side by side inside overtaking opportunities at Scavone and Entrada do Miolo.

 

Formula Ultimate Gen 2: AI brake very early for the chicane at Kansai - the backmarkers, anyway. Even if no AI cars are just of them.