View Full Version : Very inconsistent AI (Formula C / Brno)

15-04-2018, 15:25

I know there are a lot of threads about AI being "too slow/fast/reckless" etc, but I didn't really find a thread that matched my scenario, so here it goes:

I've been playing Career mode Formula Rookie and after getting used to it I found it was a blast. Slowly raising the AI level and ended the season on 88 (being a non-regular gamer with about 1-2 hours per week I was pleased with that level). I proceeded to Formula C and the European Championship, starting at Brno. Started a Custom Race with AI on 88 first just to get a hunch about what lap times to expect. The fastest AI in qualy did a 2:06:7xx something. With that knowledge I began to practice Brno in Free Practice mode and after a couple of weeks (remember, only about an hour or two per week for me) I was finally hitting high 2:06 and then I decided to continue with my career mode.

The first thing I noticed was that the Brno race was at the 26th of February, and I've been practicing in the spring (sometime in May i think). Oh well, at least it's the same for everyone, right? The ambient temperature was 7C and the track temp was 4C and man, my tires were COLD. After a couple of shaky laps, sliding a lot, I managed to at least get a couple of laps in without spinning and my lap times was around 2:13...

Then I noticed the AI times... The fastest AI did a 2:06:1 and they appeared to have set that on their first lap following the out lap! So from cold tires, with a track temperature of 4C they just needed ONE out lap to have what seemed ideal conditions to set a lap like they did when I ran the test in the spring/early summer.

Bummed, I tried "everything" with my setup even though it felt hopeless to find 6-7 seconds in these conditions... I gave up after a while and did some tests instead to see how the temperatures/weather affected the AI and I found something very strange:

All the tests were made from a Custom Race at Brno, 19 opponents, AI level 88. What was different was the time of year and weather so the ambient and track temperatures would be different. I made sure that all my comparisons were made with the same tire compound (slicks). All I did was to sit in the pit and monitor the AI on their first push lap.

2018-06-01 (summer)
Weather: Medium Cloud
Ambient: 26C
Track: 37C

Best AI time: 2:05:6

2018-02-26 (winter)
Weather: Heavy fog
Ambient: 6C
Track: 4C

Best AI time: 2:05:6 (same as summer)

I did similar tests on the wet tire compound, and the results were similar (roughly the same time in the summer as in the winter).

But here's the really strange thing: If I didn't sit and monitor the AI from the pit on their out lap following their first push lap, but instead simulated the session (Skip to end) straight away, the lap times were WAY slower. Around 5-6 seconds slower compared to when I let them drive a couple of laps while monitoring...

I went back to my Career Mode again (remember, I was only managing around 2:13 on the very cold track compared to the 2:06:ish by the AI) and simulated the session... Sure enough, the fastest AI car did 2:13:1... So, I would've been fighting for a good position if only they didn't drive supernaturally good when I was on the track (or watching from the pits).

So that's quite a bit of a conundrum for me... If I try to race, the AI destroy me on a cold track and if I instead simulate the session right away, they are driving right about my ability but then I'm last as I didn't get a time.

Version 05.02.1030 on a regular PS4.

Sorry for the wall of text, but I wanted to share this in hopes I can find some way around this as I love the game and love the concept of tight, fair races against the AI in my career mode.


15-04-2018, 23:40
You only had problems with formula c on Brno?

I'm on formula C also but just finished Panorama. It's being a pain.

16-04-2018, 13:20
I’ve only tested Formula C/Brno yet. As I said, I’m a “slow gamer” right now. I can do some tests on Panorama with the AI on different time of year and also try monitoring vs. simulating to see if it’s the same behavior.

16-04-2018, 15:48
Testing Bathurst Mount Panorama now, but only by monitoring AI and also simulating complete session (won't have time to try to learn the track myself yet).

AI still att 88, but the same thing happens with AI on 120.

First test:
2018-01-12 (Summer in Australia - Ambient: 29C, Track: 58C)
Weather: Medium Cloud
Lap time while monitoring: 2:14.667
Lap time when simulating: 2:21.106
Diff: 6.439 sec

2018-01-06 (Winter in Australia - Ambient: 11C, Track: 16C)
Weather: Heavy Fog (early in the morning to get colder temps, so actually still dark as well)
Lap time while monitoring: 2:14.665
Lap time when simulating: 2:21.180
Diff: 6.515 sec

So still almost identical lap times even if the conditions were vastly different. And Heavy Fog in the dark didn't seem to matter a bit for the AI...

For fun, I did one more test on the winter settings:

2018-01-06 (Winter in Australia - Ambient: 8C, Track: 11C)
Weather: Blizzard (6am, so still dark as well)
Lap time while monitoring: 2:24.565
Lap time when simulating: 2:21.429
Diff: -3.136 sec

In this case, the lap time while monitoring was slower than when simulating (AI slower on Wet tyres?), but what's weird is that the lap time while simulating during a Blizzard in the early morning while it's still dark is basically the the same as the ones in way better conditions...

Not sure what to make of this...

16-04-2018, 16:33
Did more testing but this time with a GT3 car and now the time while monitoring vs. the time while simulating the session is roughly the same. However, the AI still doesn't seem to be affected one bit when driving on a track that's 2C vs. a track that's 30+ C which feels a bit like "cheating" to me...

16-04-2018, 19:01
Ok... I just learned: Don't use "Slicks" when it's cold, they won't get any temperature at all. Tried to use "Sport" instead on the cold track and sure enough, the difference in temperature was huge and I was able to once again set competitive times!