Is Sophy 2.0 no longer competitive?

[Sturdy]

It certainly isn't as competitive as the original release, simply because in a similar car I can now win, which was never going to happen against the initial version we had!

I'm happy with it at this stage though, it is competitive enough, it means I can race stock cars and have a good time (which I enjoy), and also the grids change so the races aren't as curated / processional which keeps it interesting.

An extra difficulty level above would be good, but as a first implementation, it really isn't bad, albeit a shame it's not in Custom Races, which would be brilliant.

 

[NJ72]

Standard AI will use the car as you have tuned it except for using the tyres it was acquired on.
To me this looks like to be bugged instead.

The key difference here is that the AI doesn't choose the car or tuning. PD chose the car and tuning and have coded it in. Which is why the opponents always drive the exact same cars and are in the exact same positions.

With Sophy the grid is randomised every single time, based on the stock PP and tyres of your car. This leads to WAY more repeatability of the races, and much more challenging races.

It is not bugged, it is by design. Rubber banding has been removed, so it is down to you to tune the car to make the race challenging. It isn't perfect, but if you get it right racing against Sophy is absolutely mega compared to the normal "AI" in the other modes.

 

[StuntmanRich]

From what I’ve played so far I’ve enjoyed the new AI. I don’t know how people are beating it on the hardest difficulty over 3/4 laps while also still making mistakes.

My one observation of the AI is they have no regard for track limits at Le Mans, particularly down the Mulsanne straight.

 

[Mafio]

From what I’ve played so far I’ve enjoyed the new AI. I don’t know how people are beating it on the hardest difficulty over 3/4 laps while also still making mistakes.

Conveniently leaving out the part of them being in an overpowered car. On easy I barley made it to first place on the second lap in a bone stock, fresh off the lot 911. I still feel like the Sophy doesn't immediately yield to the passer. She gets a little spicy when its a tight pass.

 

[benjoi84]

For me, the fact we can set up a reasonably challenging & fun race in many different cars (albeit at only 9 circuits), without having to faff around with Custom Settings is a massive step forward for Single Player.

 

[Bojador]

Conveniently leaving out the part of them being in an overpowered car. On easy I barley made it to first place on the second lap in a bone stock, fresh off the lot 911. I still feel like the Sophy doesn't immediately yield to the passer. She gets a little spicy when its a tight pass.

Have you tried it with a gr4/3/2 car? From what I’m reading people are finding it challenging when using road stock cars, but racing cars where bop is actually balanced, on some tracks at least, it is not very challenging.

 

[Meythia]

where bop is actually balanced,

I dont think the quick race mode is using BoP but instead the default settings of the Gr.? cars.
Though they are pretty close together anyway.

 

[Bojador]

I dont think the quick race mode is using BoP but instead the default settings of the Gr.? cars.
Though they are pretty close together anyway.

When you try to tune the cars it says bop is on. Unless it’s bugged, I’d say it is on.

 

[meekrab]

Another big improvement: Sophy cars are not magnetically anchored to the track, if you punt them in revenge for their clueless driving they actually spin.

 

[newmedia_dev]

I'd imagine that the earlier test in the year they were testing which Sophy fit the majority of players and we got green/blue but not purple.

Apologies if this has already been said.

 

[Mafio]

Have you tried it with a gr4/3/2 car? From what I’m reading people are finding it challenging when using road stock cars, but racing cars where bop is actually balanced, on some tracks at least, it is not very challenging.

Im going to have to try it tonight.

 

[Met Entso]

The only use I see for the current Sophy is to practice close quarters driving using stock cars and hard tyres. Specially useful for VR probably.

 

[BallPtPenTheif]

So this thread is a revelation to me that I suck.

I see you guys leaving them behind with 5 seconds difference while I'm here racing Sophy on Tsukuba with a bone stock GR86 and I'm struggling to catch up with the 2nd and 1st position, but other than that I do feel like they're not as fast as the one from the preview build.

You don't suck. People are just using suped-up cars against the AI and then wondering, "Why is the AI so slow?"

In the Sophy Preview, your own car was faster in easy and worse or stock on the harder Sophy levels. Now you just have to choose worse tires to make it more challenging.

I don't understand why people think that Sophy itself is slower when the difference between Sports Hard and Race Soft tires is so massive.

 

[meekrab]

You don't suck. People are just using suped-up cars against the AI and then wondering, "Why is the AI so slow?"

In the Sophy Preview, your own car was faster in easy and worse or stock on the harder Sophy levels. Now you just have to choose worse tires to make it more challenging.

I don't understand why people think that Sophy itself is slower when the difference between Sports Hard and Race Soft tires is so massive.

We all suck, that's why we're racing virtual cars and not stealing Kevin Estre's job.

But don't assuming people are all using OP cars, I've been racing group 4 and group 3 on hard tires. If I can go from tenth to first in 3 laps the other cars are just slow. In most daily races I go backwards, not forward.

 

[i386]

I did a few races against sophy on hardest difficulty in le mans with the 750vgt gr1 (with full downforce mode ) and it was fun. Best place was p2 for me (p1 is usually a group c car)

 

[Bojador]

You don't suck. People are just using suped-up cars against the AI and then wondering, "Why is the AI so slow?"

In the Sophy Preview, your own car was faster in easy and worse or stock on the harder Sophy levels. Now you just have to choose worse tires to make it more challenging.

I don't understand why people think that Sophy itself is slower when the difference between Sports Hard and Race Soft tires is so massive.

They are slower. I tested gr2 cars, same tyres, stock (doesn’t really matter because bop is on) and the AI top speed is much lower than my car. You may get better experience nerfing your own car, but if you use race cars where bop is on that’s not even an option.

 

[RacinLei]

Sophy right now is the right balance and could be released for the hole game. On hard difficulty and same hard tire, she gives you good and fair battles, dosnt divebomb anymore and no blockracing needed. In the test phase you have to block sophy to much to stay in front.
Thumbs up 😃

 

[RapidRacer720]

What I discovered is that it's really fast at the start of the race, around the first few corners or so. Then it gets slower, especially on some straights. This behaviour is really similar to what happens to the second player in split-screen races.

My assumption is that this might actually be a bug. It seems that when the frame rate is lower due to the high demands of Sophy and the memory being full with assets, the opponents movement per frame doesn't properly scale up.

Did anyone experience this too?

 

[rlx]

What I discovered is that it's really fast at the start of the race, around the first few corners or so. Then it gets slower, especially on some straights. This behaviour is really similar to what happens to the second player in split-screen races.

My assumption is that this might actually be a bug. It seems that when the frame rate is lower due to the high demands of Sophy and the memory being full with assets, the opponents movement per frame doesn't properly scale up.

Did anyone experience this too?

I had the same impression: Sophy getting slower towards the end of the race, most obviously on straights. Definitely requires more testing though.

I have never used split-screen, so I don't know what that bug is, and how it might be related. If it had something to do with frame rate, then wouldn't the effects be experienced mostly during crowded starts?

 

[MroczMR]

I'd imagine that the earlier test in the year they were testing which Sophy fit the majority of players and we got green/blue but not purple.

Apologies if this has already been said.

It seems that way.

On the other hand. PD Should give another difficulty level like "alien" or "hardcore" or "pro" with Sophy in it's full potential.

 

[tofuhero]

Thinking about it some more, I suspect that the reason Sophy isn't as challenging as the preview period was is because it's a full field of them vs. the 3(?) we had before. That's probably quite a bit of CPU power, and I know they said higher difficulty Sophy takes up more horsepower. I would rather have a couple of them be the really challenging type and the backfield being rather bad, than the way it is now. I bet they will continue to adjust that mix based on feedback.

 

[RapidRacer720]

I had the same impression: Sophy getting slower towards the end of the race, most obviously on straights. Definitely requires more testing though.

I have never used split-screen, so I don't know what that bug is, and how it might be related. If it had something to do with frame rate, then wouldn't the effects be experienced mostly during crowded starts?

No, it wouldn't be just on crowded starts. All AIs are simulated all of the time, regardless whether they're close together.

The split-screen bug is really similar to this. In the first lap, the performance of both players is similar, but after that, player 2 becomes significantly slower.

I also noticed that the AI seems particularly slow on some straights, while being faster on others. This might be related to some scenes being more complex to render than others. However, I cannot say anything for sure.

 

[Bendough]

An engine swapped R34 was good enough to take the lead by Dunlop chicane on the hardest level in Le Mans . Gr2 was much better, I actually had to fight for the win. Haven’t tried other cars just yet

 

[rlx]

I've watched some Le Mans replays.

On hardest difficulty, Sophy 2.0 in a Chiron (Sports Hards) reaches 350 km/h in 7th gear before the first chicane on the Mulsanne Straight (in P1 with no other cars nearby).

Driving a stock Chiron with power restricted to 70% (Sports Hards, PP 666.65), I can reach 360 km/h in 6th gear at the same spot.

Requires further testing, but we're definitely dealing with a pretty significant performance nerf.

 

[Artremis]

That is assuming it was intentional.
Standard AI will use the car as you have tuned it except for using the tyres it was acquired on.
To me this looks like to be bugged instead.

This isn't true, on quick races the opponent selection is based on your car regardless of tuning. Quick example:



Sophy has just inherited this behavior. Presumably because it relates to how the game itself chooses opponents and not really because of the AI

I had the same impression: Sophy getting slower towards the end of the race, most obviously on straights. Definitely requires more testing though.

This is absolutely the case. An easy way to test it is in a BoP'd class on a track with very long straights like Deep Forest or Spa, on lap 1 they will run normally but in lap 2 and 3 you will be significantly quicker than any car, even those in the slip-stream with you staying outside of it. I don't know if it's them going slowly or you getting a boost, but there's absolutely a pace difference there.

The way Sophy is implemented at the moment in general seems to be as an enhanced version of the game's base AI, just much cleaner and aware of other cars, capable of fighting with each other, etc. I hope the "crawl on lap 2-3" is just behavior they can tweak and not how Sophy is actually scaled to lower difficulties, because for that first lap the battles are genuinely enjoyable even if I don't make up that many places. I'd expect them to keep this behavior for things like campaign events or the quick races as long as in terms of game design the objective is to finish in the top 3, but a toggle for it on custom races would fix it all.

 

[rlx]

This is absolutely the case. An easy way to test it is in a BoP'd class on a track with very long straights like Deep Forest or Spa, on lap 1 they will run normally but in lap 2 and 3 you will be significantly quicker than any car, even those in the slip-stream with you staying outside of it. I don't know if it's them going slowly or you getting a boost, but there's absolutely a pace difference there.

Another Le Mans replay, Gr.3, hardest difficulty.

Below is the top speed of the car in P1 (in this case the AMG '20 and the Ford GT Test Car) at four points of the track:
before the first chicane / before the second chicane / before Mulsanne Corner / before Indianapolis Corner.

Lap 1: 270 / 264 / 246 / 254
Lap 2: 242 / 238 / 234 / 250
Lap 3: 254 / 246 / 242 / 246

Obviously this is still not enough data, but the above does look like a slow-down.

 

[ShiftingGears]

I mean this is a valid critique of the implementation of a new technology, I'd argue rather than the technology itself.

We are still in the early days of this. I was surprised by the amount of tracks and full 20-car fields frankly, this is PD we are talking about. Their timelines are close to geologic time periods.

Yes, but I can only judge the game by what's implemented in the game. I would never argue against having an AI with a higher ceiling.

 

[RapidRacer720]

on lap 1 they will run normally but in lap 2 and 3 you will be significantly quicker than any car

I had the same impression: Sophy getting slower towards the end of the race, most obviously on straights.

Do you think the issue is related to frame rate in any way?

Usually, in the amount the cars move in this frame would be multiplied by the time it took the frame to render, so they have the same speed regardless of frame rate. If this wouldn't be the case, the cars would move slower at lower frame rates.

The slow-down of the AI seems pretty arbitrary. And since Sophy AI (and 2-player splitscreen where a similar problem occurs) probably require a lot of computational power, this seems like a reasonable explanation to me. They still are fast in the first lap because the memory is not yet filled, so the performance demands are slightly lower.

Even if the frame rate only drops from 60 to 55 FPS, it would still make the AI a few seconds slower every lap.

 

[tofuhero]

Another Le Mans replay, Gr.3, hardest difficulty.

Below is the top speed of the car in P1 (in this case the AMG '20 and the Ford GT Test Car) at four points of the track:
before the first chicane / before the second chicane / before Mulsanne Corner / before Indianapolis Corner.

Lap 1: 270 / 264 / 246 / 254
Lap 2: 242 / 238 / 234 / 250
Lap 3: 254 / 246 / 242 / 246

Obviously this is still not enough data, but the above does look like a slow-down.

Did the old AI opponents in Arcade Races have similar slowdowns? I wonder if this is something being applied to Sophy's vehicles by the game mode, and not Sophy itself doing this. I guess an easy test would be to see if the non-Sophy tracks in Quick Races have the same AI slowdowns.

 

[rlx]

Do you think the issue is related to frame rate in any way?

Usually, in the amount the cars move in this frame would be multiplied by the time it took the frame to render, so they have the same speed regardless of frame rate. If this wouldn't be the case, the cars would move slower at lower frame rates.

The slow-down of the AI seems pretty arbitrary. And since Sophy AI (and 2-player splitscreen where a similar problem occurs) probably require a lot of computational power, this seems like a reasonable explanation to me. They still are fast in the first lap because the memory is not yet filled, so the performance demands are slightly lower.

Even if the frame rate only drops from 60 to 55 FPS, it would still make the AI a few seconds slower every lap.

Of course I could be wrong, but I don't think frame rate is the issue here. As far as I can tell, there is a loop listening for player input, there is a loop querying the AI (at 10 Hz, at least originally), and there is a loop rendering the game state (at 60 Hz). I can't think of any good reason why - or observable cases where - rendering lag in GT7 would affect the actual performance of the cars (i.e. the game state). I have recorded a lot of telemetry data over time, and there is close to zero packet loss or delay; to me, the overall system doesn't seem laggy or prone to performance degradation.

Did the old AI opponents in Arcade Races have similar slowdowns? I wonder if this is something being applied to Sophy's vehicles by the game mode, and not Sophy itself doing this. I guess an easy test would be to see if the non-Sophy tracks in Quick Races have the same AI slowdowns.

Good point. I don't recall that this has ever been brought up by anyone, but it's worth testing.