When is it a feat - and when is it a cheat? [The Suspend/Resume Bug]

[Famine]

So, earlier today I finished the Nürburgring 24 Hour race. As you can see above, I won, after finishing one more lap than the lead AI - in fact the margin was about 1'05. As you can also see above, I won using a GT300 car - the premium Subaru Cusco Impreza '08, more specifically. Given that this car was stock (oil change and VCD set to 20:80; otherwise stock), running at 311hp, 1100kg and 528PP and the lead AI cars were 650PP and above, that's quite an achievement - a feat, in fact.

Right?

Nope. I knew I was going to win the moment I picked the car. Why? Well let's look more closely at that results screen:

I finished 157 laps. The lead AI - a McLaren F1 GTR, a Pagani Zonda LM, an Audi R10, a Peugeot 905 and a Ford GT LM - all finished 156 laps, with a sixth car one further lap back on 155. The fact we all finished so close, despite huge PP, tyre use and fuel use variances should set alarm bells ringing. Let's compare it to a few race reports:

The last few of that bunch are running pretty much of a muchness, but the lead AI are losing 2-6 laps (lest you think that trifling, it's 15 to 50 minutes)... How odd. More odd still - those are both races run partially in wet conditions, whereas mine was run entirely in the dry... The McLaren finished 156 laps in an entirely dry 24 hours - something it should be easily able to do in 22 and a half. In fact I won the race at an average speed sufficient to win the Nürburgring 4 Hour race against the Amuse S2000 GT1 on sport hards - I should have been caned by the GT1 cars here...

How is this possible? By cheating. This wasn't a feat - it was a cheat.

In fact, the reason I did this race in this manner was to highlight the cheat in the most obvious way I could possibly manage - for a little background, I was actually planning on running this race in as little car as I could manage (as usual). I'd found a 600PP candidate in the shape of a Nissan R34 TC and had run about 5 hours of quite a tight race when I discovered a problem - a problem which meant my race to that point had been utterly without merit because I could win at will and that wasn't the race I was after. I was quite disappointed (or "pissed off") at having engineered a close race and spending 5 hours on it only for this flaw to render it pointless and I decided to bring it to the attention of the community at large (and hopefully PD) by cheating the race in the grossest possible way.

By way of garnering attention, I'm not going to say what the "problem" is just yet. I'm going to ask for guesses If no-one's got it in 24 hours (appropriate), I'll reveal all of my data and the nature of the "problem". I believe it to be fixable - if it isn't it could render all attempts at certain endurance races pointless and that's quite a poor state of affairs. Particularly if, like me, you like endurance races - I don't want it publicised so people can exploit it, but so that PD can fix it.


Some people - a couple of folk with whom I've been exchanging PMs and most of the Infield - know what the "problem" is and I'll request that they do not give the game away just yet. The purpose is to get attention for the "problem".

Guess away

Feel free to ask any questions too.


Edit: Explanation in post 39.

 

[CarBastard]

Looking much forward to this Famine. I see you as one of the most knowledgeable here in GTP and your explanation of this tremendous flaw of GT5s AI is certainly one to watch. Specially given that I've been trying something somewhat similar for the 4-Hour Nordschleife race.

👍

My guess is: Rubberband AI that drives slowly on purpose.

 

[Schwartz]

Does this problem exist in all endurance races?
Or on B-Spec?

 

[Famine]

Does this problem exist in all endurance races?

No, only certain ones.

Or on B-Spec?

I haven't personally confirmed it, but I have received very good and trusted evidence that it exists in B-Spec too, yes. Sadly.

 

[Deko Wolf-GTPT]

So...once you finish the race, the game does some sort of calculations to produce the "final" result. Since the A.I. never actually finished, those results are just a calculation.

If that's not it, then...the A.I. is not exactly competing, it's just going for a Sunday drive.

I'm trying to figure it out the best I can, but I can't quite understand the information you're providing.
But, I get the sense that you are on to something here...

 

[Schwartz]

Maybe it has to do with pit stops. If you cancel a pit stop everyone else goes in and stops (because they always follow you into the pits) and you just go right on through.

 

[Famine]

So...once you finish the race, the game does some sort of calculations to produce the "final" result. Since the A.I. never actually finished, those results are just a calculation.

If that's not it, then...the A.I. is not exactly competing, it's just going for a Sunday drive.

I'm trying to figure it out the best I can, but I can't quite understand the information you're providing.
But, I get the sense that you are on to something here...

I'm being a little vague on purpose, but I hope not too much!

I'll add that I was in the lead for pretty much 75% of the race. I would get utterly trounced on Dottinger Hohe every few laps, but their lead would never last. Since I was already cheating I took the trouble to block a bit as well, but blocking six 200mph cars at once in one 155mph car is... not simple. Or possible

Maybe it has to do with pit stops. If you cancel a pit stop everyone else goes in and stops (because they always follow you into the pits) and you just go right on through.

I think it would be fair to say that pitting is part of the puzzle, but not quite in this manner.

I was on a 3-lap strategy (I'd pit every third lap, so I had an out-lap, a fast lap and an in-lap), whereas all the AI were on a 2-lap strategy.

 

[Will27]

“Suicide” laps in my race. If it wasn’t for that, there would have been less of a lead 1~2 laps I think.

What were your lap times like Famine?
I was averaging approx 8:25s (excluding pit in/out laps) on Racing hards.

I like doing the same with engineering a tight race, interested to hear the "problem"..

 

[ghsnu]

I believe the lazy AI is in place to allow anyone to be competitive, thus making wheel users or "controller gods" to be laps faster than the AI.

The AI is flawed because PD didn't implement "learning AI" (like Forza) or a difficulty system (again...like Forza and many other games).

The one endurance race I did was the Grand Valley 300km in a 390hp Ferrari SP1 before the pp system (so I didn't have anything to really base my car choice and tuning on ). I thought it would be a competitive race - turns out it wasn't that hard. I think I only had trouble with a 350Z or 370Z, other than that, the race was a walk in the park (I was also on Sports Hards or Softs I think).

Is an AI reform needed? Yes. Or atleast provide some restrictions to follow if the player chooses.

 

[g2010ag]

Would it have to do with the AI starting or running on the wrong tires?
Just my best guess.

 

[GTlondoner]

I noticed on Le mans 24 hour that the AI would wear out their tyre's and do 50mph cruise to pits making there in-laps much longer than mine.

Is it the same case here even though they doing two laps on I'm guessing racing hards and they should manage two laps as they drive smoother.

EDIT: Plus on two lap strategy the AI never get to do a genuine fast lap.

 

[Deko Wolf-GTPT]

I'm being a little vague on purpose, but I hope not too much!

I'll add that I was in the lead for pretty much 75% of the race. I would get utterly trounced on Dottinger Hohe every few laps, but their lead would never last. Since I was already cheating I took the trouble to block a bit as well, but blocking six 200mph cars at once in one 155mph car is... not simple. Or possible

You somehow forced the A.I. to stay out longer, thus using more tyeres, resulting in slower in and out laps. Thar leads me to...
Pit Strategy...you are somehow manipulating the A.I. to pit in strategic laps so that you can resume the lead.
When they stop, you resume the lead, and vice-versa, until you complete the race, and they are about to pit...

This is what I think the result would be, but how you got to it is a different matter.

 

[Famine]

“Suicide” laps in my race. If it wasn’t for that, there would have been less of a lead 1~2 laps I think.

What were your lap times like Famine?
I was averaging approx 8:25s (excluding pit in/out laps) on Racing hards.

I like doing the same with engineering a tight race, interested to hear the "problem"..

My average lap across the race was about 9'09.9. I'd tend to run outlaps of 9'12-9'30 depending on fuel, a fast lap of 8'51 and an in-lap of about 9'00, usually depending on my AI interactions.

I believe the lazy AI is in place to allow anyone to be competitive, thus making wheel users or "controller gods" to be laps faster than the AI.

The AI is flawed because PD didn't implement "learning AI" (like Forza) or a difficulty system (again...like Forza and many other games).

The one endurance race I did was the Grand Valley 300km in a 390hp Ferrari SP1 before the pp system (so I didn't have anything to really base my car choice and tuning on ). I thought it would be a competitive race - turns out it wasn't that hard. I think I only had trouble with a 350Z or 370Z, other than that, the race was a walk in the park (I was also on Sports Hards or Softs I think).

Is an AI reform needed? Yes. Or atleast provide some restrictions to follow if the player chooses.

This is more than lazy AI - this is an observable change in their behaviour due to... player cheating.

They should all run considerably more laps than they did with the cheat in place. Even with the guy in the R35 TC tooling around 17 laps ahead of them and on a wet circuit, they were running up to an hour ahead of my AI. And this is fairly consistent - you should not be able to win this race on less than 163 laps and much, much more than that if it doesn't rain. I won on 157 laps in the dry. By cheating.

Would it have to do with the AI starting or running on the wrong tires?
Just my best guess.

Although in an A-Spec race I cannot directly observe the AI tyres, I'd say that there's a big fat "yes" to that one. However, the AI will pit in on any current lap on which they are running incorrect tyres, so that would only affect them for a maximum of one lap.

I noticed on Le mans 24 hour that the AI would wear out their tyre's and do 50mph cruise to pits making there in-laps much longer than mine.

Is it the same case here even though they doing two laps on I'm guessing racing hards and they should manage two laps as they drive smoother.

EDIT: Plus on two lap strategy the AI never get to do a genuine fast lap.

Yes, some of the AI would do that on the N4 race too - a suicide lap - but I believe that was patched in 2.05.

I can also say that the lead AI - particularly the 905 - were 40s a lap faster than me on any given lap.

You somehow forced the A.I. to stay out longer, thus using more tyeres, resulting in slower in and out laps. Thar leads me to...
Pit Strategy...you are somehow manipulating the A.I. to pit in strategic laps so that you can resume the lead.
When they stop, you resume the lead, and vice-versa, until you complete the race, and they are about to pit...

This is what I think the result would be, but how you got to it is a different matter.

I'd say all of that is fair - but that it's quite difficult to force all of the AI to miss their pit stops and still run a GT300 car to 9'09 averages on the N24 course!

 

[250 GTO]

Does it somehow have anything to do with that fact that it's never an even grid of AI Cars. There is always a number of cars that clearly isn't in the same PP-class as the majority. ?? I've always found that odd.

 

[phil_75]

Something to do with being in a 4WD car?

Easier to drive on worn tyres or last longer by fitting Sport tyres?
I noticed you were going to use the R34 Skyline (4WD) first!?

 

[GTlondoner]

Seeing as you done it in 157 laps and say it's not possible to win in under 163 laps, Then you must have been blocking around the whole track slowing every AI car and only had problems holding them back on the wide straights.

But then the 905 being 40s faster makes me wrong unless it was catching up from a pitstop that made it 40's behind.

 

[Deko Wolf-GTPT]

There's something with those numbers...as if after 4h the A.I. changes the pattern.
And why is the R10 ahead the 905, and both behind the McLaren? Both are superior than the McLaren...right?

...I dunno, you got to give us more evidence Famine. Maybe by combining all of this previous guesses from other members.

 

[GTPorsche]

Would it possibly have anything to do with crossing the line at the finish with no other cars close enough to cross the line before the finish times out? That's happened with my driver in B-Spec before in the 4 hour race.

 

[Mulan]

The PS3 gets hot and slows down the ai?

 

[grashopper]

Does it have anything to do with the new save feature? Is it somehow loading the AI back improperly.

 

[Furinkazen]

It's... ~not saying because I know~

 

[crooky369]

Shocking how slow the AI are. That 905 in real life is probably a good 30 seconds a lap quicker around a small track than that Subaru... Should be 2+ minutes around the 'ring.

 

[Deko Wolf-GTPT]

Does it have anything to do with the new save feature? Is it somehow loading the AI back improperly.

Interesting...forget about that feature.
Let's assume that the A.I. is loaded correctly into the track in the same position and location on the track, BUUUT, pit in after that?

 

[HuskyGT]

It's definitely has to do with the save feature. That's why the thread is called "When is it a feature, and when is it a cheat." Some people always considered saving during an endurance a cheat because it beats the purpose of an actual "Endurance."

Could it be that if you manage to learn when does the AI pit in for tires and fuel, and then you use the save feature, then you load the game and the AI doesn't pit anymore.

 

[Catlin]

i did notice,i just done the 24hr sarthe for the first time,i was saving between every 10 or 20 laps depending on how much i felt like doing,my weather cycled differently each time,the first time i pitted in,the weather was heavy rain with 100% track,when i booted it up the next day it was 0% wet track and the rain was gone. im not sure how it calculates all this,but i guess its something to do with it.

 

[Ryk]

Gah - I was planning on doing this Sunday - hate to have a second enduro messed up---

Can you do this cheat accidently?
(Is it a vegetable...)

 

[dudeondacouch]

It's definitely has to do with the save feature. That's why the thread is called "When is it a feature, and when is it a cheat."

"feat" ≠ "feature"

 

[Cowboys965]

Is it when your losing, but after the timer is up you cross the line 1st, so the game puts you in 1st, even though you weren't leading?

And it only exists in timed races?

 

[mvpers]

It's definitely has to do with the save feature. That's why the thread is called "When is it a feature, and when is it a cheat." Some people always considered saving during an endurance a cheat because it beats the purpose of an actual "Endurance."

Could it be that if you manage to learn when does the AI pit in for tires and fuel, and then you use the save feature, then you load the game and the AI doesn't pit anymore.

a "feat" is an accomplishment, it doesnt mean feature.

 

[SicKmcNasTy]

i did notice,i just done the 24hr sarthe for the first time,i was saving between every 10 or 20 laps depending on how much i felt like doing,my weather cycled differently each time,the first time i pitted in,the weather was heavy rain with 100% track,when i booted it up the next day it was 0% wet track and the rain was gone. im not sure how it calculates all this,but i guess its something to do with it.

Judging off of what you said, I'm assuming that the second they all pit for wet tires he saves the game and when he comes back it's dry and they're still on wets, possibly forcing yet another pit.