Copying this from another thread to save the moderators the hassle. So without getting into too much history in another thread we were discussing if it was possible to have a competetive race with the AI in Gran Turismo 6. @Tenacious D claimed that you could, and said he'd had several good races with the AI in which he'd lost as evidence of this. He provided the replay files of these races for anyone to analyse:
So far me and @Johnnypenso have analysed some of these replays. Here are our findings so far:
Obviously this thread isn't just about analyzing Tenacious D or my replays, it's about analyzing the AI in general. So, let the discussion continue in this thread rather than off topic in another I guess.
So anyhow, here are my replays, hopefully. All in their proper folders and whatnot. If not whatnot, let me know and I'll see if I can sort things out. And if you're lucky, you'll se me tooting around a track in my VWs against race cars and getting flogged. Humor, narf.
https://www.dropbox.com/s/cdq3tc87ejpv5xe/Replay-1.zip
So far me and @Johnnypenso have analysed some of these replays. Here are our findings so far:
Okay that worked. I left the USB in the PS3 and was able to download it from the USB into the Replay Gallery. I watched one race, 430 Scuderia at Motegi against a field of race cars. Putting aside the fact that @Tenacious D cakewalked through most of the field of full fledged race cars on the same tires in a street car by both out cornering and out accelerating them, here's what stood out to me.
Lap 2
AI Leader2 - 2:01
TenD - 2:04
So we know the AI can run at around 2:00 pace but all of a sudden on lap 3 he's forgotten when to brake, can't find an apex and is uncertain when to hit the throttle. So...
Lap 3
AI Leader - 2:11 (10 seconds slower than lap 2)
TenD - 2:03
By now TenD is up through most of the field but still not in first place yet...so...
Lap 4
AI Leader - 2:11 (again 10 seconds slower than lap 2)
TenD - 2:05.6
By now TenD has caught the leader and a battle ensues and all of a sudden the leader can find an apex and remembers how the throttle and brakes work so...
Lap 5
AI Leader - 2:05
TenD - 2:04
I'm sure it's debatable, but the AI was behind TenD exiting the last corner and appear to lift briefly before pulling out to make a failed pass attempt at the line.
Make of it what you will, to me that's not racing. The leader simply slowed his pace by 20 seconds in the middle of the race to let the player catch up and then when he caught up he put up a half-hearted battle for the lead. I'm sure many GT players can suspend reality and rational thought and pretend they actually had a race and that's great for them. Only speaking for myself, but once I realize what the AI is doing and the race isn't "real", it's game over for me.
Ok, I've also analysed one race I selected at random from the lot. It was the Dream Car Championship Race 1, at Motegi. The lead car was in a Jaguar XJR-9 Race Car, @Tenacious D was in a Lamborghini Diablo GT-1. So first of all let's look at the stats, the facts on those cars.
Jaguar XJR-9 - 749BHP, 894KG, 680PP
Diablo GT1 - 591BHP, 1200KG, 593PP
So right away we can see there is a huge performance gulf. TenD is giving away over 150bhp in power and is much heavier. So I look at the replay. First of all one further fact, looking at the data at the first checkpoint TenD is 14.5 seconds behind the Jaguar.
- So lap one and the main thing I notice is how cautiously TenD is driving, he's very smooth and not aggressive in his driving style. Little oversteer etc. Despite this he blows by the first AI cars, flying around the outside of many with ease and ends lap 1 from 14th to 8th, a 1.53.8.
- Lap two is more of the same, easy passes with no effort despite fairly cautious driving. 1.51, he's up to 5th.
- By the end of lap three he's made two more fairly simple passes and ends it in 3rd, a 1.53.7.
- On lap four he's again, very cautious with the throttle. As he was going around the outside of a Nissan I believe he backed off and hit the brakes when putting his foot down would have got him past sooner. Despite this he ends the lap in 2nd on a 1.52.9.
- Final lap is spent attacking the lead Jaguar, he loses distance on the straight but is able to easily outbrake into the corners to make a pass into the lead, at turn 11, the 90 degree right hander. He takes the last few corners in a normal manner but doesn't really defend and coming out of the last corner pretty slowly. The Jaguar uses it's superior power to pip him on the line by a tenth or so, a final lap time of 1.51.1.
Race Summary
TenD finished second because of two main reasons, his somewhat timid, cautious driving style and the fact the AI was significantly faster on the straights, because of the higher power.
My attempt
- Out of curiosity following this I decided to try the race myself. I hadn't played the game at all for weeks and I didn't own the Lambo, so I went and bought it. Did nothing to it, went in stock and blind. I drove with a DS3, all aids off except ABS1. I don't know how you drove. I made sure the Jag was in first but obviously the entire field wasn't identical. First lap was a struggle getting used to the car and track, ended with a 1.55.1 in 9th, so a little behind TenD at this point.
- Second lap got better into the flow, a 1.50.9 and I'm in 5th, same as TenD (but faster lap) and same experience, flying past the AI with no trouble, usually around the outside.
- Third lap is a 1.51.9 and I'm up to third, again matching TenD in position but faster lap.
- Fourth lap is a 1.51.5 and I'm 1st on the home straight but passed by the Mercedes (Which passes the Jaguar) in a very similar manner to TenD, outpowered, and just end the lap in 2nd.
- I very easily pass the Merc in turn 1 to take the lead and spend the rest of the lap defending fairly (IMO). I take a defensive line on the straight where he catches up but fly away in the corners. I end the final lap with a 1.50.1 and finish 1st, a few tenths ahead of the Merc. Final race time 9.19.
Of course it would only be fair to share my replay file, so here you go - https://mega.co.nz/#!hUEyFbYY!SlK9YY5qBAxadwQWZh9859kNqyoBveI1-5HNpRrh8xI
I personally believe I drove fairly, there were a couple of bumps but nothing over the line IMO. If you have different thoughts on my race I'd be happy to hear them.
Final thoughts
So yes, I agree with you on one aspect TenD, the AI do not always let you win. Absolutely they wanted to win those races.
However I personally would not call that real racing and I do not find it fun. The only reason it was remotely close is because of the power advantage the AI had over us. They were painfully slow in the corners making any overtaking simple, with no challenge. I had to defend slightly on the final lap but the AI braked so early and were so slow in the corners it was no real challenge to keep the win, and I didn't find the experience particularly fun. I blasted past the AI with relative ease driving a new car for the first time in months and so did you. As I say you only lost because the Jaguar had so much more power and outdid you on the final straight, plus I personally thought your driving was a little timid. That isn't a slight on you before you think it is, I accept we're all at a different skill level and all drive differently. I'm certainly nowhere near the best drivers, and you're not awfully slow.
I did actually find that analysis and comparison very interesting though, truly, it does show how different people play the game and how their skill greatly affects the outcome when there is no difficulty settings in game. I may look at some others later.
EDIT: I realise I forgot to note the AI laptimes on my analysis. I will do that now and add it later.
Anyway, the laps:
TenD :
1. 1.53.8
2. 1.51
3. 1.53.7
4. 1.52.9
5. 1.51.8
Jag:
1. Impossible to say exactly because it starts past turn two, but approximately 1.56/7 based on simple maths
2. 1.57.7
3. 1.56.0
4. 1.54.4
5. 1.53.3
So the Jag did get progressively faster as the race went on however it was still well under it's potential and still slower than TenD for every lap.
Here are the numbers for my effort:
Me:
1. 1.55.1
2. 1.50.9
3. 1.51.9
4. 1.51.5
5. 1.50.1
Jag/Merc
1. Again an estimated 1.56/7
2. 1.57.4
3. 1.57.0
4. 1.54.8
4. 1.50.4
So again it got faster towards the end of the race when I got closer but started out very slowly, and perhaps compensated for my slow first lap by having a slower third lap but since I got past it the final lap was considerably faster.
@Tenacious D , do you have any thoughts on the analysis? I also watched another replay, I chose this time one with equal cars to see how that went and to be honest without giving a full breakdown it does seem the simple reason you have more fun with the AI is that the locked in default difficulty is closer to your skill than it perhaps is for other people, in both directions. I watched the race with the Phoenix Audi GT3 at Nurburgring, again you seemed quite cautious and I did the race afterwards, and lapped 2.5 seconds quicker than you, even though the stock handling of that car I find awful.
So yes, to me there are four clear points to improving the AI and single player experience for everyone.
1. Removing all forms of rubberbanding/AI slowing/speeding up on purpose
2. Removing the staggered start, either add standing starts or realistic double file rolling starts
3. Adding some difficulty levels, perhaps two more above the current one and two below. Certainly at least one either way.
4. Make sure all AI cars are equal and as close to the race PP limits as possible
If you imagine the first race I analysed with 1 and 2 in place you'd have started mid way down the grid perhaps but only a few seconds behind the Jagaur. However rather than catching it the Jaguar would've performed to it's ultimate ability and flown away, neither of us would have got close to it. Therefore we could've entered the race with a car similar to the Jaguar and had a true close race, with similar performance in both the straights and corners. With 3 in place I could have increased the difficulty to have a harder race and with 4 in place passing the rest of the cars wouldn't have been such a breeze for either of us. Like I said in the analysis, flying around the outside of other cars with no effort is not fun.
Obviously this thread isn't just about analyzing Tenacious D or my replays, it's about analyzing the AI in general. So, let the discussion continue in this thread rather than off topic in another I guess.