Dirt Rally 2.0 difficulty depends on car selection

[TurbodTalon]

I finished Dirt Rally 1's campaign and am now working through Dirt Rally 2.0. It took me 30+ hours to graduate out of Open (yes, you read that correctly). I chose and fully upgraded several cars and simply was not able to ever rank high enough to get out of Open.

I started to notice something, and I would like to see if anyone here in the forums can offer some logical input. A clean run with the Cooper or the Lancia Fulvia puts me comfortably in first place. A run with a few mistakes +penalty for reset still puts me in the top 5, if not still in first place. And I am not talking about a balls to the wall/I have memorized the stage type of run. Just clean and smooth, hitting good apexes, ect.

I finally got out of Open using a fully upgraded Lancer EVO VI, but it took a few tries. A clean run with good speed and good apexes would net me a top 5 spot, which allowed me to win overall due to AI placings.

A clean run with the Subaru Legacy RS or Subaru R4 Rally and I'm around 25th place. I have to get a break-neck, super clean run to even break into the top 10.

In Clubman I am using a Peugeot 205 T16 Evo 2 because I wanted to try a higher powered AWD car. A clean run nets me anywhere between 25th-30th place, so I am sure I will have to move to a lower power car to get out of Clubman.

Long story short, I feel that although there's no actual difficulty slider in career mode, there is absolutely different difficulty built into the game that depends on what car/car class you've chosen. I remember that I was only able to beat Dirt Rally 1 Master's difficulty with the Cooper. I could not make the cut with a more powerful car. I was playing on a controller back then, and I do recall the game being frustratingly difficult. Rewarding, but difficult.

[Psychotic_Frog]

Just sounds like you're having trouble being as precise, fast and consistent with higher powered cars. Since the cars themselves are much faster you probably have the impression you're going fast, when in reality you're not carrying enough speed in corners.

You could compare your runs with the leaderboard in time trial to see if it's the case; if your runs in Fulvia are always top 5% but you can't break top 15% in the leaderboards that might be the problem.
Or check out some people on Youtube to see how they drive the faster cars and see if you're doing the same thing.

[Xenial Jesse]

I experience the exact same as you Talon. I haven't reached Pro but I believe I could in the H1s; except I can't be bothered sitting through it in a H1. I've just assumed for a long time that part of the game's design was higher class = higher difficulty in the artificial times, but I could be wrong and it could be just me (or us).

So it's a good idea of Frog of using the real leaderboards to have a genuine quantitative measure of things across the different classes.

If the difficulties are out of sync, you'd expect to see the top Clubman AI time for a H1 over a sample of stages to be in the top X% of the real player times, and then the top Clubman AI time for a NR4 be in the top Y%, and the top Clubman AI time for an R5 be in the top Z%, where Z < Y < X.

The big problem however is that the leaderboard interface sucks. There isn't a way to know a percentage for your position, since there isn't a way to see how many records there are on any leaderboard in DR2. You can know that you're 200th in a leaderboard, but you can't tell if that's out of 1,000 or 10,000 other racers.

As far as I can think of, the best we can do is look at the position of the little slider on the right and make very rough guessimates.

[TurbodTalon]

That is actually a really good idea guys. I never thought to use the leaderboards as a comparison.

I did one run each at Norinbee Ridge Ascent / Dry with the Cooper, EVO VI, and Peugeot 205 T16 Evo 2.

- Thrustmaster T300RS GT wheel/pedals
- Soft tires
- Manual sequential
- 1 traction control, 1 stability control

The leaderboard is already divided into 12ths, and that little slider is a visual representation of your placing in comparison to the entire leaderboard. Using some rudimentary algebra, here's what I was able to come up with:

COOPER
Time: 4:07:730
38,769th place out of 259,498 total on leaderboard = TOP 14.94%

EVO VI
Time: 3:33:215
11,936th place out of 41,087 total on leaderboard = TOP 29.05%

PEUGEOT 205 T16 Evo 2
Time 3:43:343
8,666th place out of 23,127 total on leaderboard = TOP 37.47%

My "clean" runs with the Peugeot are clearly not clean, and driving the car is much like wrestling an alligator down the road.

The Peugeot runs are 10 seconds slower than the EVO VI, which is a car with half the horsepower.

It's pretty easy to see now why I can graduate classes so easily with the Cooper.

Hopefully this helps someone else who "thinks" they're making clean runs. Numbers don't lie.

[Psychotic_Frog]

Hey, great that you actually verified! Nice! And good luck trying to tame that Peugeot haha it's quite intense

[TurbodTalon]

The Peugeot is a car that requires a lot of practice that I unfortunately don't have the time to commit these days. I may try a couple more runs with the assists dialed up to max, but then the game starts to feel too "arcade", if that makes sense?

[DevestatinGirth]

hilariously enough I did a full breakdown on this for dirt 4 that I posted on reddit, so I 100% believe it existing in this game aswell.
The biggest thing I remember was if you picked the BMW M3 for H3, the AI ran 12 seconds faster then control leaderboard time I used (which was not the fastest, but had to have atleast 5 different drivers post similar times) at max difficulty. On the easiest difficulty the AI ran within 2 seconds of the control time and if you picked the Sierra Cosworth (the slowest of the H3) they ran 10 seconds slower on max difficulty vs the control time.

[TurbodTalon]

What caused you to do the research?

[DevestatinGirth]

Originally posted by TurbodTalon:

What caused you to do the research?

Just seeing several posts on reddit and the codemaster forum basically like you have done here. Being bored one night I came up with the premise for a trial test, it came back showing there was evidence to the claim so I away I went.

[TurbodTalon]

It seems like Dirty Rally 2.0 is fairly balanced, based on my personal research. If I'm using a car that's difficult to keep on the track (the Peugeot for example), I place lower on the global leaderboard and I lose races.

I wasn't winning any stages with my Mitsubishi Evos at first. But as I got better with them and got them upgraded, my times started to go down, I placed better on the global leaderboard, and I started winning stages. It seems to be pretty linear based on how much time you're willing to dedicate to learning how to drive the more powerful cars.

I should have done a bit more research before I ever made this thread, but I never thought to use the global leaderboard as a test bed.

[Xenial Jesse]

Hi Turbod, testing appreciated because I am too lazy to do this kind of thing myself.
However I am quite confused here so allow me to ask a heap of questions. :)

1.

The leaderboard is already divided into 12ths

I'm not sure what you're referring to here. Where do you see it divided into 12ths?

2.

38,769th place out of 259,498 total on leaderboard

How did you know how many rankings were on the leaderboard?

3.
In this post where you wrote:

Originally posted by TurbodTalon:

I did one run each at Norinbee Ridge Ascent / Dry with the Cooper, EVO VI, and Peugeot 205 T16 Evo 2.

...

What conclusion are you actually saying? I think you are saying that your relative skill is the culprit, not the game's difficulty across vehicles. In other words I think you're saying that the game doesn't make the leagues harder to pass based on vehicle, but rather, you personally just get worse at passing them. Is that right?

Frog appears to be saying "It's you", and Girth appears to be saying "It's the game". I can't tell which angle you're going with.

4.
Not so much a question here but to confirm, your numbers don't appear to directly address the topic as there are no AI times in them.

Your numbers do make a suggestion about your personal skill changing you use different classes, and this does suggest that it's possible that AI difficulty isn't changing, only you.

But to directly question whether the AI does or doesn't run at different difficulties (in a given league) for different cars... I think actual AI times need to be involved.


5.
I think it's worth checking to see if we're thinking the same thing about what difficulty means.

• Note: Since reading Girth's post we have a suggestion that even inside one class (I'll refer to this as intra-class), the difficulty is changing in response to the player's specific vehicle selection. So I think there are 2 ways to question difficulty now: intra-class and inter-class.
  Intra-class difficulty questions difficulty changes going from Sierra Cosworth to BMW M3.
  Inter-class difficulty questions difficulty changes going from H3 to 2000cc.

So do you agree with these understandings in this code box?

For intra-class difficulty, we are saying something is more difficult if smaller times are being set by the AI.

For inter-class difficulty, we are saying something is more difficult if a smaller proportion of human players are beating it.

Both sentences above assumed a fixed stage, wet/dry condition, Career Rally league, and class (for intra-class questioning).

It is note-worthy that we can determine whether there are intra-class difficulty adjustments completely objectively; no human results required. But not so when questioning whether there are inter-class difficulty adjustments.

I wrote up some much wordier definitions that I don't want to delete now! :D If you can be bothered with this wall of text, see if they still sound good to you.

Intra-Class (Per-vehicle)

Inside a single class, AI difficulty is changing in response to the player's selected vehicle if the AI times on the given stage are consistently better when the player selects one vehicle, compared to when the player selects a different vehicle.

 --> Different difficulty means different AI times in response to different vehicle selection.

Inter-Class (Across classes)

Across different classes, AI difficulty is changing in response to the player's selected class if the AI times on the given stage are consistently beating a lower proportion of players in one class than in another class.

 --> Different difficulty means different proportions of human players beating the AI times over different classes.

Special Notes:

1. If we have already found that intra-class difficulty is changing with vehicle selection, then determining whether inter-class difficulty is also changing with class selection becomes much messier, and requires comparisons across multiple vehicles in each class; enough to ensure that we aren't simply mistaking intra-class difficulty adjustments for inter-class difficulty adjustments.
   
2. Assuming same Career Rally league, same stage, same wet/dry conditions.
   
3. Note that Career Rally uses inconsistent environments on the same leaderboards. Full degradation in the rain at night uses the same leaderboard as zero degradation in the wet with no rain in the day. This may or may not matter when trying to test the above definitions. It also may or may not matter because the above definitions are already probably too much for anyone to bother testing. :D

@Girth

Just to confirm one thing...

With the H3s in DiRT 4, on max difficulty, with a control time already set, you found that:
* if the player selected the BMW H3, the AI ran 12 seconds faster than the control time
* if the player selected the Sierra Cosworth, the AI ran 10 seconds slower than the control time

Have I got that right? Interesting if so. I was thinking there was a question of difficulty change per class selected only, not per vehicle selected.

[Xenial Jesse]

Ok I think I'm actually done editing that now.

[Robininthehoods]

Originally posted by Xenial Jesse:

Frog appears to be saying "It's you", and Girth appears to be saying "It's the game". I can't tell which angle you're going with.

I was thinking there was a question of difficulty change per class selected only, not per vehicle selected.

tldr, I'm not sure if the difficulty changes per class or per vehicle but I'm pretty sure that both Frog and Girth are right. Its you, also, its the game :))

[TurbodTalon]

Originally posted by Xenial Jesse:

Hi Turbod, testing appreciated because I am too lazy to do this kind of thing myself.

However I am quite confused here so allow me to ask a heap of questions. :)

1.

The leaderboard is already divided into 12ths

I'm not sure what you're referring to here. Where do you see it divided into 12ths?

1A. It only shows 12 rankings, and it also shows your relative position (little slider bar) to the rest of the leaderboard. It would help a lot if I could attach a screen shot.

2.

38,769th place out of 259,498 total on leaderboard

How did you know how many rankings were on the leaderboard?

2A. Based on my ranking and the position of the little slider bar. For example, if I was ranked 1,000th place, and the slider bar was exactly 50% between the top and bottom, we can assume that there are 2,000 total on the leaderboard. Again, it would help a lot if I could post a screen shot. If you've got some time to burn, you can also scroll down to the bottom of the leaderboard manually.

3. In this post where you wrote:

Originally posted by TurbodTalon:

I did one run each at Norinbee Ridge Ascent / Dry with the Cooper, EVO VI, and Peugeot 205 T16 Evo 2.

...

What conclusion are you actually saying? I think you are saying that your relative skill is the culprit, not the game's difficulty across vehicles. In other words I think you're saying that the game doesn't make the leagues harder to pass based on vehicle, but rather, you personally just get worse at passing them. Is that right?

Frog appears to be saying "It's you", and Girth appears to be saying "It's the game". I can't tell which angle you're going with.

TURBODTALON: My skill level is the main culprit, and I think that more testing with AI stage times will further enforce that.

3A. The stage times got progressively worse in direct relation to the car I was driving. I did not delve into AI stage times, only myself versus the leaderboard. Therefore, my conclusion is that MOST of the problem is my lack of skill with the more powerful cars.

4. Not so much a question here but to confirm, your numbers don't appear to directly address the topic as there are no AI times in them.

Your numbers do make a suggestion about your personal skill changing you use different classes, and this does suggest that it's possible that AI difficulty isn't changing, only you.

But to directly question whether the AI does or doesn't run at different difficulties (in a given league) for different cars... I think actual AI times need to be involved.

4A. Yes, I am saying that, according to my limited research using only human leaderboards, my skill level is the culprit. As I mentioned in 3A, I did not do any research using AI times. I do plan on following up with AI times at Norinbee Ridge Ascent though.

5. I think it's worth checking to see if we're thinking the same thing about what difficulty means.

• Note: Since reading Girth's post we have a suggestion that even inside one class (I'll refer to this as intra-class), the difficulty is changing in response to the player's specific vehicle selection. So I think there are 2 ways to question difficulty now: intra-class and inter-class.
  Intra-class difficulty questions difficulty changes going from Sierra Cosworth to BMW M3.
  Inter-class difficulty questions difficulty changes going from H3 to 2000cc.

So do you agree with these understandings in this code box?

For intra-class difficulty, we are saying something is more difficult if smaller times are being set by the AI.

For inter-class difficulty, we are saying something is more difficult if a smaller proportion of human players are beating it.

Both sentences above assumed a fixed stage, wet/dry condition, Career Rally league, and class (for intra-class questioning).

It is note-worthy that we can determine whether there are intra-class difficulty adjustments completely objectively; no human results required. But not so when questioning whether there are inter-class difficulty adjustments.

I wrote up some much wordier definitions that I don't want to delete now! :D If you can be bothered with this wall of text, see if they still sound good to you.

Intra-Class (Per-vehicle)

Inside a single class, AI difficulty is changing in response to the player's selected vehicle if the AI times on the given stage are consistently better when the player selects one vehicle, compared to when the player selects a different vehicle.

 --> Different difficulty means different AI times in response to different vehicle selection.

Inter-Class (Across classes)

Across different classes, AI difficulty is changing in response to the player's selected class if the AI times on the given stage are consistently beating a lower proportion of players in one class than in another class.

 --> Different difficulty means different proportions of human players beating the AI times over different classes.

Special Notes:

1. If we have already found that intra-class difficulty is changing with vehicle selection, then determining whether inter-class difficulty is also changing with class selection becomes much messier, and requires comparisons across multiple vehicles in each class; enough to ensure that we aren't simply mistaking intra-class difficulty adjustments for inter-class difficulty adjustments.
   
2. Assuming same Career Rally league, same stage, same wet/dry conditions.
   
3. Note that Career Rally uses inconsistent environments on the same leaderboards. Full degradation in the rain at night uses the same leaderboard as zero degradation in the wet with no rain in the day. This may or may not matter when trying to test the above definitions. It also may or may not matter because the above definitions are already probably too much for anyone to bother testing. :D

@Girth

Just to confirm one thing...

With the H3s in DiRT 4, on max difficulty, with a control time already set, you found that:
* if the player selected the BMW H3, the AI ran 12 seconds faster than the control time
* if the player selected the Sierra Cosworth, the AI ran 10 seconds slower than the control time

Have I got that right? Interesting if so. I was thinking there was a question of difficulty change per class selected only, not per vehicle selected.

[TurbodTalon]

Here's more data with AI and Top leaderboard times included. The AI strangely seems to lag behind the fastest human on the leaderboard by ~20%, regardless of car or class.

In conclusion, I have no reason to believe there is a best car or car class; Dirt Rally 2.0 simply requires that you learn how to drive your car. Pick a car and purchase its upgrades. With enough time and practice you will start winning races.

And OH MY GOODNESS...that R5 class just handles like a dream, even with no assists or upgrades. I highly recommend getting one once you can afford to. It'll probably help in the higher classes.

Thrustmaster T300RS GT wheel/pedals @ 720° rotation
Manual sequential (paddle shifters)
Soft tires
No assists are used unless otherwise noted

MINI COOPER (H1 FWD)
NOORINBEE RIDGE ASCENT, AUSTRALIA DAYTIME/DRY
BEST LEADERBOARD TIME: 3:16:981
MY TIME: 3:54:567 (1ST PLACE)
BEST AI TIME: 4:08:076
AI is 20.6% slower than #1 on the Global leaderboard.
****************************************************
OPEL ASCONA 400 (H3 RWD)
MOUNT KAYE PASS DAYTIME/DRY
BEST LEADERBOARD TIME: 7:16:082 436
MY TIME: 9:52:295 (28TH PLACE)
BEST AI TIME: 7:52:523 473
AI is 7.9% slower than Top leaderboard.
****************************************************
MITSUBISH EVO VI (GROUP A)
NOORINBEE RIDGE ASCENT, AUSTRALIA DAYTIME/DRY
BEST LEADERBOARD TIME: 2:52:629
MY TIME: 3:28:462 (3RD PLACE)
BEST AI TIME: 3:27:999
AI is 16.8% slower than #1 on the Global leaderboard.
****************************************************
PEUGEOT 205 T16 EVO 2 (GROUP B 4WD)
NOORINBEE RIDGE ASCENT, AUSTRALIA DAYTIME/DRY
BEST LEADERBOARD TIME: 2:47:308
MY TIME: 3:35:351 (4TH PLACE)
BEST AI TIME: 3:30:014
AI is 20.5% slower than #1 on the Global leaderboard.
****************************************************
MITSUBISHI EVO X (NR4/R4)
NOORINBEE RIDGE ASCENT, AUSTRALIA DAYTIME/DRY
BEST LEADERBOARD TIME: 2:57:29
MY TIME: 3:26:370 (1ST PLACE)
BEST AI TIME: 3:39:788
AI is 19.5% slower than #1 on the Global leaderboard.
****************************************************
MITSUBISHI EVO X (NR4/R4)
KREUZUNGSRING REVERSE, GERMANY WET
BEST LEADERBOARD TIME: 2:44:068 164
MY TIME: 3:09:275 (1ST PLACE)
BEST AI TIME: 3:19:403 199
AI is 17.6% slower than #1 on the Global leaderboard.
****************************************************
MITSUBISHI EVO X (NR4/R4)
VALLE DE LOS PUENTES, ARGENTINA DAY/DRY
BEST LEADERBOARD TIME: 5:25:670
MY TIME: 6:27:897 (2ND PLACE)
BEST AI TIME: 6:25:305
AI is 15.4% slower than #1 on the Global leaderboard.
****************************************************
VOLKSWAGON POLO GTI R5 (R5)
NOORINBEE RIDGE ASCENT, AUSTRALIA DAYTIME/DRY
BEST LEADERBOARD TIME: 2:51:536 172
MY TIME: 3:26:271 (1ST PLACE)
BEST AI TIME: 3:31:224
AI is 18.5% slower than #1 on the Global leaderboard.
****************************************************
CHEVROLET CAMARO GT4.R (R-GT)
NOORINBEE RIDGE ASCENT, AUSTRALIA DAYTIME/DRY
STABILITY CONTROL = 2
BEST LEADERBOARD TIME: 3:00:170
MY TIME: 3:49:307 (1ST PLACE)
BEST AI TIME: 3:49:384
AI is 21.4% slower than #1 on the Global leaderboard.

[TurbodTalon]

It appears that I again spoke too soon.

In Clubman with the Mitsubishi Evo X, the AI lagged roughly 20% behind the top time on the global leaderboard. I have a ton of data to back that up.

Now, in Pro, also with the Mitsubishi Evo X, the AI consistently lags 10-12% behind the top times on the global leaderboard. I've got a decent bit of data now to back this up as I take screenshots at the end of every couple of stages. Needless to say, my placings are getting progressively worse as well.

I'll update the percentages for Elite and Masters when I get to them, although I may need to move to a less powerful car.

I also had to switch to the Cooper to finish Dirty Rally 1 on Masters difficulty.