Now about Variation

So ever since I was first introduced into GP4 I have had a nagging question over and over.

Variance: how does it work exactly?

And I ask this knowing that possibly no one has the answer. But here goes nothing

Does it work on a 0->x scale?
So if base skill is 16000, 250 Var= Min. 16000, Max. 16250?

Or is it similar to variance in a bell curve (mathematics) style?
base skill 16000, 250 Var=Min. 15750, Max. 16250?

Or, and this one probably scares me the most, am I completely wrong and it is neither of those?

Now I am scared that I've missed something fairly basic somewhere :|

I would have thought it would be, as in your example above, base skill of 16000 with a variance of 250, would be plus or minus 250 from 16000.

So in essence, it would be in a range between 15750 to 16250.

Probably am wrong though.

Slasher Wrote:
-------------------------------------------------------
> I would have thought it would be, as in your
> example above, base skill of 16000 with a variance
> of 250, would be plus or minus 250 from 16000.
>
> So in essence, it would be in a range between
> 15750 to 16250.
>
> Probably am wrong though.


Yeah, I agree, that is how I have always thought that it worked. But I've seen some instances where it doesn't seem to have worked like that. Of course, they may have just been very lucky that they are always getting positive variation

I think it's the first. It's a variance of 0 to x.

=====================================================


Intel NUC 8i3, 8GB RAM, MS Sidewinder Wheel

Have to say I've always wondered this as well.

I've always assumed it worked on a 0 to x scale but really is there anyway of actually finding out one way or the other?

Interested in managing an F1 team? Sign up for the Grandprixgames F1 Manager game here

According to Nicola Acciarri it's the second one: [www.grandprixgames.org]

Mal claims in this post that it's the chance of breakdown: [www.grandprixgames.org]
(gitanes saying the same in the last post in the thread)


I always assumed it was the second one, not sure about that anymore...

---

Some mods
F1 1996 | F1 2002 | F1 2007 | F1 2011 | F1 2013 | F1 2015 | F1 2018

Pretty sure that three-digt number is perf variance whilst the four-digit number for a team is chance of breakdown. That's hlw I seem to remember it from using GP3Edit

---

GPGSL -
GPGSL-3 - Pizza Party Racing manager and driver
Nations Cup - Team Scotland manager

smoglessbutton4 Wrote:
-------------------------------------------------------
> Pretty sure that three-digt number is perf
> variance whilst the four-digit number for a team
> is chance of breakdown. That's hlw I seem to
> remember it from using GP3Edit


this is true. Teams have failure chance and drivers performance variation. I´m also quite sure that the variation is from 15750 to 16250, if performance is 16000 and variation is set to 250.

I haven´t tested if it´s but sometime in the past race and quali performances were in opposite way in team editor, when compared to other similar editors or to perf.txt files.

Lo2k or SDI will likely now a lot more about the performance stuff then many other do.



Edited 1 time(s). Last edit at 05/21/2015 12:28PM by vesuvius.

As you can see from here, there is a lot of speculation, not as much rock solid assurance! Here's hoping we do find out properly

I've run some very basic tests on variance just to see what the actual effect is, basically adjusting it around and seeing what laptimes come out in qualifying.

From what I can see the scale is 0 to x, whenever the variance is above 0 the laptimes don't drop below the 0 variance times.
The higher the value the less consistent the laptimes are, although I think that one was pretty obvious.
A higher variance seems to increase the base laptime, I expected it would be possible to go faster with higher variance but it seems it's guarenteed the higher the number, ie:

Driver Variance: 0 Mean Laptime: 1:34.132 Deviation: 0.087s
Driver Variance: 250 Mean Laptime: 1:33.737 Deviation: 0.168s
Driver Variance: 500 Mean Laptime: 1:33.323 Deviation: 0.180s
Drive Variance: 1000 Mean Laptime: 1:32.844 Deviation: 0.202s

Thats literally churning out numbers, I'll confess to no knowledge of the inner workings of the game!

_________________________________________________

For a list of EVERY download for GP4, look here: [docs.google.com]

I always thought it's 15750-16250

---

GPGSL - Yakuza driver

GPGSL activity check app: Direct link - Source code - Have you posted?

0->x seems to be correct. Simulated a few quali sessions, with performance of 16000 and different variance values. The higher the variance, the better the laptimes. Always. I also got the impression that it's not 0->x, but rather x. The order of the drivers was always the same.

---

Some mods
F1 1996 | F1 2002 | F1 2007 | F1 2011 | F1 2013 | F1 2015 | F1 2018

Overall, I think the big takeaway from these simulations is that increasing the variance not only increases the spread in a given driver's laptime, but also increases their mean laptime as well - so one has to adjust for this in the base skill value as well.

A huge thank you for truecrysis and kedy89 for carrying out the actual tests.

---

My workthread - [www.grandprixgames.org]
Full of classic F1/non-F1 track layouts

My blog about F1 performance analysis - [thef1formbook.wordpress.com]

This is some serious procrastination for me now because screw revising for exams

Some rough estimates for how variance affects performance:

Average pace increases by about 2/3 of the variance value. So a variance of 500 equates to about 330 increased performance.
From this point the times seem to vary about 1/7th each way, with perhaps an ever so slight bias towards slower (i think variance is logarithmic which would explain this, ie raising the number has less of an effect the high you go)

So in figure terms, 16000, 500 equates to about 16280-16380
Similarly 16000,250 gives you 16140-16190. This will probably change at different performance levels as perf isn't exactly a linear scale.

Maybe i'll start studying now...

_________________________________________________

For a list of EVERY download for GP4, look here: [docs.google.com]

@truecrysis: If you'll be doing as well on the exam preparations and on the actual exam as you did here, then you'll surely pass with a fine grade.

EDIT: Thank you, once more.

---

My workthread - [www.grandprixgames.org]
Full of classic F1/non-F1 track layouts

My blog about F1 performance analysis - [thef1formbook.wordpress.com]



Edited 1 time(s). Last edit at 05/21/2015 03:58PM by Atticus..

Well that is food for thought now, very nice work gentlemen, pat on the back. :D

after reading some comments about this on racedepartment..
shouldn't this be called "agressiveness" then?

higher value = The more the driver pushes the car, he is faster but more prone to making small mistakes, hence the faster laptimes due to drivers pushing a lot more, and higher variance due to the small mistakes.

---

GPGSL - Yakuza driver

GPGSL activity check app: Direct link - Source code - Have you posted?



Edited 1 time(s). Last edit at 05/23/2015 02:57PM by miki2000milos.

Makes sense.

---

My workthread - [www.grandprixgames.org]
Full of classic F1/non-F1 track layouts

My blog about F1 performance analysis - [thef1formbook.wordpress.com]

This is what I can read from the executable:

For GP4 it's the first one, for GP3 it's the second one.

In GP4 the variance is multiplied by a random number between 0.0 and 1.0, and then added to the ability. 25% of the random part is determined at the start of the weekend (using the qualify variance), 75% for each other session (in qualify even every time you return to the pits).

In GP3 the random number is between -1.0 and 1.0, and 62.5% is already determined at the start of the weekend.

The resulting ability is then multiplied by the track grip factor.

Note that 16384 (0x4000) means 1.0. The game uses fixed point calculations with a precision of 1/16384 instead of floating point, hence these numbers show up in editors.

René Smit, Independent Software Developer.
Download my GPx tools here.



Edited 1 time(s). Last edit at 06/21/2015 01:45PM by SDI.

That is.... interesting, thought provoking...
If anything there is more questions out of that, but thanks very much Rene!