BGWiki PDIF Clarification

I've been trying to fully understand the pDIF calculations recently and in doing so I think I've found one error and an additional lack of clarity in the BGwiki PDIF page. I wanted to ask for a more seasoned eye to double check me before making any changes.

The error is at the bottom of the qRatio section where it has you reduce any value above 3 to 3.
Seeing as the caps are all above 3 now I have to assume that this is a remnant of old pDIF calculations. Should it be altered to refer to the 3.25 1H / 3.5 H2H / 3.75 2H chart?


My second point has to do with when the cap is applied in the qRatio calculation. The current layout suggests that they are applied immediately to Upper Limit (UL) and Lesser Limit (LL), but the DPS Spreadsheets suggest otherwise.
I think they are only applied once after "randomly select a value between LL and UL."

Let's say you are using a 1H weapon with a wRatio of 3.25. Without an immediate cap your UL will be 3.625 and your LL will be 2.875. When you go to select a random number between those two, you'll end up with a 50% chance (in this example) of getting a number >= 3.25, which would be reduced to 3.25 by capping on the next step. Conversely, if you cap the UL to 3.25 and then select the random number, you'll only ever hit 3.25 when you get a rare max roll.


If I'm correct in my above points, my proposed change to the page would be the following.

Thanks in advance on this admittedly technical topic.

yes it is.

not quite, neither lower or upper limit can go above the cap for the weapon. let's say you have a ratio of 3 for 1H, then your range of values is 2.625 and 3.375, if you have a ratio of 3.25 then it's 2.875 to 3.25, since you can't go above 3.25. If your ratio is 3.625 or greater, which is the ratio effective cap for 1H (excluding new neck,) then your values are between 3.25 and 3.25. You then multiply your base damage by the random number between the lower and upper values.


Which if attack capped is 3.25, you floor it, then multiply by a random value between 1 and 1.05, so that should be changed as well, since I believe it says to multiply the value between lower and upper by the random when it should be floored beforehand.

Do you have any evidence to support this? I feel like it's unclear on the page and there's no distribution data referenced (or if it is, it's buried in some old topic.) I've more or less forgotten which way it works or if it was ever verified.

I'm also not sure you truly see what he's asking about. Take your 1 handed ratio of 3, for example. There's two models you could use to calculate pDIF based on the page.

min(rand()*(3.375 - 2.625) + 2.625, 3.25)*(rand()*.05 + 1)

or

(rand()*(3.25 - 2.625) + 2.625)*(rand()*.05 + 1)

The former approach would produce higher average pDIF values although you've suggested it's the latter.

its more likely the latter, motenten had numbers somewhere when they initially changed. i can't find a link offhand, but i do recall him saying he'd do 1.02 instead of 1.025 for his spread sheet averages to try to replicate it. and considering the several hours i spent smacking level 1 mobs the other day and didn't see as many max rolls as the first would suggest, I'd also lean towards the latter. also the random 1 to 1.05 is after you floor your damage.

floor(floor(x * 3.25) * uniform(1, 1.05))

It's not going to matter on level 1 mobs, you're way past the effective ratio cap. It's only on circumstances like above where your attack/defense ratio is near the effective ratio cap but not there.

i was on a level 8 monk recently to confirm that damage is floored before the secondary random number after getting results that it is on a 99. i can go out and get data off greater colibri this evening, though

I think you're still missing the point. In order to answer this question you will a good sample size in controlled near but not capped conditions and you're not looking at caps you're looking at averages.

no i get your point. i can easily be near capped on greaters

Alright. Seems like it would require a large sample size. With a cratio of exactly 3, you'd be trying to identify between a mean pDIF of 3.064~ and 3.012~, less than a 2% difference.

Thanks for the discussion guys. I figured the 2nd bit might not have a known answer since it only comes into effect when UL is at (or above) cap and LL is not.


I've noticed the 1.02 in Motenten's spreadsheets and wondered about that too. To me, that would suggest a non uniform spread in the 1 to 1.05 multiplier considering how he calculates things before that.

In case it's helpful I'll detail the way he does it below.

((3.25 * (1-Cap Range/Full Range)) + (1-(1-Cap Range/Full Range)) * ((UL + LL) / 2)) * 1.02

• The 3.25 is weapon cap (1H in this case).

• Full Range = uncapped UL - LL, generally equaling 0.75.

• Cap Range = capped UL - LL. In case it wasn't clear, he has cells with both uncapped UL and capped UL.

• ((UL + LL) / 2) uses capped UL & LL values.

Doing it this way should give the same result as the method I first suggested, it just gets there in a different way. Taking the same 3.25 wRatio example from before you'll end up with :
((3.25 * 50%) + 50% * ((3.25 + 2.875) / 2)) * 1.02

Any actual testing would be appreciated Austar.