A matter in rdps calculation

Hi there, thanks to maintainers’ efforts to rdps statistics in FFlogs. The method that divides rdps contribution among damage buffs is truly smart.

Nevertheless, a player discovered a matter in the method that divides rdps between damage buffs and crit (or direct hit) buffs.

In conclusion, current method tends to underestimate contribution from crit and direct hit buffs while overestimate damage buffs when they are combined. He presented here.It’s in Chinese, I will summary here if you would.

Considering two conditions

C1.Mary, whose crit rate is 160% and base crit chance is 30%, casts two Glare, both would hit for 1000 points if Mary had no buff. But a Glare is buffed by Trick Attack, the other Glare is buffed by Chain Strategem and is crit due to the buff. So Mary hits for 1000×1.1+1000×1.6=2700 in total.

C2.Mary also casts two Glare, this time a Glare is buffed by both Trick Attack and Chain Strategem and is crit while the other one recevied none. So Mary hits for 1000×1.1×1.6+1000=2760.

In C2 Mary hits 60 points more, it explains why we should allign several buffs —— the extra hit points are from combination of buffs, thus they should be divided among these buffs. But current method owes
them all to Trick Attack.

That says, Trick Attack takes 1760/1.1×0.1=160 rdps first, Chain Strategem takes 150 rdps then in C2. But in C1, where Trick Attack and Chain Strategem take effect separately, Chain Strategem still takes 150 rdps while Trick Attack only takes 100 rdps. Trick Attack takes all extra 60 points from combination of two buffs. It seems Chain Strategem suffers unfair treatment in such method. So as other direct hit and crit buffs when they are combined with damage buffs.

We put forward a modificatory method when a crit or direct hit event happens.

In C2, when the 1760 hit occured, we divide it into two parts.

The first part is 30%/(10%+30%)=3/4 of the 1760 points (1320), which are independent of Chain Strategem, it is caused by Mary’s own crit attributes. So Trick Attack takes 1320/1.1×0.1=120 rdps here.

The second part is the other 1/4 of the 1760 points(440), which shall be regarded as crit Chain Strategem brought. In this part, we take Chain Strategem as a 60% damage increse buff. So Chain Strategem takes 440/1.76×0.76×lg(1.6)/lg(1.76)=157.97 rdps here, while Trick Attack takes 440/1.76×0.76×lg(1.6)/lg(1.76)=32.03.

In this method, Trick Attack takes 120+32.03=152.03 rdps in total, Chain Strategem takes 157.97, Mary has 1450 remains as her own rdps. By the means it looks as if we succeed in alloting extra points
between Chain Strategem and Trick Attack.

Look forward to your comments

Yeah, I agree with this math and was already considering making this change.

Note that I can’t really recrawl all of Eden Savage at this point, so this would be a change to consider for the next savage tier in 5.2.

1 Like

In a similar logic, let me introduce Gary to you. Gary have a 1-min CD that increases his damage by 20%.

G1. Gary strikes for two times: one is buffed by trick shot and another is buffed by his own CD. Gary deals 1000 * 1.1 + 1000 * 1.2 = 2300 damage.

G2. Gary learns to align buffs and again strikes for two times: now, one is buffed by both trick shot and his own CD. Gary now deals 1000 * 1.1 * 1.2 + 1000 = 2320 damage.

Gary and the ninja worked together for an extra 20 damage, but with current rdps calculation, the ninja would take all 20 as his/her rdps and brags about it.

Gary is upset. :frowning:

It is much trickier to deal with personal buffs. Some personal damage buffs have very high uptime (e.g., Greased Lightning), and it becomes nonsensical to try to normalize a personal buff like Greased Lightning with external buffs.

So while the personal buffs could possibly be included, there would need to be a careful evaluation of every single personal buff to decide if it should actually be included in the normalization or not. This was tricky enough that I opted not to for now.

It’s cheerful to hear that. May the Twelve bless you.:grin:

I have implemented these changes (both the crit/dh changes and the inclusion of personal buffs in the normalization). I left out a handful of personal buffs that are always up.

I am targeting 5.1 with these changes, so when 5.1 goes live, logs uploaded with it will have the new calculations.

1 Like

Don’t really agree with this. Different jobs have different ways of personally aligning with raid buffs. While some classes may do this with a personal buff, other classes do so by pooling a resource (mana, job gauge), or by saving ogcds. It is impossible to normalize all these different methods of self-buffing/bursting, so I’d rather rDPS be left consistent and accurate. To reward self-alignment in the proposed method would make rDPS inaccurate and inconsistent.

I think the buff giver(s) always deserve full credit for multiplicative damage.

The rationale is:

  1. rDPS, by design, does not reward a player for hitting positionals, saving ogcds or using higher potency skills during external buffs. Doing more damage under the buff rewards only the buff giver.

  2. Self buffs are a part of a job’s kit that result in higher potency skills. So self-buffed attacks are equivalent to hitting a positional or using a higher potency attack etc.

  3. It would be inconsistent to reward the player for using higher potency attacks in #2 while not rewarding them for the higher potency attacks in #1.

I am inclined to agree with mint here and revert the change to include personal buffs in the normalization in 5.1.

As he has pointed out, people have completely different ways of increasing DPS during an external buff window, and it seems better not to reward one kind of increase (a personal % buff) over another.

I’ll wait and see if anyone else has feedback on this though.

It’s really hard to form a fair index that demonstrates “real” dps.
Usually we refer to “dps” in such situations:

  1. Compare damage contribution from different jobs.
  2. Compare damage contribution from different players in same party (but different jobs).
  3. Compare damage contribution from different players in same job.

Current rdps works well in #1, while adps is more suitable to #3. As for #2……let it be their party’s “internal affairs”?:joy:
Dose FFLogs need to tell people use different index according to their cases?