WoW rogue formulae

These have been gleaned from various sources, but most I have been verified personally. I'm always interested in confirmation of course, but remember to do lots of testing and not just complain that in one fight your DPS differed etc.

Normalised_delay here means 1.7 for daggers or 2.4 for swords and maces. Otherwise the standard weapon delay is used. Damage is a hit within the range quoted on the weapon itself. Eg 54 to 101 inclusive for Felstriker. For DPS calculations I use the average. WoW looks to pick a flat distribution across the entire available range.

Standard hit:
damage + attack*delay/14

Sinister strike:
(1 + .02*Aggression) * (damage + attack*normalised_delay/14 + 68)

Hemorrhage:
damage + attack*delay/14 + 7

Note: Here I add the +7 assuming that if you've done one hemorrhage and it's landed then you'll likely be doing more and keeping the +7 topped up. Constant spamming of Hemorrhage instead of Sinister Strike therefore should be viewed with the +7 added. Also note that the real delay is used and not the normalised delay. This is why some rogues are switching to hemorrhage for big dps and slow weapons.

Backstab:
(1 + .04*Opportunity) * (1.5 * (damage + attack*normalised_delay/14) + 210)

Ambush:
(1 + .04*Opportunity) * (2.5 * (damage + attack*normalised_delay/14) + 290)

Note: Backstab and (untested) Ambush have the extra 20% from Oppertunity rank 5 applied to the *entire* formula. This means the tooltip of 180% + 210 for backstab in game is wrong - it is 180% + 252. I have parsed this against many level 1 mobs with various attack powers and weapons to verify that this is indeed how it works.

Swing delay:
I decided for ease of displaying average and max hits to model the Sword Specialisations extra hits as an averaged way of reducing the weapon delay. I am not sure this is entirely true as I hear that possibly the extra hits always come from the main hand even if the "proc" (if you view the random chance in that manner) may be triggered from the offhand. I do not use this talent myself currently so this is untested.

How to interpret the results

It takes into account your average weapon speed, attack power and the type of attack to predict an average non-crit attack, possibly boosted by talents too (eg Opportunity). From this it then works out what the avg crit is and hence, given the crit percentage a total average hit with that type of attack. It's pretty straight forward to then work out the DPS for that type of attack.

Points to consider:

There's no time for Feint here. Do too much DPS and you're dead, or at the very least for a dagger rogue you'll have aggro and be doing less DPS.
This is sustained dps. Therefore it's no use commenting that your favourite ambush + cold blood backstab isn't factored in; it's not what I was trying to calculate. Think of it as PvE stats for raid bosses.
There's no armour mitigation. I think the formula for this is AC/(AC+85*level+400), but as it's a fixed percentage it's just a scale factor so it has no relevance on comparing one talent setup to another.
It compares throwing all combo points into Eviscerate vs building up 5 point SnD and throwing any remaining combo points we can build up into Eviscerate. In the second mode it won't spend any more points on Eviscerate than it can, allowing the appropriate time for building up SnD again so as to maximize the time it is up. For many weapon combinations the exact maximum of DPS is somewhere between the two models (eg 3 point SND and the rest on Eviscerate, or 5 point SnD and 5 point Eviscerate ignoring the fact that SnD runs out part way).
Damage isn't everything. Sometimes control (eg stuns, interrupting spells, etc) is more important.
It may all of course be completely wrong. I don't think so, but feel free to study the code too if you care.