Jump to content


Photo

SimulationCraft for Rogues


  • This topic is locked This topic is locked
268 replies to this topic

#261 Brone

Brone

    Glass Joe

  • Members
  • 6 posts

Posted 30 July 2010 - 05:11 PM

Well, I believe I have resolved the issue. It looks like SimulationCraft is using a 2 roll structure to evaluate special attacks, where the first roll decides whether the attack is a crit, and the second roll decides whether the mob dodges the attack. The toon that I was simming had 0 expertise, and thus was being dodged 5% of the time (against a lvl 80 mob). Roughly half of those dodges were consuming a crit determined on the previous roll, thus leading to my lower than expected crit rate.

I looked up the two-roll theory on Wowwiki and noticed that it claims that crits are evaluated on the SECOND roll. So is Wowwiki wrong or SimulationCraft wrong?

-Brone

#262 Aldriana

Aldriana

    Mike Tyson

  • Moderators
  • 13,510 posts

Posted 30 July 2010 - 05:27 PM

Short answer is that it doesn't actually matter which is which. Consider:

For simplicity, lets say our crit rate is 50% and our miss/dodge rate is 10%. If we do the miss/dodge roll first, we then get 90% hits and 10% dodges; then half of those 90% that hit crit, so we get 45% crit, 45% regular hit, and 10% dodge.

If, on the other hand, we do the crit roll first, we get 50% crit and 50% hits, and then 10% of each miss, leaving us with... 45% crit, 45% regular hit, and 10% dodge.

Hence: they're both right, and are saying the same thing. You can think about it in either order, and the math works out.

#263 Anthropology

Anthropology

    Glass Joe

  • Members
  • 20 posts

Posted 08 August 2010 - 03:17 AM

From the GUI: When you Import a character from the Armory or from Wowhead, it moves you to Simulate tab where you will see a wall of text. Included here will be a talent link, a list of glyphs, a list of gear, and a list of lines that start "actions+=".

Comment out the line of interest by inserting a '#' character at the beginning of the line.


I got very strange results with this, and found that deleting "&dot.rupture.remains" from two of the eviscerate lines brought the DPS much closer to that of rupture cycle simulations.

#264 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 08 August 2010 - 10:52 PM

I got very strange results with this, and found that deleting "&dot.rupture.remains" from two of the eviscerate lines brought the DPS much closer to that of rupture cycle simulations.


Excellent point. I had not considered that the Rogue action pseudo-AI had rupture-related conditionals on the eviscerate logic.

#265 Brone

Brone

    Glass Joe

  • Members
  • 6 posts

Posted 29 September 2010 - 08:23 AM

What is the difference between the variables 'hat_donor' and 'honor_among_thieves' in SimulationCraft? They are both listed as procs.

I am running a simulation with only 1 person (myself). hat_donor seems to correctly list the number of HAT procs under the Count column, when I am specced 3/3 in HAT. However, if I go 1/3 or 2/3, the count doesn't seem to get reduced correctly. Alternatively, the honor_among_thieves variable count seems too high to represent the number of HAT procs. I have tried really hard to figure this out on my own and haven't been able to. In the source code it reads as if hat_donor counts the number of specials and honor_among_thieves counts the number of HAT procs but that is not what I am seeing.

#266 todemax

todemax

    Piston Honda

  • Members
  • 127 posts

Posted 29 September 2010 - 09:29 AM

There's a one second cd on HaT procs. So hat_donor would be the abilities that actually can proc HaT. And as far as I can see the honor_among_thieves variable is used to track the one second cooldown. The 1/3 chance of proccing HaT with 1/3 points are included in the rng to determine if a combo point is gained and not in the hat_donor.

#267 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 30 September 2010 - 07:54 PM

What is the difference between the variables 'hat_donor' and 'honor_among_thieves' in SimulationCraft? They are both listed as procs.

I am running a simulation with only 1 person (myself). hat_donor seems to correctly list the number of HAT procs under the Count column, when I am specced 3/3 in HAT. However, if I go 1/3 or 2/3, the count doesn't seem to get reduced correctly. Alternatively, the honor_among_thieves variable count seems too high to represent the number of HAT procs. I have tried really hard to figure this out on my own and haven't been able to. In the source code it reads as if hat_donor counts the number of specials and honor_among_thieves counts the number of HAT procs but that is not what I am seeing.


Looks like the code simply checks for the existence of that HAT talent without actually looking at the number of ranks. This is a bug. Please open an Issue on the site.

Note that "hat_donor" refers to the actor generating HAT procs for Rogues in his party. In single-player sims, the option "critical_strike_intervals=A/B/C/D" represent average intervals between critical strikes for four "virtual" players. If you run with multiple players and put them in the same party, then the "critical_strike_intervals" option is ignored and the Rogue receives HAT procs from his party members.

#268 todemax

todemax

    Piston Honda

  • Members
  • 127 posts

Posted 01 October 2010 - 08:31 AM

Doesn't
if ( ! rogue -> rng_honor_among_thieves -> roll( rogue -> talents.honor_among_thieves / 3.0 ) ) return;

    add_combo_point( rogue );
check for the ranks in HaT?

#269 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 02 October 2010 - 03:39 AM

Doesn't

if ( ! rogue -> rng_honor_among_thieves -> roll( rogue -> talents.honor_among_thieves / 3.0 ) ) return;

    add_combo_point( rogue );
check for the ranks in HaT?


Argh.... Yes. That code still gets triggered even with critical_strike_intervals.

I have no explanation then. I'll have to exercise the code before I make any more snap judgements.




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users