SimulationCraft Model Development - Page 21

foopeemoa · 10/20/09, 11:46 AM

Originally Posted by dedmonwakeen

Actually, the weapons will be sync'd by default...... I need to add an option to go the other way....

Interesting, the sim is apparently much better at managing rage than I am in that situation. Doesn't the game by default delay the offhand attack by 0.5 second (value needs verification) if you start attacking while in melee range of the target? I believe if you right click the target when out of melee range, you will end up with both weapons hitting when you immediately get into range, but if you right click when in melee range it will delay the offhand to keep them desynced.

I just finally got around to messing with this sim, I didn't realize that it modelled Fury & Arms, since the main page doesn't list warriors. I get very similar results to Landsoul's spreadsheet for my current gearset, however this lets me fine-tune rotation details and rage tolerances. I also get very comparable results to the sheet's SEP values from calculate_scale_factors, setting all values to 20 (gem level) and doing 10000 iterations.

Have you considered using seed-based RNG to allow for repeatable results? I deal a lot with simulations for my work, and was surprised that it uses non-repeatable RNG. It seems like that would allow for faster gear/stat comparisons, as you wouldn't need to do as many iterations for each gear set to get consistent mean values.

Also, is there some hidden documentation somewhere? The code has very few comments. Like if I want an action to occur only if the cooldown of an ability is higher than some value, I haven't figured out how to specify that in the action list.

dedmonwakeen · 10/20/09, 12:37 PM

[quote=foopeemoa;1430684
I just finally got around to messing with this sim, I didn't realize that it modelled Fury & Arms, since the main page doesn't list warriors. I get very similar results to Landsoul's spreadsheet for my current gearset, however this lets me fine-tune rotation details and rage tolerances. I also get very comparable results to the sheet's SEP values from calculate_scale_factors, setting all values to 20 (gem level) and doing 10000 iterations.[/QUOTE]

That is remarkable...... I would have thought that scale-deltas of that size would have required many more iterations. I haven't exercised the Warrior module enough to get a feel for its determinism. I fear you may have gotten "lucky".

Have you considered using seed-based RNG to allow for repeatable results? I deal a lot with simulations for my work, and was surprised that it uses non-repeatable RNG. It seems like that would allow for faster gear/stat comparisons, as you wouldn't need to do as many iterations for each gear set to get consistent mean values.

I guess the short answer question to your first question is "yes". Simply set the parameter "seed" to a numerical value.

It is very very very tricky to add determinism into the RNG without creating edge conditions that result in discontinuities in the scale factor generation. One of our developers (Nenad) invested a considerable amount of time creating so-called "smooth" RNG packages. (See sc_rng.cpp for details.) He successfully reduced the number of iterations required for convergence by two orders of magnitude. This can be experimented with by setting smooth_rng=1.

However....... There were certain stats for certain classes (Shadow Priests and Haste, for example) that did not behave well with smooth_rng=1. Discontinuities began to appear in the DPS/Stat curve that resulted in step-functions in scale factor generation. I still believe that this functionality is vital to gear exploration....... but it needs more investment of my time. Until that happens, I have to turn it off by default.

Also, is there some hidden documentation somewhere? The code has very few comments. Like if I want an action to occur only if the cooldown of an ability is higher than some value, I haven't figured out how to specify that in the action list.

The code itself will never have a ton of comments. If you find something obscure, then I'll invest the time to change the code until it reads like your native tongue. User documentation........ That is a real concern. With the recent GUI release I've been spending more time in that area, but it still needs a ton of work. The bulk of the existing documentation can be found in Welcome.html, Legend.html, and Examples.simc.

Regarding your specific question: As action conditionals have been required we have been implementing them. We have an infrastructure for generic conditionals (also done by Nenad) but we have not yet tied each of the modules into it. In a couple weeks you will be able to refer to any buff, debuff, cooldown, talent, glyph, etc as an action conditional and combine them using logical expressions.

Fizzl · 10/24/09, 6:46 AM

I have a couple of issues with the current version in the download section (simc-322-0-win32.zip)

1
Because SIMC.BAT is not in the same directory you can't drag and drop files from the profiles folder (says can't find simc.exe in this directory).
Editing the .bat to be ..\simc.exe makes this work, I don't know if it is used by anything else though?

2
More importantly your fire Mages don't have combustion! (Yes I know I once said combustion was crap but they buffed it in the last patch)
I think T9_20_51_00 should be using this: The World of Warcraft Armory
No ranged talents but 35 yards is enough (Scorch will only be 30 yards but that is what warlocks are for!)

This change gets me this: Simulationcraft Results (5000 Iterations)

dedmonwakeen · 10/24/09, 8:00 AM

Originally Posted by Fizzl

I have a couple of issues with the current version in the download section (simc-322-0-win32.zip)

1
Because SIMC.BAT is not in the same directory you can't drag and drop files from the profiles folder (says can't find simc.exe in this directory).
Editing the .bat to be ..\simc.exe makes this work, I don't know if it is used by anything else though?

2
More importantly your fire Mages don't have combustion! (Yes I know I once said combustion was crap but they buffed it in the last patch)
I think T9_20_51_00 should be using this: The World of Warcraft Armory
No ranged talents but 35 yards is enough (Scorch will only be 30 yards but that is what warlocks are for!)

This change gets me this: Simulationcraft Results (5000 Iterations)

Thanks, Fizzl. The wiki output is pretty stale at this point. I need to put out a new release and update the pages.

I'll experiment a little with SIMC.BAT but to be honest I hope it goes away real soon given the new simcqt gui app. (I do need to add some support for drag-and-drop into the Simulate tab though.

Maje · 10/26/09, 6:54 AM

Something I can't understand in the mage output, Living Bomb's explosion average numbers seem lower then they should be. That might be because the minimum explosion damage is listed as 0 which isn't possible when hitcapped, I even changed the action list to only use LB when time to die is >= 12 (which probably should be in the default profiles).

The 'seems lower' is from comparing to the average explosion damage done on the usual fights with no wierd multipliers/special mehcanics.

dedmonwakeen · 10/26/09, 8:17 AM

Originally Posted by Maje

Something I can't understand in the mage output, Living Bomb's explosion average numbers seem lower then they should be. That might be because the minimum explosion damage is listed as 0 which isn't possible when hitcapped, I even changed the action list to only use LB when time to die is >= 12 (which probably should be in the default profiles).

The 'seems lower' is from comparing to the average explosion damage done on the usual fights with no wierd multipliers/special mehcanics.

I can ask Lhivera, but if you have some screen shots out in the wild showing damage and your character tab, that would make it easier. If not, no worries, we'll double-check.

Maje · 10/26/09, 9:48 AM

I'm not sure if I'm not misinterpreting it but Living Bomb info contains this for spell data:

static rank_t ranks[] =
    {
      { 80, 3, 759, 759, 690, 0.22 },
      { 70, 2, 568, 568, 512, 0.22 },
      { 60, 1, 336, 336, 306, 0.22 },
      { 0, 0, 0, 0, 0, 0 }
    };

For level 80 from what I can understand it says the direct damage is 759-759 and tick is 690, however LB r3 is 690 direct damage while the tick is 1380 spread over 12 seconds/every 3.

Also taking from my mages profile in simcraft raid buffed spellpower is 3238, so DD hit should be 690 + 3238 * 0.4(coef) * 1.13(CoE) * 1.1(firepower) = 690 + 1609.93 = 2299.93 with 6% partial resist 2161.9

EDIT: or rather firepower multiplier should also be applied to the base damage so 2299.93 + 690*.1 = 2368.93 with resist 2226.7942

However the output shows:
Hit=1373| 0|3488

Again I'm not really familiar with how mage simcraft is coded/implemented so I hope I haven't botched too much up.

dedmonwakeen · 10/26/09, 11:11 AM

Originally Posted by Maje

I'm not sure if I'm not misinterpreting it but Living Bomb info contains this for spell data:

static rank_t ranks[] =
    {
      { 80, 3, 759, 759, 690, 0.22 },
      { 70, 2, 568, 568, 512, 0.22 },
      { 60, 1, 336, 336, 306, 0.22 },
      { 0, 0, 0, 0, 0, 0 }
    };

For level 80 from what I can understand it says the direct damage is 759-759 and tick is 690, however LB r3 is 690 direct damage while the tick is 1380 spread over 12 seconds/every 3.

Also taking from my mages profile in simcraft raid buffed spellpower is 3238, so DD hit should be 690 + 3238 * 0.4(coef) * 1.13(CoE) * 1.1(firepower) = 690 + 1609.93 = 2299.93 with 6% partial resist 2161.9

EDIT: or rather firepower multiplier should also be applied to the base damage so 2299.93 + 690*.1 = 2368.93 with resist 2226.7942

However the output shows:
Hit=1373| 0|3488

Again I'm not really familiar with how mage simcraft is coded/implemented so I hope I haven't botched too much up.

Maje, you're spot on. We'll get it fixed ASAP. Time for another release....

Maje · 10/26/09, 11:50 AM

Thank you, also I made a mistake in my previous post and CoE should ofcourse be also applied to the base damage but that is probably already handled in simcraft.

In any case if I change the code to have the correct values the numbers are even lower then what they are supposed to be, for instance for the said mage WoL shows LB tick hits to be ~1243.6 while the changed simcraft shows Tick=887|829|1087. Which is obviously wrong, I used the following for ranks[]
{ 80, 3, 690, 690, 345, 0.22 }, //ie. 690 is the base explosion, 345 is the tick which is 1380/4.

So I'm either fiddling with something in a wrong way or the spellpower coefficients/multipliers are wrong.

Also explosion damage is still showing 0 as the minimum...

One other thing, if I look at the debug output and count the number of explosions it's 20 however, the normal output shows 40 Living Bombs, which can't really be because 40*12seconds per bomb = 480 seconds and the simulator runs for 300 seconds. Perhaps for some reason it calculates each LB as two spells the first is the application which does 0 damage and second the explosion which would explain the 0-s.

dedmonwakeen · 10/26/09, 2:03 PM

Originally Posted by Maje

Thank you, also I made a mistake in my previous post and CoE should ofcourse be also applied to the base damage but that is probably already handled in simcraft.

In any case if I change the code to have the correct values the numbers are even lower then what they are supposed to be, for instance for the said mage WoL shows LB tick hits to be ~1243.6 while the changed simcraft shows Tick=887|829|1087. Which is obviously wrong, I used the following for ranks[]
{ 80, 3, 690, 690, 345, 0.22 }, //ie. 690 is the base explosion, 345 is the tick which is 1380/4.

So I'm either fiddling with something in a wrong way or the spellpower coefficients/multipliers are wrong.

Also explosion damage is still showing 0 as the minimum...

One other thing, if I look at the debug output and count the number of explosions it's 20 however, the normal output shows 40 Living Bombs, which can't really be because 40*12seconds per bomb = 480 seconds and the simulator runs for 300 seconds. Perhaps for some reason it calculates each LB as two spells the first is the application which does 0 damage and second the explosion which would explain the 0-s.

The zero min value and double-counted applications are a reporting issue due to the DD portion being applied at last_tick(). It's on my plate to fix.

Looks like this spell has seen a bit of churn recently. Seems the spell power coefficient for each tick was reduced from 0.20 to 0.10. I'm not sure that is correct......

Maje · 10/26/09, 3:30 PM

Looks like this spell has seen a bit of churn recently. Seems the spell power coefficient for each tick was reduced from 0.20 to 0.10. I'm not sure that is correct......

As far as I know the DOT is still 0.8 coefficient with 0.2 per tick. The numbers on WoL agree with it as well.

The zero min value and double-counted applications are a reporting issue due to the DD portion being applied at last_tick(). It's on my plate to fix.

Hmm perhaps that's what causes the lower average damage to be displayed.

Maje · 10/27/09, 4:38 AM

I checked the latest changes, the ticks seem to fit. The average explosion damage (hit and crit) is exactly 50% of where it should be; I'm guessing it's a side effect of how the spell is handled.

This however doesn't change the value of the spell since total damage done during one LB is what it should be.

dedmonwakeen · 10/27/09, 8:23 AM

Originally Posted by Maje

I checked the latest changes, the ticks seem to fit. The average explosion damage (hit and crit) is exactly 50% of where it should be; I'm guessing it's a side effect of how the spell is handled.

This however doesn't change the value of the spell since total damage done during one LB is what it should be.

I'm in the process of changing stats_t::add_result() to detect a zero value RESULT_HIT which should -finally- solve this reporting discrepancy.

EDIT: r3682 has the fix.

Maje · 10/28/09, 3:21 AM

Just checked numbers from r3682, the output looks good and explosion/tick damage fits WoL.

I'm checking Reign of the Dead output, at first look it seems the crit damage is slightly on the lower side. I'll try to run more tests on live and see if I can get more info. How is it's damage amplified currently; CoE and 1.5 multiplier for crit?

EDIT: checked some WoL logs, the damage is modeled correctly ie. CoE * 1.13 sanctified retribution * 1.03 for criticals generic * 1.5 and * 1.03 due to meta gem. However it seems at least mages benefit from Playing with Fire - Spell - World of Warcraft increasing the damage by another 3%.

I also saw some wonky numbers on Faction Champions which for some reason at times show higher then possible hits/crits but I'll attribute it to wierd mechanics since there are no partial resists on the fight either.

dedmonwakeen · 10/29/09, 1:50 PM

SimulationCraft version 322.1 is available for download.

The GUI has probably moved from "alpha" to "beta" now.

Note that the .BAT files have been deleted. Drag-n-Drop .simc configs on top of the GUI exec simcqt.exe instead.

EDIT: As usual, supports all currently known 3.3.0 mechanics (and T10 bonuses). Just specify the patch level to get PTR mechanics.