I was just wondering if anyone else was having issues with shadowcraft giving them different results for the same gear set. I equipped new weapons and ran my gear through shadowcraft and got this:

http://i44.tinypic.com/5ov1tw.jpg

So I reforged as it told me, but when I ran it again with those 'optimal' reforges, it tweaked it again to give me a 1.1 DPS increase:

http://i39.tinypic.com/dfg8zn.jpg

I've run the second one again, and it doesn't suggest any changes from that point on. It's quite a minor difference but I'm curious how the engine can give different results for the same gear set, and only after I'd already reforged to what it had suggested the first time.

Hey folks. Sorry I've been negligent in keeping this up; I've merged the latest changes from Aldriana and Dazer, and deployed them to the repo. The server interface still need a few tweaks, but at least that's done.

The site has been of great use for me but especially with Sub I've noticed 3 issues:
First off, I've seen EP values that shouldn't fluctuate much because they scale fairly linearly (Crit & Mastery) fluctuate by up to 50% just by reforging, I couldn't really isolate anything particular causing this so this probably won't help you much.

The other two problems involve the Hemorrhage glyph; One issue that you're definitely aware of is that you still can't choose to weave in Hemorrhage when using the Hemorrhage glyph. The second issue is that its value seems way too high - for my gear it shows as a ~1400 DPS increase over my ~42k DPS. That's almost exactly 3% of total damage while in practice I've never seen the DoT do more than 2% of total damage - e.g. in my last fight it was only doing 1.5% of total damage, that's half the value shown by Shadowcraft.
How is its damage calculated? Because in reality the damage is dependant on the actual damage Hemorrhage did (-> crits cause a much stronger DoT that can crit again then) and sadly higher damage stacks seem to get overridden by lower damage ones, thus the only way you can keep up the DoT based on crit damage (which would come close to the value shown in Shadowcraft) is to use Hemorrhage until it crits, switch to BS until it runs out, rinse and repeat.

That's the last comment I left in the model when dealign with this glyph. The figure you are getting is that of a only hemo cycle which will be higher than that of a mixed cycle, as the modeler runs into several issues (for instance the fact that it asumes the dot keeps being refreshed during shadowdance, which is not always the case); if you are checking against a live mixed cycle, then the numbers won't match, and if you are actually using a hemo-only build, then the modeler will make asumptions that are not true, favoring the glyph to a certain extent.

See that the model was exclusively intended to project the dps of the interweaved hemo cycle, which, as you said, cannot be selected from the UI (I'm not sure what needs to be done in the front-end to make that possible). In fact, when running the engine with the mixed cycle, the highest projected dps for the dot I've run across is ~2.4%. That said, studying the live logs I've been checking, the dot is accounting for 1.8%-2.2% (I discard logs with with less than 95% dot uptime) of the total dps so, yes, I do think the model is a bit too optimistic but still a good estimate.

A couple points: the subtlety model is not as largely optimized as those for combat and assassination; ever since release it's shown a clear skew and has overestimated the general output compared to live logs. The model, as of now, is using an oversimplified shadowdance model, and lacks a proper distribution model for find weakness; while that is not directly related to the glyph, it adds up in comparisons like the one you are trying to make: global dps percentage ones, that is. On top of that, the spec is still underplayed and I wouldn't (yet) draw such conclusions from the few public live logs we have: I do think the dot is accounting for more than it should be, but based purely on my knowledge of some of the issues with the model, and I insist: the interweaved cycle (and intended purpose) do resemble the outcome of those live logs. Lastly, I would expect similar 'margins of error' (more like simple overestimates, really) in the other glyphs too, so I'd say the comparison between them all should be about right.

To adress your previous question regarding the model of the dot: the engine figures an average spacing for the ticks (ticks per second) based on the amount of 'replaced' backstabs (with some energy adjustements) to then apply the expected amount of damage to those ticks, averaging the 4 possible outcomes of the hemorrhage strikes that produce them (hit from hit, hit form crit, crit form hit and crit from crit). I does not 'fish' for crits and it doesn't model any sort of crit distribution, effectively refreshing strong dots with weak ones if hemorrhages are performed more often than 24''.

All in all, I do share the concern and I've monitored the cycle extensively; but the 'formula' cannot be adjusted to reflect what we see in live, since what I believe to be causing this issue is related to the model as a whole. If anything, I personaly would rather not have the only-hemo option at all if the interweaved cycle cannot be selected.

The problem you're probably missing is that with current high end gear, Hemorrhage only builds are actually very close to BS builds, especially if you're not interweaving the glyphed DoT. Thus it would make sense to at least be able to compare both builds of similar effort, e.g. Hemorrhage only with Hemorrhage glyph vs. Backstab only with Backstab glyph.

Anyway, for some reason Hemorrhage glyph's worth has dropped by ~300 DPS in the Shadowcraft model for my gear now and is at least way closer to reality than before now (it's still another ~300 DPS too high though).

However, just a question regarding the Hemo glyph in a BS build and a Hemo build (not directly regarding Shadowcraft):
Wouldn't the only difference in DPS gains be that you lose a BS for a Hemo every 24 seconds in a BS build?
Assuming average buff/debuff uptime distribution (which should be the case for a 24 seconds interval), I don't see any reason why the DoT itself would do more DPS for a BS build than for a Hemo build.
The only way a BS build could benefit more from the DoT is if you fish for a Hemo crit (as mentioned earlier), otherwise you should have the same crit DoT and noncrit DoT uptimes for both cases.

I hope I'm not being bothersome here, I can perfectly understand why such a complex model that wasn't designed for modelling Sub to begin with can have errors when it comes to such calculations but I think when there's a margin of error of 40-80% of a single DoT, the formula for that DoT could use some small adjustment to make it come close to what you see in practice.

Something I would like to have added to Shadowcraft would be a simple date of when the back end engine that the web interface use was last updated.

I think it's only normal that these discussions arise when two radicaly different options seem to output just as much. I'm fully aware of what the problem is, as it has been posted here in the sub compendium, in some other sites and in a couple mails I've received; but ultimately it's not a problem with the model itself but rather, the problem that we need to choose between two options without really having a clear way to distinguish what's the winner. I'm not saying the model is perfect, as it could use some tweaking: there's always room for improvement, but so far, that's the best estimates that we have.

As to why the intertwined cycle and the hemo-only one are significantly different, well, there are a lot of players in place. For starters the formula for a single tick of the dot and the hemo strike is based purely in our in-game testing (as shown in the cataclysm mechanics testing thread) and has been updated with the changes in the previous patch; I would expect that to be pretty accurate, so the discrepancies lie within the model in either or both of 1- the computed ticks per second or 2- the averaging proccess of the 4 different ticks and the buffs/debuffs appied after.

1-The tick count: if hemo is the main cpg, then it goes through the normal script and uses the energy cost of hemo. If a mixed cycle is selected it'll run a backstab-only cycle, replace some (as many as we want) backstabs and compute how much more energy is there to expend. The amount of hemo strikes per second is what let's us figure the dot ticks per second. This sounds right, right? well, no: subtlety is subject to buff/debuff synergy a lot more than the other two specs.

2- Averaging process: The crit/non-crit logic is fairly simple (Bayes and conditional probabilities do the job), but buff/debuff averaged uptime is not that easy. ShadowCraft has no concept of synergy and concludes that every buff has a 100% uptime (everything overlaps) but with a 'watered down' buff: a 1000agility proc with 50% uptime would be a 500agility proc with 100% uptime. So, at the end of the day, some of those hemorrhage ticks will be benefiting with buffs/debuffs that don't belong to them. This usually is not very problematic, as it will be compensated with some other ticks that benefited less than they needed. Subtlety, however, has a buff with a very specific distribution (find weakness) that has a very large impact in this issue.

Find weakness occurs during most of ShD and 10'' after it's done (plus vanish/shadowmeld/oppener, but those are little less frequent). During that period, we don't really cast many backstabs (let alone hemorrhages), so the damage we compute in ShadowCraft for the overall damage by backstab takes more find weakness uptime than it really should do (in the same fahion, ambushes receive less than they should). Now, hemorrhage ticks (through their original replaced backstab), for the most part, should benefit a lot less than they are from find weakness; not really a big deal if spaced well appart but, in a build where hemo is the only cpg, the likeliness of ticks benefiting from something they shouldn't is larger. Thus, initial discrepancy becomes more apparent the more frequent the hemorrhages are.

Note that this *only* explains the overall dps percentage differences. Find weakness is actually applied after all the counting has been done: it's not like we compute the hemo strike, apply its find weakness and figure what the dot should tick for: in reality we compute them separatedly and apply find weakness to them both. Also, this find weakness factor is applied in a backwards engineering kind of a way if you're inclined to revise the code. I simply hope going through this explanation of how the engine works gives you an idea of what might be going on. Again, I find the model to be not too far off, but I wouldn't mind you telling me exactly where the formulator is failing compared to our -uncontrolled live log- testing.

Edit: I forgot to go over the total dps difference. There's non: the interweaved cycle (with 24'' appart strikes) does output the same exact dot dps than the hemo-only cycle -save for the very marginal higher uptime on procs due to the fact that we'd be pushing more strikes in any timeframe-. It's just that the closer you get to a smoothed out cycle (with abilities spaced out in a more flattened fashion), the closer the projected damage breakdown of the whole thing will be to that we find in live. If there are overestimations, there will be in both cycles: they'll be less aparent, and the glyph comparison will work better for the interweaved cycle. Also, the model asumes full usage of find weakness through all of the ways we have to put it up; but the short cd on ShD clashes heavily with that notion; for instance we need to delay the first usage until 10'' into a fight, so we are bound to a certain 'leekage' in FW uptime somewhere.

Does this include the fix to No'Kaled? I seem to remember it being valued at about 600 dps higher than The Sleeper at one point, but now it's at about 300 higher after reforges.

I'm not sure if this was an issue before or just started with the new code push, but in addition to the leg enchants not showing up on armory import, chest enchants (or at least the Peerless Stats enchant) now also does not import and must be manually selected.

deleted