I found my primary problem with my testing. I was going to put this in my other post, but it wouldn’t let me respond, and it seemed big enough to be worth putting in a new post.
Recount’s DPS calculations are very off. When I run them both at the same time, on the TV build, both show (8 minute test, on raid training boss dummy [single target]) 1.53M damage. But Recount shows 3617dps while Details! shows 3.1kdps. I think this is because Recount is trying to estimate downtime and remove it from the dps calculation… but I don’t know. It doesn’t seem to be a rolling average… so it’s not like “at the end I was doing 3617”. It tends to barely move after the first 1M damage, so I think it’s an overall average.
Since I know my tests are 8 minutes (and I verified damage done by watching life on the dummy before/after with nobody else hitting it), I can calculate dps, and I come up with 3198dps. I think Details! would have it perfect except that combat doesn’t end right at the time my timer and I stop (only requires an 8 second error to go from 3.2 to 3.1 on Details! with about 1.5M damage in 8 minutes).
(On another test with TV, I hit 3.4k - using my calculator, but Recount showed 3.9k, while with GS, I hit 2.8k using calculator, but Recount showed 3.8k… so Recount is in error 500dps on TV, and 1000dps on GS builds.)
So the good news, for me, is that now I can actually figure out what is wrong and optimize my build and rotation. It’s Recount. (There are some other things going on, because I can probably eek another 100-150 dps out of TV build, and now I can explore RoP and alternate leggo better.) Recount is way worse on GS builds than TV builds. On GS build, it has me doing about 3% more damage than TV, or about 3900 dps vs. 3800. But in reality it’s more like 2850 for GS and 3300 for TV.
At least I have closure on this now. LoL…
Thanks everyone for all the feedback and review of the various parts of what was wrong. I can trust Quick Sim again! Yeay!
Unless recount has been fixed recently, it was never very accurate. I think you’re right in that recount doesn’t track downtime. Like if you did a 20k opener on a boss and then died immediately, you’d still have 20k dps on recount even if the fight went on for 5 more minutes.
I think it’s trying to show you the ceiling of your build, if only you could press your keys fast enough and not cancel casts by moving around like a dolt.
But the errors I’m getting are with literally well over 60 minutes of target cast time practice in each spec, while not even holding the mouse. I suspect I may have about 5% downtime because of latency or hesitation (actually that’s one of the things still in my repertoire to try to fix… but it feels like modifying precast timing won’t do anything with my current latency - usually about 60 ms). Most of my casts are at about 60-70% of the GCD swirl.
I was getting a major discrepancy when trying to test suggested builds from Quick Sim testing. The problem was recount, but it messed me up for the better part of this week.
I use Skada for my meter, and it cannot figure out the Huntsman fight if its life depended on it.
And by that, I mean tripling my actual DPS and that of about half our raid. It had me pulling 13k for the entire fight, and I don’t even do that in my dreams.
Not sure how Skada works, but it’s likely showing the boost from the damage increase on the dogs. Details and logs don’t show that. Same applies for fights like Council, where Details and logs both reset some damage when a boss dies because padding isn’t counted when the bosses return to 100%.
But yeah, I wouldn’t really trust any of the damage meters personally. Logs are going to be more accurate.
The DPS meters (Details, Skada, Recount) have been like this since Legion. All will very in the actual numbers for the statistics. However, the percentage of damage done will be very similar.
After reading your other posts I went and tested GS on huntsman a couple times and compared it to my prior kills. GS definitely sucks with the glacial fragments legendary but I did find a bug between GS, glacial fragments, and the void elf passive. It appears to cause my passive to do massively increased damage. Using all 3 on hunstman causes my passive to do around 280 dps whereas with fevered incantation on as frost in the same fight with 20s less time on the boss it does 60 dps overall. Even with this dmg it is still around 1k lower than my TV build. Granted I did 1k more in mythic with a flask but without a pot. I would still consider it worse by metric.
Yeah, the big lesson learned is to not trust them. You can sorta trust the damage totals (not exactly). The damage totals will show you how one spec compares to another, but not necessarily tell you what actual baseline dps you’re doing.
The comments about Huntsman Altimor are interesting, but not quite right. You are probably doing 13k or whatever. I had one fight where I did I think 11k on Huntsman (based on Recount, so who knows wth I actually was doing). I know of one, but there may be more than one mechanic where DPS goes way up. The Shades of Bhargast take something like 5x+ normal damage if you’ve let them stay up through almost an entire cycle to get more shades. Then when you nuke them you do very high dps. For things like parses, I believe the debuff on the dogs is factored out when calculating your base DPS. Similarly, I believe if your dps is getting mitigated by a shield that cuts 30% (e.g.), the parsers will correct for that as well. You can’t really expect Recount, Details!, or Skada to fix these on the fly (though they could).
I can do a test (I did, in fact), that allows me to do a screenshot that shows Details nailing my actual DPS, while Recount is 500 too high (for my TV build).
https://imgur.com/a/mxVUxWT
Just to be clear, the mistake Recount is making has nothing to do with the standard meters not agreeing because some people aren’t in combat for some of the fight. It’s some sort of deadtime correction Recount tries to do that I believe is intended to correct for lag or some other dead time issue.
This is true. You get much lower DPS totals when you’re a DOT class, because the dot’s keep you in combat.
In some ways it gives a better picture because it doesn’t hold decursing, being CC’d, eating and joining late, or running from volcanoes against you, but often times as arcane I’ll be 2k higher than the other dps, but we’ll all have similar total damage.
Look at warcraft logs for the huntsman fight,it blocks all the damage you do to the adds that inflate your dps so instead of seeing 12k+ damage it will be more like 5k damage.
The problem is the logs for that fight don’t make the same calculations. Immediately after Huntsman the other night, Skada had a rogue with 23k DPS, then a Havoc DH at 17k DPS and then me at 13k DPS, then a couple of others over 10k DPS.
When logs were inspected later, out of a pack of 16 DPS I was midpack at a much lower number, and parse percentage. Everyone was under 10k on the fight.
If everything was just high by a specific percentage, it would be easy for that one fight just to lop off the extra and compare, but in addition to inflating our DPS it’s also not inflating it at the same percentage for every class/player. It basically makes the meter damage useless on that fight.
Recount has always been off, just for a while back in the day it was either the only option or the least off of the other options.
Most people I know stopped using Recount in MoP when details was made, it’s a crap damage meter. It’s pretty much only good for stroking epeen as it shows your dps higher than it really is. If you want actual usuable info to improve your rotation or something you don’t want Recount.
There’s always been an argument that it’s useful info for a raid lead or a m+ group to know. E.g one of their dps is doing good dps, when they’re alive, but they keep dying.
Instead of just looking like utter trash on the meters because they were dead for x amount of time.
But the reality is that you just end up with inaccurate meters. Every time someone has thought they were doing ok when they weren’t, it’s been because they’re using recount in my experience.
Details is far superior.
There is an option somewhere in details to make it behave like recount by the way. I forget where, it’s kinda hidden away.
Can be useful when you’re trying to help bad dps. Find out if their issue is their rotation, or just mechanics.