In Progress Pet Melee Lag

[Sancus]

I re-parsed all of this on Bristlebane over the weekend outside of raids, and outside of the guild hall. Margin of error must be taken into consideration. If you are seeing gains or decreases < +/- 5%, then that's all well within margin of error in which you can also argue that there's risk of intentionally decreasing DPS by removing spell/weapon procs. I'd argue that ~ +/- 10% is still within the margin of error. Some of the deltas Sancus is quoting above are closer to 2-3% rather than 12%, so context is not really sufficient. Guild Hall lag also is not consistent with some lag swings seen in dynamic zones.

While +/- 10% is far too large of a margin of error for parses with >10k hit attempts, every proc test had a decline in attempt rate >10% vs not using procs. It doesn't matter the deltas of tests with procs versus each other, it matters that the impact of procs is significant vs avoiding them:

That's a decline of 15.8% and 14.2% respectively.

declines of 16.4% and 14.7%

declines of 13.7% and 12.5%

___________________________________​

However, arguing against running 15K procing pet weapons and Night's Perpetual Terror procs does not make sense or correlate with the margin of error in the data.

15K base damage pet procs do negligible damage because pets can't crit procs. My largest base damage proc, Companion's Elemental Strike IV with a whopping base damage of 37,500, contributed 1,875 DPS (0.5% of the total)

In fact, proc damage in its entirety was only 1.87% of my pet's total damage on the above parse. That's within that +/- 2% margin of error delta you quoted. Why would we declare a >10% decline in melee rate as potentially spurious in favor of a damage category that in its entirety contributes negligibly to damage? If you really want to chase that extra 1.87%, that's your prerogative, but I think most are comfortable forgoing it given clear evidence that the degradation in pet melee it causes is larger than the damage it adds.

Ultimately the consensus is: If you're raiding, remove every single pet/spell proc, as the effect on pet hit attempts is overwhelmingly detrimental and obvious. However if you're grouping, only remove pet/spell procs that are exceptionally minimal in damage output; or, that activate very often (e.g., BST 2.0 epic) that are also damage-minimal in nature. The goal in the group game would be to only eradicate high proc-per-minute (PPM) pet spells/procs, or very low damage effects that are essentially unparseable.

You're welcome to do that, but it's certainly not a consensus that I'm signing on to.

My contention, supported by hours of parses actually posted in this thread: in all content for DPS purposes, you're best off avoiding procs. The benefit of doing so is far more acute in raids, but procs degrade DPS in all contexts. There are some procs (notably, Prism Skin) that, given the less severe degradation outside of raid content, I think are worth using as needed. If your only goal is to maximize DPS, though, avoid all procs on your pet.

 

[Lodestar]

I never stated a 2% margin of error anywhere in my posts. A 2% margin does not exist in any content, anywhere in Everquest unless you're on a localized server. You are not referencing statistics correctly.

The issue is that no one's parses are able to pragmatically prove or disprove the exact extent of pet melee lag in group content. The margin of error is too large, and the accuracy of parse analyses are too slim. However in raid content, as you stated the results are acute, which exceed the margin of error giving us a repeatable result and a simple solution.

 

[Iribabh]

Great work @snactucs!

 

[Barton]

So it appears there is no consensus.

 

[menown]

There may not be consensus, but there sure is an answer. The more pet procs you have, the less melee swings your pet will have in all content. Yes, this is expounded in laggy raid environments. The amount of DPS that pet procs provide is really not worth the tradeoff of their higher DPS output from melee swings.

 

[Barton]

There may not be consensus, but there sure is an answer. The more pet procs you have, the less melee swings your pet will have in all content. Yes, this is expounded in laggy raid environments. The amount of DPS that pet procs provide is really not worth the tradeoff of their higher DPS output from melee swings.

sorry, I was trying to be funny
I think the advice of not using procs in either setting is closer to a consensus.

 

[bobokatt]

Hi Sancus. Might you explain WHY this pet lag/proc issue DOES not affect our actual toons?
My 120 ranger has when all out, the reptile druid, the chanter terror line, the Shaman Melancholy, then the list of type 18/19 I have, then bard, then self, then this and that. So why does THIS not affect us as toons and only affects the pets? Thank you.

 

[bobokatt]

While +/- 10% is far too large of a margin of error for parses with >10k hit attempts, every proc test had a decline in attempt rate >10% vs not using procs. It doesn't matter the deltas of tests with procs versus each other, it matters that the impact of procs is significant vs avoiding them:

___________________________________​

15K base damage pet procs do negligible damage because pets can't crit procs. My largest base damage proc, Companion's Elemental Strike IV with a whopping base damage of 37,500, contributed 1,875 DPS (0.5% of the total)

In fact, proc damage in its entirety was only 1.87% of my pet's total damage on the above parse. That's within that +/- 2% margin of error delta you quoted. Why would we declare a >10% decline in melee rate as potentially spurious in favor of a damage category that in its entirety contributes negligibly to damage? If you really want to chase that extra 1.87%, that's your prerogative, but I think most are comfortable forgoing it given clear evidence that the degradation in pet melee it causes is larger than the damage it adds.

You're welcome to do that, but it's certainly not a consensus that I'm signing on to.

My contention, supported by hours of parses actually posted in this thread: in all content for DPS purposes, you're best off avoiding procs. The benefit of doing so is far more acute in raids, but procs degrade DPS in all contexts. There are some procs (notably, Prism Skin) that, given the less severe degradation outside of raid content, I think are worth using as needed. If your only goal is to maximize DPS, though, avoid all procs on your pet.

Hi Sancus. Might you explain WHY this pet lag/proc issue DOES not affect our actual toons?
My 120 ranger has when all out, the reptile druid, the chanter terror line, the Shaman Melancholy, then the list of type 18/19 I have, then bard, then self, then this and that. So why does THIS not affect us as toons and only affects the pets? Thank you.

 

[Lodestar]

It would be Daybreak's responsibility to answer why it affects pets, and not players. It has to do with the differences in either/and/or programming, code, servers, database handling, etc. of players vs. pets. We know that part of the impact is due to certain DB servers/databases being significantly lower performance or slower than others (there's another post on that somewhere, in being able to identify whether you are in a high or low-performing raid instance). Ultimately not worth our time trying to determine why, since Daybreak hasn't been able to fix it anyhow.

All that matters is that there is a known difference with pets on raids, to the extent that a rule can be stated to not run any spell/weapon procs on pets in raids. For groups, all that can be stated definitively is that you should not run any high PPM or unparseable procs (like 200dmg) on your pets. It's at your own risk or opinion, as to what point high damage procs cause even more degradation in melee output in group content.

 

[Barton]

It would be Daybreak's responsibility to answer why it affects pets, and not players. It has to do with the differences in either/and/or programming, code, servers, database handling, etc. of players vs. pets. We know that part of the impact is due to certain DB servers/databases being significantly lower performance or slower than others (there's another post on that somewhere, in being able to identify whether you are in a high or low-performing raid instance). Ultimately not worth our time trying to determine why, since Daybreak hasn't been able to fix it anyhow.

All that matters is that there is a known difference with pets on raids, to the extent that a rule can be stated to not run any spell/weapon procs on pets in raids. For groups, all that can be stated definitively is that you should not run any high PPM or unparseable procs (like 200dmg) on your pets. It's at your own risk or opinion, as to what point high damage procs cause even more degradation in melee output in group content.

Seems like there is a known difference in groups that should disallow procs as well unless you have parses that show dramatic differences than the ones Sancus provided.
EDIT: and you can provide those parses

 

[Lodestar]

There is a known impact in group content, but it's not feasible to quantify the point or extent at which high-damage procs fail to outweigh the loss in melee swings. There is a substantial margin of error in group zones due to +/- lag from players entering/leaving, the quality of the zone or instance instance you have, and errors in parseability since you can't hit the same mob for 2 hours straight. When you combine high error margin with minimal parse deltas, you enter a spectrum of analyzing parses that don't exceed the margin of error. I re-parsed group content over the weekend to ensure this was still the case, which it is. Further, parses from Guild Lobby on a combat dummy don't suffice for group zones and instances.

However in raids, the results are astronomical in that several procs will cause your pet to not swing for multiple seconds at a time. This has an overwhelming impact on parses obviously, in that your pet's hit rate completely tanks. Anyone can literally see pet animations cease upon procs, with no swing animations for extended durations. The situation is not the same for groups: The known impact is exceptionally minimal and compounded by a similarly overlapping margin of error. This is why raids are clearly substantiated, while group content is not substantiated as to the exact extent one should not use certain high-damage procs or adps procs (such as Dire Bite). Again, it's entirely fair to assume that low-damage or high PPM procs should be removed from pets in group content.

 

[Svann2]

Maybe they could add a training dummy to pok so you could parse it for hours. The arena is too low population to include effects of lag.

 

[fransisco]

I would like to point out that EVERY new expansion, the devs make new pet weapons and proccing spells. All of which are always strictly wrong to use, because in every case they decrease pet dps.

If the bug wont be fixed, dont waste spells on garbage that is a trap.

 

[Metanis]

I would like to point out that EVERY new expansion, the devs make new pet weapons and proccing spells. All of which are always strictly wrong to use, because in every case they decrease pet dps.

If the bug wont be fixed, dont waste spells on garbage that is a trap.

Except you can tell instantly if someone is a noob by simply looking at what their pet has equipped and buffed.

 

[Rael]

People need to start doing their own parses, or review data more precisely.

For raid content, the evidence is undeniable as to not using any procs on pets. This is a function of known issues with poor performing servers/databases for certain raid instances spawned. Parse results exceed any margin of error because of how overwhelming the impact from pet procs clearly are.

For group content, the dynamics are very different. As mentioned earlier in this thread, the margin of error exceeds that of any statistically significant results (i.e., the resulting lag from pet procs in group content is not significant enough to produce an identifiable result that outweighs inherent sampling error). You're talking very small % +/- swings in group content, against continuous +/- 5% or more margin of error due to unstable control variables. No dice.

Lodestar is accurate in suggesting that high proc-per-minute weapons or effects of minimal damage should not be used in group content (specifically effects far above average PPM); however, trying to quantify legitimate procs against a remarkably subtler amount of lag in group content is not something anyone has shown yet. It's unlikely it will ever be shown until lag becomes as excessive as it is in raids.

The only exception I've ever seen to this, with any consistency, would be Great Divide in TOV with high populations. There was enough noticeable lag to likely cause a significant enough impact from pet procs to outweigh margin of error. It would need to be parsed in era with an active population where said zone lag was occuring.

 

[GrandOpener]

IMO Sancus did provide convincing parse-based evidence that using procs with pets for DPS purposes in group content is incorrect and decreases overall DPS beyond any reasonable margin of error.

However, even if you find that evidence unconvincing, the reverse is also true. There is no concrete evidence (at least, none that I've seen in this thread) that pet procs contribute significantly to DPS. In fact there is parse-based evidence that procs generally add a trivial amount of DPS, even not considering the bug at hand.

If there's no evidence that procs add significant DPS, but there is some evidence that procs may decrease overall DPS, it seems like the safe choice is pretty obviously to not use pet procs in any content. (Unless, as noted previously, you're using them because of a non-DPS reason.)

 

[Rael]

If there's no evidence that procs add significant DPS, but there is some evidence that procs may decrease overall DPS, it seems like the safe choice is pretty obviously to not use pet procs in any content. (Unless, as noted previously, you're using them because of a non-DPS reason.)

Did Sancus provide the margin of error due to zone lag differentials, in-zone lag swings, player population swings in zone, etc.? No he didn't. Therefore, your judgment would be outright questioned by any statistician by stating Sancus provided conclusive evidence. Statistics doesn't work when you ignore the margin of error in a game like Everquest, that has a serious well-known issue with it. Blanket statement: Statistics doesn't work in any application when ignoring sampling error.

Further, assuming procs don't add relevant DPS is blasphemy. As you approach infinity towards adding more procs, more spells, or anything else, you will eventually hit an inflection point in which this statement is categorically false. Procs do add a small % of DPS; however the challenge is that pet procs decrease melee swing DPS by a relatedly small % (in group-game only), while also compounded by a small margin of error % that makes results consistency difficult to nail down. In essence, and as only an approximate example, you're talking group DPS impacts in the single digits of:

~ +2.5% procs
~ -2.5% pet swing loss
~ +/- 5-10% margin of error.

Good luck with that. Plug it in to SPSS and let me know what comes out. Doesn't matter if procs increase slightly more or less than this, or pet swings; either way against a comparable margin of error, it's a wash with impacts being so low and error being relatedly substantial enough to hamper definitive results. In raids, the pet swing loss is >50% so there's no debating raids' proven results.

Considering these truths, here's the known facts for the group game:

• The impact of pet procs is vastly less in the group game, to the point of being single digit %s and difficult to parse on their own.
• The margin of error due to different zones' lag at any given point, in-zone lag shifts, player population shifts, etc. induces sampling error, causing a small % error margin impacting the accuracy and repeatability of the aforementioned.
• If you run a parse yourself, while taking both aforementioned bullets into accounts, you'll come to find that raid results are consistently conclusive, while group results are inconsistently inconclusive.

 

[GrandOpener]

The Test Server update for 11/28 contained these notes:

- Reduced pet combat lag due to spell casting and weapon procs.

Is this issue finally getting addressed?

 

[Lodestar]

The Test Server update for 11/28 contained these notes:

“- Reduced pet combat lag due to spell casting and weapon procs.”

Is this issue finally getting addressed?

"Reduced". Let's see.

 

[NyteShayd]

Magic 8 Ball says "Signs point to No."