Diablo® III

SNS CD PING and you

Posts: 13,645
View profile
I don't need perfect identical stats. As humans we can interpolate. When someone has both high CC and CD than I and my multiplier is low and with higher ping I know culprit.
Edited by Aimless#1700 on 1/18/2013 6:15 AM PST
Reply Quote
Did 5 runs with nearly identical results:

Standard SNS - 6th slot unused
167109 DPS
58.5% cc
2.84 IAS
70 ms

7.22 x
Reply Quote
01/18/2013 02:22 AMPosted by Alesso
IMO this test would be better and more controlled if every participant used the exact same gear. Extra variables deteriorates the quality of the test.

You are right, the scientific method is best reserved when you control everything, but even then there are errors that you simply cannot control no matter what.

However, the best anyone can do is use is either:

A) Use legendaries in all slots, because they roll very close to each other,
B) Use a particular set and give it to each person to test, or
C) Find the exact same gear

Even B isn't going to work, because varying paragon levels mean different stats, however small the differences are.

That's why every scientific paper calculates error as part of what they're studying. While Aimless didn't do that here, he didn't really need to, that can be an exercise to anyone else who wants to do it, it's quite easy.
Edited by KhalARon#1159 on 1/18/2013 7:30 AM PST
Reply Quote
Posts: 4,560
View profile
Let's all just agree that lower ping = higher multiplier. It may also be worth mentioning that using a macro vs button mashing may also help a bit. My multiplier pre macro was 7.3 based on reviewing a bunch of saved videos.
Reply Quote
@KhalARon: Sure we did on the error analysis. That's the analysis I was doing for you guys with the correlation coefficients, bootstrapped correlation coefficients (to protect against outliers), and line of best fit (providing p values for the correlation coefficient and the norm of the residuals). ;)

Still I agree with Worloega and you about the study. Yes, it's messy and we didn't control for everything and that's not necessarily ideal, but it doesn't immediately invalidate the study. From a practical perspective, we can still learn something from a study that isn't as carefully controled. That is so long as the effect you are studying is not being dwarfed by the outside variables. We can clearly see the relationship is linear between latency+effective DPS multiplier.

And yep there's quite a bit of variance, but at the population analysis the variation in gear is just noise. The only way we'd have problems is if there was a systematic relationship between good gear and latency. But just look at poor novice. He's got decent gear, but despite all that has the absolute worst latency and multiplier. Stuff like that is very compelling. Now better methodology would be cleaner and potentially give us a more accurate picture, but I'm dubious it would change the result. And that's the real question, how perfect do we need the results before we believe it's true. I think it's good enough for the question of knowing if there's a linear relationship between latency and effective DPS multiplier, as well as a rough formula approximating the relationship. But beyond that we can't say much else.
Edited by TekkZero#1963 on 1/18/2013 8:31 AM PST
Reply Quote
01/18/2013 08:15 AMPosted by KillerElite
Let's all just agree that lower ping = higher multiplier.

It should be accepted as fact, absolutely.

01/18/2013 08:17 AMPosted by TekkZero
@KhalARon: Sure we did on the error analysis.

Not quite. You did get p <0.001. However, a true error analysis would include an upper bound and a lower bound surrounding the best fit line when the linear regression was done. Yes, it's a little nitpicky, but I just want to be constructive, that's all.

I must say, though, that the work done on this issue by TekkZero and Aimless was top notch in terms of statistical analysis. This is why I visit the Wizard section more often, you guys have much more intellectual horsepower than certainly the General forum. If this was a problem set, you'd both get A's
Edited by KhalARon#1159 on 1/18/2013 9:15 AM PST
Reply Quote
Did 3 runs with no follower and using standard SNS/bone chill and have the following results

Mp10 Ghom 138425408
DPS: 185634 AP: 2.76 CC:52.5% CD: 315%
Time: 100 sec
Ping: 68 ms

Reply Quote
I sincerely hope it didn't seem like I don't trust this test or recognize the obvious qualitative trend correlating higher ping with lower multiplier -- yes, cliche as it is to say it, we're human and can see this, interpolate, etc.

My comment stems from a quantitative interest in the relationship between ping and multiplier. In this regard the test as it is falls short, while remaining a good qualitative trend identifier.
Reply Quote
@KhalARon: That's not a bad idea to add some confidence intervals to the regression. I'll see about crunching that. :)

@Alesso: I'm 100% there with you. It all depends on your goal. I'm most interested in the trend myself than actually quantifying the actual relationship. Though I was kind of thinking it'd be cool to use something like NetLimiter to artificially slow down the connection speed of Diablo 3 (or artificially increase the bandwith that windows reserves for the system). This is probably the most ideal method, since it can be done on the same computer, same person, and same gear.
Reply Quote
01/18/2013 12:04 PMPosted by TekkZero
Though I was kind of thinking it'd be cool to use something like NetLimiter to artificially slow down the connection speed of Diablo 3 (or artificially increase the bandwith that windows reserves for the system). This is probably the most ideal method, since it can be done on the same computer, same person, and same gear.

Agree with this ^^

I think this is the best overall method to come up with a real answer that removes the most variables. Makes sure the same gear is used, that the play style is the same (mashing vs. macros of all sorts), etc. That said, its really irrelevant because the whole point was just to identify a trend.
Reply Quote
I've been playing around with this, doing my own testing, changing gear to change dps, aps, cc, and cd, did it with full scoundrel, naked scoundrel and no scoundrel, and I must say that it doesn't change a whole lot of anything....the best I did was with full scoundrel at 149k dps, 2.78 aps, 59.5 cc, 289 cd and that was 6.62, and worst I did was no scoundrel and 138k dps 2.74 aps. 50.5 cc, 239 cd, and that was 6.54. Yeah gear does makes a little difference, but less than I thought it might....
Reply Quote
First of all, great thread!

I run between 250-300ms and still working on gear. What ias should i shoot for to perma freeze single targets/ubers? Currently at 2.61aps and looking to hit 2.73
Reply Quote
Posts: 5,376
View profile
from what i understand from others is at ur ping level ull get an improved freeze at 2.73 but it'll be more of a stutterlock then someone with 100ms ping or less
ie at 30-50ms which is aimless' ping he gets permalock i believe
Reply Quote
so only way for someone like me on 250/300ms is to gun for 3aps to attain aimless's permalock?

Reply Quote
tested on the ptr at mp7

(note i've commented b4 the ptr feels smoother as though I'm getting more ticks out of ww viz meteor)...

250-300ms ping
2.99 aps/57%cc (ie: 2.74 breakpoint)

able to very close stutter freeze uber magda by herself.

250-300ms ping
3.24 aps, 57%cc (3.01 breakpoint)
bonechill, evocation,pinpoint
able to down rakanoth/ghom (well ghom stopped casting clouds eventually)
able to bring sb/zk down to 20% before getting killed from pickup + whirlwinds.
Edited by chrisloup#6305 on 1/21/2013 11:39 PM PST
Reply Quote
this was a fun challenge. i was very surprised by the results.
i did 3 runs, nearly identical results.

Standard SNS:
78ms ping
220k dps
78sec, 8.04 multiplier (lol, that's 1.7m dps)

Next I tried the same test, this time w/o casting Shocking Aspect at all.

Same gear and stats, took 122sec. That's a 5.14 multiplier. Crazy how much dps that spell does.
Edited by BeerMan#1244 on 1/22/2013 8:08 PM PST
Reply Quote
so only way for someone like me on 250/300ms is to gun for 3aps to attain aimless's permalock?


I have tested all the different aps up to 3.1. You will never get permalock like aimless. With that ping, even at 2.73aps, you will not register all the casting.

I was playing with 250/300ms for a long time when I have Foxtel running on the same cable. I have disconnected Foxtel, and now my ping is around 200ms, 2.73aps seem to works better.

If you consistently getting 300ms, then probably 2.51aps is the max you can go.

With all that said. With higher aps, you do get the benefit of more ticks per second, but just not the extra twister from the increased attack speed.

I think it better if you try for higher CC at 2.73aps.
Reply Quote
MP8 Ghom 70,625k Latency 59-62ms

#1 62.2 seconds
2.74 APS BP
62.5% CC
75.5% ET CC
123k DPS sheet
1,135k DPS actual
9.2x multiplier

#2 68.5 seconds
3.01 APS BP
61.5 CC%
68.5% ET CC
116K DPS sheet
1,031K DPS actual
8.8x multiplier
(time stretch)

#3 60.9 seconds
3.01 APS BP
62.5% CC
69.5% ET CC
123k DPS sheet
1,159K DPS actual
9.4x multiplier

#4 57.9 seconds
2.74 APS BP
62.5% CC
69.5% ET CC
123k DPS sheet
1,219k DPS actual
9.9x multiplier

1,136KDPS with 9.3x multiplier
Edited by MasterJay#1651 on 2/5/2013 9:07 PM PST
Reply Quote
In conclusion of these testings it actually tells me how reaching 3 aps breakpoint might not be as important as we thought on a dps scale. Sure for mitigations / returns it's likely superior to 2.74 however with enough CC (especially through ET CC) you would be refreshing fast enough for a permafreeze anyway and more defense through shield spam that 3 aps really isn't needed or even optimal for defensives as well as offensives.

The breakpoint is very clutch compared to just purely driving CC up. Venom of course wins because Ghom is a single target, Electrify would win on elites. Time stretch would win on groups, not solo at all.
Reply Quote
7.84x @ about 130 ms, standard sns, no 6th skill

2.75 aps
55% cc
388% cd
Edited by Ryan#1899 on 2/5/2013 9:53 PM PST
Reply Quote

Please report any Code of Conduct violations, including:

Threats of violence. We take these seriously and will alert the proper authorities.

Posts containing personal information about other players. This includes physical addresses, e-mail addresses, phone numbers, and inappropriate photos and/or videos.

Harassing or discriminatory language. This will not be tolerated.

Forums Code of Conduct

Report Post # written by

Explain (256 characters max)