Diablo® III

SNS CD PING and you

Posts: 13,645
View profile
Well at least ur a beliver now. Thanks for another data point. Pretty good ping coming from NE Canada.
Reply Quote
if your gona use my numbers for some calcs you should take note that i did use the scoundrel, also measured time with a stopwatch oldschool style. So at the very least id take .1 from the multi just for those extra 3cc if ur gona compare to the others who didnt use scoundrel.
Reply Quote
im pretty amazed that im having better ping here in spain than most of you guys in us.
Reply Quote
For what is worth I just did mp10 ghom with 100ms ping (in spain, with ADSL (no fiber optics in my damn building thanks to the administrators grrr)).

It took 143 seconds, so with 124k sheet dps that makes around a 7.7 multiplier.

Edit: That was with 2.77 APS (and 55.5 CC). And farming build (bone chill, wormhole)

Out of curiosity I did another try with 3.01 APS:
178 seconds with 106k dps (the rest is the same, just switch the chant): 7.3 multiplier.
Edited by Molitro#2566 on 1/16/2013 2:10 PM PST
Reply Quote
ximae: Actually, I'm using a stop watch too for those numbers. I think it'd be very nice to standardize the variables better to get the most accurate picture, but so far we have a pretty good relationship. I'm actually very surprised by how good the fit is, despite the measurements being taken by different players and on different gear.

Here's the updated stats (I added Killer Elites+Molitro plus altered my measurement based on my revised multiplier):
n = 7
peopleContributed = [novice, worx, tekkzero, killer elite, molitro, ximae, aimless]
latency = [500 300 129 116 100 55 24];
multiplier = [3.4 4.95 6.9 7.76 7.7 8.1 8.8];
Correlation Coefficient: r = -0.9861, p<0.0001
Best Fit for Predicted DPSMultiplier: DPSMultiplier = latency *-0.0113 + 8.7832 (with normR = 0.7873)
Edited by TekkZero#1963 on 1/16/2013 2:12 PM PST
Reply Quote
im with cable :D
Reply Quote
I did mp 10 three times and killed him in 141, 142, 142.....
142469 dps, 58cc, 249cd, 128ping

Puts me at 6.84 multiplier....
Edited by Vikesfan353#1980 on 1/16/2013 2:44 PM PST
Reply Quote
Posts: 4,967
View profile
I should also add that my ping varies from 95 up to as high as 118. At the end of the 116ms video I looked at my ping again and it had dropped to 105ms.

@Aimless

4700km to the servers. So I guess my ping is pretty good based on that :)
Reply Quote
Posts: 13,645
View profile
ximae: Actually, I'm using a stop watch too for those numbers. I think it'd be very nice to standardize the variables better to get the most accurate picture, but so far we have a pretty good relationship. I'm actually very surprised by how good the fit is, despite the measurements being taken by different players and on different gear.

Here's the updated stats (I added Killer Elites+Molitro plus altered my measurement based on my revised multiplier):
n = 7
peopleContributed = [novice, worx, tekkzero, killer elite, molitro, ximae, aimless]
latency = [500 300 129 116 100 55 24];
multiplier = [3.4 4.95 6.9 7.76 7.7 8.1 8.8];
Correlation Coefficient: r = -0.9861, p<0.0001
Best Fit for Predicted DPSMultiplier: DPSMultiplier = latency *-0.0113 + 8.7832 (with normR = 0.7873)


English? Still fiting model you assuming earlier
Reply Quote
Oh, those posting multipliers should probably dismiss the follower also. Some might be using a cold based scoundrel, which will inflate the number because of higher CB uptime, while others might be using chantress. Also, I'm not sure what the dps of the followers is but that can also impact results.

Basically, if the idea is to compare CD and ping impact on effective multiplier, all other things need to be normalized as well as possible. The main thing that won't be normalized is crit chance, but if enough people have about the same crit chance it should be fine.
Reply Quote
@Aimless: Yes, it's the same model as earlier with more datapoints. In english, all the math is saying is that there's a very high inverse correlation between latency and paper dps (98.6% to be exact and a perfect line would be 100%). The effect is statistically speaking very significant and the equation best-fitting the data right now is PredictedDPSMultiplier = Latency (in ms)*-0.0113 + 8.7832. Still the sample size is small and there is a lot of variation in how each data point was obtained.

@Loroese: Agreed, we should standardize measurements if we seriously want to do this beyond the rough correlation that we have between latency and the effective DPS multiplier.
Reply Quote
Posts: 13,645
View profile
Thanks tek. So what's happening? just loses packets you think?
Reply Quote
Posts: 5,376
View profile
im almost exactly 4 thou km away from blizz in irvine (assuming thats the server space)
would it help if i added some runs?
Edited by SteelPhantom#1820 on 1/16/2013 6:01 PM PST
Reply Quote
Nice little data gathering we got going on here.

I'll chime in some numbers.

Ghom MP8 70.6m HP

Bone chill + Evocation

34.8s / 36.2s / 35.4s = avg ~ 35.46s

237k DPS with scoundrel @ 302% CD and 4% elite bonus

= ~ 8.40x

@ ~50-60 latency
Reply Quote
Standard SNS
Naked Scoundrel w/ Crit Passive

3.03 Atk/Sec
51.5% Crit
130 Crit Dmg
109,364.58 DPS

25,738,048 HP
~35s x3 Trials

735,372.8 eDPS

--------

6.724x Multiplier
96-100 AVG Ping
Reply Quote
Ghom MP8, standard SNS, no follower
2.77 APS, 49.5% CC, 18 APOC

3 trials avg 6.72x @ 95-105 ping
Reply Quote
Okay, I updated the values:
n = 10
peopleContributed = [novice, worx, tekkzero, killer elite, molitro, Worloega, cykikvisage, ximae, boozor, aimless]
latency = [500 300 129 116 100 100 98 55 55 24];
multiplier = [3.4 4.95 6.9 7.76 7.7 6.72 6.72 8.1 8.4 8.8];
Correlation Coefficient: r=-0.9555, p <0.0001
Bootstrapped Correlation Coefficient: r = -0.7760, p = 0.0083
Best Fit for Predicted DPSMultiplier: DPSMultiplier = latency*-0.0110 + 8.5645 (with normR = 1.4677)

P.S. As a heads-up to protect ourselves from the dataset being somewhat lop-sided (ie we now have a lot more data points <150ms latency), I'm also now performing a boot-strap of the small data-set (10,000 iterations). I'm taking the value at 5th percentile as our value of the correlation coefficient. This just means that 95% of the times that I computed the correlation coefficient (from the re-sampled data), I got a more significant value. The value has the same meaning as before, but now I'm protecting us against a few data points skewing the correlation spuriously upwards.
Edited by TekkZero#1963 on 1/16/2013 7:11 PM PST
Reply Quote
Posts: 5,376
View profile
Ghom MP8
oops messed up original count, corrected it now

80s / 79s = avg ~ 79.5s

70,625,208 HP / 80 seconds = 882815.1 eDPS. 882815.1/ 122399 = 7.212 multiplier.

122399 cause i have 6%bonus to elites and with scoundrel but i took off scoundrels gear to remove his dps as an effect

did i calc it rite tho my multiplier seems very ok for 95-120ms ping

i made sure to bring ghom to top left corner of map too before i turned on sa and started timer, etc

2.74aps, 53cc, 217 crit dmg
Edited by SteelPhantom#1820 on 1/16/2013 8:14 PM PST
Reply Quote
Did 3 mp10 trials, averaged 148s with 134.4k dps, no follower.
Avg 6.97 multiplier, 103 ms.
A/s: 3.03
ChC: 47%
ChD: 198%
APoC: 20
Reply Quote
Did it again at mp 8 w/o a follower..
77 seconds
120-130 latency
138k dps
2.76aps
cc 55.5
cd 249%

puts me at 6.62
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

Reason
Explain (256 characters max)

Reported!

[Close]