Diablo® III

CM Proc Rate (Testers needed)

Another note, just doing some testing myself and occasionally I get APoC returns with no yellow number onscreen, is this normal apo? Also having trouble getting my video quality good at slow speeds, AP seems to lag occaisionally, what framerates do you use?

My test was 1 APS 5CC 20 APoC, got an APoC return of 13 with RoF? Ill try get some more values

EDIT: few more tests, getting either 12 (incorrect sorry) or 13 APoC returns. Also I havnt yet seen a yellow number show with an apoc return? but lets not try and understand the display at this point haha

EDIT: just recorded a 14 APoC return...
Edited by Tobes#1649 on 2/28/2013 3:15 PM PST
Reply Quote
Another note, just doing some testing myself and occasionally I get APoC returns with no yellow number onscreen, is this normal apo? Also having trouble getting my video quality good at slow speeds, AP seems to lag occaisionally, what framerates do you use?

My test was 1 APS 5CC 20 APOC, got an APOC return of 13 with RoF? Ill try get some more values


RoF shouldn't display yellow numbers. All spells like that and WW, so called "DoT type spells" do not display crits like that. Instead they just display the amount of damage done over the last 0.5s, which may or may not include more than 1 actual damage tic in that timeframe.
Reply Quote
Ah that would explain it cheers
Reply Quote
This could explain the better performance of APoC since 1.07. But it can't be the only thing responsible for what we previously saw with CM and SA. Because some of the results we got there were so far off that even a proc coef of 1 couldn't explain them.
Also, all APoC tests were done with a single spell and at 1 APS only. So we can't know if the APoC coefficient is a completely separate value or based on the LoH coefficient, if it scales with IAS and so on.


Any possible chance of running that same exact test with the same apoc values at double the Aps.

I.e. as I understand it you tried to get exactly 1.000 aps correct? I would like to see if maybe they added an internal scaling with higher aps (thinking 1.500 or 2.000 if possible). Would like to know if they altered Apoc with aps to help with scaling attack speeds. This also could explain the better performance of 10 apoc now.
Edited by Harrowing#1449 on 2/28/2013 2:22 PM PST
Reply Quote
@Apo: I guess it's possible, but the more compelling issue is why you didn't see the '3s' so easily in the 6APoC case. Depending on your initial APoC return value you could flip it (ie >0.5), where you have mostly high values. But can you still predict your high results of 3APoC, along with the 6APoC results? As you said in your above post, we would expect a close sequence of similar numbers over time. This should be stereotyped.

So for example, let's try 1.8x as our APoC return, as that would correspond to a proc rate of 0.6. This would match your 3APoC results of 1, 2, 2, 2, 2, and 1:
Tick 1: 1.8+0.0-1=0.8 remainder
Tick 2: 1.8+0.8-2=0.6 remainder
Tick 3: 1.8+0.6-2=0.4 remainder
Tick 4: 1.8+0.4-2=0.2 remainder
Tick 5: 1.8+0.2-2=0.0 remainder
Tick 6-10: We then repeat the sequence exactly...

BUT let's say we double the APoC to 6, this would presumably make the 1.8 return now 3.6:
Tick 1: 3.6+0.0-3=0.6 remainder
Tick 2: 3.6+0.6-4=0.2 remainder
Tick 3: 3.6+0.2-3=0.8 remainder
Tick 4: 3.6+0.8-4=0.4 remainder
Tick 5: 3.6+0.4-4=0.0 remainder
Tick 6-10: We repeat the sequence exactly...

Thus, the sequence we'd predict would be something close to 3, 4, 3, 4, 4... So with the 6APoC case you would have had to see the 3 tick come up very frequently. Based on what you said, it was mostly 4s and you were having a hard time finding a 3. Even if we say the proc rate is not exactly 0.6, it would have to be pretty close to this series of numbers. Any thoughts?
Edited by TekkZero#1963 on 2/28/2013 2:28 PM PST
Reply Quote
@Tobes: I use the free version of Bandicam for recording. With the default settings = 30 FPS.
I don't have any lags, but I did have one strange game where I just couldn't get my AP globe to properly sync up. With the same test setup I was suddenly not getting 5 AP cost ticks, but 4 instead. It was still constant over a long period of time, so the balance of AP costs and regeneration was still intact. But it kinda freaked me out, so I started a new game and everything was back to normal again.
The AP globe tooltip is changing very frequently, so I can see how it might be influenced by high ping or low framerate.

AP return of 13 @ 20 APoC is in line with what I saw. But as we don't know what the APoC coefficient is for RoF (assuming this is really what's happening here - there is still no proof), you could be dealing with a fractional AP return. Which means you could be seeing values other than 13 with the same setup.
Reply Quote
@apo yea did a few more tests and I found a 14 APoC return also, got the video running a little smoother just had to slow it down better on VLC. Also what framerate are you running at in game?

@Harrowing its pretty easy running at 1 APS, you just sit in hell with defensive gear and no weapon and you can kill zombies for ages. Also the AP globe is really stable with a spell cost of 10 AP/s and same regen.
Edited by Tobes#1649 on 2/28/2013 2:36 PM PST
Reply Quote
@TekkZero I managed to get 12, 13 and 14 APoC returns from 20 APoC gear. Can you see any way that would fit your rounding theory? Its quite hard to get a good string of returns together to get the exact sequence btw

EDIT 12 was a mistake, so theory looks alright
Edited by Tobes#1649 on 2/28/2013 3:16 PM PST
Reply Quote
@Harrowing: The problem is, for my method to work best, you need to sync up AP costs and AP regeneration. At 2 APS, RoF costs 20 AP per second, so I would be constantly losing AP. That would be severely limiting test duration as well as complicating video analysis. I don't know how to work around that. You could try with skills like Diamond Skin - Prism and/or Astral Presence. But it has to sync up exactly, or it will screw up analysis. So that requires some serious planning.

@TekkZero: You are probably right. But I didn't really pay attention to the sequence, so I can't say much about that. For example one possible reason for me not finding the 3 AP, is that I mostly started channeling first and then hit the recording button. So it's entirely possible that it was always the first and I just missed it.
I don't have a clear overview of how the sequence looks like exactly and if it's the same every time.
It's definitely something we should look at more closely.

@Tobes: 60 FPS in game, sometimes falling a little below that while recording. I got vsync active, so it can't get any higher than 60.
Now to your test results: 13 and 14 means the proc coef would have to be higher than 0.65 and lower than 2/3
Because a coef of 0.65 and lower would result in AP returns of 13 or lower in your setup. And 2/3 or higher would mean I couldn't get 1 AP @ 3 APoC or 3 AP @ 6 APoC.
That fits and gives us a pretty good estimate. But you also mentioned a 12. And that actually makes the theory fall apart. Because it shouldn't be possible to get three different values.
Reply Quote
Just had another look back at the video and it looks like I miscalculated the 12 return, my AP was flicking very quickly to 18 before 19 and it often wasn't showing up show the starting value was too high. So far only got about 5 good readings, most were 13 and one 14
Reply Quote
Tried astral presence with 1.2 APS, 5CC 20APoC.

Got a string of 13, 14, 13 and then 13, 2 crits for 27 total.

Looks like attack speed isn't affecting how much AP you get per crit?
Reply Quote
That's good news. I can't think of any scenario in which three different return values make sense.
I'll try to verify with 20 APoC tomorrow.

Another thought for TekkZero and the sequence discussion: I always started channeling, then used energy armor several times to initially bring down my AP pool. So that further increases the chances of me missing the first proc.
We really need dedicated sequence tests. My data is not suitable for this kind of analysis.

[edit]Great! Thanks for testing. Didn't think of that simple solution for a different APS value. It would be nice to get some more values, but that would involve AP cost reduction. And I have absolutely no idea how AP cost reduction is calculated for channeled spells. Always wanted to test this, but never got to it.
Edited by apo#2677 on 2/28/2013 3:37 PM PST
Reply Quote
Another problem is I don't think you can get AP reduction on RoF. Could have a look at disintegrate though, you can get at least 14 reduced AP cost making it only cost 4 AP! AP cost reduction for channelled spells works the same as if it was a normal spell at your attack speed I believe. So if you got if you got 8 reduced cost to disintegrate you could run at 1 APS and be stable AP. Tomorrow if I have time and can be bothered!

EDIT you can actually reduce disintegrate to 1 AP so that is definitely the spell to play with!
Edited by Tobes#1649 on 2/28/2013 3:50 PM PST
Reply Quote
We have Diamond Skin - Prism (probably difficult to get 100% uptime) and Storm Armor - Power of the Storm for global AP cost reduction.
What I'm not sure about is the following:
AP costs are paid per tick. So the 10 AP cost listed in the RoF tooltip is actually a cost of 5 per tick.
At 1 APS, this results in 2 ticks/s * 5 AP = 10 AP/s.
At 1.2 APS it's 2.4 ticks/s * 5 AP = 12 AP/s. And so on.
So now what happens when we reduce AP costs by 3? Is it reducing the AP costs to 7 at 1 APS? That means costs of 3.5 per tick. So now we got a fraction again, great...

There was a post on diablofans.com several months ago, detailing AP cost with reduction gear for Disintegrate at several different attack speeds. And as far as I remember the test results didn't make as much sense as the poster would have liked. I think it never got explained properly.

Also, you have to pay attention to the respective breakpoints of each skill. RoF for example scales with APS exactly like WW (see https://us.battle.net/d3/en/forum/topic/6794871641 ). So tickrates are not always simply APS*2, it's a little more complicated than that.

Disintegrate should use the same tick rate progression algorithm, but roughly based on APS*3. So AP cost per tick is 6 and tick rate is based on the table below:
APS min - APS max Ticks/s
0.95239 - 1.00000 3.00000
1.00001 - 1.05263 3.15789
1.05264 - 1.11111 3.33333
1.11112 - 1.17647 3.52941
1.17648 - 1.25000 3.75000
1.25001 - 1.33333 4.00000
1.33334 - 1.42857 4.28571
1.42858 - 1.53846 4.61538
1.53847 - 1.66666 5.00000
1.66667 - 1.81818 5.45455
1.81819 - 2.00000 6.00000
2.00001 - 2.22222 6.66667
2.22223 - 2.50000 7.50000
2.50001 - 2.85714 8.57143
2.85715 - 3.33333 10.00000


So yeah, it's certainly possible to find some more test setups that work. But it might require some effort.
Reply Quote
@Tobes: If you're only seeing two values, then I'm inclined that it's closer to what apo has observed before and that it's not a sliding threshold (rather one of two values).

@apo: Well, the sequence doesn't have to start exactly right on the money. Mathematically, it should repeat, unless the new proc coefficient isn't 0.6. And even if it isn't exaclty that then it would just have some outliers, but the general sequence (if you observed about 3-4 times) should show a clear periodic nature (if it exists). I don't know how long you looked. And even if it's exactly repeatable, that would just suggest maybe the game uses a different method to decide whether to round up or down the APoC value.

But is it really a change in proc-coefficients? Maybe it's that weird 2x multiplier we were theorizing about earlier with CM (in 1.0.6). But I'm still having a hard-time stomaching that one, especially in light that we have only observe one tick per half second. But I don't see how our clumping hypothesis would even explain what we are seeing here (assuming it's relevant). One way we could see if there is a common unifying factor, would be to see if the APoC returns across some different skills all scale with the same modifier.
Edited by TekkZero#1963 on 2/28/2013 4:20 PM PST
Reply Quote
02/28/2013 04:19 PMPosted by TekkZero
One way we could see if there is a common unifying factor, would be to see if the APoC returns across some different skills all scale with the same modifier.


That should be pretty easy to test with MM - Seeker or some similar skill that has a proc coefficient of 1. Just drain your AP, mouse over your bubble, use the ability and see how much AP you gain.
Reply Quote
02/28/2013 04:19 PMPosted by TekkZero
@apo: Well, the sequence doesn't have to start exactly right on the money. Mathematically, it should repeat, unless the new proc coefficient isn't 0.6. And even if it isn't exaclty that then it would just have some outliers, but the general sequence (if you observed about 3-4 times) should show a clear periodic nature (if it exists). I don't know how long you looked. And even if it's exactly repeatable, that would just suggest maybe the game uses a different method to decide whether to round up or down the APoC value.

You are correct. But again: I need to repeat the test and focus on the sequence this time. In my opinion, the data I have right now does not allow me to draw any conclusions yet.

But is it really a change in proc-coefficients? Maybe it's that weird 2x multiplier we were theorizing about earlier with CM (in 1.0.6). But I'm still having a hard-time stomaching that one, especially in light that we have only observe one tick per half second. But I don't see how our clumping hypothesis would even explain what we are seeing here (assuming it's relevant). One way we could see if there is a common unifying factor, would be to see if the APoC returns across some different skills all scale with the same modifier.

It's still a theory. Everything we did so far was collecting evidence. And we didn't find anything disproving the theory. But that doesn't mean it is correct.
And even if there was a way to prove it, we still can't know if there's more to it. The CM stuff might as well be completely unrelated. Or it's a combination of the alleged proc coefficient change and a different, yet unknown bug/change.

I think looking at Disintegrate might give us some hints. But any other than those two spells are probably much harder to test :-/
Reply Quote
It's still a theory. Everything we did so far was collecting evidence. And we didn't find anything disproving the theory. But that doesn't mean it is correct.
And even if there was a way to prove it, we still can't know if there's more to it. The CM stuff might as well be completely unrelated. Or it's a combination of the alleged proc coefficient change and a different, yet unknown bug/change.
Agreed it's still a theory. I wasn't saying it wasnt, as we still need much more data to be sure. But it's a leading candidate, so it's worth testing explicitly. I agree the Disintegrate spell will be a good test method. I have a -5 Disintegrate Soj, so if I get time tonight I'll try to do a test.
Reply Quote
Forgot about break points, thats actually a good thing as it will make it easier to match APS to AP regen for disintegrate.

IMO AP should keep track of fractions, just like it is possible as a monk to get 2.33 spirit regen on piece of gear. And it also seems sensible that based on this we should see patterns in the amount of AP returned, so I guess we just need to get some good tests to confirm that something else isn't going on.
Reply Quote
trying to match AP regen and cost for arcane torrent and it isnt working as expected... Have a look at my profile, I should be stable at 12 AP cost per second, 12 AP gained per second while channelling but I seem to be losing AP at about 0.5 AP per sec.

Anyone know why? Might try disintegrate instead...
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)
Submit Cancel

Reported!

[Close]