Since the testing showed that glancing does in fact proc enchants, that column is essentially meaningless. It was just an attempt to see if that explained the boss-level reduction in PPM that DMM was seeing. Since that's not the case, you can basically ignore that column and we still have to come up with a theory about why the boss-level dummy is causing a reduction in PPM.
While it's possible that the answer is just "bosses cause a reduction in PPM because they are bosses," the scientist in me really wants to find an explanation, since it seems odd that the mob you are fighting would affect inherent properties of your weapon.
Since the testing showed that glancing does in fact proc enchants, that column is essentially meaningless. It was just an attempt to see if that explained the boss-level reduction in PPM that DMM was seeing. Since that's not the case, you can basically ignore that column and we still have to come up with a theory about why the boss-level dummy is causing a reduction in PPM.
Yes, that's kinda my point. Showing that column with the wrong data while it has been proven that glancings do proc berserking could confuse some people. But I do get the point of having it there before we knew.
Yeah, I simply was too lazy to change my script, which automatically generates the [table] output.
Some new data, this time will almost full equip, only empty slots were both trinkets.
Again, Mongoose and Berserking look similar. I think it's quite safe to assume that both enchants have the same PPM.
PPM did go up a bit, but I didn't search for my notes explaining how to calculate statistical certainty, so might still be perfectly in line.
Last edited by sp00n : 03/15/09 at 8:51 AM.
Reason: added combat log
sp00n, I assume the 2nd set of data is also on the Boss Level dummy? If that's the case, then it actually looks like these small data sets are disputing the fact that PPM adjusts with haste rating on gear. If you look at both sets, while the actual PPM is about 20% different, the PPM at Base Speed stays very close to 1 in all cases.
Looks like 1 PPM to me? Still haven't found my notes for statistical analysis, could somebody help me out here? I'm really really bad at reading formulae, so googling didn't help me much.
Want to write that down (in plain words!), so that I can refer to it and include it in my script.
Further questions: does the haste procced by Mongoose affect PPM? Does hit rating affect PPM (or rather, is PPM calculated against a theoretical 100% hit chance)?
I think expertise cap is not working well... I readed that the cap is 26 ( 214) but If I have more than 148 spreadsheet marks expertise red. I turned all buff to FALSE.
Is a bug or spreadsheet is asuming something I don´t know?
I think expertise cap is not working well... I readed that the cap is 26 ( 214) but If I have more than 148 spreadsheet marks expertise red. I turned all buff to FALSE.
Are you sure you don't have the talent Weapon Expertise turned on?
I've finally found the calculation for 95% confidence intervals (one I could understand at least). Hope they are correct, and if they are, it seems the data collected so far isn't enough.
Not sure if you can combine the tests in full equipment versus level 83 and 60. Certainly not if there is a reduction in PPM effects. However, the data doesn't really seem to support this - if I read it correctly.
I spent some time talking to Spoon about his ppm calculation and I think we need a third opinion.
His formula is as follows:
Observed PPM = (1 / Actual Time per swing) * 60 * Number of Procs / Number of landed attacks
So for instance, in his lvl 60 Full calculation, the hasted/actual weapon speed was 1.1628, number of procs = 276, and number of landed attacks was 13308, if you plug the numbers, you obtain 1.0702 PPM, according to Spoon's formula.
I would think that PPM should be calculated simply as Number of Procs/Time spent attacking measured in minutes. Because he spent some time repairing and getting ported to Dalaran, you can back out the time spent attacking from number of swings made and time per swing.
So if actual weapon speed is S, number of swings is N, number of procs is P, then time spent attacking is S * N seconds, or S * N / 60 minutes. Consequently, PPM is P*60/(S*N) or
PPM = (1 / Actual Time per swing) * 60 * Number of Procs / Number of swings.
Note the difference from Spoon's formula is Number of swings versus Number of Landed Attacks. Anyone can clarify which approach is correct?
Note the difference from Spoon's formula is Number of swings versus Number of Landed Attacks. Anyone can clarify which approach is correct?
I would say that Spoon is right, because the proc is a chance on HIT. If you don't connect, you can't proc. If you let all swings count, your method will yield different "proc rates" at different levels of hit. As far as I know, misses, dodges, and parries don't give you procs. I suppose this should be tested.
From what we know (or think we know) about PPM mechanics, the "PPM" is used to set a simple percentage based chance to proc depending on your weapon's speed. It doesn't cover for avoided swings. If you stood in the miss chance cloud on archavon and attacked for 10 minutes, you certainly wouldn't get anywhere near the number of procs you would get from attacking for 10 minutes outside the cloud.
This is not about whether dodges and misses can proc, in all likelihood they can't. But proc rate could still be fixed as proc per swing, so actual theoretical ppm is independent of your hit rating and experitse and is only determined by the type of the enchant. But observed ppm is decreased by your misses and dodges.
In another thread it came to my notice that the spreadsheet thinks that Hit gems are superior to AP when you're below 237 even when Heroic Presence is active (Ie Poison Cap is actually 210)
Is this a oversight from the time spellhit and hit were seperate?
I'm noticing this as well. I'm sitting at 213 hit rating with Misery and Heroic Presence in raid so my poisons are capped. Sheet still says hit is my best option, which if it is applying the buffs correctly would mean it is the best stat till the hard cap.
If I disable Heroic Presence and then play around with gems until my hit percentage is the same as it was before hit drops to an EP of ~1.4 instead of ~2.2 so it seems that Heroic Presence is not being applied to poisons.
This is not about whether dodges and misses can proc, in all likelihood they can't. But proc rate could still be fixed as proc per swing, so actual theoretical ppm is independent of your hit rating and experitse and is only determined by the type of the enchant. But observed ppm is decreased by your misses and dodges.
I fail to see what you're getting at. Let's say that a proc rate is fixed at X PPM across all swings. This translates into a per-swing proc rate of Y% based on your base weapon speed. Your suggestion that you would divide the experimental data by total swings (instead of total attacks landed) suggests that no matter what percentage of attacks miss/are dodged during the experiment, you should still experience about X PPM. That would imply that Y must increase if your chance to miss/be dodged increases. Although I haven't parsed any recent data, I'm fairly certain that we've not observed such an effect in the past at any time.
This is not about whether dodges and misses can proc, in all likelihood they can't. But proc rate could still be fixed as proc per swing, so actual theoretical ppm is independent of your hit rating and experitse and is only determined by the type of the enchant. But observed ppm is decreased by your misses and dodges.
Using your proposed method, going from 0%->5% dodge ignore, should generate more procs without changing number of swings. Using Sp00n's, both the number of procs and the number of landed swings change by the same amount. Of course the proc chance per hit could vary with you expertise and hit (which is what I am reading you propose in the quoted post), then it should effectively be the same as Sp00n's proposed formula. You will get different PPM's but they would have to be implemented differently in a spreadsheet yielding the same result.
This is not about whether dodges and misses can proc, in all likelihood they can't. But proc rate could still be fixed as proc per swing, so actual theoretical ppm is independent of your hit rating and experitse and is only determined by the type of the enchant. But observed ppm is decreased by your misses and dodges.
The thing is, if you say that the rate is fixed as "proc per swing", it seems logical to assume that all swings have an equal chance to proc. However, the reality is that when a swing doesn't connect, it can't proc, so we can't measure the proc rate that the misses and dodges "would have had". If all swings have an equal chance to proc, then the subset of swings that DO connect should exhibit the same proc rate as the swings that don't connect would have if they connected.
Given this information, we can say that sp00n's formula is in fact calculating what the per swing rate should be before your data gets muddled with misses.
I think the sticking point is the terminology. When he says "Observed PPM", he means "PPM number as calculated from observed proc rate", not "Procs observed per unit time". The first remains the same regardless of avoidance. The second obviously decreases as you are dodged, parried, etc.
Let's just look at the specific example, so that we can make sure we are talking about same thing. This file has an example of 10000 swings, with 1.1 actual speed, under two settings, one with 10% misses and another with 20% misses: PPM_Hypotheses.xls Download File on FileFront.
Then I consider two hypotheses. I'd like to figure out two things out of this, first is which of the two hypotheses is supported from previous tests people have done. Second is which formula should be used to estimate PPM. In particular if WOW mechanics fixes proc per minute rate at some theoretical value of say (1.2 for mongoose), how can we extract that value from data.
Under hypothesis A, wow fixes PPM at a rate of 1.2, which determines chance to proc per landed hit. If you increase your chance to hit, there are more landed hits per minute, which means chance to proc per landed hit must decrease to keep PPM constant. The number of procs stays the same, as hit rating changes. Spoon's formula calculates values of 1.5 and 1.33, while my formula which is based on procs/time gives a number of 1.2.
Under hypothesis B, wow fixes PPM at a rate of 1.2, but this time it determines the chance to proc per swing (however only connected swings have an ability to proc). Under this hypotheses, when you increase hit rating, chance to proc per swing stays same while the number of observed procs increases. In this case, Spoon's formula returns PPM of 1.2, and observed PPM based on procs/time gives lower values of 0.96 and 1.08 respectively.
If what you are saying is right, and hypothesis A is correct, then I think it's wrong to use Spoon's formula to estimate the theoretical parameter fixed by wow (the 1.2 fixed ppm rate). If you don't mind, please look at the file and tell me if I am wrong in my conclusions. Thanks.
Under hypothesis A, wow fixes PPM at a rate of 1.2, which determines chance to proc per landed hit. If you increase your chance to hit, there are more landed hits per minute, which means chance to proc per landed hit must decrease to keep PPM constant. The number of procs stays the same, as hit rating changes. Spoon's formula calculates values of 1.5 and 1.33, while my formula which is based on procs/time gives a number of 1.2.
Under hypothesis B, wow fixes PPM at a rate of 1.2, but this time it determines the chance to proc per swing (however only connected swings have an ability to proc). Under this hypotheses, when you increase hit rating, chance to proc per swing stays same while the number of observed procs increases. In this case, Spoon's formula returns PPM of 1.2, and observed PPM based on procs/time gives lower values of 0.96 and 1.08 respectively.
It is my understanding that all of the data collected so far seems to indicate that Hypothesis B is the correct way that WoW handles PPM mechanics. This is why sp00n is using the formula that he is, as it gives you the correct PPM ignoring the fact that misses and dodges take away from the actual "Procs Observed Per Minute."
Originally Posted by Trazhenko
I think the sticking point is the terminology. When he says "Observed PPM", he means "PPM number as calculated from observed proc rate", not "Procs observed per unit time". The first remains the same regardless of avoidance. The second obviously decreases as you are dodged, parried, etc.
I suspect this might be part of the confusion, as Trazhenko mentioned, when sp00n says "Observed PPM" it doesn't mean "How many Procs did I actually observe on average in a given minute?" it means "based on the observed data, what is the PPM number that WoW uses to figure out the chance to proc this effect per swing?"
If what you are saying is right, and hypothesis A is correct, then I think it's wrong to use Spoon's formula to estimate the theoretical parameter fixed by wow (the 1.2 fixed ppm rate). If you don't mind, please look at the file and tell me if I am wrong in my conclusions. Thanks.
Data collected so far indicates that Mongoose, Berserking, and our awesome new poisons on the PTR work like B.
When I say that the PPM determines a fixed rate, I mean that the PPM number in combination with the weapon speed determines a percentage proc rate for the effect. This percentage is the number that is rolled against to determine a proc.
Example: Say mongoose is "1 PPM" effect, and I have a 3.0 speed weapon. WoW assigns my weapon a 1 * 3.0/60 = .05 = 5% chance to proc mongoose.
PPM is a horrible term. We should call it a speed normalized effect. Even that is ambiguous, because it could mean that it's normalized to base speed (3.0 in my example), static hasted speed(accounting for haste rating on gear), or current hasted speed(accounting for temporary effects like WF and SnD).
I think some warrior effects, unbridled wrath maybe, used to be normalized to *current* hasted speed. That would be similar to Hypothesis A, but not quite, because as far as I know, avoidance has never been used to scale the assigned proc %. Only haste.
Data collected so far indicates that Mongoose, Berserking, and our awesome new poisons on the PTR work like B.
When I say that the PPM determines a fixed rate, I mean that the PPM number in combination with the weapon speed determines a percentage proc rate for the effect. This percentage is the number that is rolled against to determine a proc.
Example: Say mongoose is "1 PPM" effect, and I have a 3.0 speed weapon. WoW assigns my weapon a 1 * 3.0/60 = .05 = 5% chance to proc mongoose.
PPM is a horrible term. We should call it a speed normalized effect. Even that is ambiguous, because it could mean that it's normalized to base speed (3.0 in my example), static hasted speed(accounting for haste rating on gear), or current hasted speed(accounting for temporary effects like WF and SnD).
I think some warrior effects, unbridled wrath maybe, used to be normalized to *current* hasted speed. That would be similar to Hypothesis A, but not quite, because as far as I know, avoidance has never been used to scale the assigned proc %. Only haste.
Ok it makes more sense now. Hypothesis B is what I personally viewed as more plausible.
Just one more comment, since you brought in poisons, I actually thought that ppm poisons still work in yet another way differently from Mongoose. When you stack haste, the number of Mongoose procs per minute does not increase. At the same time, the new ppm poisons are normalized to 1.4 base speed. So stacking haste will increase their ppm. See for instance Asuah's findings on this page 3.1 Rogue Changes - Early Preview.
Since you seem to agree with the formula I'm using, I have a new data set with >50k swings versus a level 60 dummy. Only berserking and only one weapon, so no misses and doges at all.
Talents
0/0/0
Equipment
Full
(no trinkets)
Target
Level 60
Training Dummy
Hit Rating
255
7.7768%
Haste Rating
341
10.3995%
Weapon
Speed
Enchant
Adjusted Speed
% of Attacks
MH
1.4
Berserking
1.2681
100%
OH
0
none
0
0%
Type
Amount
Per cent
Attacks
53828
100%
Hits
38024
70.64%
Crits
15804
29.36%
Glancing
0
0%
Misses
0
0%
Dodges
0
0%
Landed total
53828
100%
Berserking
1.4 speed (1.2681)
Landed attacks
Lower
Upper
Swings
53828
Procs
1305
1235
1375
Proc % per swing
2.42%
2.29%
2.55%
PPM
1.1471
1.0855
1.2086
PPM with base speed
1.039
0.9833
1.0948
Confidence intervals are still too widely spread for my taste, hmpf.
03/13/09, 2:26 PM
Similar Threads
