[Satorri]

So, it make sense that parry is great for DK's though the presiding non-class-centric theory is still dodge > parry. DK's get the same cap and diminishing returns and DK's get parry from strength that pushes us along the curve (though surpasses the non-diminishing parry from Warr/Pally talents in %). And for pure survival/avoidance, it's not really in question.

What I want to get a more accurate sense of though, is just how much does parry buff our threat through parry-haste. What can we expect to see in response to a parry?

As a quick review for those who don't know or haven't looked in a while, when a player parries an incoming attack, it *can* take a cut out of the remaining swing timer thus making their next hit come sooner ("haste"). The mechanic is slightly complicated though as it has regions of effect. Here's how that works specifically:

When you parry an attack, the remaining time on your current swing is reduced by 40% of your weapon speed, unless this would result in a reduction to less than 20% of your swing time remaining

This is direct from WoWwiki's formula (though of course, they get their info from us, ha ha). One descriptor that seems to be missing in that is that if it would reduce your swing below 20% it will only go to 20%.

This means that if you just hit, and you parry, your swing will be knocked down to 60% remaining. If you have 60% left on your swing it'll be knocked down to 20% remaining. If you have 30% remaining, it will still only be put down to 20% remaining. 20% or less left and nothing will happen.


Here's where math gets rocky. First, depending on your swing time and the bosses, you could fall into a constant phase alignment (the boss always hits you when you're in your last 20%), or a shifting alignment (hits you at a different point in your swing each time) with a wide variety of shift values. Then add the complication of which swings are parried, another random element, and it gets even murkier. For simplicity sake though, let's imagine it is over a GARGANTUAN number of incoming swings and out-going melee. That means that we'll say:

• 40% of the swings will hit you when you can get the full 40% cut
• 40% of the swings will hit when you will get 0-40%, so we'll average those at 20% cut
• 20% of the swings get nothing

We'll also assume that on that giant average we'll get an equal count of each type. That means that on (GIANT) average each parry will represent 24% reduced swing time (this is where the value everyone cites comes from).

If the tank has a 1.5 sec swing timer (i.e. Warrior/Pally), 24% off is 1.14 swing time after a parry. So, if a warrior had 20% parry and a boss is attacking with a 2.4 sec attack speed:
No parries = 40 swings/min
20% parries = 41.2 swings/min (3% increase)

Let's now take a DK who has a 3.5 sec swing timer (i.e. 2h DK). 24% off is a 2.66 sec time, or almost a full second shaved off. With the same 20% parry and a 2.4 sec swinging boss (note: two parries during a single swing can both reduce the remaining time):
No parries = 17.1 swings/min
20% parries = 18.3 swings/min (7% increase)


Now let's look, using the same formula, at a range of parry ratings for the DK:
10% parry = 17.7 swings/min
15% parry = 18.0 swings/min
20% parry = 18.3 swings/min
25% parry = 18.6 swings/min
30% parry = 18.9 swings/min

More succinctly, each 1% parry is worth 0.06 extra swings/min, or 0.35% haste.

So, some composite values for tanking stats:

10 Str = 20 AP, 2.5 Parry Rating (0.05% parry before DR)
10 Agi = 20 armor (0.56 AP w/ 5/5 Blade Arm), 0.16% crit, 0.14% dodge
20 Armor = ~0.14-0.18% reduced physical damage taken (well-geared tank 26k-30k armor), 0.11 AP per point in Blade Armor
10 Def Rating = 0.08% dodge, 0.08% parry, 0.08% miss (0.24% total)
10 Dodge Rating = 0.25% dodge
10 Parry Rating = 0.20% parry, 0.07% haste
10 Hit Rating = -0.30% chance to miss with melee-based attacks, -0.38% chance to miss with spell-based attacks
10 Expertise Rating = -0.30% chance to be dodged AND -0.30% chance to be parried

(PS peese Blizzard, add Agility as a base stat to my t10 tanking gear kthxbye your BFF Satorri)

[Splug]

While I agree it's worth noting the parry-haste contribution is going to be higher for a tanking class with a higher ratio of incoming swings to outgoing swings, it still leaves open the question of whether the stat becomes more efficient than combining other raw stats such as dodge and expertise (or even dodge and haste, as a more direct comparison). While providing a silver lining to the hard-itemized parry rating on several items, unless the stat can actively become stronger than an amalgamation of subsidiary avoidance/threat stats, it still is not worth pursuing directly. A second point to note is whether the skew on those numbers changes appreciably with a faster boss attack speed (IE: Algalon), particularly when attacks are at or exceeding the frequency of 1h attacks and thus exceeding a 2:1 ratio against the 2h attacks. Also, I'm not sure you can disregard the effects of multiple parries on a single autoattack; as a subsequent parry against a hasted attack will by defintion occur in closer proximity to the weapon's swing, there is a higher probability of landing in the 20% parry-haste dead zone and no chance at all to land in the full-benefit region. In fact, for the example given a back-to-back parry would have to land within the dead zone: a full swing timer would be reduced from 3.5 to 2.66, meaning 2.4 seconds later when the next swing occurs the swing time remaining is .26, which will never benefit from parry haste. Though if my understanding is correct, at a faster boss attack speed or against multiple opponents, it would be possible to at least generate a partially-hasted swing.

That said, due to the diminishing returns on dodge rating (and the hard caps on accuracy ratings), it is reasonable that a point where parry surpasses dodge+haste does actually exist, though unlikely at a reasonable WotLK gear level. I also haven't sat down and crunched out the relative values of accuracy versus haste for tanking either; that may be interesting, particularly considering Rune Strike's increased threat and differing reaction to both expertise and haste relative to normal rotation abilities.

-Splug

[Satorri]

Thus why I added the small stat value breakdown at the bottom, but ooooo good idea, hit and expertise should be on that list.

I also did mention the wishywashy nature of these calculations. If the incoming swings and outgoing swings fall into an unfavorable phase you can lose the value heavily, while if they fall into a beneficial one it can really inflate the value. Boss swing speed surely makes a difference and I think it is worth highlighting your point:

The value of Parry for improving threat is not strictly absolute, but improves with the ratio of outgoing attack speed vs incoming attack speed, the higher that ratio (on the outgoing side > incoming) the bigger Parry's value becomes.

I hadn't put numbers to the idea yet, so this feels good. =)

[Molohk]

There's still a lot of questions about parry's contribution to threat, so I wanted to make estimations beyond the effects in auto-attack haste, and maybe do a comparative analysis between dodge and parry while I'm at it. The premise is: as a DK tank, dodge and parry contribute to threat through extra Rune Strikes; parry has the additional benefit of parry hasting which contributes to threat through extra Auto-Attacks; my theory is that parry-haste could have an additional benefit for slower weapons, by recovering lost Rune Strikes whenever a DK's weapon is slower than the boss's swing speed, and the DK dodges/parries twice before he has a chance to swing again; if the first avoided hit is a parry, then parry-haste could "recover" that lost RS oportunity.

Because I find it easier to build an auto-attack simulation than it is to make a probabilistic analysis of back-to-back parry/dodge combinations, I decided to run a few simulations with the next goals:

1. Quantify the auto-attack benefit of parry, and compare the results from model in the OP.
2. Quantify the rune strike benefit of parry and dodge, and make a comparative analysis.
3. Estimate the TPS contribution of parry and dodge, and make a comparative analysis.
4. Estimate the TPS contribution of parry rating and dodge rating, and make a comparative analysis.

The simulation follows the next rules:

• Simulation ran for 100 days of combat time (8640000 seconds).
• The boss' swing speed is constant (set at 2.4), there are 3.6 million boss swings during combat time.
• Each boss swing will be labeled with 3 possible values, through a single-roll simplified combat table RNG:

• Swing is parried (procs Rune Strike, and parry hasting accorging to #5).
• Swing is dodged (procs Rune Strike).
• Swing is a hit/miss.
• The player swing speed is constant (set at 3.5), but every time a player parry occurs the scheduling for the next swing will be re-evaluated (see #6).
• A Runic Strike is recorded whenever a player swing follows a parry or dodge.
• When the player parries, the next swing will be recorded as a hasted attack, and it's scheduling will be affected in the following way:

• If the next swing is scheduled to occur in more than 60% of the player's swing speed, the scheduling will be shortened by 40% of the player's swing speed.
• If the next swing is scheduled to occur within 60% and 20% of the player's swing speed, the next swing will be re-scheduled to occur in 20% of the player's swing speed.
• If the next swing is scheduled to occur in less than 20% of the swing speed, the scheduling is unchanged.

I ran two sets of simulations: the first set had a constant chance to dodge (25%) and simulations were run for each percentage of parry chance from 0 to 40; the second set had a constant chance to parry (25%) and simulations were run for each percentage of dodge chance from 0 to 40. The two sets allow comparative analysis of the benefits gained through dodge and parry. Since parry can yield threat through additional auto-attacks and additional rune-strikes (in theory), I evaluated both elements independently. The goal is to quantify both ways in which a tank's threat could benefit from parry: gaining extra auto-attacks, and gaining extra rune strikes.

Extra auto-attacks
This element is probably better analized trhough modeling, rather than simulation, but I thought it would be interesting to see if the simulation supports the model shown in the OP, and the conclusion that 1% parry is worth 0.06 extra swings or 0.35% haste. Here's the summary of the results obtained through simulation:

The graph shows how many extra auto-attacks the player gained with each 1% increment of parry chance. Interestingly enough, the returns on parry start around where the model in the OP suggested (0.06 swings per minute), but the simulation shows increasing returns of extra auto-attacks with each increment of parry chance. The only possible reason I can think of for these increasing returns is the fact that by accelerating swing speed the player is decreasing the chances for 2 parries to happen between swings, in which case the second parry is likely to yield a less significant gain. Any thoughts on this are welcome.

Extra Rune Strikes
Because dodge also yields extra Rune Strikes, I had to do a comparative analysis to figure out if in fact parry can yield additional Rune Strikes over dodge as proposed by my theory, the simulation should show these results. Here's the summary of the results obtained through simulation:

The results are also interesting, dodge appears to have diminishing returns of extra Rune Strikes, perhaps because of the posibility of dodging/parrying twice between swings, while parry seems to have a stable return, perhaps because the increased swing speed and the "recovered" rune strikes are just enough to counter for the diminishing returns. This means the more dodge you have, the more extra rune strikes you could get from parry (as oposed to dodge), however the difference is not very big, if a DK is sitting at 20% parry and 25% dodge, a % of parry would only yield around 0.035 RS per minute over a percentage of dodge.

Hypothetical TPS calculations
Lets see how that could translate into actual TPS. If a player has 4,500 AP, 15% crit chance, and is wielding one of the U10 3.5 weapons (to stay in line with the weapon speed used in this thread), the numbers would look like this:

Auto-Attack hit: 1943
Auto-Attack crit: 3886
Avg AUto-Attack (85% hit, 15% crit): 2234
Avg Auto-Attack threat: 4633 (edit: used to 9373 which was wrong, thanks Satorri)

Rune Strike hit: 3616
Rune Strike crit: 7232
Avg Rune Strike (75% hit, 25% crit): 4520
Avg Rune Strike threat: 14059

Using the trend lines from the graphs and the threat values listed above, we can estimate the tps gains per percent dodge/parry at different levels, and come up with this data:

Dodge    RS gained  TPS from  Parry    AA gained  RS gained  TPS from
percent  per min    dodge     percent  per min    per min    parry
15%      0.206      48.27     15%      0.074      0.229      59.37
20%      0.198      46.39     20%      0.078      0.230      59.91
25%      0.19       44.52     25%      0.082      0.231      60.46
30%      0.182      42.65     30%      0.086      0.232      61.00

As expected, chance to parry has a higher return of TPS than chance to dodge. But lets see what happens when we take into consideration the rating budgets and diminishing returns for each stat, by calculating how much rating it takes to gain those TPS values at the differente parry/dodge levels (using Whitetooth's formulas).

Dodge	 Rating per   TPS per   Parry    Rating per  TPS per
percent  Dodge %      D-rating  percent  Parry %     P-rating
15%      42           1.15      15%       71         0.84
20%      48           0.97      20%       93         0.64
25%      54           0.82      25%      128         0.47
30%      63           0.68      30%      189         0.32

The estimation shows that the increased TPS gains from parry in most cases are not enough to offset the higher rating cost and deeper DR for parry. In fact, a player could only gain more TPS from parry rating over dodge if his parry chance was bellow 19% and his dodge chance above 26%, and even within that range. More importantly, the 19% limitation makes it impossible to stack parry for the purpose of gaining threat.

Conclusion
The actual TPS gain from parry hasting seems very underwhelming (I'm a bit disapointed, lol). In fact, even though parry yields extra auti-attacks and extra rune strikes, under most circumstances dodge rating is still better than parry rating in terms of threat. I guess the old adage "never gem for Parry" is still valid, and I'd just complement it to "never gem for Parry, even if you want more threat".

PS: If you're interested, you can download the simulator here: http://parry-haste-simulator.googlec...s/parry_sim.pl. It's in Perl, and very simple.

[Molohk]

These are the data points I gathered with the first set of simulations, used to quantify the auto-attacks gained by each percentage of chance to parry.

Parry                 Player       Gained    Swings per
Chance    Parries     Swings       Swings    Minute Gained
0%            -      2,468,572          0      0.00
1%         36,081    2,478,550      9,978    0.0693
2%         72,225    2,488,599     10,049    0.0698
3%        107,574    2,498,529      9,930    0.0690
4%        143,711    2,508,738     10,209    0.0709
5%        179,615    2,518,710      9,972    0.0693
6%        216,613    2,529,195     10,485    0.0728
7%        252,406    2,539,393     10,198    0.0708
8%        287,942    2,549,308      9,915    0.0689
9%        322,906    2,559,491     10,183    0.0707
10%       360,528    2,570,409     10,918    0.0758
11%       395,997    2,580,823     10,414    0.0723
12%       431,275    2,591,228     10,405    0.0723
13%       468,488    2,601,801     10,573    0.0734
14%       504,048    2,612,355     10,554    0.0733
15%       538,744    2,623,012     10,657    0.0740
16%       574,310    2,633,836     10,824    0.0752
17%       611,395    2,644,755     10,919    0.0758
18%       647,949    2,656,123     11,368    0.0789
19%       683,571    2,667,196     11,073    0.0769
20%       719,642    2,678,206     11,010    0.0765
21%       758,264    2,690,594     12,388    0.0860
22%       792,031    2,700,863     10,269    0.0713
23%       829,463    2,713,277     12,414    0.0862
24%       862,720    2,723,773     10,496    0.0729
25%       899,821    2,735,857     12,084    0.0839
26%       937,387    2,748,431     12,574    0.0873
27%       972,162    2,759,815     11,384    0.0791
28%     1,006,930    2,771,262     11,447    0.0795
29%     1,042,675    2,783,102     11,840    0.0822
30%     1,080,254    2,795,956     12,854    0.0893
31%     1,115,209    2,807,978     12,022    0.0835
32%     1,151,174    2,820,168     12,190    0.0847
33%     1,187,298    2,832,814     12,646    0.0878
34%     1,223,439    2,845,096     12,282    0.0853
35%     1,260,256    2,858,297     13,201    0.0917
36%     1,294,339    2,870,476     12,179    0.0846
37%     1,332,948    2,884,819     14,343    0.0996
38%     1,367,372    2,896,504     11,685    0.0811
39%     1,404,638    2,910,153     13,649    0.0948

These are the data points I gathered with the second set of simulations, used to quantify the rune-strikes gained by each percentage of chance to parry and dodge.

Gained   ---- with parry ----   Gained   ---- with dodge ----
Parry    # of RS   RS per min   Dodge    # of RS   RS per min
0%       829,881        -       0%       900,041        -   
1%       863,099    0.2307      1%       932,737    0.2271 
2%       896,291    0.2305      2%       965,433    0.2271 
3%       927,785    0.2187      3%       999,321    0.2353 
4%       961,595    0.2348      4%     1,030,971    0.2198 
5%       993,459    0.2213      5%     1,063,401    0.2252 
6%     1,028,531    0.2436      6%     1,093,925    0.2120 
7%     1,059,964    0.2183      7%     1,125,986    0.2226 
8%     1,092,888    0.2286      8%     1,157,109    0.2161 
9%     1,126,588    0.2340      9%     1,188,562    0.2184 
10%    1,159,257    0.2269     10%     1,221,172    0.2265 
11%    1,192,710    0.2323     11%     1,252,042    0.2144 
12%    1,225,404    0.2270     12%     1,282,364    0.2106 
13%    1,258,927    0.2328     13%     1,311,417    0.2018 
14%    1,292,670    0.2343     14%     1,340,119    0.1993 
15%    1,324,829    0.2233     15%     1,371,427    0.2174 
16%    1,357,383    0.2261     16%     1,401,535    0.2091 
17%    1,391,805    0.2390     17%     1,429,134    0.1917 
18%    1,424,425    0.2265     18%     1,459,233    0.2090 
19%    1,458,362    0.2357     19%     1,487,666    0.1975 
20%    1,490,076    0.2202     20%     1,516,915    0.2031 
21%    1,525,171    0.2437     21%     1,544,438    0.1911 
22%    1,557,316    0.2232     22%     1,573,912    0.2047 
23%    1,590,487    0.2304     23%     1,599,371    0.1768 
24%    1,622,531    0.2225     24%     1,627,345    0.1943 
25%    1,654,378    0.2212     25%     1,655,273    0.1939 
26%    1,689,245    0.2421     26%     1,683,795    0.1981 
27%    1,723,042    0.2347     27%     1,710,281    0.1839 
28%    1,755,253    0.2237     28%     1,735,462    0.1749 
29%    1,788,551    0.2312     29%     1,762,760    0.1896 
30%    1,822,755    0.2375     30%     1,789,240    0.1839 
31%    1,855,891    0.2301     31%     1,816,959    0.1925 
32%    1,888,730    0.2280     32%     1,841,072    0.1675 
33%    1,920,483    0.2205     33%     1,868,136    0.1879 
34%    1,954,805    0.2383     34%     1,893,842    0.1785 
35%    1,988,879    0.2366     35%     1,917,813    0.1665 
36%    2,020,768    0.2215     36%     1,943,333    0.1772 
37%    2,056,443    0.2477     37%     1,967,368    0.1669 
38%    2,088,374    0.2217     38%     1,992,062    0.1715 
39%    2,124,106    0.2481     39%     2,016,920    0.1726

[Satorri]

Well done, Molohk! Looks solid, only one nitpicky detail that shouldn't make a big difference is that it is possible for you model to have the simulated DK swing almost twice in a moment. On the grand average though it should average out, not a big deal.

2 other curiosities:
1.) Did you add the +50% threat factor on your threat values (I wonder from looking at your sample swing vs RS threat values)?
2.) The TPS from Dodge/Parry affecting RS should be rather higher, no? Since the two combined should sum up to the full TPS value of RS? Maybe I'm just mind-eyeballing the numbers wrong.

Very nice to see this put to numbers in a way I don't have the software to calculate. =)

[Splug]

While that's a very detailed report, I do not understand how you are calculating several things, particularly rune strikes per minute. I don't understand the formula used to arrive at the values in the dodge RS/min category in the second chart. If they're calculated in terms of (X/3,600,000*60), then a higher value of X should yield a higher result. In that chart, it does not. If you were instead going for change in rune strikes gained per minute (which would be the derivative with respect to parry chance, RS/(min*parry)), it would be ((X1-X2)/3,600,000*60), but a handful of test points show everything being off by a factor of about 2.4.

The overall trend seems appropriate either way since from what I can see, that 2.4 factor is in every calculation. But I'm curious where it came from.

-Splug

EDIT: Hrm, it'd be an approximation of the derivative with respect to parry/dodge chance, not with respect to time.
EDIT2: Explained below: discrepancy is an error on my part, caused by using the wrong coefficient.

[Molohk]

Source: Satorri Well done, Molohk! Looks solid, only one nitpicky detail that shouldn't make a big difference is that it is possible for you model to have the simulated DK swing almost twice in a moment. On the grand average though it should average out, not a big deal. 2 other curiosities: 1.) Did you add the +50% threat factor on your threat values (I wonder from looking at your sample swing vs RS threat values)? 2.) The TPS from Dodge/Parry affecting RS should be rather higher, no? Since the two combined should sum up to the full TPS value of RS? Maybe I'm just mind-eyeballing the numbers wrong. Very nice to see this put to numbers in a way I don't have the software to calculate. =)

What do you mean that the DK can swing twice in a moment in my Model? If that's so, it's an error that I'd like to fix.

1) Thanks for spotting it, I'm not sure why I wrote the wrong value for Auto-Attack in the post. The graphs and calculations were actually done with the correct threat value of 4,633 though, there error is only on the post.

2) Yes, that's correct. Look at it this way, a DK with 3.5 weapon speed will auto-attack 17 (rounded down) times in a minute (not counting parry-haste), normally around 55% of those swings should be RS which translates into 9 (rounded down) RS. If you add all the RS per minute gains from dodge 1%-25% and parry 1%-20% (assuming 25% dodge and 20% rating) you get 9.44 RS per minute. With an average threat value of 14056 per RS, that's a total of 132716 threat per minute or 2211 TPS. If you split that 2211 between each of the 45% of parry+dodge, you end up with an average of 49 TPS per second for each percentage of parry or dodge, which is consistent with the TPS values listed.

Source: Splug While that's a very detailed report, I do not understand how you are calculating several things, particularly rune strikes per minute. I don't understand the formula used to arrive at the values in the dodge RS/min category in the second chart. If they're calculated in terms of (X/3,600,000*60), then a higher value of X should yield a higher result. In that chart, it does not. If you were instead going for change in rune strikes gained per minute (which would be the derivative with respect to parry chance, RS/(min*parry)), it would be ((X1-X2)/3,600,000*60), but a handful of test points show everything being off by a factor of about 2.4. The overall trend seems appropriate either way since from what I can see, that 2.4 factor is in every calculation. But I'm curious where it came from. -Splug EDIT: Hrm, it'd be an approximation of the derivative with respect to parry/dodge chance, not with respect to time.

I'm not sure why you'd like to divide by 3.6 mill. The simulation is 8,640,000 seconds longs which translates into 3.6 million swings for the boss (swing speed of 2.4) and a variable number of swings for the player because of parry haste. And the idea is to calculate the incremental gains per each % of parry (like you said, the derivative) so the RS per minute are calculated ((X2-X1)/8,640,000*60).

I used the 2.4 boss attack speed to stay in line with Satorri's OP, which is a base speed of 2 swings per second, affected by imp TC or imp IT.

[Satorri]

Maybe it didn't work this way in the model but you describe that:

100%-60% of swing remaining reduces swing time by 40%
60%-20% of swing time remaining reduces swing time by 20%
20%-0% of swing time remaining reduces swing time by 0%

That's fair for napkin math and big averages, but remember that in that set of conditions, if the swing timer is at 21% the model will cut the remaining time to 1%. The way the game works makes it so it will never be cut to less than 20% (thus the average, at 60% it'll be a full 40% off, 50% it's only 30% cut, 40% only 20% goes, and at 21% only 1% is cut).

Like I said, I don't know that it would really mess up the numbers, since you're doing such a high sample size it should come out in the wash, but there it is if you want to make adjustment to a more complicated conditional. =)

[Molohk]

Oh, maybe I wasn't clear when I explained the rules: between 60% and 20% the simulator cuts the remaining time down *to* 20%, not *by* 20%. If the swing timer has 60% left it will go down to 20%, and if it has 21% left it will also go down to 20%.

[Satorri]

Perfect, like I said, I figured it was more likely a matter of the description than the actual simulator. One doesn't go through the trouble of setting it up without some good attention to detail.

=)

[Splug]

Source: Molohk I'm not sure why you'd like to divide by 3.6 mill. The simulation is 8,640,000 seconds longs which translates into 3.6 million swings for the boss (swing speed of 2.4) and a variable number of swings for the player because of parry haste. And the idea is to calculate the incremental gains per each % of parry (like you said, the derivative) so the RS per minute are calculated ((X2-X1)/8,640,000*60). I used the 2.4 boss attack speed to stay in line with Satorri's OP, which is a base speed of 2 swings per second, affected by imp TC or imp IT.

Ok, that explains it. Somehow I transposed the time with the swing count when I was reading it (probably had to do with subconsciously correlating 3600*10^x with some factor of hours due to including 60^2). Since the multiplier between the swing count and the time was 2.4, that explains where my missing factor came in.

The numbers look correct to me then, with one other question: what did you use as non-diminished portions for parry/dodge, or were those numbers just looking at the component from rating? For example, swordshattering and anticipation would not contribute to DR if I remember correctly, meaning the values presented would not be the parry/dodge values on the character sheet. Either way the model still holds valid, though the gap between diminished parry values and diminished dodge values may be a bit closer/further than it appears. The ultimate lesson holds true either way: parry rating is no more viable for threat generation than dodge rating, and less effective for avoidance.

-Splug

[Satorri]

Isn't that disappointing? I'd love to get a Blizz rep/dev's feedback on this, maybe even a veiled suggestion to something we're missing that could account for this.

[Molohk]

Source: Splug The numbers look correct to me then, with one other question: what did you use as non-diminished portions for parry/dodge, or were those numbers just looking at the component from rating? For example, swordshattering and anticipation would not contribute to DR if I remember correctly, meaning the values presented would not be the parry/dodge values on the character sheet. Either way the model still holds valid, though the gap between diminished parry values and diminished dodge values may be a bit closer/further than it appears. The ultimate lesson holds true either way: parry rating is no more viable for threat generation than dodge rating, and less effective for avoidance.

I considered a 11% chance to dodge and 7% chance to parry that is not subject to diminishing returns. I basically got the numbers from my naked unbuffed stats (I had 10.03% dodge and 5.95 parry), and I added the 1% you get from RotSG.

For example, to calculate how much it takes to get from 15% parry to 16% parry, I used the following math:

After Diminishing Returns:
15% parry = 7% base parry + and 8% from gear
16% parry = 7% base parry + and 9% from gear

If you solve the DR formula to know how much you need before DR, you get:
15% parry = 7% base parry + 9.2167% from gear
16% parry = 7% base parry + 10.6416% from gear

That means you need 1.4249% to get from 15% to 16%, and 1.4249% parry = 71 parry rating. Since I already knew I gain 59.37 TPS by going from 15% to 16%, you can say it costs 71 parry rating to gain 59.37 TPS, and therefore you gain aproximately 71 / 59.37 = 0.84 TPS with every point of parry rating.

Source: Satorri Isn't that disappointing? I'd love to get a Blizz rep/dev's feedback on this, maybe even a veiled suggestion to something we're missing that could account for this.

Absolutely. Before I ran the sims I was expecting (hoping?) TPS from parry to be higher than it's normally regarded. So I was really disapointed when I realized that TPS from parry rating is actually lower than TPS from dodge rating (unless you find yourself with very low parry and very high dodge).

The bottom line is that parry rating costs as much as dodge rating (from an item budget point of view) yet parry rating gives DKs less avoidance and less TPS than dodge rating. So what's the point of parry? I think DKs were a good oportunity for Blizzard to make something out of parry, and I also wonder why Blizzard felt the need to punish parry with such harsh DR and such an expensive rating conversion.

[Satorri]

Well, like I said, I'm sure there is some rationale behind it, and I suspect there is some larger scale phenomena we aren't seeing or thinking of where Parry would become really too powerful. I want to believe that anyway.

Generally, I think it's just a need for a bit of tweaking and they tuned it a little too much in the direction of toning it down.

[Molohk]

Yeah, I know what you mean. I'm sure Blizzard didn't use /roll to come up with the parry cap, and rating values. But WTF.

[Molohk]

From PTR patch notes:

Dodge Rating: The amount of dodge rating required per percentage of dodge has been increased by 15%. This is before diminishing returns. Combined with other changes, this makes dodge rating and parry rating equally potent before diminishing returns apply. ... Parry Rating: The amount of parry rating required per percentage of parry has been reduced by 8%. This is before diminishing returns. Combined with other changes, this makes dodge rating and parry rating equally potent before diminishing returns apply. Parry still diminishes more quickly than dodge.

If these changes go through, the new coefficient for both dodge and parry would be 39.34798813*1.15 = 49.18499*0.92 = 45.25019. For fun, lets re-calculate TPS contributions from dodge and parry rating with this new proposed value and see how it looks.

Since the actual number of RS and Auto-Attacks gained per percentage of dodge and parry chance isn't affected by the new coefficient, and we only need to update the raiting points needed to increase parry/dodge by 1%, we only have to re-calculate the last of my tables (assuming they haven't changed the DR formula).

The TPS contributions, with the new coefficient taken into consideration, look like this:

Dodge	 Rating per   TPS per   Parry    Rating per  TPS per
percent  Dodge %      D-rating  percent  Parry %     P-rating
15%      48           1.15      15%       65         0.84
20%      55           0.97      20%       86         0.64
25%      62           0.82      25%      118         0.47
30%      72           0.68      30%      174         0.32

The new graph looks like this:

To continue the disapointment, at each percent, dodge rating is still better than parry even after the proposed coefficient changes, but they are much close now, close enough to make parry worth considering if you already have a higher chance to dodge (as we most do). To sum it up, if the change goes through, you would actually gain more TPS from parry rating if your parry chance is lower than 22.5% and your dodge chance is higher than 30%.

For my personal stats, just by looking at the chart, since I currently have 26.5% dodge and 19.4% parry, I'd be gaining just about 0.7 TPS with 1 point of dodge rating and 1 point of parry rating, however, because of DR I'm still gaining less avoidance from parry. It seems like the change is intended to be an avoidance nerf, rather than a balancing of parry rating vs dodge rating.

As a side note, because of DR and depending on your personal stats, you still need anywhere from 35% to 141% more parry rating to achieve 1% of avoidance, than the dodge rating needed to gain the same 1% of avoidance.

[GravityDK]

Source: Molohk As expected, chance to parry has a higher return of TPS than chance to dodge. But lets see what happens when we take into consideration the rating budgets and diminishing returns for each stat, by calculating how much rating it takes to gain those TPS values at the differente parry/dodge levels (using Whitetooth's formulas).

I couldn't tell if you'd used the new item ratings:avoidance ratio that will come into effect in 3.2, or the current live ones?

[Satorri]

I've been rather busy, I'm going to have to hop on the PTR tonight to see how this will work out for my avoidance.

I like what I'm seeing though, since they haven't said anything about changing parry's itemization value (since we get slightly more than we would of dodge rating). I do recall in playing with the gear comparison toy on WoWhead that trading Naxx gear into Ulduar gear was losing a LOT of parry and gaining a lot of dodge.

Still, it should be a less painful reduction for DK's than for the other classes, since we always carry a lot more parry rating just from Forceful Deflection (man, that name just makes parry sound fun, doesn't it?).

[Molohk]

Source: GravityDK I couldn't tell if you'd used the new item ratings:avoidance ratio that will come into effect in 3.2, or the current live ones?

The numbers in my post from 6-30 were calculated with the current rating conversion coefficients (39.34798813 for dodge and 49.18499 for parry). The numbers in my post from yesterday (7-7) were calculated with the current 3.2 PTR coefficient (45.25019 for dodge and parry).

Source: Satorri I've been rather busy, I'm going to have to hop on the PTR tonight to see how this will work out for my avoidance. I like what I'm seeing though, since they haven't said anything about changing parry's itemization value (since we get slightly more than we would of dodge rating). I do recall in playing with the gear comparison toy on WoWhead that trading Naxx gear into Ulduar gear was losing a LOT of parry and gaining a lot of dodge. Still, it should be a less painful reduction for DK's than for the other classes, since we always carry a lot more parry rating just from Forceful Deflection (man, that name just makes parry sound fun, doesn't it?).

I haven't had a chance to play around in PTR, but I'm thinking the nerf to overall avoidance shouldn't be too bad, and it's a step in the right direction for balancing both avoidance stats. But I still don't see why anyone would pick parry rating over dodge rating, using the PTR coefficient, I would get the same TPS from both stats but would I get around 30% more avoidance from dodge, just because of DR. (I'm also assuming itemization values are staying the same)

I remember there was a blue post implying that gear from Ulduar, even if iLvl was the same, would have "more optimized stat distribution". Maybe having some stats intentionally inferior to others (parry inferior to dodge) is part of it, that way they can replace parry rating with dodge rating to make an item "more optimized" under the same itemization budget.