Diminishing Returns - Miss Research

We know from blue that Miss avoidance gained from defense is subject to diminishing returns. The issue in discovering the rate at which it diminishes is that the current rate isn’t available through either the game UI or API commands.

No, the only way to work out ones Miss rate is to do the extremely tedious job of data collection. Namely to take a vast amount of attacks at various levels of defense and record the results.

So I took all my plate gear up to karabor and set to work. The Supplicants and Initiates up there have a high attack rate (dw), are level 70 and survivable for long periods as a protadin. Luckily I’ve collected a fair bit of gear specifically to farm here, so I was able to build a set that would survive versus ~30 of them for extended periods at about 450 def and beyond. The biggest issue became shield durability, which limited data to ~10k sample sets. I’m considering buying a set of Dawnforged Defender’s for further testing.

By nature this data is erratic and will have a high margin of error. Without an infinite data set I won’t get the exact number, what I’m aiming for is a zone, an area it falls into, which further data will gradually reduce.

Using Whitetooths data and formulas: http://elitistjerks.com/f31/t29453-combat_ratings_level_80_a/

Assuming K is a constant, the formula for a paladin versus a dual wield mob should be: 1 / ((1 / Cm) + (0.956 / Miss)) + 24 and we can calculate the Cm for each sample (which will differ due to the margin for error)



So, we have a range of 10.07 to 17.24. Both extremes come from lower sample sizes, and as we can assume that the larger the sample, the more accurate the data, its highly probable the exact figure falls between these points.

As a weighted average this equates so far to: 13.756


I'll attempt to collect more data before northrend distracts from such mundane tasks!

So does this imply defense is nearly always favorable to dodge/agility/parry?

Do the supplicants suffer from the dual wielding penalty?
_________________
http://armorylite.com/us/frostmane/fridmarr/

yeah there's tons of missing data here like: dw for npcs, lv80 avoidance stats not being equal to lv70, player level+3 mobs being the most important. probably best to test on beta or wait until thursday.

I believe that defense is now more valuable than the other ratings for one simple reason. Diminishing returns rewards a more balanced rating increase and punishes stacking individual ratings.

Previous to the addition of diminishing returns, every point of avoidance stats was worth more than the last. Now, it seems to be tuned more like armor, where the absolute increase in survivability is more linear. The problem with this in terms of avoidance versus armor, is that you have an interaction between three avoidance numbers that are each on a separate DR scale. This means that if you focus on dodge it will be severely diminished in its effectiveness per item point. The natural solution to this is to raise all three equally, which is accomplished via defense rating. Additionally, as miss is only gained through defense rating, it is required to remain balanced (as opposed to matching up dodge and parry rating amounts).

In the old days of crushing blows, defense rating was the superior stat at low item levels, followed (and surpassed at 102.4) by dodge rating. Now with diminishing returns in the picture, I think that defense rating will end up being superior even once you are block capped.

There is, of course, and additional complication in that the diminishing returns coefficients do appear to be different. So there may, in fact, be a certain proportion of dodge and parry rating to defense rating that would be optimal. Perhaps when sufficient data has been collected I can crunch some numbers to determine the optimal ratios of the three ratings.
_________________
Shadow Council - US - Horde
Xandreth (70) - Prot Paladin - MT
Tyrenn (80) - Frostfire Mage
Saege (70) - Resto Druid in training

So I threw together a quick spreadsheet and threw some sample numbers into it. On a strict 1 to 1 basis (assuming that all ratings carry the same item values) dodge is strictly better than defense which is in turn strictly better than parry. This is purely on the basis of avoidance (ignoring the block contribution from defense). At level 80 uncrittable values dodge gives slightly more avoidance than defense and remains larger even after DR through stacking over 2000 rating.

This calculation was done using the Cm derived by the OP. I think that the extremely low value of Cm is the reason that defense seems to scale more poorly than I anticipated. If this value of Cm is in fact accurate, then we will still use the old BC model of defense to uncrittable, then a mix of dodge rating and EH stats. The only real difference is that there will probably end up being more emphasis on EH and less on avoidance.

Can anyone tell me if the relative item values are the same between defense, parry, and dodge ratings? I'm at work so I have trouble looking up the details. If I can add all of the details and polish it up a bit, I can post the spreadsheet and any graphs if they appear to be of any use.

[edit] I went back and found the point of inflection. The point when defense becomes more valuable than dodge is around 4200 defense rating. (uncrittable + ~3500 dodge or defense rating) I highly doubt that we will be seeing ratings that high in the time frame of WotLK (it would put us in the realm of 75% avoidance).

[edit 2] I tweaked numbers a bit and it looks like an equal balance of defense rating and dodge rating may give the highest avoidance. I need to ponder this a bit to see if I can figure out how to model this to run calculate the optimal values. So far I have only been plugging numbers in to a simple calculation spreadsheet to see which ones give the max avoidance. Also, it looks like parry rating is still a pretty bad choice, not that I'm surprised.
_________________
Shadow Council - US - Horde
Xandreth (70) - Prot Paladin - MT
Tyrenn (80) - Frostfire Mage
Saege (70) - Resto Druid in training