paladins behaving badly

April 10, 2010

A commenter by the name of Tankadin asked me to look at some problems with my Protection Paladin stats page. First off, apologies for the time it has taken to respond to this; I’m always happy to check for errors.

Tankadin asked whether 1.2 was a reasonable minimum mainhand attack speed for prot pallies. That is what the data shows. Level 80 prot pallies come equipped with a variety of weapons in the mainhand slot. In my sample we have this for mainhand speed:

Min Avg Max Weapon Type Count
1.35 1.633 2.90 One-Handed Swords 14094
1.24 1.95 2.80 One-Handed Maces 2849
1.41 1.76 2.90 One-Handed Axes 1752
2.70 3.31 3.80 Two-Handed Axes 115
1.50 2.76 3.00 Miscellaneous 72
2.66 2.98 3.00 Fishing Poles 67
2.12 3.32 3.80 Two-Handed Maces 62
2.93 3.26 3.80 Two-Handed Swords 34
2.85 3.30 3.50 Polearms 14

The other question was: why do they show any offhand dps, when they should be equipped with shields. The answer to this one is a bit more disturbing. The stats shown in the armoury are all calculated on the server side (presumably to prevent people reverse-engineering the equations involved) and there is no direct connection in the XML between the stats and the items that the character is equipped with.

No doubt, somewhere in the bowels of Blizz’s servers, the stats are being computed by taking into account the bonuses granted by each item, gem, enchant et cetera, but when the data gets to your browser it is just a set of disconnected numbers.

What is happening is that a paladin who has nothing equipped in the offhand slot gets reported as having a non-zero offhand dps. It is a purely theoretical number since there is no weapon in the slot that could attack anybody using that dps. But the number is there in the XML.

So, regrettably, that table is… misunderestimated… There may be other examples of this in some of the stats reports and I’ll try to set aside some time to see if some better sanity checks can be put in place. Unfortunately there are limits to what can be done since I use a lot of pretty generic SQL to do the work.

Tankadin’s final questions were about whether the stats were being influenced by outliers. Well, outliers there certainly are. I found one one level 80 prot pally wielding this in the mainhand slot. If you’re ever tempted to equip it for the next guild raid, you’ll be pleased to know that it has an attack speed of 2.0!

The reason why that doesn’t bother me too much is that the outliers form a relatively small proportion of the sample set. Most characters adopt one or another of a very small subset of items. We might call them the “popular” group. We’ve remarked on this interesting distribution before, but its worth looking at a chart again. Here is a plot of the number of level 80 protection paladins equipped with each distinct item in the mainhand slot:

Mainhand Item Count for Level 80 Prot Paladins

You can see the “long tail” of outliers which stretches to the right hand axis of the chart, but the reality is that there are about 1800 toons in the long tail, compared to 22,000 toons in the popular camp.


9 Responses to “paladins behaving badly”

  1. tankadin Says:

    Thanks for looking into this.

    I’m not sure how the armory data numbers are so skewed, but there are no weapons in the game that a paladin can use that have a 1.2 weapon speed. In fact, the lowest actual weapon is 1.4 speed so the minimum cannot be lower than that for any weapon type. There is a 1.3 speed quest item sword, but it’s part of a Death Knight-only chain. Look for yourself…;cr=36:127;crs=4:2;crv=1.4:0

    Could the offhand dps be due to a shield spike, or possibly having a two-hand weapon equipped?

    As for the whelpling main-hand. That’s impossible. You cannot equip it in any slot. It’s a vanity pet. I’d suspect an error in Blizzard’s data. Possibly a database corruption of an similar item number that’s off by one digit or something.

  2. Owen Says:

    Could it be possible that the offhand damage is coming from Holy Shield?

  3. zardoz Says:

    There’s some odd things going on with the armoury data and some of these things need further investigation.

    A few points for now:
    1. The mainhand speed is not the speed assigned to the weapon (ie not the speed you see in the item tooltip for a weapon) but the value that is applied in game to the character. I’m no theorycrafter, but I’m sure the weapon speed is being combined with other values to produce the number I’m quoting. My number comes from the character stats part of the XML and not from the items part. I can’t see any example in the data where the character’s mainhand speed exactly matches what is quoted for the item in the mainhand slot.

    2. The offhand dps only shows up when the offhand slot is empty – ie no shield at all. But I’ll recheck that on Monday. I thought of the two-handed weapon thing as well but I ran out of time to set up a query to confirm it. I’ll have a look at that on Monday too.

    3. Yes I should have looked more closely at the whelpling page. The suggestion that the item ID is corrupted is a good one. I might see what other items IDs are close to the number, since I use the Blizzard item ID as my database key.

    4. I’d expect Holy Shield to show up in the defensive stats rather than dps, if temporary effects like that are included in the character stats.

    There is no doubt that there is bad data in the armoury. How much of it is bad is an interesting question. Maybe a lot of those items in the “long tail” are just noise; they’re all equipped by one toon only.

  4. Dyr Says:

    Discovered your site while looking for ways to scrape the armory.

    I am an actuary student looking for data to massage to keep my skills sharp while waiting for employment.

    If you need any help, please contact me.

    Great site. Keep up the good work.

  5. Cabal Says:

    Off-topic: Have the graphs on the app site been updated since 1/11/10? That seems pretty stale, given the changes in 3.3.3. Thanks.

  6. zardoz Says:

    I’m just refreshing my database for 3.3.3 now. Shoud have the site updated by the end of next week.

  7. Ozzard Says:

    I think the choice of the whelpling as a weapon is inspired… if you’re tanking Onyxia. I mean, what better way to keep aggro than to repeatedly slap her around the face with one of her own offspring?

    (Disclaimer: big smiley here. And I agree re the bad stats, though it’s difficult to see quite what process might corrupt so many)

  8. zardoz Says:

    It’s an interesting question – I’ve come across some clearly well-out-of-range stats in the armoury data. Maybe they really are in the game itself and the players who benefit (or lose out) aren’t noticing, or aren’t saying anything. Maybe the armoury code isn’t at the same standard as the game code. Blizz are clearly first-rate engineers so you’d think they’d have database constraints and other checks in place for the game itself.

  9. tankadin Says:

    I think Haste Rating is the culprit.
    Haste increases attack speed…

    So people like me that run PvP battlegrounds in Prot spec with Ret gear (but with a 1H and shield) are probably polluting your armory data if we forget to put our normal gear on when we log out. There’s a LOT of us.

    The only way to get reliable data for a given spec+gear combination is to figure out some way to filter these people. Just looking at their spec isn’t enough. I think we’re entering cat and bear territory here.

    I imagine it would be common for some tanks to have two, possibly three, pieces of Ret gear in their tanking set for extra threat, especially when running heroics, so you can’t really filter on that.

    Here’s a proposal…

    For each character:
    1) Determine their average item level – X
    2) Scale each stat based on X to “normalize” the data
    3) Remove people that are too far away from the average on any stat (or combination of stats)

    By scaling the stats with item level, you won’t be tossing out the very best and very worst geared level 80s.

    It will also naturally remove PvP players because the average resilience for a PvE player should be close to zero (though a lot of PvE tanks use the PvP shoulder enchant).

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: