Monday, November 8, 2010

API WTF?

I don't know what an API Hook is, I'm just going to get that out there first thing. Today I'd like to talk a little bit about using our battleground scoreboard, as limited as it currently is, to pull out useful data when you're running a premade, or a fairly consistent rated-BG group. This is a thinking post, with the following caveats :

1) Right now, this is too much effort for too little reward, because it's nearly impossible to get a consistent group together, and the PUGs you're likely to be running with are not going to care about performance to the same level that you would hope out of a more dedicated cataclysm approach.

2) You currently have to gather this data YOURSELF, which is irritating. I'm not exactly sure what an API Hook is, like I said, but from the context of cynwise's post I'm going to assume that it means "addons can access the data from the scoreboard and manipulate/report it in various ways". So we can all dream and pray that that will happen someday, but in the meantime you have no tools but your vid-capping software of choice, and manual entry of the final scoreboard into a spreadsheet program like excel to manipulate them yourself.

Ok, those caveats out of the way, and being on hiatus and therefore unable to pick up a scoreboard screenshot myself, I have elected (with gracious permission) to analyse this WSG game that Cynwise participated in :



I have input that data into excel, and made the following spreadsheet:

Overall Competence

The most important thing here, obviously, is who won; hopefully you don't need a spreadsheet to tell you that much. But we are also looking for data on how efficiently you won -- I remember running a premade and having somebody quit after the first game (a win in EOTS, that we 4-capped) because we struggled a little at first and to him it was not enough that we win, we had to win smashingly and efficiently. It's an extreme example but people do expect that you will win solidly and efficiently -- if they wanted to grind out a 27-minute WSG that they won because they capped the last flag, they'd just risk it and queue normally. There are a number of stats that you can pull out from the data you're given that can help you analyse this:

Total HKs / Total KBs : You receive an honorable kill credit if you are within range of somebody dying, regardless of whether you contributed anything or not; hence there are going to be a lot more HKs than KBs. HKs are awarded to whoever is in range, even though it's all for the same kill, and that's the key to understanding this stat. HKs/KBs gives you a measure of group cohesiveness, as compared to the opponents' -- how well did your team move in a group, versus how many people wandered off alone? In this example, Cynwise's winning ally team had a "cohesiveness factor" of 6.40 compared to a Horde CF of 6.24. Statistically, a slight difference, but on a battlefield it's usually not a hugely disparate difference -- your victory (or loss) is the sum of a number of incremental factors, and here I would say that the Alliance was slightly better than Horde in this game at keeping together and moving as a group.

HKs Average : What you want here is to get an average of the HKs on your team, and find the standard deviation of it; statistically, assuming a normal distribution, 68% of your team will be within 1 st.dev. of that mean. I would stress here that we are looking at a matter of cohesion here, not strategic placement. This is an individual indicator of how well that person stayed with the group, and passes no mathematical judgment on where that group is on the field (maybe they're all farming in the mid, for example). In this example, the Ally HK average +/- 1 st.dev goes like 36.85 - 48.94. There are two people who come out above that (a warrior and a mage) and one below (pally). The warrior and the mage are good at scenting where the fighting is and being on top of it. The pally *may* have a problem -- the numbers give you a guide to finding potential problems, you still need to use your intelligence to work out if it actually is or not. It could also mean that he was a latecomer to the party and only got the last 5 minutes of the game. One may also note that his HKs scored / deaths comes in at a very low (9th out of 10) 3.4 -- and that stat is independent of time spent in the BG. Gear is also not necessarily a factor here; it contributes to more deaths but it should not stop you from moving to get with a group. So if I were running this as a premade group, I might keep any eye on those warning signs and watch him more closely in future runs. Speaking of that stat --

HKs / Deaths : What does this stat mean? Obviously you have to take into account some considerations when you're comparing to the other team, but presumably everyone on your team suffers from the same limitations regarding surfeit/lack of healers, so it's something of a level playing field. It's also not unfairly balanced against aggressive players -- people defending a quiet node will get less HKs, but also correspondingly less deaths. In this game the ally average for this ratio was about 5.18, with a standard deviation of 2.06. I take an average/stdev of 9 here, because they had a mage with them in this game with a ratio of 49 HKs to only 2 deaths, which does tend to have a large skewing effect. So if we can all just acknowledge his greatness, and then calculate without it, we'll be better off ;-). I would like to observe here that being on the bottom of this list DOES NOT MEAN YOU ARE A BAD PLAYER, necessarily. As Cynwise reminds, it's about individual excellence *in a team setting*, and the #10 worst performer in this category, a shaman, also capped a flag and returned one. I would venture to guess, with little knowledge to go on, that this person knew the objectives of the game and was one of those people (like me on occasion, *blush*) who goes repeatedly running off to Return The Flag! without checking to see if they have backup before so doing. So they've got the basis for solidness, but maybe their patience / part-of-the-group mentality needs a little work.

Healer Tools

A point near and dear to my heart, obviously, is the performance of your team's healers. Cynwise used this example, I presume from the name of the file, to demonstrate that it is not heals which win a BG, and that is true to some extent. Notwithstanding that, here are some useful stats that you can pull out of the scoreboard --

Healing Done / Opposing Damage Done : This stat is a measure of your healers' effectiveness at gauging (and getting to) hot spots on the field. Alliance here managed to heal 25% of the damage done, whereas horde managed to heal 50% of alliance damage done. Considering that the alliance *had no healers*, 25% is a fairly respectable number. Looking quickly at the other two scoreboards mentioned in that post of Cynwise's :

Game 1 (twink WSG, Alliance victory) : Alliance with 4 heals covered 60.12% of the damage; Horde with 4 heals covered 54.98%

Game 2 (80 EOTS, tie) : Unfortunately I can only see part of the scoreboard but alliance with 0/11 heals covered 18.6% of the damage shown, horde with 0/8 heals covered 11.96%. Horde's low here b/c a greater percentage of their people are not shown -- and alliance is high similarly. Now with no way of knowing what those invisible people are, I'm going to take a stab in the dark here and say that with no healers, self- /off-spec- / bandage- healing ~15% of the damage maybe would be appropriate with no heals.

So the 25% is quite standoutish! I would bet that there were a few people who were willing to heal even though that wasn't their primary job description, once they realised that their team had a significant lack. Another of those incremental factors.

Deaths / healer . Again, paucity of data prevents me from doing a detailed analysis of this but I would venture to say that a higher number here means either a) your healers were overtaxed and you needed more, or b) they weren't getting very good coverage (all 4 of your premade healers somehow ended up together at BS, for example). In our star example, the alliance number is undefined, horde had 67 deaths / 2 healers = 33.5 deaths per healer. One more piece of data I shall try to start tracking in cataclysm, I suppose :-).

A Last General Stat

Look at the number of objectives attacked (successfully or not) divided by the time of the BG -- this provides a rough measure (very rough) of how much activity there was. For example here,

Alliance : 2 flag caps, horde returned 6 times = picked up the flag 8 times in 27 minutes; there was *some* kind of activity around the horde flag every 3.375 minutes, in other words.

Horde : 1 flag cap, alliance returned 4 times = 5 flag events in 27 minutes, some kind of activity every 5.4 minutes.

See the difference? We can parse that out a little further actually, and make a general supposition that alliance was more aggressive on offense (they managed to at least pick up the flag, 3 more times than horde did) but that horde had a slightly stronger defense (they returned the flag 6/8 times which indicates a perhaps higher percentage of suicide 1-2 man grabs out of the EFR for alliance). But the alliance controlled the flow of that match -- they did more action, and horde was doing too much re-acting.

Welp, that's my story! Look forward in cataclysm to in-depth vid-capping / excruciating analysis of games such as this one; If raiders can spend hours poring over WoL to dissect success and failure in raids, there's no reason PvPers can't hold themselves to the same standard!

So what would you look for? What do various combinations of those stats suggest to you? I know I for one would love to see some kind of assist column for if you were within X yards of a flag capped, node assaulted / defended...but for the meanwhile we're stuck with what we've got!

And if cynwise shows up in the comments to tell me that my read of that game is 100% off-base then...I guess disregard everything I just wrote :-)