<Trained Warlock>
After the ‘death pie’ the other day I mentioned I wanted to do a graph showing damage taken in 3v3 arena. I settled on what you see below, a line graph with a seperate line for each character showing the total amount of damage taken in a single arena match along the X axis with the number of times that total occured (ie number of matches) along the Y axis. I know it’s not a perfect graph because the data we have access to is kind of limited (mostly because people keep fighting after a match is effectively over), but it’s still fairly interesting.
Notice the huge spike of games which ended after 20k damage (roughly a single health bar for each of us) where presumably some serious CC chain happened right off the bat or (more likely) the rogue got opened on out of LOS and died in two globals. Quite a lot of the games Azuba dies in are games like this, most often from hard-to-peel cleave teams like rogue/dk/healer that blow cooldowns to get a kill at the start.
Protip: don’t ask Ina about that one game where he took 220000 damage or you’re liable to be death coiled. Yes, it was double melee.
Here’s the same thing for wins instead of losses:
Note that we win a huge amount of games in which our warlock is left alone. I don’t think that will surprise anybody! Note also that there are games where our warlock gets ridiculously trained and we still win, like this – Same cleave as before, check the ret’s damage! Apparently we can also win games where I get crapped on all match, like this double caster priest on RoV.
You can’t see it from the graph, but there’s also wins where nobody takes any damage, usually when some 2k cleave champion warrior charges behind a Nagrand pillar, trinkets KS and dies in a dismantle duration to tricks conflag lava burst.
On data visualisation: I think what you want here are histograms, with stacked bars and maybe 10k buckets.
On 3s team names: Next season I think it should be “Queueing Into Cleave” – in the hope that, like “Our Shaman Has No Gear”, it eventually becomes ironic as opposed to painfully true.
On Flawless Victories: Or some 2.7k MMR mage overextends on Dalaran, and explodes in a garotte silence.
On anagrams: So I went looking today; “Plate Nadir” struck me as pretty accurate in the case of Ret Paladins (though I’m sure “Lard Pantie” would go down much better on the forums, hurr hurr.). And I think “Worried Occult Tanks” sums Destruction Warlocks up pretty well. That or “Tard Clickers Won Out,” fffff.
You’re right, and that’s essentially what I was going for – the data is quantized into 10k blocks. The reason I used a line graph was because I couldn’t figure out a way to get the google chart API to nicely present multiple data sets on a bar graph. I’ve had another go at it and this looks better; though perhaps a bit cluttered:
As a bonus this includes the games we played last week since they are now on the armory. I’ve also changed the buckets so the damage is rounded to the nearest 10k rather than truncated, so the first bar is 0-5000 damage, the second is 5000-15000, etc. I don’t know if that’s the most sensible way to do it.
I’m definitely no statistician so I’m keen to hear any other criticisms you have.
(I believe gchart has a stacked bars option which I will have a look at when I get some more time)
Something like this, perhaps?
I’m not sure if that’s the right way to do a stacked bar graph, seems to make the vertical scale kind of massive, and it’s sort of difficult to compare between the different data sets.
Thoughts?
That’s what I was thinking of, but looking at it it’s obviously not as useful since you can only really eyeball the lowest series (also, histogram bins are truncated by definition afaict.)
I think some kind of step chart showing frequencies would be the best of all possible worlds – but if you were to use Google Charts for that, you’d have to hax up your own via cht=lxy as it doesn’t seem like there’s an option for cht=lc.