The Oracle of Death - Page 8

maradine · 2013-08-11, 01:07 PM

Here's the new data set and template applied to the LMG pile. Note - less than a day of data, so there's some wonky stuff going on in the quartiles.

And updated with a quick run this evening, plus the HA specials.

Raw sheet for copypasta here: http://bit.ly/1eCq4Wr

Goliith · 2013-08-11, 01:23 PM

Ohh these new stats are delicious, I can't wait until you get a week or two of data to run off of. Here's to hoping you don't encounter another reason to Nuke'n'Pave

maradine · 2013-08-11, 01:32 PM

I shouldn't. I also wrote a subset bot that just does the collection so I can work on the Oracle without interrupting data flow. The main collector is now up in the source repo should anyone want to replicate the pull type and timing engine.

On the topic of nuke and pave, I'm thinking of aggregating all kill data at the end of a GU into its own table and wiping the raw kills to save space. Basically, about 100 million rows down to about 500. Can anyone think of a reason to have timing data more granular than a GU, since that's the main increment of balance?

Goliith · 2013-08-11, 02:18 PM

I can't think of a realistic reason to want to keep the older raw data, or for having a larger timescale aside from just catering to the folks that dont care about GU balance changes and just want the raw lifetime stats of things, but there's PS2-Stats for that.

EvilNinjadude · 2013-08-11, 03:44 PM

The Quartile KPU is slightly off because of...?
And could it be partially improved by showing which percentage of the total each quartile made up? (Since they're split by BR and not by number of people, like actual Statistical Quartiles)

Also, I'm really tired, so I can't think straight.

Storm Frog · 2013-08-11, 03:55 PM

Originally Posted by maradine

Can anyone think of a reason to have timing data more granular than a GU, since that's the main increment of balance?

There are occasionally hotfixes that impact weapon balance. That's the only situation I can think of though.

Goliith · 2013-08-11, 06:03 PM

Originally Posted by EvilNinjadude

The Quartile KPU is slightly off because of...?
And could it be partially improved by showing which percentage of the total each quartile made up? (Since they're split by BR and not by number of people, like actual Statistical Quartiles)

Also, I'm really tired, so I can't think straight.

Not sure if you read the legend while looking at the picture, or if you're asking something different about it?

But the Quartile KPUs are different because the different Quartiles represent different Battle Ranks, 1-25, 25-50, 50-75, and 75-100. So naturally, players of higher BR would have a greater familiarity with the weapons they are using, and the tactics and equipment of their opposition. So, you are more likely to see higher Kills per Unit from more experienced players, who have already found and settled on a given weapon to be their "Go-to" weapon, at which they excel most with, and therefor earn the most kills with.

maradine · 2013-08-11, 07:04 PM

Originally Posted by EvilNinjadude

The Quartile KPU is slightly off because of...?
And could it be partially improved by showing which percentage of the total each quartile made up? (Since they're split by BR and not by number of people, like actual Statistical Quartiles)

Also, I'm really tired, so I can't think straight.

I don't really care about the distribution of players inside BR - I pulled AVGBR of the group as a decent summary. What we care about is their performance, not so much how they're distributed.

camycamera · 2013-08-11, 09:25 PM

Guyz the tr op nerf plz

exohkay · 2013-08-11, 09:28 PM

Originally Posted by camycamera

Guyz the tr op nerf plz

I don't think anyone is saying specifically this. There are avenues where TR are outperforming the other factions rather drastically, and there are certain areas where VS are outperforming.

We're asking for these to be remedied.

Balance is a two way street, when statistics indicated Mags were the best tank, when ZOE was the best option (and perhaps still is).

maradine · 2013-08-11, 11:17 PM

And, as a friendly reminder, take any requests for action based on any of this somewhere else. This thread is solely focused on the method and providing access to the data.

Temple · 2013-08-11, 11:45 PM

O Oracle...

Can you get this info.

By Server
By Faction
Team Kills per player (or 1000 kills)

So we can see per head (or 1000) which faction TK's the most per server.

Eg, I'm on Connery and playing NC, I'm 100% sure that we are the worst...but I'd love to be proven wrong. If per 1000 deaths, 5 were TK's on the TR, 5.554 were VS and NC had 3.558 for example, that would be pretty cool to know.

maradine · 2013-08-12, 12:05 AM

Not currently resolving and storing victim faction_id as an API load courtesy and space-saving measure. I'll put it on the wishlist for when I'm low on inspiration and can take the collector offline to free the database for a table alter. I suspect, for what it's worth, that you're right.

edit: just touched up the LMG charts and added the HA specials. I'll tackle carbine country over lunch.

EvilNinjadude · 2013-08-12, 01:18 AM

Originally Posted by maradine

I don't really care about the distribution of players inside BR - I pulled AVGBR of the group as a decent summary. What we care about is their performance, not so much how they're distributed.

Kay. If we don't care about that, I'm fine with that. AVGBR is good enough for most things.

DirtyBird · 2013-08-12, 01:29 AM

Great work Maradine, thanks for sharing the results.