The Oracle of Death - Page 4

Chewy · 2013-08-09, 01:23 AM

Originally Posted by maradine

If I had to draw conclusions from what I've pulled so far, it'd be that TR has some massive advantages, VS has a few useful gems, and NC is left holding the bag. That said, there are plenty more things to examine, and I'm hoping that as we look at more situations, everyone gets some love in different places.

I'd also like to thank everyone for not using this thread to dive into the various arguments that generally accompany these topics. I do encourage discussions and suggestions on getting more interesting data, however.

When/if you make your system public Im going to devote a weekend or more if needed no another round of MAX balance issues. Might even dip my hand into vehicle weapons or just go class by class if the want hits me.

Iv already gotten data on how the MAX AI weapons work in theory and under perfect conditions that prove NC MAXes are lacking in both range and sustainability. Once I get real in-game data on how things are playing out under the thousands upon thousands of things that one can't account for in tests then who knows what will turn up.

Im not even certain if the devs know what is really happening with details. They may have all the raw data in the world but if someone doesn't think to look for something and know how to use that data then it wouldn't be seen even if it was sitting in front of someones face. We've all done that in some way and working all day on other things the devs might not have the time to dig up stuff like this until someone brings it to them.

Just thinking about what I might find is putting my mind into overdrive. I love doing things like this for giving me something to focus on and putting my ADHD to work. Can just let my mind run wild and put its speed and ever changing nature to some good without having to hold it back by thinking about thinking. I work so much better if I can just let the thing run on a subconscious level free of things that can grab my attention. Without a focus then my mind just wonders about over everything I happen to see or hear. That's why I type such wordy posts, I have to many thoughts to type and end up changing/adding to the post many times over.

It also makes falling asleep hell when you can't stop thinking, even after laying in the dark for over an hour or 2 smelling the smells and hearing the sounds of night. Thus the reason for me typing this at 2:23AM.

Sledgecrushr · 2013-08-09, 09:34 AM

Originally Posted by maradine

The LMGs:

Code:

id 78 NC6 Gauss SAW
133,189 Kills
13,649 Unique Killers
9.7 kpu

id 79 T9 CARV
128,296 Kills
12,258 Unique Killers
10.47 kpu

id 80 Orion VS54
128,924 Kills
12,142 Unique Killers
10.62 kpu

Given that this is only 2 days of data, I'd call that a statistical dead heat.

This is kind of how I was thinking it would play out after using all of those weapons quite a bit. Thank you Oracle of Death.

typhaon · 2013-08-09, 11:02 AM

Originally Posted by maradine

The LMGs:

Code:

id 78 NC6 Gauss SAW
133,189 Kills
13,649 Unique Killers
9.7 kpu

id 79 T9 CARV
128,296 Kills
12,258 Unique Killers
10.47 kpu

id 80 Orion VS54
128,924 Kills
12,142 Unique Killers
10.62 kpu

Given that this is only 2 days of data, I'd call that a statistical dead heat.

Hmm... the Orion is almost 10% more effective than the Gauss Saw. It may fit OK in the context of asymmetrical balance (as could other discrepancies) - but I wouldn't call it a statistical dead heat.

A 5-10% difference in the main weapon of one of the most popular classes, IS a 5-10% difference.... It exists. If it is allowed to persist there, it has to be made up elsewhere.

I'd think 'good' balance in this game would be reflected by more popular weapons in a faction having lower KPU - with the less popular weapons having greater KPU (because the KPU is a result of them being used in a more specialized manner).

Problems are evident when you see very prolific weapons with high KPU,

I'd be interested to see comprehensive LMG stats. I'm pretty sure we'll see some of the balance issues I described.

maradine · 2013-08-09, 11:27 AM

Originally Posted by typhaon

Hmm... the Orion is almost 10% more effective than the Gauss Saw. It may fit OK in the context of asymmetrical balance (as could other discrepancies) - but I wouldn't call it a statistical dead heat.

You missed the first part of the sentence, which I felt was very important - it was only two days of data, with the collector going up and down for hours at a time as I worked on it. Many outfit Op nights - such as mine - aren't reflected in a 2 day pull. So you're probably going to want to trust me when I say my data is statistically interesting or not - or at least wait until we have a week or two before calling a <10% gap interesting.

If you're looking for something to start a thread over, here's a little bit more of a statistical anomaly:

Code:

Harasser AI Weapons - by kpu, NS weapons aggregated

id 6122 P525 Marauder-H
31,626 Kills
2,190 Unique Killers
14.4 kpu

id 6115, 6116, 6117 M40 Fury (Combined)
20,476 Kills
1,964 Unique Killers
10.46 kpu

id 6125 Proton II PPA-H
3,990 Kills
666 Unique Killers
5.99 kpu

id 6100, 6101, 6102 M12 Kobalt (Combined)
1,399 Kills
346 Unique Killers
4.04 kpu

id 6124 Enforcer C85 Modified-H
317 Kills
92 Unique Killers
3.44 kpu

id 6112, 6113, 6114 M60-G Bulldog (Combined)
2,602 Kills
780 Unique Killers
3.33 kpu

id 6103, 6104, 6105 M20 Basilisk (Combined)
7,470 Kills
4,358 Unique Killers
1.71 kpu

maradine · 2013-08-09, 11:40 AM

New Empire-specific Pistols

id 7389 LA3 Desperado
3,804 Kills
682 Unique Killers
5.58 kpu

id 7413 Cerberus
1,306 Kills
253 Unique Killers
5.16 kpu

id 7401 TS2 Inquisitor
1,665 Kills
352 Unique Killers
4.73 kpu

I think it's the ESF rockets and then I'm caught up. Yell if I missed you.

Not expecting much difference on the ESF rockets, but you never know!

Stomps · 2013-08-09, 12:05 PM

Originally Posted by maradine

Pulled that one back in post #15, though with the caveat that it might be wrong due to inconsistencies in their item database. I'll run it again when I get some clarification.

Thanks for getting back to me, yes you covered it but I was confused as to exactly what those were relating to, theres no timestamp or date range with the figures.

There was a change to both the claymore laser brightness as well as changes to pattern and strength in GU13 so im only interested in data from that date onwards. If you could consider that in the next run id be grateful.

Thanks a lot!

typhaon · 2013-08-09, 12:09 PM

Maradine - I didn't miss what you said. I'm just discounting (because of the sample size) the likelihood that we'll see significant swings in the ratios.

Been a while since statistics... but the margin of error must be pretty low and the confidence interval pretty high for this data.

Certainly you need to be consistent in your measuring of the data... I suppose I can only say this data is likely super accurate for the circumstances under which you measured it (times/populations).

More data is always welcome, however.

* hehe...

The stunning thing in that Harasser AI data... I can't believe anyone... ever... has used the C85-modified H to kill an enemy. I sit with my jaw on the floor.

maradine · 2013-08-09, 12:25 PM

Originally Posted by typhaon

The stunning thing in that Harasser AI data... I can't believe anyone... ever... has used the C85-modified H to kill an enemy. I sit with my jaw on the floor.

I know, right? It hilarious on like three levels.

The reason I'm pretty adamant on waiting a week for close judgements is that a lot of usages are asymmetric. For instance, if two days doesn't cover an EXE drunken squad, frac/burster averages will be different.

Long story short, the pull interval once we're live will be by GU. Doesn't get much broader than that.

typhaon · 2013-08-09, 12:33 PM

Can you give us a taste of some more LMG stats across factions? I have a suspicion the SV-88 is OP beyond all others... and need some fuel for my burning rage.

Also, very curious about comparisons of NC LMGs.

Sledgecrushr · 2013-08-09, 12:40 PM

Maradine you are an absolute genius with this thread and your ability to look into the api.

Great Oracle of Death, would you look into your crystal ball and give me the rundown on MBT AP turrets cross faction?

maradine · 2013-08-09, 12:42 PM

Originally Posted by typhaon

Can you give us a taste of some more LMG stats across factions? I have a suspicion the SV-88 is OP beyond all others... and need some fuel for my burning rage.

Also, very curious about comparisons of NC LMGs.

It shall be so!

Also:

Originally Posted by Stomps

Thanks for getting back to me, yes you covered it but I was confused as to exactly what those were relating to, theres no timestamp or date range with the figures.

There was a change to both the claymore laser brightness as well as changes to pattern and strength in GU13 so im only interested in data from that date onwards. If you could consider that in the next run id be grateful.

Thanks a lot!

Ah, I see the confusion. The entire dataset starts at epoch 1375832653, or Tue, 06 Aug 2013 23:44:13 UTC. Therefore, everything in the pulls so far is GU13 by default. I sped the collector up by about 3x last night, so I have enough headroom to start walking backwards in time now - assuming SOE retains the data.

Originally Posted by Sledgecrushr

Maradine you are an absolute genius with this thread and your ability to look into the api.

Great Oracle of Death, would you look into your crystal ball and give me the rundown on MBT AP turrets cross faction?

I'm not doing a ton more than ps2-stats was doing - just collecting slightly more data for a slightly better interpretation of what we're seeing. But flattery will definitely get you places.

Re: MBT AP turrets - did you mean main guns, ala a few pages back? Or the AP secondaries?

Wahooo · 2013-08-09, 12:49 PM

Originally Posted by maradine

Not expecting much difference on the ESF rockets, but you never know!

Though the rockets are the same the question will be do the differences in kill stats reflect the platform?

I'm actually interested to see if there are differences in some items that are the same, not just NS weapons but say the high velocity bolt drivers. Longshot, Parallax, and RAMS.50. Should be the same but any differences would be user based and not have anything to do with empire imbalances people are looking for.

maradine · 2013-08-09, 12:52 PM

Originally Posted by Wahooo

Though the rockets are the same the question will be do the differences in kill stats reflect the platform?

I'm actually interested to see if there are differences in some items that are the same, not just NS weapons but say the high velocity bolt drivers. Longshot, Parallax, and RAMS.50. Should be the same but any differences would be user based and not have anything to do with empire imbalances people are looking for.

Very good point - should be able to do that with the humble Basilisk all over the place, too. If the same bias appears in all of those like-for-like items everywhere, then you get something resembling an empire skill bias that you can subtract back out of everything else. Maybe. There's still all sorts of intangibles going on. For instance - if an empire is consistently outnumbered on a server, will their snipers do better on average, or worse? Would there be any effect at all?

THESE QUESTIONS LEAD TO MORE QUESTIONS </inception>

Adding:

Code:

ESF RoflPods by kpu
======

id 4903 Hellfire Rocket Pods
54,461 Kills
4,532 Unique Killers
12.02 kpu

id 4601 Breaker Rocket Pods
30,704 Kills
3,282 Unique Killers
9.36 kpu

id 4301 Dual Photon Pods
30,721 Kills
3,545 Unique Killers
8.67 kpu

typhaon · 2013-08-09, 01:16 PM

Generic guns across platforms would be fantastic, also.

I feel like one of the areas where balance has fallen off is in underestimating the impact of maneuverability on both the offensive and defensive ends.

maradine · 2013-08-09, 01:23 PM

Code:

Really Big Sniper Rifles by kpu

id 26004 Parallax VX3
51,527 Kills
2,929 Unique Killers
17.59 kpu

id 25004 RAMS .50M
59,260 Kills
3,536 Unique Killers
16.76 kpu

id 24004 EM4 Longshot
47,515 Kills
2,867 Unique Killers
16.57 kpu