Greedy Goblin

Tuesday, June 17, 2014

Goons laugh off losses (or not)

Here is the last nail to the coffin of the evil that plagues New Eden. Their last myth destroyed. I've already proven that they take much more damage from "shitlords" than from their "relevant enemies". However they stick to their magic: they are simply immune to losses. It doesn't matter how many trillions of ISK goes boom, they just laugh and carry on like nothing happened. Well, it's bizarre for first glance, but hey, they are a special kind of people, they aren't like the "pubbies" who cry over losses.

If you claim something bizarre enough times, people start to believe it. And they did. Most commenters believed that it's no point killing CFC as it does nothing. They even conjured up some ideology that fleet losses are paid from the "infinite" war chest and the individual losses are on the individual and it has zero effect on the fleet.

Instead of logic (which isn't really working on them), I looked for simple facts, like the 2013 GSF kill and loss data. Since there was no GRR project in 2013, at least they are right in one thing: I'm totally irrelevant in this dataset.

I filtered out those pilots who had less than 100M ISK activity (de-whored kills + losses) in the first half of 2013. They are either cyno alts, or simply weren't active in the first half. 3601 pilots remained. Then I checked the correlation between the first half activity and the second half activity. Correlation coefficient is calculated by a program (EXCEL for me) and gives the connection strength between the variations of two datasets. The H1-H2 activity correlation was 0.55. What does it mean? You grab a large sample of Goons who were one sigma more active in the first half of the year than average. In the second half, some get more active, some less active, but their average activity will be 0.55 sigma above the big average. It makes sense, those who were above average active in the first half, are likely remain above average. 0.55 is not a bad correlation, but far from perfect either. There must be another factor.

Since H1 activity didn't explain H2 activity fully, I looked for a heretic idea: "net ISK". It is dewhored_kills minus losses. It must be totally uncorrelated, as Goons are totally immune to losses. Instead, I've found 0.42 correlation between H1 net ISK and H2 activity. Yep, it's almost as high as the H1-H2 activity correlation. So the current ISK ratio of a GSF member is almost as good predictor of his future activity as his current activity. What does it mean numerically? Since the variance of the total activity is larger than of the net ISK (5612M and 4729M), if you gank an 1B ship of a GSF member, he will have 1B*5.61/4.73*0.42 = 0.5B less activity (kills+losses) in the next half year on average than he'd have without the gank. "On average" means over a large group. Some of them indeed laugh off the loss, some get scared and stick to bombless bombers and some will rage quit.

Who would have guessed? Goonies aren't some superhumans, they - like ordinary players - get mad at losses. Have you considered giving them some?

Update: for those who are bad at maths or intentionally trolling, I made this table:
The pilots are placed into groups by their 2013 first half activity and balance. The top 500 active pilots (holding 62% of total H1 activity) got into "Activity: 2". The next 1500 pilots (holding 33% of the total H1 activity) got into "Activity: 1". The remaining 5736 pilots got into "Activity: 0". Those who had 1B more de-whored kills than losses, got into "Balance: 1". Those who had 1B more losses than de-whored kills, got into "Balance: -1". Everyone else got to "Balance: 0". Please note that "Activity: 0" had 700M top, so they couldn't get into any other balance group than 0.

The data inside the table is the de-whored kills of the second half of the year. You can see on the "Balance Ignored" column that the more active a pilot was in the first half, the more kills he had in the second half. So activity can predict kills alone. The "Activity Ignored" can't do the same, the middle group had the least kills. However using both, we get a much better prediction than by using activity alone. Within the same activity group, lower balance means significantly less kills. It's still true that activity is the better predictor (Activity 2, Balance -1 > Activity 1, Balance +1), but it's hard to ignore Balance.

6 comments:

Anonymous said...

Although I appreciate the work you put into all the data processing, I really wonder how much impact your ranting about goons has.
True, you were unsatisfied with the speed of growth of your little lemmings project, understandable, but with only 30-40k people playing eve (worldwide!!!!!!!), every single RL member of your corp is as precious as can be.

Anyway, your position is quite understandable for me as well as the tears of goon-related people trying to drown your rants with their tears and their we don't care propaganda.
I am almost convinced that some goon players were impacted by your campaign.
In fact, you are metagaming right now, against goons.
Expensive losses always "hurt" one way or other, even if one is trying to laugh away losses, it still is a forced laugh...
They have in impact on "fun"!!!
SO!!!!
What I would be interested in most is the followong information:
Do goons lose members? Are less goon players logging in? Will that impact future fleets, like when people who could fly titans do not log on when desperately needed?
You already shwoed that there is a correlation between suffering losses and playing less afterwards...
Is there already a trend showing?

EXE Grunt said...

You can do better than that. You are forgetting important factors.

First of all, when "de-whoring" kills you count only DPS ships. Logistics tend to not even get on killmails, neither do interdictors or heavy interdictors (even when the enemy aggresses you by trying to warp out of the bubble, you get 0 damage dealt. So de-whored, you didn't do any damage to the target.)

Second of all, goons have good SRP. Expensive losses in battle don't generally affect them, since they get 100% reimbursement (and 200% in some cases, I believe. Yes, they're paid to get blown up.)

The "second half of the year" drop in activity is very easy to explain: summer. Summer always brings with itself low numbers. On top of that, the winter war in 2013 was a painful sov grind that burned out a lot of people - so here's another reason numbers were low for not just GSF, but CFC whole.

Gevlon said...

Logistics and interdictors doesn't get kills in the first half of the year either, so they don't even get into the dataset.

SRP doesn't cover ganks, which is most of the CFC losses.

Overall drops are filtered out, as correlation sees deviation from the average.

EXE Grunt said...

And? They're somehow not important in the fleet, their presence/lack of it is irrelevant?

You're either sampling the wrong goons, or misinterpreting your outcome. From dealing with them, those who rat in ships that are very blingy can replace them and laugh it off when they lose them.

Others don't rat in blingy ships, just use disposable ishtars.

Anonymous said...

Part of the issue us that you are assuming a flat, static role for all. In other words, no growth or change in skill points or training plan.

I start the year as a basic BS pilot. I have t1 large weapons and basic fitting skills. As the year progresses, I may enjoy my role and improve to T2 guns. Such a person's activity by your metrics would stay about the same.

On the other hand, say I hate being part of the F1 blob and instead want to find a specialist role that I can enjoy. I could become a logi pilot, tackle (HIC, DIC, or intercepter), or ewar. In all cases, though I am of great value to the fleet, my dps contribution will drop considerably.

Yet there is no path save for carrier or dread where my DPS will increase significantly to be visible in your chart. And carrier or dread contributions are limited by capital deployment doctrine, meaning even if I do train into dread, I might not fly it on every op, where as logi, tackle, and ewar, i can.

So how do you account for this in your metrics, especially since you can count for loss of activity, but since you limit to a defined set of pilots, you don't open room for replacements when a pilot changes roles?

Anonymous said...

> SRP doesn't cover ganks, which is most of the CFC losses.

Goons also tell their members, "Don't go to highsec." Everything they need can be obtained in null, provided at good prices from the traders that stock their markets.

The Goons that are getting ganked in high sec are the dumb ones. They're the gazelle that decided to go drink from that stream instead of the pool where the rest of the herd is drinking. That they get ganked and quit Goons and/or EVE is not a great loss to the collective. In many ways, their loss strengthens the group.

Actual Goon commerce consists of an enormously-stocked alliance fund, and anonymous alt traders that are whitelisted so as to be able to safely jump freighter in and out. The blinged ratters that get blown up are doing so because they can, and they're not hurting because of it. The average line goon might be cash poor, but has access to a full line of doctrine fitted ships that they don't have to pay for, and reap the benefits of.

Furthermore, "dewhoring" logistics and bubblers just proves your elementary understanding of the modern EVE combat environment. Many, many, many doctrines are constructed such that they do not operate in a vaccuum, but rather have logibro support. By separating "tank" and "DPS" onto different hulls, both can be better at their jobs given the modern combat environment; local tanks are no longer required for many doctrines. By your flawed methodology, they're not only worthless, but decrease the efficacy of modern doctrines, as well.