Correlation is usually sufficient for “prediction” (as in modeling/regression). In fact, I can imagine cases where highly correlated but non-causal predictor variables are more useful for the model than the causal, but less correlated predictor variables (particularly if the cause is spread among a number of such variables).
This fact, however, does not make the highly correlated predictor variable/factor any more likely to be the cause.
And “prediction” using such methods is suspect if applied outside the range of the data used to fit the model (not sure this applies, just sayin…)
Predicting seasonal business trends are sometimes driven by obvious causes, but often not. And the rates we predict are typically not something we can causationally describe.
We still predict them with correlations.
Global warming is almost all based on correlational data.
Most anything that uses statistics is murky intnhe causational area… which is why we use statistics.
Ops research, as a degree… applied statistics to big data problems… you’re looking for correlations that describe what will happen.
Causation is nice when you can get it, but it’s not critical.
I’ve been researching image extrapolation for a couple years now… so I clearly like trying to “draw outside the lines.” If nothing else, though, I’ve learned first-hand that it can be a very difficult problem attempting to predict beyond the range of your data.
Usually the best approach is to simply see if the model still performs roughly the same given the new data. Use the old data as “training set” to fit the model, then test the resulting model on “validation set” (the new data).
You are leaving out the rise of options and an aging player base.
Look at kids today - late teens mainly. With the choices of games that they have and all this VR stuff too, do any of them strike you as being the type of kid, or having the kind of personality even, to play Eve?
I just don’t see it. Heck I’m the crazy uncle who played Eve. “Oh that game where everybody kills each other”.
Add to this the vast variety even for older gamers. Go back to 2009. What was THE game that let you have so much customization of your character, gear, and skills?
It was Eve Online. WoW had their skill trees, which were not as free-wheeling as Eve’s .
But now? It’s like every game has this skill tree concept. Skins. Etc. Eve caught on late with skins actually, after years of being a leader in character customization.
Eve is for the older crowd, and that older crowd is busy. I got 10 things to be doing instead of shitpoasting right now - though I like your post and thought out presentation - I think the overall picture is being missed here. Eve is an old game, played by older people who are busy. The “kids” are not interested. Heck after messing around with my nephews occulus contraption thing, that young wippersnapper had it set up last time I visited, I can’t imagine anybody younger than 25 paying for a sub for “spreadsheets in space”.
I complain that Eve is becoming more like a “browser game”. Possibly they are doing this because they have to. PI is almost Farmville now that I think about it.
It’s just the way it is.
Only if the goal is not to change the prediction.
Example, we can put more money in firefighters in the areas that are frequent subjects to fires, because we assume the prediction will remain the same. Because we assume that money is best invested in places that need them on a regular basis.
Now if we want to prevent the fires, we need to find the causes of the fires, eg are they from human cause ? Can we act to prevent them ? Then we need causations.
example double blind tests. That’s why we are saying correlation is not enough, we need further investigation.
You don’t predict. You assume the trend will remain the same.
Global warning is based on more than correlation. Sure there are a lot of assumptions, but there are a lot of model that are used which can make predictions. Yet they may be all wrong, that’s why people may or may not accept them. Those models were however good enough to predict the effect of a reduction in CFC gas usage… but they may still be wrong.
But I still say it’s like this. If the game’s NPE hasn’t changed, and the standard distribution of player’s acceptable experiences hasn’t changed (think Gaussian distribution along a pve-pvp axis)…
We would expect the % of players quitting to remain essentially unchanged.
Ie- if we change nothing we should get the same result (accounting for random variation).
“if we change nothing”. Eve is constantly evolving. The goal of the devs is to make it evolve in a way that is best for them. You can’t make prediction based on statistics when you want to change the environment. It’s the opposite of what you do on market analysis.
My first post in this thread was not correctly stated. I would delete or modify it if it didn’t have replies. Since then we’ve wandered into causation and correlation and… stuff.
But this is basically what I was trying to get at. When you see a significant change in the rate of player retention compared to most recent years at this time, you can safely assume that SOMETHING has changed (even if it’s just a range of the data where the ‘model’ behaves differently). Without more data we don’t know what has changed from a purely statistical perspective, but we are not ignorant to all relevant factors. If we allow certain assumptions, we can then make educated guesses. Those guesses will of course be contingent on our assumptions and influenced by our individual biases, but this is a very common thing in such analysis when one doesn’t have all the data and can’t get it. Just state your assumptions.
For example, I doubt we can prove causation for the 2008 financial crash (Perhaps ancient aliens placed economy bombs set to go off just at that time?). That said, if I go research it right now I doubt I’ll find papers refusing to even analyse the event because they can’t isolate factors and prove causation. I’m free to disagree with what will likely be opinions backed by statistics, other historical events that are similar in some way, economic principles, etc.
Similarly, if we assume the rate of new players entering the game is unchanged, and they are the same type of players, etc., then we can probably make some good guesses as to which changes to the game or existing players might have caused this reduction in player retention. One of us might claim it’s old players naturally growing tired of the game… another that it was the war dec. changes… another might claim our assumptions are horse ■■■■, it’s a change in the type of new players…
If you demand certainty, then almost none of the discussions on these forums are going to meet your standard. Hell, you better stick to reading math books…
In the absence of provable causation, uneasy consensus is the best we can hope for. Every one of us relies on such constantly, every day. We can’t get there, though, without some speculation and discussion.
This is the central problem of this game. With friends and objectives, Eve is the best game you’ll ever play. Without, it’s only boredom and pain, and no updates will ever change that.
Since the correlation/causation discussion has collapsed due to “EVE denial”, here’s some cause and effect data.
Highsec wardec griefing was about 5% of the reason I stopped playing EVE last time.
I wasn’t in a Corp, had considered either creating one or joining a highSec Corp, checked them out, and discovered that “wardec griefing” of rookie and social Corps was well established and quite active.
It removed one more potentially interesting activity off the list of “potentially fun and/or helpful things to do in EVE” due to game design that didn’t consider “new player retention”.
Note: As it happened, this was the “last straw”, but it was far from the only one. Wardecs took me from approx “95% stop playing” to “100% stop playing”. A similar factor (indirectly related to “bumping”) took me from approx “40% stop playing” to “50% stop playing”. So it was a more important factor, but it came earlier, so it wasn’t the “trigger”.
Much of the remaining 85% was of a different nature (more nearly comparable to the (for me) new plague of widespread botting).
Correlation is something which we think, when we can’t see under the covers. So the less the information we have the more we are forced to observe correlations. Similarly the more information we have the more transparent things will become and the more we will be able to see the actual casual relationships.
If you care to do a multidimensional analysis of the price of different types of goods in the last 3 years, it would all be clear to you. I have tried to write a few times about what is happening but I have been banned so many times that I gave up …
Ok. So the source isn’t anything to do directly with player retention numbers and just concurrent log ins.
Yes retention obviously impacts that but we have no idea if we are doing better with player retention as a percentage from those numbers. Because we don’t know how many actual players are starting and how many are staying.
But there is a reason that player activity is a widely used metric for overall game health.
The wardec change was made to improve player retention. So if it works, player retention, and by extension activity, should show a positive shift. But it does not seem to have worked. In fact it seems to have had the opposite effect with a downward shift since the change.
This is even more interesting given how people have described the effect of wardecs on player retention to be so devastating. If it was so devastating, its removal should make a significantly positive effect on player retention. If that wasn’t the intention of the change, why did we bother?
It isnt so easy. There are multiple factors in such a complicated system:
no wars
underwhelming events
price of PLEX for paying extra alts
banning bots
people dont come back because they dont know something changed, or they found out there is better game already for them
And obviously few others I dont have any idea they exist, but probably there are some.
That the PCU when summarized is lower, doesnt mean that it wasnt some slight possitive effect with those wardeck changes, but can be undermined by other effects.
A question about the actual measurements CCP collect. Is it:
Total EVE player activity?
Player activity by different classes of character and player (e.g: Player activity by character age")?
Nana
EVE has an extremely high startup threshold, especially by modern gaming standards. I’d expect that many ex-.players wouldn’t come back even after modest improvements (such as the wardec change).
A clear commitment from CCP to reducing anti-rookie play would influence me. The wardec change alone isn’t enough though.
Note (not for you, but it’s an EVE forum, so this is needed /sigh):
“Anti-rookie” means exactly what it says. If I meant “non-consensual PvP”, or “one-sided non-consensual PvP” I would have said so.
It happens a lot, and unsurprisingly rookies aren’t impressed.