Server turnaround

[Iron adam]

Don't forget to ::vote

[Patel]

Yesterday we had over 180 players online. That's just brilliant stuff.

On that day we peaked at 184 according to our records. Yesterday we peaked at 194, that's the highest we've seen in a year. Fingers crossed for 200+ next weekend.

How many are old players - how many are new?
Hopefully not too much to search haha. Just curious.

[Rapsey]

How many are old players - how many are new?
Hopefully not too much to search haha. Just curious.

I'm afraid we don't have that information. The system just knows which accounts are online. Those accounts aren't linked to some unique player identity containing when this individual first started playing. The only way to find that out would be to go digging through logs for each account and then using human interpretation to figure out who is who.

I could give you a distribution of when the accounts were created that were online on that day but it wouldn't tell you anything about how old or new the player actually is.

[Fungamer]

How many are old players - how many are new?
Hopefully not too much to search haha. Just curious.

I'm afraid we don't have that information. The system just knows which accounts are online. Those accounts aren't linked to some unique player identity containing when this individual first started playing. The only way to find that out would be to go digging through logs for each account and then using human interpretation to figure out who is who.

I could give you a distribution of when the accounts were created that were online on that day but it wouldn't tell you anything about how old or new the player actually is.

I guess it wouldn't provide the perfect representation but at least it'd somewhat indicate which players are coming back

E.G theres more players on who created their account 2 years ago than 3 years, this does not fully prove, but at least shows us that 3 years for many players might be enough to "forget" about PKH. Knowing this in turn makes it nice to know that returning players are expecting stuff from 2 years ago and can't really relate to how things were earlier

Sure, we'll have people who created accounts 8 years ago playing now or people who've played for 8 years playing on a new account, but surely those are just outliers?

[Rapsey]

Sure, we'll have people who created accounts 8 years ago playing now or people who've played for 8 years playing on a new account, but surely those are just outliers?

Um no. The times when the data you get is actually correct would be the exceptions. Most of the time it would be way off. This isn't the kind of game where 90% of the people create 1 account when they join and then only play that account forever. On here that sort of thing is the exception. If you look at the accounts that are online at any given time, more often than not it is not their first ever account on pkh. People use alts, honors, get banned, share accounts, do services etc etc.

Even the people who still play on their original account tend to play on other accounts as well. If your 8 year veteran logs onto a few of his alts as well that day the system won't see that as 1 player who has been here for 8 years. It will see it as a player of 8 years, then another one of 3 years, then another one of 2 years etc etc. It will just see each account as a different player and assume this was their first account. So when you're looking at that data you'd think you are seeing plenty of players who have joined in recent years but really the data you're looking at is just bogus. The vast majority of "new" accounts belongs to old players. Similarly it would not be able to give you an accurate representation of how many old players there are. If I have multiple accounts that are 8+ years old it would see me as being multiple old players.

With this level of inaccuracy there's no point in even looking at the data. Might as well look at the output of a random number generator.

EDIT: and if you say "just make it look at IP's" you deserve a paddling.

[Patel]

I'm afraid we don't have that information. The system just knows which accounts are online. Those accounts aren't linked to some unique player identity containing when this individual first started playing. The only way to find that out would be to go digging through logs for each account and then using human interpretation to figure out who is who.

Dang, but true.

I could give you a distribution of when the accounts were created that were online on that day but it wouldn't tell you anything about how old or new the player actually is.

Well I wouldn't go that far, unless you've already looked at this and that's indeed the case (linear/uniform even with multiple days of binned data).

But it wouldn't be a great indicator at face value for sure. I imagine (if binned properly) you'd be able to see several peaks corresponding to periods of major growth. It'd be super weird if it were uniform/linear on long timescales. You can't say if it's a player returning every time there's an update and then getting banned, but I don't think that's most players. The odds of it not being a new player at a period of growth would be slimmer than it not being a new player the rest of the time. That's not to say that old players don't come back and make new accounts during periods of growth, but moreso to say that if there's a logical condition for isolating a period when brand new players are joining, it's when the server is growing. Then looking at the ratio of new accounts during those periods to the ratio of new accounts during "regular" periods could give a rough estimate, given that former is true. The hard part is quantifying how many are new, which I'd be happy to dig into but cannot. A couple periods of growth an I could characterize the correlation, then we'd be set for life! For life I say.

You'd probably have to integrate several days worth of data to get decent granularity.

Damn, now I wanna see lol. Even a table/file for a few days (If you have the time)?
I can save you the trouble of making graphs and finding modes (though it would probably take like 5 minutes if you're gonna do it anyways).

Welcome to the world of statistics Rapsey, where there is no such thing as a perfect solution (outside of simulation data).

[Church]

Sure, we'll have people who created accounts 8 years ago playing now or people who've played for 8 years playing on a new account, but surely those are just outliers?

Um no. The times when the data you get is actually correct would be the exceptions. Most of the time it would be way off. This isn't the kind of game where 90% of the people create 1 account when they join and then only play that account forever. On here that sort of thing is the exception. If you look at the accounts that are online at any given time, more often than not it is not their first ever account on pkh. People use alts, honors, get banned, share accounts, do services etc etc.

Even the people who still play on their original account tend to play on other accounts as well. If your 8 year veteran logs onto a few of his alts as well that day the system won't see that as 1 player who has been here for 8 years. It will see it as a player of 8 years, then another one of 3 years, then another one of 2 years etc etc. It will just see each account as a different player and assume this was their first account. So when you're looking at that data you'd think you are seeing plenty of players who have joined in recent years but really the data you're looking at is just bogus. The vast majority of "new" accounts belongs to old players. Similarly it would not be able to give you an accurate representation of how many old players there are. If I have multiple accounts that are 8+ years old it would see me as being multiple old players.

With this level of inaccuracy there's no point in even looking at the data. Might as well look at the output of a random number generator.

EDIT: and if you say "just make it look at IP's" you deserve a paddling.

Just make it look at IPs

okay i'm ready daddy rapsey

[Rapsey]

This isn't really a problem of statistical analysis. It's more a problem of filtering the data. Every day around 100 new accounts are created, of which maybe 1 is an actual new player. The new accounts coming from existing players completely drown out the ones from newcomers. So when you try to analyse this, all you see are the activity trends of your existing players. It's all nicely proportional to your other metrics like playercounts, and any spikes you see are due to events prompting your existing players to make new accounts (like the release of a new game mode) or the result of random things happening (like one person deciding it's time to make 100 merching accounts).

The only way you're gonna be able to spot trends in the prevalence of new players is by first being able to identify those new players with some reasonable degree of certainty. That's not really a question of applying the right statistical analysis.

One simple way would be to take the logs of every single login since 2010. So then you'd have a list of like 50 million timestamp-username-ip records. Then you could write a program to go through the entire list and for every new account name it encounters, it checks if there have been any prior logins from that IP. If there aren't we can assume that new account + new IP = new player. It would take a long time but you could do that.

But here's the thing: how accurate would that be? The majority of players have dynamic IPs that change every day. Of those "new account + new IP"'s you identified, 9 out of 10 were probably existing players that just got a new dynamic IP. So once again you are left with data that's so polluted you can't draw any conclusions from it.

If you see a solution to this I'm all ears. Solving this would mean solving everything from PKP boosting to an unavoidable IP ban.

[Raj]

@Rapsey

Spoiler: show

Is it possible to write a script that takes the last login date for each player that logs in on a given day and get a distribution for that? I think that would at least try to numerically capture returning vets but it wouldn't be very accurate (Mr Chuck logs in every day and afks vs. Mr Chuck starts doing slayer again), analyzing this way wouldn't account for weird cases like that. However it is the best I could come up with while trying to use relatively simple and filterable data points.

Although I'm gonna be honest, I don't think we'll get any meaningful data from anything like this. There's way too many exterior factors that will create a ton of noise and skew the distribution. Probably with the model I suggested the number of vets actually picking the game up again with intentions of playing will be grossly skewed, either underrepresented by Chuck cases or overrepresented by people logging back in just to say hi, hard to tell though without seeing it. And that's under the assumption that there is a reasonable model which works on easy to collect data, which there probably isn't.

There is a thought I'm having you could add variables to reduce skew on the data. I.e if you condition the distribution on people that spent less than 50% of their time in an afk chair, and condition further on people who spent over an hour logged in, to reduce outliers or unrelated data coming from vets. On the opposite end you could condition the count of new accounts (probably the group corresponding to a last login of 0 or NULL) on accounts which were created on unique IPs. I know some other network data is actually collected on login however the uniqueness of that data is debatable but I wouldn't consider it likely, maybe a combination of that and recognized IP could be used to determine new players to the most accurate degree possible. Obviously people trying to look new to an analysis algorithm will probably manage to do that but that's fairly rare I think. However, again, this works under the major assumption that data for last login, time spent AFK, and time logged in are easy to collect for Rapsey and that the initial predictor variable I chose (last login date) is actually a good one

[Fungamer]

EDIT: and if you say "just make it look at IP's" you deserve a paddling.

Use the users webcam (if available) linked to a facial recognition script. Gotta get that accuracy.


Edit:

If you see a solution to this I'm all ears. Solving this would mean solving everything from PKP boosting to an unavoidable IP ban.

My beautiful idea would solve this too. After we do this, we could start logins with facial recognition. The future is now, old man. Y'all hiring a new sysadmin?