Menu
What's New
By:
Filters

Hit rate testing round 2: Claim respawn rates

I did some tests over the weekend and observed spawn rates of three minutes.
 
kingofaces said:
It makes sense to a degree. The way those confidence intervals work is that if you repeated this experiment 100 times, 95 of those times you'd expect the average hit rate to fall within that range. In that paper, the confidence intervals around their average of 27.1% would be between 26-28%. We were working with smaller sample sizes, so our confidence intervals were much wider, but 30 is normally considered a high enough sample size to run statistical tests without problems associated with say a sample size of 10.
Click to expand...

What about a variable sample size ..... Say of up to when the first NRF occurs in a mining run ?

Here's my hypothesis, when mining, you are either removing or adding claims to the pool in the form of peds. On average you can expect to get I think it was 95-98% of what was used back as usable resources..

If I am correct this value can be "skewed" by simply "quitting while you are ahead" so to speak and letting other avatars spend their probes to "recharge" the pool.
 
R4tt3xx said:
What about a variable sample size ..... Say of up to when the first NRF occurs in a mining run ?

Here's my hypothesis, when mining, you are either removing or adding claims to the pool in the form of peds. On average you can expect to get I think it was 95-98% of what was used back as usable resources..

If I am correct this value can be "skewed" by simply "quitting while you are ahead" so to speak and letting other avatars spend their probes to "recharge" the pool.
Click to expand...

The analysis can handle variable sample sizes, but I don't see that helping anything here. If you're just going up to the first NRF, that's not really allowing for randomness either, and you're going to have tiny sample sizes you can't infer anything from.

There's also a caution with the "quitting while you are ahead" approach. You need to differentiate between randomness just giving you a lucky 40% HR in a few test drops versus the true mean HR being 40% for that given location/time. In some sequential samples, you can get wide swings even if the mean was 28% where you get up to 50% HR for 30 drops, then nothing for the remaining ones. If the true mean was 28% in that case, you aren't quitting while you're ahead (unless you quit the game entirely) because your returns will roughly approach the mean over time (called regression toward the mean)

We can't really easily test how the loot pool works, just how HR varies or doesn't under controlled conditions since we can generate data for that.
 
R4tt3xx said:
I did some tests over the weekend and observed spawn rates of three minutes.
Click to expand...

HR is very different than the "usual" for the moment, (MA adjusted it) with lower HR and bigger claims for people to try those boxes from the "mining" event cause they need a VII claim to get a box.

Also spawn rates of resources vary a lot, it's normal that belk/blaus are there in a few minutes and rare ones take longer.
 
kingofaces said:
The analysis can handle variable sample sizes, but I don't see that helping anything here. If you're just going up to the first NRF, that's not really allowing for randomness either, and you're going to have tiny sample sizes you can't infer anything from.

There's also a caution with the "quitting while you are ahead" approach. You need to differentiate between randomness just giving you a lucky 40% HR in a few test drops versus the true mean HR being 40% for that given location/time. In some sequential samples, you can get wide swings even if the mean was 28% where you get up to 50% HR for 30 drops, then nothing for the remaining ones. If the true mean was 28% in that case, you aren't quitting while you're ahead (unless you quit the game entirely) because your returns will roughly approach the mean over time (called regression toward the mean)

We can't really easily test how the loot pool works, just how HR varies or doesn't under controlled conditions since we can generate data for that.
Click to expand...

I understand what you are saying. But what controlled condition improves HR ? Does not repeatably mining at the same coordinate set improve HR ? I would say yes...

Is the system absolute ie does it rely on certain coordinate sets to generate claims or is it more relative ?

Does it create an array like Project Entropia did ?

Where do we even start ?
 
LeelooM said:
HR is very different than the "usual" for the moment, (MA adjusted it) with lower HR and bigger claims for people to try those boxes from the "mining" event cause they need a VII claim to get a box.

Also spawn rates of resources vary a lot, it's normal that belk/blaus are there in a few minutes and rare ones take longer.
Click to expand...

I don't understand why MA would put in a system that works like this... It encourages miners to wait, ie not to participate in the game rather than causing tools to decay. It makes no sense...
 
R4tt3xx said:
I don't understand why MA would put in a system that works like this... It encourages miners to wait, ie not to participate in the game rather than causing tools to decay. It makes no sense...
Click to expand...

So you want all resources to be TT food?
 
kingofaces said:
The analysis can handle variable sample sizes, but I don't see that helping anything here. If you're just going up to the first NRF, that's not really allowing for randomness either, and you're going to have tiny sample sizes you can't infer anything from.

There's also a caution with the "quitting while you are ahead" approach. You need to differentiate between randomness just giving you a lucky 40% HR in a few test drops versus the true mean HR being 40% for that given location/time. In some sequential samples, you can get wide swings even if the mean was 28% where you get up to 50% HR for 30 drops, then nothing for the remaining ones. If the true mean was 28% in that case, you aren't quitting while you're ahead (unless you quit the game entirely) because your returns will roughly approach the mean over time (called regression toward the mean)

We can't really easily test how the loot pool works, just how HR varies or doesn't under controlled conditions since we can generate data for that.
Click to expand...

What I can tell is how many peds I am taking out of the loot pool. The issue with the game and I am noticing it more and more is how "front loaded" it is. The game gives good loot when you have been offline for a while in order to encourage you to play.

It's like a form of entropy (I know, what a concept) with each avatar that is active shedding off some of their heat, in this case peds, to those that are less active ie colder.
 
R4tt3xx said:
What I can tell is how many peds I am taking out of the loot pool. The issue with the game and I am noticing it more and more is how "front loaded" it is. The game gives good loot when you have been offline for a while in order to encourage you to play.

It's like a form of entropy (I know, what a concept) with each avatar that is active shedding off some of their heat, in this case peds, to those that are less active ie colder.
Click to expand...

We're getting off the topic of this thread, but even if that, those claims still need data with formal testing to back it up. This thread isn't really a theories thread, but is focused on how hit rate varies based on other miners in the area or related testable data.
 
kingofaces said:
We're getting off the topic of this thread, but even if that, those claims still need data with formal testing to back it up. This thread isn't really a theories thread, but is focused on how hit rate varies based on other miners in the area or related testable data.
Click to expand...

Sorry, yes you are correct...

Was the test conducted using a 110m hexagonal layout ?
 
Could it be that getting mentioned respawn times happened only because game server decided to "refresh" claims globally at the exact moment?
While its true respawn interval is higher somehow.
 
stc44 said:
Could it be that getting mentioned respawn times happened only because game server decided to "refresh" claims globally at the exact moment?
While its true respawn interval is higher somehow.
Click to expand...

There are just so many unanswered questions that will lead to more unanswered questions with regards to mining, it's structure and mechanics, etc. But this is not the thread for it....
 
I think my question is legit and not-offtopic as I suggested an alternative hypothesis that could possibly explain mentioned test results.
 
This coincides with an observation of mine, the respawn being around 20 mins, when I cared to measure. That is going back on the exact same spots (via LBML), to hit a significant where a small was before. There would be an interesting discussion about double dropping (I mean repeating the drop in same radius/or at the claim itself) in this context, I think that is alot more important in the long run than if other miners been there, because is about your own habit rather than external incontrollable input.
 
R4tt3xx said:
Was the test conducted using a 110m hexagonal layout ?
Click to expand...

Essentially, yes.

stc44 said:
Could it be that getting mentioned respawn times happened only because game server decided to "refresh" claims globally at the exact moment?
While its true respawn interval is higher somehow.
Click to expand...

The multiple treatments and replications would have captured this to a degree. Basically, the testing is designed to handle a random claim "refresh" whether it's a bunch of claims refreshing at once every 10 minutes or if claims are just randomly regenerating at a given rate per minute. Whatever the refresh rate is and potential variation around it could be measured with more replication and more fine-tuned testing, but it seemed like we were at the point of diminishing returns in terms of what we'd gain from additional testing. There's still the question of how this varies in other areas, rarer resource types, etc. too, so this was meant to mostly address whether you can go mining after seeing another person in the area (or yourself possibly) in a relatively short period of time.
 
Doing 50 drops at the same places on every run with 1 hour in between.... so enm respawn are taking a lot more time
Mined resources : Oil - lyta - dianthus - typo - garcen

1st run : HR 40% (TT return 110%)
2nd run : HR 25% (TT return 67,28%)
3th run : HR 26% (TT return 72,88%)
4th run : HR 29% (TT return 80,64%) here oil and lyta where coming back again finally
after 6 hours I had 110% TT return again

For ores you need to test this, some are back in 5 min, others after a day, and some are lyst first, then after a few hours they change ...
so when mining for a specific ore, first round I had 100% from that ore, mining after 30-60-90 min gave me nothing but 100% lyst and after 120 min 100% specific ore again :)

All this testing was done when nobody else was around, I just stay there
 
Last edited:
LeelooM said:
Doing 50 drops at the same places on every run with 1 hour in between.... so enm respawn are taking a lot more time
Mined resources : Oil - lyta - dianthus - typo - garcen

1st run : HR 40% (TT return 110%)
2nd run : HR 25% (TT return 67,28%)
3th run : HR 26% (TT return 72,88%)
4th run : HR 29% (TT return 80,64%) here oil and lyta where coming back again finally
after 6 hours I had 110% TT return again

For ores you need to test this, some are back in 5 min, others after a day, and some are lyst first, then after a few hours they change ...
so when mining for a specific ore, first round I had 100% from that ore, mining after 30-60-90 min gave me nothing but 100% lyst and after 120 min 100% specific ore again :)

All this testing was done when nobody else was around, I just stay there
Click to expand...

Interesting, so resource nodes can be considered to "mature" over time if left alone. Problem is that you do not know if during the time that your avatar is offline, the resources in the area that your are "scouting" are being depleted.
 
kingofaces said:
Essentially, yes.



The multiple treatments and replications would have captured this to a degree. Basically, the testing is designed to handle a random claim "refresh" whether it's a bunch of claims refreshing at once every 10 minutes or if claims are just randomly regenerating at a given rate per minute. Whatever the refresh rate is and potential variation around it could be measured with more replication and more fine-tuned testing, but it seemed like we were at the point of diminishing returns in terms of what we'd gain from additional testing. There's still the question of how this varies in other areas, rarer resource types, etc. too, so this was meant to mostly address whether you can go mining after seeing another person in the area (or yourself possibly) in a relatively short period of time.
Click to expand...

There is still the question of depth, can one find multiple claims at different depths in the same 110 x 110m cell ?
 
R4tt3xx said:
There is still the question of depth, can one find multiple claims at different depths in the same 110 x 110m cell ?
Click to expand...

I put that testing on hold because you don't really need multiple people to do that. I'll be trying out some testing on depth that'll address that later this summer in terms of general hit rate and maybe an example of a specific ore/enmatter's hit rate as you increase depth.
 
kingofaces said:
Each person will do a minimum of 30 drops (60 total using both ore and enmatter), though I might see if a couple want to do 50 dual drops instead.
Click to expand...

Way too small sample-size.
do 10000 consecutive drops, then it is somewhat scientifically sound.
 
Alukat123 said:
Way too small sample-size.
do 10000 consecutive drops, then it is somewhat scientifically sound.
Click to expand...

That's not proper sampling statistics and actually something that gets taught day 1 in experimental design courses not to do through brute force like that. There's no reason to do 10,000 samples when less than 100 will tell you the same thing if you're using the correct statistical tests.

The statistical tests are telling you what the averages would be at a high sample size like 10,000 drops. On a per treatment basis, a sample size of 30 for a binomial response like hit rate is generally considered pretty decent (n=10 even gets by in some papers). All increasing the sample size does from there is shrink the confidence intervals with the average bouncing around somewhere within those intervals for the most part. An example of that would be the first three treatments here between 0 to 30 minutes. Their averages are not significantly different at this sample size, so if you increased the sample size, the test is saying there's a high likelihood those averages are still not going to be different.

Check out the thread for the first round of testing for more background on this where people asked the same question about using too high of sample sizes.
 
kingofaces said:
That's not proper sampling statistics and actually something that gets taught day 1 in experimental design courses not to do through brute force like that. There's no reason to do 10,000 samples when less than 100 will tell you the same thing if you're using the correct statistical tests.

The statistical tests are telling you what the averages would be at a high sample size like 10,000 drops. On a per treatment basis, a sample size of 30 for a binomial response like hit rate is generally considered pretty decent (n=10 even gets by in some papers). All increasing the sample size does from there is shrink the confidence intervals with the average bouncing around somewhere within those intervals for the most part. An example of that would be the first three treatments here between 0 to 30 minutes. Their averages are not significantly different at this sample size, so if you increased the sample size, the test is saying there's a high likelihood those averages are still not going to be different.

Check out the thread for the first round of testing for more background on this where people asked the same question about using too high of sample sizes.
Click to expand...

let's say, leeloo does solo run, 100 drops, gets higher than average claim rate, say 40% claim rate, now joins your testing, normalisation happens and leeloo gets lower than average claim rate

Example:
assuming 30% is average claim rate.
130*0,3 = 39 claims
40 claims have happened during the first 100 solo drops, normalisation kicks in during your test and leeloo gets only 2 claims in those 30 drops (6,67%HR).

If you extend it, like leeloo doing another 30 drops after your test with 6 hits.
160 drops * 0,3 = 48 claims.
40 claims + 2 claims + 6 claims = 48 claims.

So with such a small sample size, there's the risk that the normalisation after a higher than average run just kicks in during your test and you're getting a false-positive for your theory.

That's why you need to run big consecutive sample-size, the impact of solo runs before joining your test, will become neglectable in large sample size, giving you a more realistic and valid result.
 
Last edited:
LeelooM said:
Doing 50 drops at the same places on every run with 1 hour in between.... so enm respawn are taking a lot more time
Mined resources : Oil - lyta - dianthus - typo - garcen

1st run : HR 40% (TT return 110%)
2nd run : HR 25% (TT return 67,28%)
3th run : HR 26% (TT return 72,88%)
4th run : HR 29% (TT return 80,64%) here oil and lyta where coming back again finally
after 6 hours I had 110% TT return again

For ores you need to test this, some are back in 5 min, others after a day, and some are lyst first, then after a few hours they change ...
so when mining for a specific ore, first round I had 100% from that ore, mining after 30-60-90 min gave me nothing but 100% lyst and after 120 min 100% specific ore again :)

All this testing was done when nobody else was around, I just stay there
Click to expand...

Yeah i would go by rule of thumb 3 hours, before mining same area again for specific stuff. Like Terra that needs good depth and small amp.
 
kingofaces said:
That's not proper sampling statistics and actually something that gets taught day 1 in experimental design courses not to do through brute force like that. There's no reason to do 10,000 samples when less than 100 will tell you the same thing if you're using the correct statistical tests.

The statistical tests are telling you what the averages would be at a high sample size like 10,000 drops. On a per treatment basis, a sample size of 30 for a binomial response like hit rate is generally considered pretty decent (n=10 even gets by in some papers). All increasing the sample size does from there is shrink the confidence intervals with the average bouncing around somewhere within those intervals for the most part. An example of that would be the first three treatments here between 0 to 30 minutes. Their averages are not significantly different at this sample size, so if you increased the sample size, the test is saying there's a high likelihood those averages are still not going to be different.

Check out the thread for the first round of testing for more background on this where people asked the same question about using too high of sample sizes.
Click to expand...

I work with 100 drops (so 50 ped enm / 100 ped ores/150 ped trea / amped or not) runs, and even after doing the same runs over and over again (so thousand of drops), the results stay about the same as my first run, so yes need to agree on that 100%.
But after doing 1 run I can already say if that zone is gonna be ok or not ^^
 
LeelooM said:
I work with 100 drops (so 50 ped enm / 100 ped ores/150 ped trea / amped or not) runs, and even after doing the same runs over and over again (so thousand of drops), the results stay about the same as my first run, so yes need to agree on that 100%.
But after doing 1 run I can already say if that zone is gonna be ok or not ^^
Click to expand...

and if the zone is not okay, you leave, and you try around untill you confirmation biased your way to make it fit your theory?

Anyway, for the testing the OP wants to do, that sample size is simply far too small.
 
Last edited:
Alukat123 said:
and if the zone is not okay, you leave, and you try around untill you confirmation biased your way to make it fit your theory?

Anyway, for the testing the OP wants to do, that sample size is simply far too small.
Click to expand...

I do not have a theory ... I uses common sense + I can count.

Do read this Einstein

EDIT : and do not forget that there is the economical fact also so doing the same area's thousents of time is when economical is stable and here it isn't, so you need to adjust so a long term average is a NO GO

https://courses.lumenlearning.com/boundless-statistics/chapter/the-law-of-averages/
 
Last edited:
You must log in or register to reply here.
Back
Top