Hit rate testing round 2: Claim respawn rates

[kingofaces]

I'm going to keep these requests for miners and results threads together from now on and add to this thread when the testing is done.

After our first round of testing, we found that previous miners recently in the area can affect your hit rate. The next question is how long does that effect last?

This can be tested much the same as before. I have a set of coordinates ready to go, and we'll use the same setup of F-101 with no attachments or enhancers (you just need enough skill to max its depth and radius) to make sure as many variables we can account for are controlled. What we'll do instead is have the first person drop at each of the coordinates in order. Then, we'll wait a set amount of time x for the next person to go, a slightly longer period of time for the next, and so on. 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.

We don't need the claims to "fully" regenerate (though it would be nice to document), but just measure the hit rate change over wait time. These results won't be an ultimate guide to when you can get the best hit rate for a specific area since that can likely change due to other factors, but we can at least figure out how long you should wait to at least avoid significantly overlapping on "emptied" areas.

So first, I need to know how many people are interested tentatively in joining in with the first miner on March 9, likely at 16:00 UTC again as a start time. I'd like to get everyone to meet up just to confirm we're all ready to go at our preset individual times.

Last time, we shared the loot to in order to spread out loses if they happened, so we can do that again if people agree to it. I'm looking to fill this rough setup of treatments right now, but that can be modified next week as things develop, we get more people, etc.:

Miner #	Wait time (minutes) after previous miner	Avatar	Start time (UTC)
1	NA (drop until radius filled)	Kingofaces	16:00
2	15	Leeloo	16:15
3	30	Casper	16:45
4	60	Leeloo	17:45
5	120	Casper	19:45

This one will be a bit more time sensitive, so if sometime next week you realize you can't make it at an assigned time, let me know before that Saturday so I can modify the order and let others know since it will change their start time.

This is only going to be done at one location for now as an initial experiment. It's possible claims may regenerate quicker in some locations, but that'll be secondary variation to talk about afterwards. Depth testing will probably be the next experiment after this one later in March.

 

[kingofaces]

We had some interesting results from this today. Methods are all above mostly, so I'll go straight into results.

In short, I went as the first person to "clear out" the full finder radius area, so I dropped both ore/enmatter and kept dropping at each locations until I no longer found a claim. The others did just one drop of ore/en at each of the 30 locations. We had a 5 minute buffer last time when we found that previous miners can affect your hit rate, so we increased the initial start time after the first miner to 15 minutes.

However, we found no effect of time since last drop at a location in the treatments listed above (p = 0.429, well above the threshold of 0.05 where you'd say wait time affected hit rate).

As a reminder, time 0 above is when the first person went, not 0 minutes after another miner. We were watching at least 20 minutes before hand in the area, and no other miners were present. The remainder of treatments show minutes since the last miner. All the confidence intervals overlap, so none of those runs had significantly different hit rates, also shown in the high p-value. My hit rate at time 0 also only shows the initial first drop at each location for comparison. Just as background, my hit rate was 27% instead of 30% when I add in those extra drops (18 extra combining ore/en), but this wasn't testing the strategy of dropping until your finder radius is filled up, so take difference that with a grain of salt.

So why no effect here, but a strong effect in our last round of testing? It looks like whatever the "respawn" rate is (if it's truly just random respawning or something else) functionally replenishes claims in less than 15 minutes, but doesn't come into effect too much at 5 minutes. That's a much narrower window than we expected, but we also showed that people mining within the area at those longer time frames definitely do not affect your hit rate. We couldn't calculate actual respawn rates, but if it takes less than 15 minutes, it's probably not worth worrying about more than what we've looked into.

In short, if you see someone on your radar mining, it's probably best to just wait 15 minutes, at which point your hit rate isn't going to be significantly different than what the previous miner got.

We did record TT, but due to a global and near-global as outliers in the first two runs, this is again something to be cautious about interpreting since someone might be tempted to say there's a slight difference at 30 minutes. In short, pretty standard TT variability with no trend or significant statistical tests:

So hopefully that mostly closes the book on mining whether you know other people are around. If you're worried about reduced hit rate and low TT, just wait 15 minutes or at least more than 5.

 

[R4tt3xx]

Personally I do not think that there are pre-generated claims, its more like a set value of resources that exist in an area that can be extracted upon request.

 

[LeelooM]

I'll be there and agree to split

 

[kingofaces]

Personally I do not think that there are pre-generated claims, its more like a set value of resources that exist in an area that can be extracted upon request.

If that were the case (at least for finding a generic claim, not specific resource type), we likely wouldn't have found this in the last round. Maybe MA does something different on the back end and it's not really claims "respawning" or pre-generated before your drop as you say, but since the end result is that your hit rate and TT are affected, it's functionally as if they were "in the ground" rather than generated at the drop request.

I used to be in the claims aren't in the ground camp before, but I'm getting convinced the other way with the data so far. We'll see how this and the later depth testing turn out.

 

[Casper]

I'm in and willing to split.

 

[R4tt3xx]

If that were the case (at least for finding a generic claim, not specific resource type), we likely wouldn't have found this in the last round. Maybe MA does something different on the back end and it's not really claims "respawning" or pre-generated before your drop as you say, but since the end result is that your hit rate and TT are affected, it's functionally as if they were "in the ground" rather than generated at the drop request.

I used to be in the claims aren't in the ground camp before, but I'm getting convinced the other way with the data so far. We'll see how this and the later depth testing turn out.

Other active miners within the same area as you effect each other's hitrate, I agree.

This is my hypothesis and always has been, but due to the idea's complexity, it is very difficult to verify. The game's mining areas are broken down into blocks or fields of the same size, when a miner drops a probe and misses, it creates a set of "virtual claims" around that avatar that populate that field, those claims exist in that field for as long as no miner in that area misses, once a miss occurs, that new avatar is the origin of the new set of "virtual claims".

 

[kingofaces]

That's three people so far. I would be great to have two more for next weekend.

 

[kingofaces]

Just another reminder to sign up for this weekend.

 

[R4tt3xx]

If that were the case (at least for finding a generic claim, not specific resource type), we likely wouldn't have found this in the last round. Maybe MA does something different on the back end and it's not really claims "respawning" or pre-generated before your drop as you say, but since the end result is that your hit rate and TT are affected, it's functionally as if they were "in the ground" rather than generated at the drop request.

I used to be in the claims aren't in the ground camp before, but I'm getting convinced the other way with the data so far. We'll see how this and the later depth testing turn out.

Claims are not in the ground.... They are generated upon a miner getting a miss when mining and spawn around that avatar, any avatar that is within radar range of that event will have their claim spawning layout altered to match the miss event's coordinates. Any avatar that did not witness that event will not have their layout altered.

When a request is made, a timestamp is used to generate the array based from the coordinates of the last miss that the avatar last witnessed. If that avatar's request is within range of a found "virtual claim", it spawns.

Being aware of other avatars actions is crucial.

 

[Idiocracy]

Claims are not in the ground.... They are generated upon a miner getting a miss when mining and spawn around that avatar, any avatar that is within radar range of that event will have their claim spawning layout altered to match the miss event's coordinates. Any avatar that did not witness that event will not have their layout altered.

When a request is made, a timestamp is used to generate the array based from the coordinates of the last miss that the avatar last witnessed. If that avatar's request is within range of a found "virtual claim", it spawns.

Being aware of other avatars actions is crucial.

schroedingers claim?

 

[R4tt3xx]

schroedingers claim?

Actually the opposite, if I see someone mine at a set location and miss, do you think I am going to mine there ?
HELL NO !!!

But if someone misses and I do not see it, it does not impact me.

 

[kingofaces]

Claims are not in the ground.... They are generated upon a miner getting a miss when mining and spawn around that avatar, any avatar that is within radar range of that event will have their claim spawning layout altered to match the miss event's coordinates. Any avatar that did not witness that event will not have their layout altered.

When a request is made, a timestamp is used to generate the array based from the coordinates of the last miss that the avatar last witnessed. If that avatar's request is within range of a found "virtual claim", it spawns.

This isn't really the place for speculating to that degree without data, and the whole intent of this series of experiments is to focus on data. Without good data, claims about knowing how mining claims work generally need to be dismissed as it would for any other subject when you start scientifically testing things.

That being said, something like you're claiming would have been apparent in the last round of testing, which it was not. I was outside of radar range most of the time, though did catch up to people sometimes too. We'll see what the next round of testing shows, but it looks like what data there is doesn't really support this. As I said before, there's a difference between knowing what actually goes on at the back-end at MA and describing what functionally matters to us on the front end. Functionally, claims in the ground are affected by other miners, but it's pointless to speculate beyond that without specific data tailored to the specific question.

 

[kingofaces]

Bump for additional testers. Two more would be ideal. Let me know sometime on Friday at the latest.

 

[R4tt3xx]

This isn't really the place for speculating to that degree without data, and the whole intent of this series of experiments is to focus on data. Without good data, claims about knowing how mining claims work generally need to be dismissed as it would for any other subject when you start scientifically testing things.

That being said, something like you're claiming would have been apparent in the last round of testing, which it was not. I was outside of radar range most of the time, though did catch up to people sometimes too. We'll see what the next round of testing shows, but it looks like what data there is doesn't really support this. As I said before, there's a difference between knowing what actually goes on at the back-end at MA and describing what functionally matters to us on the front end. Functionally, claims in the ground are affected by other miners, but it's pointless to speculate beyond that without specific data tailored to the specific question.

I understand that, I have been participating in this little adventure since open beta so I do have a bit of observational experience with regards to how the game's systems have been constructed. Of course that means nothing without the correct data and the correct data means nothing without the correct context and variables.

You are looking at statistics at the moment which is what you are good at, making sense of things after the fact. When I play, I have ideas racing in my head about how this little game works and it has always come back to the word that I pm'ed you.

I understand that this is just from personal experience, but I do not have enough of a scientific mind yet to start to perform proper tests, would you be willing to help me with this in order to get my ideas and hypotheses properly tested ?

 

[kingofaces]

I understand that, I have been participating in this little adventure since open beta so I do have a bit of observational experience with regards to how the game's systems have been constructed. Of course that means nothing without the correct data and the correct data means nothing without the correct context and variables.

You are looking at statistics at the moment which is what you are good at, making sense of things after the fact. When I play, I have ideas racing in my head about how this little game works and it has always come back to the word that I pm'ed you.

I understand that this is just from personal experience, but I do not have enough of a scientific mind yet to start to perform proper tests, would you be willing to help me with this in order to get my ideas and hypotheses properly tested ?

I responded over PM, but for any interested, this line of questioning gets into whether claims are truly random, or clumped in some fashion. There are ways to assess that, but I'll leave that lower on the priority list for now.

 

[kingofaces]

We had some interesting results from this today. Methods are all above mostly, so I'll go straight into results.

In short, I went as the first person to "clear out" the full finder radius area, so I dropped both ore/enmatter and kept dropping at each locations until I no longer found a claim. The others did just one drop of ore/en at each of the 30 locations. We had a 5 minute buffer last time when we found that previous miners can affect your hit rate, so we increased the initial start time after the first miner to 15 minutes.

However, we found no effect of time since last drop at a location in the treatments listed above (p = 0.429, well above the threshold of 0.05 where you'd say wait time affected hit rate).

As a reminder, time 0 above is when the first person went, not 0 minutes after another miner. We were watching at least 20 minutes before hand in the area, and no other miners were present. The remainder of treatments show minutes since the last miner. All the confidence intervals overlap, so none of those runs had significantly different hit rates, also shown in the high p-value. My hit rate at time 0 also only shows the initial first drop at each location for comparison. Just as background, my hit rate was 27% instead of 30% when I add in those extra drops (18 extra combining ore/en), but this wasn't testing the strategy of dropping until your finder radius is filled up, so take difference that with a grain of salt.

So why no effect here, but a strong effect in our last round of testing? It looks like whatever the "respawn" rate is (if it's truly just random respawning or something else) functionally replenishes claims in less than 15 minutes, but doesn't come into effect too much at 5 minutes. That's a much narrower window than we expected, but we also showed that people mining within the area at those longer time frames definitely do not affect your hit rate. We couldn't calculate actual respawn rates, but if it takes less than 15 minutes, it's probably not worth worrying about more than what we've looked into.

In short, if you see someone on your radar mining, it's probably best to just wait 15 minutes, at which point your hit rate isn't going to be significantly different than what the previous miner got.

We did record TT, but due to a global and near-global as outliers in the first two runs, this is again something to be cautious about interpreting since someone might be tempted to say there's a slight difference at 30 minutes. In short, pretty standard TT variability with no trend or significant statistical tests:

So hopefully that mostly closes the book on mining whether you know other people are around. If you're worried about reduced hit rate and low TT, just wait 15 minutes or at least more than 5.

I just posted this on the first page (2nd post), so just leaving this here as an update.

 

[R4tt3xx]

We had some interesting results from this today. Methods are all above mostly, so I'll go straight into results.

In short, I went as the first person to "clear out" the full finder radius area, so I dropped both ore/enmatter and kept dropping at each locations until I no longer found a claim. The others did just one drop of ore/en at each of the 30 locations. We had a 5 minute buffer last time when we found that previous miners can affect your hit rate, so we increased the initial start time after the first miner to 15 minutes.

However, we found no effect of time since last drop at a location in the treatments listed above (p = 0.429, well above the threshold of 0.05 where you'd say wait time affected hit rate).

As a reminder, time 0 above is when the first person went, not 0 minutes after another miner. We were watching at least 20 minutes before hand in the area, and no other miners were present. The remainder of treatments show minutes since the last miner. All the confidence intervals overlap, so none of those runs had significantly different hit rates, also shown in the high p-value. My hit rate at time 0 also only shows the initial first drop at each location for comparison. Just as background, my hit rate was 27% instead of 30% when I add in those extra drops (18 extra combining ore/en), but this wasn't testing the strategy of dropping until your finder radius is filled up, so take difference that with a grain of salt.

So why no effect here, but a strong effect in our last round of testing? It looks like whatever the "respawn" rate is (if it's truly just random respawning or something else) functionally replenishes claims in less than 15 minutes, but doesn't come into effect too much at 5 minutes. That's a much narrower window than we expected, but we also showed that people mining within the area at those longer time frames definitely do not affect your hit rate. We couldn't calculate actual respawn rates, but if it takes less than 15 minutes, it's probably not worth worrying about more than what we've looked into.

In short, if you see someone on your radar mining, it's probably best to just wait 15 minutes, at which point your hit rate isn't going to be significantly different than what the previous miner got.

We did record TT, but due to a global and near-global as outliers in the first two runs, this is again something to be cautious about interpreting since someone might be tempted to say there's a slight difference at 30 minutes. In short, pretty standard TT variability with no trend or significant statistical tests:

So hopefully that mostly closes the book on mining whether you know other people are around. If you're worried about reduced hit rate and low TT, just wait 15 minutes or at least more than 5.

Am I correct to assume that the avatars participating in the test did not see each other's actions ?

If so then fantastic, I have one less variable to worry about...

Just a quick side-note on the respawn timer, I know that you are not going to believe me but I am going to place it here just as proof... The respawn timer should vary depending on what tool your are using and the lvl of your avatar.

From the observation of a 5 minute respawn cycle, I can deduce that no one was above lvl 80. Of course that statement means very little without proof and is most likely to be true due to the sheer amount of time it would take a acquire such a level.

 

[kingofaces]

Am I correct to assume that the avatars participating in the test did not see each other's actions ?

If so then fantastic, I have one less variable to worry about...

Just a quick side-note on the respawn timer, I know that you are not going to believe me but I am going to place it here just as proof... The respawn timer should vary depending on what tool your are using and the lvl of your avatar.

From the observation of a 5 minute respawn cycle, I can deduce that no one was above lvl 80. Of course that statement means very little without proof and is most likely to be true due to the sheer amount of time it would take a acquire such a level.

Correct on not seeing others while mining aside maybe a split for a second in the first (drop 15) and second (drop 1) runs, but no one else was mining at the same time aside from that. Just to be clear, the previous experiment before this showed respawns did not cycle back in 5 minutes. This is showing more like 15 minutes when other miners are involved.

We don't know though if the dynamics are different if you're competing against yourself (e.g., a personal timer where you get reduced hits in areas you recently dropped). That's an area I'll leave for individuals to test though since that can be done solo. The main thing here was whether we had to worry about other miners, which the answer is only a little bit.

As for the rest, all the tools were the same with the same radius and even depth (next round of testing will be on depth affecting HR between miners and maybe individually too). In any profession, skill hasn't really been shown to affect anything beyond the ability to use a tool to max ability (aside from things like agility, strength, or special new skills like skinning etc.), so I wouldn't really expect that to be different here all of a sudden.

 

[R4tt3xx]

We have a ton of info at the moment... Time to put it all together...

Average hitrate % 27, Lowest hitrate % 5
Perceived respawn time 15 minutes..

Here is some quick math

I make a carpet bomb run consisting of x double drops (so every point is both entmatter and ore).

A line consisting of 1 point consists of one unique state, this is an assumption as the lowest hitrate was not 0 %
A line consisting of 2 points consists of 2 unique states, namely 0 which represents a miss and a 1 which represents a hit
A line consisting of 3 points consists of 3 unique states namely 1,11 and 111
A line consisting of 4 points consists of 4 unique states namely 1,11,111 and 1111

Lets make a grid for line 1

State - Array
1 - 0 0 0 1
2 - 0 0 1 1
3 - 0 1 1 1
4 - 1 1 1 1

Whats the average hitrate of that state ?

(sum of all the states)/(Max(state)^2)/2*100

If we perform this simulation on line 15, we get a hitrate of 26.66666667% with a min of 1/15*100 = 6.666666667%.

The above is a possible model to actual recorded observations.

 

[Kosh]

Good job and thanks for posting the results and your analysis!

I was a bit surprised by the first round of analysis because I never experienced lower hit-rates in areas that are mined heavily. Your tests indicate that the reason is probably that the "recovery rate" is just a few minutes (5-15), so unless you are closely following another miner you will not be affected too much.

 

[LeelooM]

Correct on not seeing others while mining aside maybe a split for a second in the first (drop 15) and second (drop 1) runs, but no one else was mining at the same time aside from that. Just to be clear, the previous experiment before this showed respawns did not cycle back in 5 minutes. This is showing more like 15 minutes when other miners are involved.

We don't know though if the dynamics are different if you're competing against yourself (e.g., a personal timer where you get reduced hits in areas you recently dropped). That's an area I'll leave for individuals to test though since that can be done solo. The main thing here was whether we had to worry about other miners, which the answer is only a little bit.

As for the rest, all the tools were the same with the same radius and even depth (next round of testing will be on depth affecting HR between miners and maybe individually too). In any profession, skill hasn't really been shown to affect anything beyond the ability to use a tool to max ability (aside from things like agility, strength, or special new skills like skinning etc.), so I wouldn't really expect that to be different here all of a sudden.

As I remember correct HR for every miner was above 30% so that was ok and a timer of 15 min is pretty fast
Question is also if claims where maxed and I think that was the case cause looking at TT return, can't say where, but this area I know for having very good or very bad returns, nothing in between and ofc I have no idea what makes those differences.

But was great fun again and I'm in for the next round !!!

 

[Ace Flyster]

You may / or may not find this interesting. I did a similar test many years ago.

Rgds

Ace

https://www.planetcalypsoforum.com/...ores-refresh-rate-in-mining&highlight=refresh

 

[kingofaces]

We have a ton of info at the moment... Time to put it all together...

Average hitrate % 27, Lowest hitrate % 5
Perceived respawn time 15 minutes..

Here is some quick math

I make a carpet bomb run consisting of x double drops (so every point is both entmatter and ore).

A line consisting of 1 point consists of one unique state, this is an assumption as the lowest hitrate was not 0 %
A line consisting of 2 points consists of 2 unique states, namely 0 which represents a miss and a 1 which represents a hit
A line consisting of 3 points consists of 3 unique states namely 1,11 and 111
A line consisting of 4 points consists of 4 unique states namely 1,11,111 and 1111

Lets make a grid for line 1

State - Array
1 - 0 0 0 1
2 - 0 0 1 1
3 - 0 1 1 1
4 - 1 1 1 1

Whats the average hitrate of that state ?

(sum of all the states)/(Max(state)^2)/2*100

If we perform this simulation on line 15, we get a hitrate of 26.66666667% with a min of 1/15*100 = 6.666666667%.

The above is a possible model to actual recorded observations.

I'm not really sure what you're getting at there, but if you mean predicting the true average hit rate for a given time or area, the much simpler terminology would be a sequential probability ratio test looking at how often you get hits and misses. That's tricky at best though, and isn't really relevant to the other miners question, but rather how do you know you've had enough samples to make a decision. The true mean of the hit/miss random number generator can vary, and if you get more samples, you get a better idea of what that mean is. That's basically how the confidence intervals in the graphs work too.

 

[kingofaces]

You may / or may not find this interesting. I did a similar test many years ago.

Rgds

Ace

https://www.planetcalypsoforum.com/...ores-refresh-rate-in-mining&highlight=refresh

I hadn't read that one in years and forgot about it. Your test two was kind of in the ballpark of our first test for hit rate, and you waited 8 minutes instead of 5. That was a little while ago (there's no way 2010 is already almost 10 years ago) and single avatar, but if nothing has really changed, it's probably safe to say you need to wait at least 10 minutes to start getting back to normal HR.

 

[Ace Flyster]

I hadn't read that one in years and forgot about it. Your test two was kind of in the ballpark of our first test for hit rate, and you waited 8 minutes instead of 5. That was a little while ago (there's no way 2010 is already almost 10 years ago) and single avatar, but if nothing has really changed, that probably safe to say you need to wait at least 10 minutes to start getting back to normal HR.

I vaguely remember 'feeling' that after the 8 mins my hit rate wasn't too good. I did have plans to greatly expand on it, but got bored lol

Rgds

Ace

 

[R4tt3xx]

I'm not really sure what you're getting at there, but if you mean predicting the true average hit rate for a given time or area, the much simpler terminology would be a sequential probability ratio test looking at how often you get hits and misses. That's tricky at best though, and isn't really relevant to the other miners question, but rather how do you know you've had enough samples to make a decision. The true mean of the hit/miss random number generator can vary, and if you get more samples, you get a better idea of what that mean is. That's basically how the confidence intervals in the graphs work too.

Are you aware of this paper done in 2009 looking at mining in Entropia Universe ?

Take a look if you have not seen it.

I will reply with a bit more information later.

Thanks again..

 

[kingofaces]

Are you aware of this paper done in 2009 looking at mining in Entropia Universe ?

Take a look if you have not seen it.

I will reply with a bit more information later.

Thanks again..

It's been years since I've seen that and I forgot about it. A lot of that is actually the type of analyses I do in my line of work too, so a lot of the concepts there apply to what we're doing and vice versa.

 

[R4tt3xx]

It's been years since I've seen that and I forgot about it. A lot of that is actually the type of analyses I do in my line of work too, so a lot of the concepts there apply to what we're doing and vice versa.

Amazing how the average hitrate you got matches the one in the PDF.

 

[kingofaces]

Amazing how the average hitrate you got matches the one in the PDF.

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.