Sign in to follow this  
joedobo

results of creation spamming for butcher knife blades of rarity

Recommended Posts

I made butcher knife blades for about 9.5 hours. For the first two hours I got 3 MOI's which is good. The following approximate 7 hours I got nothing. Although, in that time there were many missed windows. That 7 hours doesn't look good to support the average of1 MOI every 2 hours on SFI. But then again, about 9.5 hours I got 3 MOI's and I'd expect 4 to 5. It's not that far off.

 

What do you think?


 For the following string of characters each character represents a 30 second window. I could use suggestions on better ways to present the data which the strings are trying to communicate.
X is an action was started in the window. W is no action was started; I missed the window. R is a rarity MOI occurred.

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXWXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXWWWWXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXRXXXXXXXXXXXXXXWXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXRXXXXXRXXXXXXX
XXXXXXXXWWXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXWWWXWXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXWWWWWWWWXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXWWWWWWXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXWXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXWWWWWWWWXXXX
XWWWWWWWXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXWWWXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXWWWWWWWWWXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXWWWWWWXXXXXXXXXXXXXXXXXXXXXXXXXXXXWWWWWWWWWWWWWWWWWXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXWWWWWWWWW
WXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

 

Share this post


Link to post
Share on other sites
33 minutes ago, joedobo said:

I made butcher knife blades for about 9.5 hours. For the first two hours I got 3 MOI's which is good. The following approximate 7 hours I got nothing. Although, in that time there were many missed windows. That 7 hours doesn't look good to support the average of1 MOI every 2 hours on SFI. But then again, about 9.5 hours I got 3 MOI's and I'd expect 4 to 5. It's not that far off.

 

I recently had a similar experience - rare spamming horse shoes on 6 accounts, I got 9 MOIs within the first ~3 hours, and for the next 3 nothing (6 accounts! No MOI!), until I changed to spamming hammer heads briefly at the end of the session and got one.

 

I wouldn't be surprised if there's something to stop repetitive spamming from giving MOIs at some point (particularly if it's of the same item), but this needs much more testing. It could have also been bad luck.

Edited by Docterchese

Share this post


Link to post
Share on other sites

@Docterchese Agreed. I've done couple other runs with similar outcomes.  I also thought about swapping after a MOI between a few random choices.  

Share this post


Link to post
Share on other sites

Well, as i said before, there is no way to determine anything with a sample so small, it's pretty much all speculation.

In the past days, anyway,  i randomly made about 50hours of spamming on non-prem (72% creation chance - 50 total hours, it was multiaccount)

And while i got 6 rares in the first ~20hours, which is absolutely as expected, i got nothing,  in the last 30hours, even though i havent' checked for mois on failed creations, and i did switch some tools to other ones, so it might be worth doing some more investigation/testing to determine if it's anything but RNG.
 

Share this post


Link to post
Share on other sites

Ok so stats time:

 

Assuming you did exactly 1 action per 30 seconds in your "X" or "R" period you had 986 active periods.

 

The chance you are in a valid rare window, assuming 10s deed bonus is 0.4316% = (1- (1-1/3600)^30) *  (1- (1-0.5)*(1-0.04)*(1-0.000103)

 

Huzzah this looks binomial to me.

Number of trials: 986

Probability: 0.4316%

Successes: 3

 

From the binomial distribution you achieved a bottom 38.46% result (i.e. 61.54% chance of a better result)

 

So we cannot reject the null hypothesis of the stated rarity chances at any reasonable confidence level.

 

Edit: for context  often a reasonable confidence level would be 0.5%-5%, so you'd have to repeat this with 0 rares to claim  below a 5% confidence level or get around ~1200 datapoints with no rares to hit 0.5%. Please do continue to update your sample if you keep testing though, larger the sample the more reliable the testing.

Edited by Lethyria
  • Like 1

Share this post


Link to post
Share on other sites
34 minutes ago, Lethyria said:

Assuming you did exactly 1 action per 30 seconds in your "X" or "R" period you had 986 active periods.

I'm curious how you'd handle the W where I missed a window? Hitting every single window isn't going to happen for me.

 

My times aren't exactly 30 seconds and they vary. Above was 22 to 29 with and average of about 25 seconds. I understand the probability model you gave me in another post. The problem is that my times aren't exact and they never will be. This is why I switched to working with the idea that as long as I start an action in the 30 sec window I just use 30 second actions.

 

Since you're familiar with stats, do you know of a better graphic for X-R string? I think it's time series data. It seems I need to add a binomial analysis too.

Share this post


Link to post
Share on other sites
2 minutes ago, joedobo said:

I'm curious how you'd handle the W where I missed a window? Hitting every single window isn't going to happen for me.

 

My times aren't exactly 30 seconds and they vary. Above was 22 to 29 with and average of about 25 seconds. I understand the probability model you gave me in another post. The problem is that my times aren't exact and they never will be. This is why I switched to working with the idea that as long as I start an action in the 30 sec window I just use 30 second actions.

 

Since you're familiar with stats, do you know of a better graphic for X-R string? I think it's time series data. It seems I need to add a binomial analysis too.

 

I'm running on the assumption that each creation is essentially an entirely independent trial, so the chance that each is in a rare window is only affected since the time since the last action. As such I ignore the "W" data as it is untested, in theory you could calculate the chance for each action, which would be 30s in any period directly following a "W" where the others would just be the time between the two periods. I think that breaking down each action is probably too much work and makes the stats a bit more complicated, the simple approach would be just 3 numbers: Average time between queueing actions, Number of rare rolls, Number of actions. If you wanted to get really in depth you could output each action with the time since last action and the result (rare roll/not), we could then break down the total number of potential rare seconds and use that as the test stat, but I think this is easier to understand.

 

The probability used would be  0.3599% if you assume 25s per action, and 30s can be a bit sketch as your action could fall just outside the window or not, so 29s is probably the best chance per create.

 

I don't think this is really timeseries data, other than that its interesting to see your breaks and things :P. When looking at timeseries data its usually only relevant when the data is dependant on the previous point (like stock prices which move from one point up and down) by treating your data as independent throws the time element becomes less important other than to clarify how much time you've spent trying to figure it out.

 

  • Like 2

Share this post


Link to post
Share on other sites

I got some more data and a bit of good news that was shared in Discord.

 

First off, from Darklords  in Discord

Quote

Alright so just to be open about that we did indeed reduce single action rare creation rates on NFI only on its launch.
But it has not changed at all since the launch, or at all on sfi servers.

link to discord post

 

Onward to the addition data.
Using WA to search my event logs and I looked for "You create a butchering knife blade." And so far for this month I get a count of 15,055. I have made 28 rare butcher knife blades, no supreme and (not surprisingly) no fantastic.
I'm using an estimated 25 sec average time between hitting the create button. I haven't done anything different than above, just more days of it. This should be 0.3599% for our probability model.  You can look at the picture for the math, but basically it's really bad. If it wasn't for what Darklords said in Discord, I'd swear something is broken. I'm on Xana, SFI. Maybe I did something wrong in all this. I'm not sure if I should give up on the supreme butcher knife blade. That was part of what I wanted to achieve here. The other part was to investigate the rarity creation chances.

 

using this tool: Binomial Distribution Calculator - Binomial Probability Calculator, Binomial CDF calculations (gigacalculator.com)
 

Spoiler

LM1GVHc.png

 

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this