Masterpieces - the numbers game

[Jenkz]

Originally posted by old.Jessica
It takes less time to make 49 hinges and 1 torso than it does to make 50 torsos.

My problem with this, and what my code hopefully proves is that, you are no more likely to get an MP torso after making 49 hinges than you are making an MP torso straight off.

the maths proves that your "rinse and repeat" method will just be repeated X times until you get a MP, making hinges for absoluetly no reason - and still averaging 50 remakes on a torso before getting a MP

thats the MATHS, as i keep stressing. the practicalities ingame may be different if mythics RNG is incorrect, which i'd like to investigate further...

 

[Artemls]

All I can say is GEEKS!

 

[godfried]

1. Mathematics is the language of nature. 2. Everything around us can be represented and understood through numbers. 3. If you graph these numbers, patterns emerge. Therefore: There are patterns everywhere in nature.
When your mind becomes obsessed with anything, it will filter everything else out, and see that thing, everywhere in nature!

When I was little, my mother told me not to look at the sun. So, once when I was six, I did.

 

[Addlcove]

Originally posted by godfried
When I was little, my mother told me not to look at the sun. So, once when I was six, I did.

and it hurts like hell

 

[Caerdhros]

Originally posted by Roalith
I'm fully aware of the average chances and median results; I'm the one that introduced the median result information into this thread.

Probability says we are more likely to get a result within the first 35 attempts since the last MP (Caerdhros, that's what the median result was based on).

Average says 1 in 50, because of the erroneous results over 35 attempts 'raising the bar', so to speak.

Hinge-method crafting attempts to avoid the erroneous results over 35 attempts. Doesn't always work. Seems to work more often than not, in practical, in-game terms, though.

Roalith, I've looked into the median, and found that the formula for the median (in the perfect system) is ln(.5)/ln(.98), rounded up. This is 35. Therefore, your observed results match a 'perfect' system.

I have also observed this experimentally, I have written a perl script that makes 65000 masterpieces. The median is 35 tries, the average is very close to 50 (masterpiece only AND hinging method, using 10 tries). The only difference between methods is that you will only make 18% as many masterpieces using the hinging method.

This was written using the simple rand function provided by perl, and I would expect the same type of PRNG to be used by Mythic.

Looking at the results from my script, I came across something I didn't expect. If you make 65000 masterpieces, you will make the most (1300, which is 2%, 65000*0.02) after only 1 try, i.e. consecutively. You will make the next most (1274 = 65000*0.02*0.98) after 2 tries. Each consecutive try makes 0.98 as many masterpieces as the last one. I'd like to re-iterate that despite the above, you cannot alter the odds in any fashion. If you hinge using whatever method, you will still take, on average, 50 tries.

I am increasingly convinced that there is nothing wrong with the PRNG used by Mythic.

Formula link: http://www.wku.edu/~neal/statistics/geometric.html

Caerdhros.

 

[krait]

Whatever is supposed to be happening with the rng it's still odd that using the hinge method ,after 157 hinges this happened...

It's not always on the following hinge ,but it does seem to happen quite a bit ,in fairly quick succession ,after a long gap to make a MP hinge.

<shrug>

 

[Caerdhros]

Originally posted by krait
Whatever is supposed to be happening with the rng it's still odd that using the hinge method ,after 157 hinges this happened...

Krait, it's not that odd. It'll happen, on average, once for every 50 masterpieces you make. It's just that everyone remembers when it happens.

Caerdhros.

 

[mark-]

Unfortunately I can't be arsed to read all this thread, but here's my experience (parsed chat.log of lots of hinging with perl/excel):

create.txt was created by test.pl run on my full chat.log

wc -l create.txt
9334 create.txt

9334 Hinges

cat create.txt | grep -i -c "normal"
9151

9151 Normal hinges

cat create.txt | grep -i -c "masterpiece"
183

183 Masterpieces hinges

(183/9151) * 100 = 1.999% chance of masterpiece

Worst 'space' between masterpieces: 304 [happened 1 time]
Best 'space' between masterpieces: 0 [happened 7 times]
Masterpieces that required less than 50 attempts from last masterpiece: 120 (65.5%)

Some statistics from the data:

'number of non masterpiece hinges between masterpieces'

Minimum 0
Maximum 304
Average 49.9
Median 35
Mode 35
Std Dev 47.6

Files: test.pl create.txt mpdata.xls

 

[krait]

I've made countless hinges and only lately started jotting them down.....after a long gap 2 MP's do come in fairly quick succession.
This was my first attempt to make a MP though using the same method and getting a MP on the 1st try saved me an absolute fortune.............thats what it's all about.
If i'd just made Arc BP's the normal way then 157 attempts later i'd have got one (with enough plats)..... I know which way i prefer.

I realise it wo'nt work everytime,because the "2nd Mp" will not follow quickly enough ,but if the gap before a MP is a very long one the next does appear to follow close behind.


p.s. I used bronze hinges and waited for an extra long gap before starting the Breastplate.

 

[mark-]

(The RNG works perfectly)

 

[Caerdhros]

OK, here's the skinny:

Short version: HINGING WASTES YOU TIME AND COSTS YOU MORE.

Long version: I made 3000 bronze brackets, and combined that with mark's data he posted earlier. I wrote a script that read the data, and tried different methods. Here are the results:

No method:
251 MPs in 12095 tries: average 48.1872509960159 tries per MP

10 hinge method:
36 MPs in 2280 tries: average 63.3333333333333 tries per MP

20 hinge method:
80 MPs in 4202 tries: average 52.525 tries per MP

30 hinge method:
110 MPs in 5749 tries: average 52.2636363636364 tries per MP

40 hinge method:
144 MPs in 6995 tries: average 48.5763888888889 tries per MP

The script is here: http://www.whitestaff.co.uk/summary.pl.txt
The data is here: http://www.whitestaff.co.uk/masterpiece.txt

The data is taken from my chat log, and from mark's data (also from the chat log). Each line of the file contains the number of tries between masterpieces. A 1 indicates that the masterpieces were consecutive (mark used 0 as a base; I have incremented his data by 1)

Caerdhros.

 

[SoulFly Amarok]

Originally posted by krait
Whatever is supposed to be happening with the rng it's still odd that using the hinge method ,after 157 hinges this happened...

It's not always on the following hinge ,but it does seem to happen quite a bit ,in fairly quick succession ,after a long gap to make a MP hinge.

<shrug>

omfg

 

[Jenkz]

1. mathematics have proven that hinging wont effect anything assuming the mythic RNG is working properly
2. mythic RNG is working properly based on actual ingame masterpiece data.

doesnt look like there is any way to reduce time or remakes or anything else. for those people who dont believe that, when you run into 70+ remakes even with the hinge method, on really expensive items dont say we didnt warn you....

 

[krait]

Whatever the mathematics say ,it's still an anomaly that a MP does seem to follow rapidly after a MP following a long gap of making any object.

Try it and see.

If it saves gold what's the harm ?

 

[umilard]

1. Mathematics is the language of nature. 2. Everything around us can be represented and understood through numbers. 3. If you graph these numbers, patterns emerge. Therefore: There are patterns everywhere in nature.
When your mind becomes obsessed with anything, it will filter everything else out, and see that thing, everywhere in nature!

When I was little, my mother told me not to look at the sun. So, once when I was six, I did.

From the movie Pi, I love that movie :> But then, i'm can be quite a mathgeek at times :/

 

[Caerdhros]

Originally posted by krait
Whatever the mathematics say ,it's still an anomaly that a MP does seem to follow rapidly after a MP following a long gap of making any object.

Try it and see.

If it saves gold what's the harm ?

Krait, the harm is it COSTS you gold. Look at the stats above: it will take you more tries to make a masterpiece if you use hinging.

Caerdhros.

 

[chretien]

That's because you'll make the same number of MP items but a high percentage of them will be hinges.

 

[old.linnet]

This whole thread reminds me of people who think they can predict lottery numbers I love superstitious crafters (remembers Hendrick assuring me that you were more likely to get skill gains if you crafted in Beno than in Cornwall ...)

 

[Jenkz]

Originally posted by chretien
That's because you'll make the same number of MP items but a high percentage of them will be hinges.

you wont.

you'll just make a hell of a lot of masterpiece hinges, and a hell of a lot more non MP hinges.

and it wont effect the number of remakes of the item itself whatsoever - that remains 1 in 50 chance, and that chance (PRNG)is working correctly, hence the mathematics are sound.

 

[Alrindel]

What about the guy sitting at the forge next to you? He's using the same random number generator you are! HE'S USING UP YOUR MASTERPIECES! KILL HIM!!!

 

[old.Trine Aquavit]

What about the guy sitting at the forge next to you? He's using the same random number generator you are! HE'S USING UP YOUR MASTERPIECES! KILL HIM!!!

That deserved a ROFL

 

[chretien]

Jenkz you misunderstood me, I was agreeing with Caerdhros. You will make the same number of MPs for the total number of things you make whether you hinge or not. If you hinge though a very high percentage of the masterpieces you make will be hinges rather than useful items. (for e.g. 400 hinges + 100 gloves = 10ish masterpieces of which 8 will be hinges and 2 will be gloves; 500 gloves = 10 ish masterpiece gloves).

 

[Jenkz]

sorry

 

[iann]

RNG

Okay I read the first few pages and the last few pages.

Here is what I understand about alot of RNGs. They use a seed, and from that seed spawn a seemingly random number.

However, I know in C++ and Java if you tell it to make a random number on a really fast computer inside a loop it will occasionaly print out the same number in succession. Reason this happens is it gets its randomness from the CPU clock.

I'm guessing the only reason that the hinge thing seems to work is because of the CPU clock.

Make hinge, hinge is MP (means that the CPU is on the correct "time" for you to get a MP). Since hinges make fast, then if you start a BP I'm guessing it does the randomness at the begining of the craft, not at the end. In this case you might catch it on the same cycle...maybe.

Other than this very very small chance that this is what is happening and this is assuming that myhic is for some reason checking quality before the item is finished rather than after the timer. Otherwise I think that it's just the human mind looking for a pattern.

 

[old.Trine Aquavit]

While RNGs are sometimes seeded from the CPU clock, they aren't seeded every time the RNG is used. The seed is just used to initialise the RNG at start up. After initialisation the RNG doesn't need to be re-seeded, it just chunters away generating numbers in sequence. Basically, the last number generated provides the seed for the next number, and that provides the seed for the next, and so on.

Besides, the RNG will be shared between all players and probably all random number calculations on the server. Having a single RNG for each user makes no sense whatsoever.

Oh, and I doubt hinges are made so fast that you could get two hinges made from requests from the same player within one CPU cycle...

User: <hits craft>
Client: I'd like to make a hinge please
<TCP/IP over 15 hops or so>
Server: OK client, I'll be right with you. Hey, RNG, I'd like a random number please
RNG: OK, I've got one right here: 73409283
Server: Thanks mate. Oh, this lucky fella made a masterpiece
<TCP/IP back over the 15 hops>
Client: Oooh, a MP! Better tell the user
User: Nice a MP, better make another
User: <hits craft>
...

All within one CPU cycle

 

[Novamir]

do computer components really talk to each other like that? like in disney movies?

 

[Krissy]

I beg of u plz, let this thread die.

Its been proven wrong, so nothing more to say. start another thread about somthing more intresting