[FIXED][#174] Rarepart drop algorithm appears to be skewed (not random)

[Torg]

Not all weapons need an ATP part. So if the cell for ATP is blank, it means the part isn't required.

true. which leaves us guessing even more.

[Doc Holliday]

brilliant. i think this is as close to "valid data" as we can get. we can clearly see tech drop is not spread evenly. to my guess this is because its randomized. the initial question (thank you, William) here was: can we get a more equal, less random distribution? i just doubt this is either viable or desirable. apart from that: we dont have statistics on how many rare items have been built yet on titan. so again i doubt a cs is harder to get than a healing light or winding argument. of course: there seems to be a pattern, concerning missing ATPs for some spells and guns, but i wouldnt call this a proof.

maybe the guns listed as an example were not the best choices but clearly there are issues with what drops and where. i would swear to it.

I havent ever written down the parts i take from firemobbing in comparison to other techs but i have definitely noticed a difference in loots that drop. it seems to be the same weapons and the same parts dropping. I tried hunting hoverbots but the drop rate is significantly lower than firemob drops. Ie you dont get the guarenteed tech for every 100/100 or above killed. This made it tough to test because a few hours farming those will not produce the same volume of techs as what a few hours fire mobbing will provide.

Additionally there are still guns that i have never seen parts for drop from any firemobs. Between myself and my clan mates who i have ressed techs for and who have handed me 100s now over the time titans been up. This makes me wonder. we had a discussion about it all too. We are in agreement from all our personal experiences that the loot tables do appear somewhat imbalanced and its not as "random" as people like to believe.

Its something that needs to be looked at by the devs honestly. Yes there was always that part that you were searching for out of the set of five or six (PE Tech was my biggest bug bear in nc1 by FAR) but the sheer amount of common techs that stack up in a batch of firemobs and turn out the same makes me think something is very much off.

Finally. Yes torg i read your first post in reply too what i wrote. Paragraph the post please next time and yes i can accept an intelligent and carefully thought out opinion when its presented succinctly and is on topic and to the point of discussion. Then i will respond to said post in a similar way. This is not intended to sound sarcastic either. Please dont mistake it. I have a good grasp of english as do you so lets keep it civil and on topic.

[Brammers]

Raw data totals and pivots for Terra and Titan can be found at:-

https://docs.google.com/spreadsheet/...lE&output=html

I've now added Mars, Mercury and Neptune data. Neptune server was discontinued before 2.2, however it seems some people used the Neptune server as a 2nd rares list before it was disabled on the THN rares DB.

[William Antrim]

Any chance of some official comment?

[Torg]

Any chance of some official comment?

that's the single important question. i doubt there's an easy way to replace the randomizing algorithm governing the rare part loot table database, but who knows? any official opinion on feasibility and/or timeframe would be welcome to many of us. i, personally, dont think theres a problem at all, its just one of neocrons many inconsistencies, adding atmosphere and character to the game.

[William Antrim]

I wasnt asking you.

[Torg]

I wasnt asking you.

do you want me to care?

[Mokoi]

No need to get hostile here.

As there is no reason or intent to skew the drop chance for specific parts in one direction or the other, we shall inspect the matter as time permits.
Needless to say, that this is not a huge priority right now with a very important R#175 close to release and in need of testing.

>> moved to bug reports
From here on, please keep opinions out of this thread and respond _only_ if you have factual info to report on this issue.

[William Antrim]

Is it possible to inspect runners cabinets and their rare storage? I have never yet seen an addtech of Ravager drop but I have collected almost 4 full sets of ravager parts. Same with the CS. Everyone is asking for the same part. You will see the parts people are missing and it is always one or two parts to EACH rare. It has been this way for ten years. I have no statistical evidence to support it other than the farming I have done in that time. I have probably farmed thousands of techs by now. It is a pattern I have seen in that time.

[CMaster]

Better would be to set up a log that records when each tech part drops - what's already in cabinets is subject to a whole range of distortions.

Then wait months for any kind of result, because with a pool this big, you need huge samples before you can start coming to any kind of conclusion. My probability isn't good enough to know the actual numbers, but it strikes me that with several hundred different parts, you'd actually expect to have some dropping 100 while another has only done 10 in the same time.

[William Antrim]

Better would be to set up a log that records when each tech part drops - what's already in cabinets is subject to a whole range of distortions.

Then wait months for any kind of result, because with a pool this big, you need huge samples before you can start coming to any kind of conclusion. My probability isn't good enough to know the actual numbers, but it strikes me that with several hundred different parts, you'd actually expect to have some dropping 100 while another has only done 10 in the same time.

based on purely non-scientific experience but extensive rare part farming over the years I concur that this is what you would find. I have no proof however and THAT is the frustrating part.


What distortions? Rares should drop evenly and they dont. What can be distorted about that? If you looked in the cabs you should see a roughly even number of everything allowing for variables due to randomisation. Yeah you will find some parts drop more sure as they randomly popped up in a given day. However there should not be a hugely skewed number of CS comps etc.

2 nights ago I went rhinoing and again last night. Collected about 40 parts. In that time 6 of those parts had doubles and one even had a triplicate. so roughly 25% of my rare farming that night was the same part. I ressed the first of each to see what they were and then chucked em in the clan cabs to find out that, not surprisingly, I already had several of each part.

That is why I made the thread.

[CMaster]

What distortions? Rares should drop evenly and they dont. What can be distorted about that? If you looked in the cabs you should see a roughly even number of everything allowing for variables due to randomisation. Yeah you will find some parts drop more sure as they randomly popped up in a given day. However there should not be a hugely skewed number of CS comps etc.

Going by stuff that's currently in cabs, you have two problems: one, what people may throw away (don't know if anybody does, but it's possible) and two, what people have already built. People are more likely to trade to complete popular weapons, so you'd see less say, CS parts than you would say, EoF ones. Recording what drops gets rid of these ambiguities.

2 nights ago I went rhinoing and again last night. Collected about 40 parts. In that time 6 of those parts had doubles and one even had a triplicate. so roughly 25% of my rare farming that night was the same part. I ressed the first of each to see what they were and then chucked em in the clan cabs to find out that, not surprisingly, I already had several of each part. .

The problem with that is that due to the huge number of possible parts and the small numbers gathered (in your case 40) that's not necessarily surprising behaviour. Probability is a bit weird. Classic example is birthdays - if every person was randomly allocated a day of birth, you only need a group of 23 people before you expect there to be a shared birthday. As there are a similar number of non-MC5 techs, you'd have similar expectations by the time you had 20 or so, you'd expect a double - assuming tech drop genuinley was random. This carries on. If you looked at a collection of only say, 5000 techs, you'd actually expect to see some techs 20 times, and some 2. By the time you start getting many tens of thousands though, you'd be expecting something more consistent. That said, my probability isn't good enough to know what you would expect.


So yeah, stuff like this is a bitch to tell if is working right. Noone (not even a whole clan's) personal experience is worth a great deal. Again, this is (part of) why I dislike the idea of basing anything new off rare drops - so hard to tell if it's actually doing what you want.

Edit: Just to clarify - not trying to argue that the tech drop rate is random right now. Just that probability is weird, and it's very hard to prove either way. Neocron's certainly had plenty of weird loot issues over time (best one was when fixing grim perse loot broke stealth) that it's certainly quite possible for something to be slightly off with tech drops.

[Mokoi]

probability is weird

this. the part distribution for unidentified tech parts is indeed random.