I think what people are seeing is just how sucky low drop rates combined with the current raid loot policy really are. I'd love to see some probabilities for how many runs it'd take a static 12 to reasonably expect that each person in the run win a symbol and how many symbols the lucky ones of that 12 would have when the 12th person finally got their first symbol. Might even be a cool mobile application to have just to play around with it when bored in line at the DMV. Setting drop chances and then iterating thru with a rnd for each to see when everyone gets one. It would certainly level set expectations.
Same remark :
I farmed School about 20 times a day : absolutely nothing.
A friend captain does 4-5 runs : he got twice the "Breastplate of the Honourable Captain" !!!
Then I do a dozen of Webs with 2 friends, a warden and a mini : warden and myself got nothing and the mini got all the loots, remembrance crystals, SL crystals, teal items, tomes, relic removal scrolls, ... etc !!!
(and 2 crystals + 1 teal item in an only run, just incredible)
I really cannot explain this.
Last edited by Castorix; Mar 21 2013 at 11:46 AM.
For the second one it is even simpler. Say there were 6 loots, odds that they all went to the same guy are one third to the power 5 (first loot determines who will get everything, so it's not to the power 6). Odds of this happening are one in 243. That's small, small enough to be very noticeable, but not totally impossible. Now, if your friend won ten of those loots that would be one chance out of 19,683. It's getting serious, but not even nearly "lottery winner".
Another bias is that you include a list of loot which was not defined before running the instances, but rather after those instances were completed. If in your experiment someone else did win a relic removal scroll, I bet you would still have made the same claim here, as the mini's earnings would have already been quite exceptionnals.
So, is the remote looting bugged? probably since there are too many coincidences. Is your experience out of the norm? slightly. Is your experience enough to prove a bug? definitely not.
I Dunno if there is a bug or not but since everyone is saying these rare events probabilities wise:
- In a night, one champ played U10 for the first time and did 2x erebor with me, he got symbol and champ cloak, then left.
- Some of kinnies got 0 symbols before this 'hotfix' (i got 2, explained below) while others got 5 in the same amount of runs: its like you say, if the drop rate is 1%, well, its much probable to get 1-2 on everyone than all in the same players.
- For the inital 3-4 days i spent over 8 hours in BfE. 2 nights before the update, in 1 hour, i got golden piece, 2 symbols and 2 remembrances. This definitely looks like a flag system of sorts, maybe not intended but generated by the engine.
Elunwe Minstrel | Elunduil Hunter | Ellirin Champion | Micolo Burz |
Most of the time, doing a buch of Erebors in a row, you'd see roughly a symbol every 4-5 runs.
However, the other night we did 14 runs and pulled 8 symbols and 4 class items. All twelve of these went to the same four players. Any mathemagicians out there want to crank out the odds of that happening?
Ulver - 85 Runekeeper | Grevling - 85 Burglar
I'll put it in terms a SW person would get. Look at the first or second derivative of the results, or even the probablity density function. There are many standard test that this data set would fail on. Think of it this way, a PRNG is just a number generator, in Lotro it will likely be a 32-bit unsigned integer. To simplify the explination lets just say its 8-bits, so that a PRNG will generate a predictable sequence of numbers that "appear" random. If one were to sample this PRNG 256 time you would get the numbers 0-255 before the sequence repeats itself. If the PRNG does not produce all those values but repeats only a subset then it can not be classified as a maximal length sequence. Further, we are not looking at the full 8-bits when we decide a pass or fail test, lets say we want a random number for a raid base decision so we take our PRNG and divide it modulo 12 to get a result in the range of 0-11. That is how you break a PRNG via implementation, how you subsample the original truly good PRNG. It is not uncommon to just look at the LSB of a PRNG of significantly wider width, the wider the better as the sequence length doubles with each bit. PRNG are noise functions, but there are many types of noise, white, gray, pink, black etc... they each have certain characteristics that are desireable for different reasons. The point is that it doesn't matter if the actual PRNG that Lotro uses is good if how they use it breaks its randomness. Looking at the graph of results clearly shows there is an issue with it.
As a sample size gets larger simple statistics will mask everything. In our above example if you sample this PRNG 256 times and you get all the values as a natural counting sequence you will not have random data but you will have an even distribution of all the expected values. The graph from the web page has serious issues that only dissappear if and only if you have a very large sample set AND you evaluate solely on outcome and not randomness. Yes it's true that 10 samples is not enough but 100, 200, 300 are throwing up a glaring flag for the data set from that analysis.
The very first U10 raid I did, Flight to the Lonely Mountain, my LM received a Tarnished Symbol, a Wyrmfire token, and a Starlite Crystal. The next ~140 raids I went on using my LM, Guard, Hunter and Burglar resulted in not one Tarnished Symbol. My burglar did get the gold shoulders and my guard the gold shield. I don't remember how many of the wyrmfire, etc tokens I got because I didn't keep track and gave several to friends who don't raid often to upgrade rings. I also had four jewelled scroll cases drop.
I tutored statistics for three years in college. Casual observation of random number generated loot does not make sense. For instance, I failed to crit any of those four jewellry recipes. On the other hand I have critted eight of the ten Eastemnet Jewellers Journal recipes I have crafted. A casual observation would say that this is exactly backwards as I had a much higher crit chance on the single shot jewellry recipes than I did on the single shot scholar ones.
I've also played this game since 2007. In nearly six years of crafting and getting drops, I'd say that the random number generator is pretty accurate. Don't get me wrong, when ever I miss a crit, I don't like it. When I see someone get multiple drops while I am not getting any, I don't like it.
What I do like is some of the players in the game. On one of my many runs through BfE, a captain won back to back tarnished symbols. His response was, "Wow, didn't expect that. Okay, anybody that hasn't won one go ahead and roll for this one." He gave his second symbol to another person in the pug.
Random number generators can be frustrating. Random player integrity more than makes up for it.
Elendilmir - Officer of the Mithril Crowns (The Oldest Kinship in LOTRO)
"It doesn't matter how well you play, only how good you look while playing."
Regardless, like I already stated on the previous page, (especially) if the pRNG is indeed sufficiently random, all I know is that I am already thoroughly demotivated to keep running content at all except just to do some stuff socially, because I've simply given up on fighting the pRNG. I don't care any longer. Question for turbine is how many other people are starting to feel this way, and what that will do to their bottom line..
So it sounds like there could be two possible problems here, if I'm understanding the above comments correctly. Please correct me if I'm not.
One, the loot tables are associated in some way to a character identifier, in a way that the RNG's outputs are being distorted so that some people receive more favorable drops. Obviously this would be a bad thing.
Or two, the RNG's outputs tend to cluster in such a way that some people get streaks of great drops while others get streaks of nothing, even if in theory it would average out if those people ran the instance 1000 times. You could argue that this isn't problematic, but I'd argue it is. If you were just crafting food, for example, this kind of clustering is OK because it's relatively easy to accumulate enough outcomes to balance the successes with the failures. But when you're dealing with limited inputs (whether a rare jewelry recipe or the time it takes to complete one instance), this kind of clustering is a guaranteed bet for player frustration. The last thing Turbine should want is players ragelogging because they've run Ost Elendil thirty times with no success and then see another player get their gold class item twice in a row. And it obviously doesn't make for a productive environment on the forums either.
On a side note, I'm wondering whether using crafting as a reliable indicator is necessarily valid. It may test the underlying RNG but it doesn't test the implementation in terms of loot. I'm also not sure it's reasonable compare current loot behavior with past raid loot history either. Previously raid drops weren't done for each character individually, as is now the case for remote looting.
As to the nature of the game these days, I actually agree with you. I have a personal preference towards working towards things (via currencies) and having known grinds rather than unknown ones. That is regardless of the quality of the RNG -- a perfect one would still leave you (and me) unsatisfied once when we've ground 2 or 3 times the expected number of instances without getting a result, which is a pretty common outcome...
Except, in the cases that I've seen, there aren't really streaks. It's consistent over a number of days.Or two, the RNG's outputs tend to cluster in such a way that some people get streaks of great drops while others get streaks of nothing, even if in theory it would average out if those people ran the instance 1000 times.
Chromite (Hunter) - Grumbletocks (Guardian) on Landroval, Appendage (Hunter) on Brandywine
As people hinted at above, regardless of whether or not the RNG is being distorted by some silly factor like character ID number or something, the fact that so many people have become suspicious of it is clearly an indication that things aren't working ideally. Whether that means that the RNG itself needs investigation or whether it means that using a simple RNG at all for this kind of loot acquistion is not a very good idea is another concern.
Ulver - 85 Runekeeper | Grevling - 85 Burglar
This form's session has expired. You need to reload the page.Reload