One hundred trials later...
By
Adeon Hawkwood,
Posted on: 2011-07-13 14:17 in CoH and CoV General Discussions
Quote:
The problem with this is, when the praetorian zone events came out I worked my tuchas off with my MM; controlling pets perfectly, managing their targets, and basically being a beacon of destruction just to find out that didn't count as participating. I then went through with no pets tossing my heal and ST debuffs every time they were up, even if it would be stupid to do so, basically backpacking and viola.|
Originally Posted by Leandro
The thing is, I've been in leagues where I was on autopilot and barely doing anything, and I got a Very Rare out of it; while in other leagues I was going crazy trying to do thirty things at once, and I got commons. Whatever way participation is being weighted, it doesn't seem to do all that much.
|
So, it really depends. If there's participation scoring, you might not notice it depending on what the participation requirements are.
Murphys Military Law
#23. Teamwork is essential; it gives the enemy other people to shoot at.
#46. If you can't remember, the Claymore is pointed towards you.
#54. Killing for peace is like screwing for virginity.
Quote:
If I understand Arcanaville (and what I recall from the devs) properly, it's not your participation score, it's the league's "participation score".|
Originally Posted by Leandro
The thing is, I've been in leagues where I was on autopilot and barely doing anything, and I got a Very Rare out of it; while in other leagues I was going crazy trying to do thirty things at once, and I got commons. Whatever way participation is being weighted, it doesn't seem to do all that much.
|
Your participation score gets you out of the "10 threads" hole, but does nothing further.
The league as a whole earns a score, with bonus points for various things, that is used to weight everyone's roll for a drop table, but doesn't do anything to get you out of the "10 threads" hole. Thus, once you've done the minimum to get out of the hole, your motivation should be to make the whole league as successful as possible.
Say you're on a top-notch league, one that gets every merit (or whatever they base the league bonus on) but you're having connection problems, and spend a lot of time mapserved or even disconnected. Your leaguemates will probably get a slightly "richer" set of drop tables than a flat distribution would give...but you might well get only 10 threads. If you manage to get out of the hole, though, you get the same chances at the drop tables as everyone else on the league.
Conversely, if you bust your hump, but your league is borderline, you should be safe from the 10 threads table, but you're not going to have any better than the baseline chance at the higher drop tables.
Is that about the size of it, Arcana?
The Way of the Corruptor (Arc ID 49834): Hey villains! Do something for yourself for a change--like twisting the elements to your will. All that's standing in your way are a few secret societies...and Champions of the four elements.
Quote:
I'm not sure I understood your whole post, but am I correct in concluding that Thursdays are the best day to run trials then?
|
Originally Posted by Arcanaville
At this point I've gotten six very rare drops total. Four have been on Thursdays.
|
Very nice, Leandro!
Thank you for sharing all the stats as well as the graphs. Getting to see the information in both forms really helps with the comparison data.
Also, it helps me realize that my incarnate trail experiences are about par for length and rewards, not sub par as I had begun to suspect. What a relief!
~Yydr
Quote:
O maybe that just what they want us to think!
|
Originally Posted by Arcanaville
Of course, there's no possible way the participation system could be weighted for Thursdays.
|
Captain Den'Rath 53* Merk/Traps MM, Rivona 50Energy Blast/Time Cor,Victoria Von Heilwig 53* Dual Pistols/Traps Cor, Crab Spider Webguard 53* SOA, Accela 53* Bot/FF MM,Valkyrie's Executor 53* Broadsword/Shield Def Scrap. On FREEDOM! @Knight Of Bronze
"Hypocrisy, the human inherent." "Let not this work be wasted, apply yourself always."
Quote:
19 seconds ... haxx|
Originally Posted by Leandro
Lambda: The fastest run took 19:33 seconds; the slowest, 34:06 seconds. The average completion time was 25:08, and the median, 24:26. While it is longer than BAF, it's not that much longer; and usually, at least on Freedom, people run both back to back.
|
but seriously nice work.
Do you have any feel as to whether the faster ones gave better drops or the slower ones ? (and did you kill more in the slower ones ?) I wonder if how much you kill as a team has any effect.
It's true. This game is NOT rocket surgery. - BillZBubba
Quote:
I always figuredthat was deliberate, the emps are there for people who are screwed by the RNG, but most people will get their VR's before they need to buy them with emps.
|
Originally Posted by Oedipus_Tex
Nice work.
Math work follows that probably has some errors: if the numbers are really 10% for a Very Rare, it means that in 10 runs you would only have around a (9.^10) 35% chance to not get a V Rare. If those odds are real, then in terms of a single Very Rare, it is actually better strategy to run the BAF over and over than chase Empyreans. If you took an average amount of time to run each of the 3 trials every day based on these times, a Very Rare in Empyraen would be worth about 84 minutes of mission time across 10 days, or the equivalent of 14 hours of playing trials (not including set up time). If you ran the BAF 40 times you'd have a less than 1% chance of not having your V Rare within 11 hours, and some chance of getting more than one. Of course, you can do both. However, this also means that after 40 runs around 1% of players will not have found a V Rare. Multiply that by the number of people playing and you've got your conspiracy theories right there, and some (IMO somewhat justifiably) annoyed players. I'm still curious whether team performance changes the odds and in what way. |
"Men strunt �r strunt och snus �r snus
om ock i gyllne dosor.
Och rosor i ett sprucket krus
�r st�ndigt alltid rosor."
nice work Leandro.
one thing though.
in your uncommon salvage drop counts, did you take into account the drops from trial objectives and/or Master of objectives being completed?
note: i could be way off on that one but i've noticed i can get a LOT of uncommon salvage from those.
I never thought of timing my trial runs - that would have been an interesting bit of data.
I have drop data from 208 trials logged...I've been taking a break for the last month or so as well or it would probably be 300 by now.
Guides: Dark Armor and IOs | SS/DA | Crabbing | Fortunata
Quote:
I only wrote down the reward table drops. One trial = one drop.
|
Originally Posted by MrYukon
one thing though.
in your uncommon salvage drop counts, did you take into account the drops from trial objectives and/or Master of objectives being completed? |
www.SaveCOH.com: Calls to Action and Events Calendar
This is what 3700 heroes in a single zone looks like.
Thanks to @EnsonsDeath for the GVE code that made me VIP again!
Thanks for assembling these data, Leandro.
I find it challenging to overcome my cognitive biases regarding the frequency and distribution of drops in trials. I know, with my brain, that random is random, and that streaks and the like are to be expected. That doesn't stop me from being almost convinced that there is something else going on besides random selection of rewards.
I have a mastermind that ran 62 trials before he got a rare (as a drop; he got rares for the master completions), and never saw a very rare. I've also got a cold ruptor who has run seventeen trials, received eight rares and one very rare. My scrapper has run over a hundred and has gotten pretty close to what Leandro reports (30/40/20/10), as has one of my brutes. The other brute gets a lot more rares (over 30%) and a lot less commons. I've not kept stats on my other characters, but they don't stick out in memory as being disproportional.
When I stop looking at the numbers per-character and start looking at the total runs and rewards across all characters, the disparities basically disappear. It's still tiresome to have a favourite character (my mastermind) struggle while a side project (my cold ruptor) seems to coast through. Thank god I enjoy running these trials or I would lose my mind. Maybe I will break out my mastermind tonight and try my luck.
Quote:
Keep the ones you need for Tier 2's, downgrade the rest for Commons you also need.
|
Originally Posted by Redlynne
This would explain why every character I've run trials with always winds up FLOODED with piles and piles (and piles!) of Uncommons that I'm never going to need or use ... while I must go begging for Commons to finish out my Tier 1/2/3 slots with (and am pretty much ALWAYS forced into converting Threads for).
|
Quote:
That's the implication of what the devs have said. I cannot prove it with the data I have, but I can say if individual participation score affects your drops, it doesn't do so by enough to be noticable. The best possible bonus achievable by any means to improve your drop rates seems to be modest, and decent leagues seem to reach that limit relatively easy, and nothing an individual does can exceed that.|
Originally Posted by Balanced
If I understand Arcanaville (and what I recall from the devs) properly, it's not your participation score, it's the league's "participation score".
Your participation score gets you out of the "10 threads" hole, but does nothing further. The league as a whole earns a score, with bonus points for various things, that is used to weight everyone's roll for a drop table, but doesn't do anything to get you out of the "10 threads" hole. Thus, once you've done the minimum to get out of the hole, your motivation should be to make the whole league as successful as possible. Say you're on a top-notch league, one that gets every merit (or whatever they base the league bonus on) but you're having connection problems, and spend a lot of time mapserved or even disconnected. Your leaguemates will probably get a slightly "richer" set of drop tables than a flat distribution would give...but you might well get only 10 threads. If you manage to get out of the hole, though, you get the same chances at the drop tables as everyone else on the league. Conversely, if you bust your hump, but your league is borderline, you should be safe from the 10 threads table, but you're not going to have any better than the baseline chance at the higher drop tables. Is that about the size of it, Arcana? |
Its *possible* that a *fantastic* player on a *horrible* league that somehow manages to succeed in spite of themselves anyway might be getting a small bonus to their random drop rates, contrary to what the devs said about the system. But I don't think that possibility is strong enough to affect trial participants in a meaningful way.
Its hard to prove a negative in this case. I don't see a strong individual effect, and the devs have said there is none. Either that statement is true and there is none, or that statement is false but its too small to measure.
I can also say that almost anything you can do which helps yourself at the expense of your league mates is probably counterproductive. Doing things that help you and help your league are better than things that hurt you and help your league, certainly. But doing things that help you and hurt the league can hurt you even if you don't cause the trial to fail. Certain experiments I can't describe directly imply that helping your league mates can improve *their* score, which improves the overall average, which indirectly improves your rolls. So the counteraction: hurting league mates, and and probably does boomerang back to you. You're doing something that probably won't help you directly, but can hurt you indirectly. That's not a good trade. Because the numbers are small either way, people who do this might not even notice: they could do it and get a good drop, and assume it works. It doesn't.
Although the devs won't reveal the details of the system, even to me, one thing I can do (because everyone can) is report potential exploits. So I did make a list of all possible ways I thought I could, in general, game the system and reported that as potential exploits. Only one of them turned out to be even possible in the system, and its a weird one no player would ever attempt to do unless they actually knew it would work, because it doesn't look like it would do anything and the benefit would not be obvious without doing it repeatedly. I can't say the system is not exploitable, but I can say the devs have looked at every obvious possibility that has been discussed on the forums.
[Guide to Defense] [Scrapper Secondaries Comparison] [Archetype Popularity Analysis]
In one little corner of the universe, there's nothing more irritating than a misfile...
(Please support the best webcomic about a cosmic universal realignment by impaired angelic interference resulting in identity crisis angst. Or I release the pigmy water thieves.)
Quote:
Depending on how the "random" number is generated, it may not be a coincidence at all that you got more rares on Thursday. If the dev who wrote the random number generator took the system clock value and did some "random" arithmetic on it, it could very well generate a value that depended on the day of the week.|
Originally Posted by Arcanaville
Of course, there's no possible way the participation system could be weighted for Thursdays, so that's just what happens when you stare at raw data too long: random chance says you will find patterns if you look long enough.
|
That's the problem with generating pseudo-random numbers. You have to be extremely careful to avoid building some kind of bias into the generator. Using the system clock to seed a random number generator might be okay, but actually using the clock to generate "random" values would be a colossal error.
Add to that the fact that bugs can happen (like the uninitialized variable from years back that caused the purple drop rate to underperform), it's very difficult for us to tell from an end-user perspective whether the system is behaving as it should. We just don't have access to the basic probabilities, the inner workings of the algorithm or enough data to adequately judge the performance.
It's ironic, but the only thing that would convince the majority of players that they were getting "random" rewards would be if they got exactly the expected number of each item. That is, if they always got 30 commons, 40 uncommons, 20 rares and 10 very rares in 100 trials, they would be convinced the "random" reward system was working perfectly. That, of course, would be the furthest thing from random.
Which just goes to show that humans have absolutely no intuitive understanding of probability and statistics. We are purpose-built to detect patterns, and we find them wherever we look whether they're there or not.
Quote:
It's been said multiple times in the posts following this, but I want to go on record and say that the likelihood of qualifying for specific Component rarities depends more heavily on League participation than individual participation. I'm not sure how narcissistic someone would have to be to think that they could single-handedly award themselves a Very Rare in a League full of other people.|
Originally Posted by Leandro
The thing is, I've been in leagues where I was on autopilot and barely doing anything, and I got a Very Rare out of it; while in other leagues I was going crazy trying to do thirty things at once, and I got commons. Whatever way participation is being weighted, it doesn't seem to do all that much.
|
Having said that, individual participation is a factor, and if you just squat by the door you'll only have yourself 10 shiny new Incarnate Threads to play with afterward.
Quote:
|
Originally Posted by Leandro
This could be just random being random, but... yeah. Keyes dropped a Very Rare on me 1/3 of the time. It really bugs me.
|
So your data... well... If it's a lump of all drops for all trials, I'm gonna have to say it's an incorrect conclusion.
Quote:
The devs use rand(). And even clock hashing wouldn't amplify Thursdays: that would be difficult to do *deliberately* much less accidentally.
|
Originally Posted by Rodion
Depending on how the "random" number is generated, it may not be a coincidence at all that you got more rares on Thursday. If the dev who wrote the random number generator took the system clock value and did some "random" arithmetic on it, it could very well generate a value that depended on the day of the week.
That's the problem with generating pseudo-random numbers. You have to be extremely careful to avoid building some kind of bias into the generator. Using the system clock to seed a random number generator might be okay, but actually using the clock to generate "random" values would be a colossal error. |
[Guide to Defense] [Scrapper Secondaries Comparison] [Archetype Popularity Analysis]
In one little corner of the universe, there's nothing more irritating than a misfile...
(Please support the best webcomic about a cosmic universal realignment by impaired angelic interference resulting in identity crisis angst. Or I release the pigmy water thieves.)
Quote:
I don't think C++ has any particular algorithm specified for rand(), which would make it compiler-dependent. However, the most common algorithm behind compiler-based pseudo-random generators would be the Linear Congruential Generatior. That article lists the parameters used in various compilers.
|
Originally Posted by Arcanaville
The devs use rand().
|
Quote:
I have to wonder what would happen to Leandro's numbers if he were doing these trials with groups with less than average ability (but still capable of finishing the trials).
|
Originally Posted by Arcanaville
I will also say something else I really cannot elaborate upon, but those numbers are consistent with what I believe are the drop percentages that good leagues will tend to get assuming very good (but not necessarily difficult to get) random roll bonuses.
|
Triumph: White Succubus: 50 Ill/Emp/PF Snow Globe: 50 Ice/FF/Ice Strobe: 50 PB Shi Otomi: 50 Ninja/Ninjistu/GW Stalker My other characters
Quote:
They would go down.|
Originally Posted by Snow Globe
I have to wonder what would happen to Leo's numbers if he were doing these trials with groups with less than average ability (but still capable of finishing the trials).
|
My guess is that the floor is somewhere around 5/10/30/55, plus or minus a few percentage points here and there.
[Guide to Defense] [Scrapper Secondaries Comparison] [Archetype Popularity Analysis]
In one little corner of the universe, there's nothing more irritating than a misfile...
(Please support the best webcomic about a cosmic universal realignment by impaired angelic interference resulting in identity crisis angst. Or I release the pigmy water thieves.)
Quote:
I meant, the devs don't write their own (not with approval anyway: I think every instance where a pseudorand was written into the code it was taken out and replaced with rand() long ago). All modern C compilers have decent rand() functions. The problems with them would not be visible to MMO players and difficult to even detect with random analysis. Some of the tests that break those require more random rolls than I can reasonably collect in a lifetime, which is hundreds of billions of rolls. One of the problems with testing the rand has been that the really good tests require far more data than I can collect, so I've been forced to use lesser tests. If I can't find a problem in the rand() in a hundred million rolls, no player is going to find one in one hundred drops.
|
Originally Posted by GuyPerfect
I don't think C++ has any particular algorithm specified for rand(), which would make it compiler-dependent. However, the most common algorithm behind compiler-based pseudo-random generators would be the Linear Congruential Generatior. That article lists the parameters used in various compilers.
|
[Guide to Defense] [Scrapper Secondaries Comparison] [Archetype Popularity Analysis]
In one little corner of the universe, there's nothing more irritating than a misfile...
(Please support the best webcomic about a cosmic universal realignment by impaired angelic interference resulting in identity crisis angst. Or I release the pigmy water thieves.)
I took some time to write a random number generator in Flash that uses the estimated values from Leandro's first post. I used the values 30/40/20/10 as the estimated chance of each piece dropping. I then ran a league of 24 through a bunch of iterations to see what sort of breakdown might occur. Here is one example run (ignore the white text on the right for the most part, as I didn't finish coding it and only the first 2 reported items work):
In this situation, Players 6 and 19 have 12 Very Rares before player 9 has 1. In fact the entire league has gotten their 4 Very Rares before Player 9 got one Very Rare. These results aren't typical but they are definitely possible. I got these results by accident while testing the reporting function to see if it would correctly report the last player to get his or her V Rare, and ended up having to click 67 times. Other runs it would come up at 20 or 30 or 40 or 50; really all over the place.
The strategy? Hope you are not Player 9.
Quote:
I really don't find that acceptable.|
Originally Posted by Arcanaville
They would go down.
My guess is that the floor is somewhere around 5/10/30/55, plus or minus a few percentage points here and there. |
And I know of one player that seems to have greater odds for the higher rarities than Leandro, as well as several players with lower than you are stating here (5/10/30/55).
Triumph: White Succubus: 50 Ill/Emp/PF Snow Globe: 50 Ice/FF/Ice Strobe: 50 PB Shi Otomi: 50 Ninja/Ninjistu/GW Stalker My other characters
Quote:
Well, I'm not sure there's really anywhere to go there. That seems to be mostly a matter of preference than anything else at this point.|
Originally Posted by Snow Globe
I really don't find that acceptable.
|
Quote:
| And I know of one player that seems to have greater odds for the higher rarities than Leandro, as well as several players with lower than you are stating here (5/10/30/55). |
[Guide to Defense] [Scrapper Secondaries Comparison] [Archetype Popularity Analysis]
In one little corner of the universe, there's nothing more irritating than a misfile...
(Please support the best webcomic about a cosmic universal realignment by impaired angelic interference resulting in identity crisis angst. Or I release the pigmy water thieves.)
Thanks for the data Leandro. 
It certainly confirms the "path of least resistance" for getting incarnate rewards. Other than challenge and Emp merits, BAF will still be the trial of choice.
Since its all about doing as many trials as fast as possible due to random drops. I don't really want to suggest "trial merits" since the last thing we need is another currency. 
H: Blaster 50, Defender 50, Tank 50, Scrapper 50, Controller 50, PB 50, WS 50
V: Brute 50, Corruptor 50, MM 50, Dominator 50, Stalker 50, AW 50, AS 50
Top 4: Controller, Brute, Scrapper, Corruptor
Bottom 4: (Peacebringer) way below everything else, Mastermind, Dominator, Blaster
CoH in WQHD
I don't buy it. Participation plays a factor in what Components are rolled.
The effect was most pronounced on uncommon drops because the way random roll bonuses work they impact the more common drops more than the less common drops. Otherwise the system would be able to make a 5% chance for very rare drop into a 55% chance with a high enough score.
If *individual* participation score affects what components are rolled, I can say its effect is constrained by the data to be relatively small.
[Guide to Defense] [Scrapper Secondaries Comparison] [Archetype Popularity Analysis]
In one little corner of the universe, there's nothing more irritating than a misfile...
(Please support the best webcomic about a cosmic universal realignment by impaired angelic interference resulting in identity crisis angst. Or I release the pigmy water thieves.)