Please upgrade your browser for the best possible experience.

Chrome Firefox Internet Explorer
×

Hardmode Flashpoints; No loot variance for premade groups

STAR WARS: The Old Republic > English > Flashpoints, Operations, and Heroic Missions
Hardmode Flashpoints; No loot variance for premade groups

Novastinger's Avatar


Novastinger
01.23.2012 , 04:03 PM | #21
Yeah sorry I meant to quote Mavery above, I agree with your conclusion.

DaveMcG's Avatar


DaveMcG
01.23.2012 , 04:04 PM | #22
Quote: Originally Posted by Novastinger View Post
Yeah sorry I meant to quote Mavery above, I agree with your conclusion.
I figured as much.

Hyzerbomb's Avatar


Hyzerbomb
01.23.2012 , 04:44 PM | #23
My group of 4 experiences much the same results in HM FP's. 2 sorcs, 1 assassin and a jugg. Always the same group, and the vast majority of drops are BH/trooper loot. Even the random chests are giving aim gear. I don't expect to see suitable drops constantly, but it does seem like we are predisposed to getting aim gear.

lotj's Avatar


lotj
01.23.2012 , 06:00 PM | #24
Dave keeps giving me more work.

Quote: Originally Posted by Mavery View Post
I'll give you that 11 out of 13 drops being one token isn't probable but it also isn't unlikely in the context of a random generator. You only have a case where you can prove it is mathematically impossible to yield 11 out of 13 being one drop.
Let's say I gave you a function. I claim it randomly returns the values {1, 2, 3, 4} each with equal probabilities. I implement this function to always return 2, but you don't know that.

How many draws do you need to suspect this function isn't behaving as advertised? How many do you need to prove it is mathematically impossible for the function to behave as advertised?

If you're going for a 95% confidence, you only requires 5 draws. A 99% confidence only requires 7. To prove it's mathematically impossible you need infinitely many -- something that's not physically possible.

Are you going to conclude it must be implemented correctly, because a never ending sequence of 2's is mathematically possible based on how I described a function? No -- that's silly, but it's exactly what you're doing here.

Here's a thread where I posted our experiences and associated probabilities from a few days ago :
http://www.swtor.com/community/showt...46#post2001546

Since then we re-ran False Emp and got another smuggler jacket (which is what prompted Dave's post), and Battle for Ilum (with a slightly different group) and got the Sage off-hand.

Claiming this is "not probable" undermines just how unlikely this is, given actual randomness. Seeing this much consistency in drops across multiple flashpoints is less than a 1 in 2.5 million shot. Another way of putting it is, if every single person who purchased the game ran exactly this sequence of hardmode flashpoints, the expected number of times you'd see a sequence this probable (or worse) is less than 0.02.

In other words, you have about a 1 in 50 shot of seeing one individual encounter a sequence like this if all 2 million people who purchased SWTOR ran the same number of hm flashpoints as we have. It's also worth noting this estimate is going to be fairly high, since the game hasn't seen 2 million sales yet, and every single person most certainly isn't running hm fp's as frequently as we are.

The purpose of this thread is just raising the question -- are other people seeing the same behavior we are? Seems like some are, and some aren't. The fact that others seem to be is an indicator that something might be off with the loot system's function, given how unlikely it is to even see it once.

As was mentioned earlier in this thread -- outside of one specific device, there's no true "random" on computers. Pseudo-random number generators are simply recurrence relations that share a set of specific properties with truly random streams. This makes them incredibly difficult to test, and very error prone implementations if you're not extremely careful.

Something as simple as seeding the rng with a value that's static relative to the group would give it good behavior relative to the most commonly investigated tests (specifically, repeatedly querying a loot table to verify drops and looking at drop rates across the entire population), but cause exactly this sort of non-random behavior when faced with a static group who runs a few hm's at the same time every night. Using a rng that has poor uniformity at the bit-level can also cause significant swings in the probability of even & odd numbers occurring when used with standard binary encoding that could cause problems depending on if a (0,1) normalization or a modulus is used to bound the results.

What I'm getting at is, just because you're told it's random doesn't mean you should blindly believe it just because a sequence is possible. If it's claimed to be random and then behaves significantly differently from the supposed distributions & probabilities, then something else might be up.

Now, just to be clear -- we've tried the basic things. We've varied the time we enter the instance (but not much the time we assemble the group), who assembles the group, who loots the container, and who is the "owner"/"leader" of the story-mode. The token drops didn't seem to change. Also, while we haven't sufficiently documented the purple drops leading up to the token, we do notice they tend to be the same 1-3 drops from every boss, and (I believe) we do see completely new stuff when running with very different groups.

FtenEQ's Avatar


FtenEQ
01.23.2012 , 06:07 PM | #25
So basically, you guys won the loot lottery!

Just not in the way you would want.

Roughnesssssssss. However I would say that possibly, they could have a system attempting to tailor the gear and it is just misbehaving(misinterpreting the group makeup).

I mean HM Ops are dropping no gear at all so really it's not out of the realm of possibility. Certainly they have some bugs to clean up in their loot system and it's behavior.
Jedi Knight---Guardian Tank Extraordinaire---PVP bad-mother
Follow me on Twitter!http://www.twitter.com/fteneq
My Guardian Blog!TheORJ.blogspot.com

DaveMcG's Avatar


DaveMcG
01.23.2012 , 07:26 PM | #26
Quote: Originally Posted by lotj View Post

<SNIP A TON OF MATH AND SMART STUFF>
What he said.

Evilertoaster's Avatar


Evilertoaster
01.23.2012 , 10:18 PM | #27
anecdotal but:
Juggernaut Tank
Sniper DPS
Sorc DPS
Sorc Heal

We've had about 50 hard mode drops last few weeks and across different flash-points. About 85% of them were Aim gear... pretty annoying... Maybe they could at least make BoP gear not drop if no classes that use it are present in the group....bah

Darth_Victus's Avatar


Darth_Victus
01.23.2012 , 10:41 PM | #28
I am getting a similiar feel with my Juggernaut. Have been doing hard modes for the past 2 weeks and have only seen bounty hunter or inquisitor drops and only 1 Juggernaut drop (Columni token).

For 2 weeks, that is not a good sign for some classes. I understand the whole random loot is random, but this is just getting ridiculous.

Imapwnu's Avatar


Imapwnu
01.24.2012 , 12:58 AM | #29
OP is definitely correct. I've seen plenty of such topics, and I can vouch for it from my own experience. It DEFINITELY has to do with group composition, in a way the game basically chooses what type of gear is least useful for the group, and drops that 85% of the time. Very broken.

My HM group consisted of 2x Guardian, 1x Commando, 1x Sage. Did around 20 HM together. 75% of all drops were Smuggler, with some Trooper and the odd Sage. Just 1 JK drop from all those, with 2 guardians in group. My Doc is full PvE epic geared, just from winning Greed rolls, while my Guardian has 1 Tionese piece, bought from COmmendations. No wonder I rerolled.
Beware the beast, but enjoy the feast it offers.

DaveMcG's Avatar


DaveMcG
01.24.2012 , 01:15 AM | #30
Thanks for the updates guys.

Are there any premade groups out there that are NOT seeing this trend?