Is the mission failure chance applied separately on each follower/ship or only once?

[Deleted]

Is there any strong threorycrafting nowadays (after months of collection) to answer that, or have they made an announcement on it? It seems a lot of people are noticing abnormally high chances of failure above 80% while it might just be that each different follower or ship gets the failure chance separately so it's higher.

If I recall my 'Probability & Statistics' well (which I probably don't) it's probably the chance of "A union B" which means "a or b failing" in this case which would be..

..but wait!

In this case it might even be "A and B" which would mean "a AND b failing" which would make "RPG"-sense since you may kill the boss with one last man standing.

That would make chanceA x chanceB so say 20% chance of failure, it would be 0.2^2 for two followers = 4% chance of failure but that seems too low according to most people's intuition on the current game.

OK, I'm confused and don't have time. Does anyone have info?

[Makarena]

Each follower rolls and has the % to win, the % is the one u send the follower team to mission with. Say u got 90% success mission and 3 followers, chance of success is when ALL OF THEM hit the mark(respecitvely chance to fail is when AT LEAST ONE misses), which is 0,9^3 = 72,9%.
Have u ever wondered why your 83% missions (the 10-15k exp ones where u miss 1 counter coz ur harrison jones just doesn't have the right trait and ur newly recruited follower has wrong abilities) fail more often than they are supposed to or why u don't fail shitty 90+ solo gold or resourse mission but ur raid mission (for 3 followers) of course, fails? Coz I did, with 4 100 chars and full-fledged garrisons with treasure hunters and stuff. Well, thats because 83% in reality is 57,1% which is almost a coinflip.

[Deleted]

Each follower rolls and has the % to win, the % is the one u send the follower team to mission with. Say u got 90% success mission and 3 followers, chance of success is when ALL OF THEM hit the mark(respecitvely chance to fail is when AT LEAST ONE misses), which is 0,9^3 = 72,9%.

This is such an absurd statement that - if true - would be super trivial to prove without use of personal anecdotes.

[Deleted]

This is such an absurd statement that - if true - would be super trivial to prove without use of personal anecdotes.

What do you mean? What's your proof other than testing it? The only way to make sure is to have a large sample of test cases (at least 40 or 80 missions on a given success rate).

[Deleted]

If it were true 50% 3-follower missions would fail 87.5% of the time which would be very noticeable.

[Zelendria]

What do you mean? What's your proof other than testing it? The only way to make sure is to have a large sample of test cases (at least 40 or 80 missions on a given success rate).

Someone already did a large scale test on followers % success chance (Somewhere in the garrison section) and in almost all cases was very close to the tooltips % claim.
There's no way a 83% win chance on shipyard is 57% no way.

[Deleted]

What do you mean? What's your proof other than testing it? The only way to make sure is to have a large sample of test cases (at least 40 or 80 missions on a given success rate).

The difference between X and X^3 is super substantial when dealing with odds. Differentiating on a hunch between 96% and (96%)^3=88.5% is almost impossible. Differentiating between 90% and (90%)^3=73% will be met with confirmation bias and simply only remembering the failures. Unless ofcourse you keep track which I doubt the poster above me did. Differentiating between 50% and 12.5% should be really trivial. Claiming you need 40-80 of those missions to test is a lot better than most people would suggest (I assume you know some statistics), but probably still way higher than needed. According to R, if you do 16 missions that show 50% but would actually be 12.5%, then there's a 0.03% chance of there being 7 or more succesful. 4/10 wins (or more) would already be 0.44% chance if actual odds were 12.5%.

[Deleted]

Yeah, you need at least ~40 samples to be confident a distribution of results is more or less representative. It's standard practice in statistics. Well, you could go as low as 20 but it becomes very sketchy, at 80, 100 or above it becomes very reliable.


There's some info on "σ" (standard deviation) depended on sample size on normal distributions, but it's years now I've done that module..

[ThePants999]

The sample size you need is not fixed. If you hypothesize that all missions have a 0% success chance, I can disprove it with a single result. If you hypothesize that missions have a 1% success chance, then getting 3 successes out of 3 would be pretty close to conclusive disproof. 40 is a rule of thumb that can be useful, but like Derpette says, if you've got something that's actually a 50% chance and you're trying to test a hypothesis that it's a 12.5% chance, you won't need anything like 40 samples to get a high confidence that the hypothesis is false.

And like Zelendria said, Makarena's "hypothesis" has been thoroughly tested and conclusively disproven - the sooner people stop spreading that huge misinformation, the better. If the UI says you have an 80% chance, you have an 80% chance. The only reason lots of people are reporting their 80% chances failing is because that feels significant to them, whereas four times as many people are having their missions succeed and aren't saying a thing about it.

[Deleted]

The sample size you need is not fixed. If you hypothesize that all missions have a 0% success chance, I can disprove it with a single result. If you hypothesize that missions have a 1% success chance, then getting 3 successes out of 3 would be pretty close to conclusive disproof. 40 is a rule of thumb that can be useful, but like Derpette says, if you've got something that's actually a 50% chance and you're trying to test a hypothesis that it's a 12.5% chance, you won't need anything like 40 samples to get a high confidence that the hypothesis is false.

And like Zelendria said, Makarena's "hypothesis" has been thoroughly tested and conclusively disproven - the sooner people stop spreading that huge misinformation, the better. If the UI says you have an 80% chance, you have an 80% chance. The only reason lots of people are reporting their 80% chances failing is because that feels significant to them, whereas four times as many people are having their missions succeed and aren't saying a thing about it.

To be honest, I don't think it's that simple. Blizzard has made us used to weird "black box" "random number generators" that are really not simple random number generators, they are fixed to some gameplay guidelines that have nothing to do with basic mathematics:

For example, they have revealed in an official Interview that around WotLK they introduced a "random" chance on drops of quests that is actually increasing in chance to get a drop if you get unlucky the first few kills. Which is something people may have noticed.

I think the system may let you fail very easily if you mainly play for 100% success missions(not absolutely sure but it's possible). The reason for this is that plenty of us rarely go to <100% missions, but very rarely when we do, 95%+ easily fails.


PS. I believe the latest personal loot drops, even the guild drops on a master loot setup are not really random. It's often to notice that if e.g. you get a ton of hunters and rogues, agility trinkets drop like crazy. Stuff like that can be coded in.

[chazus]

Someone did some thorough testing, as well as some other in depth work, and debunked all of this at the beginning of WoD, and shipyards as well.

80% is 80%. There are no multiple rerolls. I had a huge discussion about it, thinking there were, and was eventually proven wrong.

[Zelendria]

Found the bloody post. By all means disprove this theory with factual figures as people always remember that shitty loss, especially if it was your cache or sea chart.

I was curious about the same thing - so I started keeping track. Stopped a while ago but recorded 2,055 mission results along with the in game success %. Tried to get as large a variety of different % missions and have at least 1 mission for each success % above 50% with the exceptions of 52%, 57% and 61%. In general they do seem to match up well with the predicted success chance, but haven't done any standard deviation calculations. I did not record how many followers were used for each mission so can't show if that had an effect such as was suggested by other posters.

To keep this list smaller will only include ones that had at least 50 missions.

% chance in game	failed	succeeded	% success observed
51%	35	39	52.70%
53%	45	45	50.00%
59%	34	63	64.95%
62%	24	41	63.08%
70%	20	39	66.10%
73%	19	68	78.16%
75%	49	156	76.10%
81%	10	52	83.87%
83%	15	130	89.66%
87%	24	123	83.67%
91%	4	93	95.88%
93%	6	69	92.00%
95%	6	143	95.97%
96%	2	56	96.55%

Highest recorded fail was 97% (1 failed, 27 success for 96.43%).
If anyone wanted more of what I recorded will keep an eye on this. Bolded the 70% questioned in the thread which turns out was a lower % observed then predicted in game, but not by much really.

A higher number of results would no doubt have brought the tooltip figures to their actual figures even closer.

[fiestatastic]

Each follower rolls and has the % to win, the % is the one u send the follower team to mission with. Say u got 90% success mission and 3 followers, chance of success is when ALL OF THEM hit the mark(respecitvely chance to fail is when AT LEAST ONE misses), which is 0,9^3 = 72,9%.
Have u ever wondered why your 83% missions (the 10-15k exp ones where u miss 1 counter coz ur harrison jones just doesn't have the right trait and ur newly recruited follower has wrong abilities) fail more often than they are supposed to or why u don't fail shitty 90+ solo gold or resourse mission but ur raid mission (for 3 followers) of course, fails? Coz I did, with 4 100 chars and full-fledged garrisons with treasure hunters and stuff. Well, thats because 83% in reality is 57,1% which is almost a coinflip.

no idea why anyone would argue against this ... if you don't have several chars with maxed followers I'd suppose your sample size and luck might be good enough for you to not notice this, but this post by Makarena seems right on. I kept track through thousands of missions and got results very similar to what makarena suggests. Another guildie who also has a lot of characters with 25 675 followers noticed the same. Our results were actually a fair amount lower for 3 follower missions in the 90%'s, but that can be chalked up to bad rng probably. We'd basically just thought that if a 3 follower mission wasn't 100% then it was 60% regardless of the %, but what Makarena suggests actually makes sense. Each follower has that % to fail, so you have 3 chances to fail a 3 follower mission.

No idea for boats since I've barely done those.

- - - Updated - - -

Found the bloody post. By all means disprove this theory with factual figures as people always remember that shitty loss, especially if it was your cache or sea chart.



A higher number of results would no doubt have brought the tooltip figures to their actual figures even closer.

I kept track of 611 97% mission with 3 followers. I failed 263 of them. Over the same course of time I had no 97% one follower missions, but out of every single follower mission in the 90%'s at all I failed one. My sample size is MUCH larger than the posted sample size, or any speculation people are having.

- - - Updated - - -

To be honest, I don't think it's that simple. Blizzard has made us used to weird "black box" "random number generators" that are really not simple random number generators, they are fixed to some gameplay guidelines that have nothing to do with basic mathematics:

For example, they have revealed in an official Interview that around WotLK they introduced a "random" chance on drops of quests that is actually increasing in chance to get a drop if you get unlucky the first few kills. Which is something people may have noticed.

I think the system may let you fail very easily if you mainly play for 100% success missions(not absolutely sure but it's possible). The reason for this is that plenty of us rarely go to <100% missions, but very rarely when we do, 95%+ easily fails.


PS. I believe the latest personal loot drops, even the guild drops on a master loot setup are not really random. It's often to notice that if e.g. you get a ton of hunters and rogues, agility trinkets drop like crazy. Stuff like that can be coded in.

this is something I noticed also, and could be true and have skewed my data into having so many high % failures. I noticed that if I had a large number of 100% missions (5 or more) and just one non 100% mission, I would fail the non 100% mission every time regardless of %, so I ended up failing quite a few 95-97% missions this way. If this is something intended that definitely would have something to do with skewing my numbers.

[chazus]

I kept track of 611 97% mission with 3 followers. I failed 263 of them.

I'ma call BS on that one. Failing nearly half of your 97% missions? Yeah, no. I've tracked WAY more than you in the past (upwards of 8000), and I don't get nearly those. Not only that, but assuming that it was based off of each person, your numbers would STILL be way off. 97% across three people comes down to about 91% total... You ought to be failing about 60, not nearly 260. So... nope. Just nope.

[Makarena]

I have 10164+9724+9039+2501 = 31428 garrison missions completed on my 4 chars combined and as u can imagine I wouldn't be chatting random shit about something as meaningless as this. I stand by my assumption.

[chazus]

The only way to make sure is to have a large sample of test cases

Which people have already done. Multiple times. And proven with that that the % shown is the % resulted. Not a multiple.

I totally agree that this is something that ought to be tested thoroughly... Except it has been, several times, over literally hundreds of thousands of iterations. We've had over a year of testing it. It was proven.

[Zelendria]

I kept track of 611 97% mission with 3 followers. I failed 263 of them. Over the same course of time I had no 97% one follower missions, but out of every single follower mission in the 90%'s at all I failed one. My sample size is MUCH larger than the posted sample size, or any speculation people are having.

The smell of bullshit in the air is so potent it's burning my eyes.

[Deleted]

By the way, on a similar subject I was almost convinced the motes of harmony on the last expact had a very weird "black box" random number generator that did not really use a flat chance. The reason for this is that it soon became obvious that if you AOE'ed like crazy a ton of mobs the mote outcome might had been similar compared to doing again a lot of mobs but about half of them only. That means it was most likely based on a chance but also on a time limit in a sense ("keep a chance but decrease it if he already got too many motes the last 10 minutes").

I do not know if WoD has something similar in some aspects of it. I haven't been playing it as much.

I believe that change was introduced because of the mass-loot option making AOE farming too easy.