Crafting mechanics and probability engine...

[jrbaetz]

I am talking about straight up syneth's no HQ involved, my high end crafting does fine by the %'s its the class I am leveling that do not. I am taking in to account the fact I go farm 100 stacks of material for 100 syneth's in that I bust ever reciepe while making something about 1 per syneth... heres the problem I us 3 syneth abilities that require an 80% chance to syneth. 1 in 3 is not 80% then we get to the multi bust syneth busting mulitple times in 1 reciepe. So..... I am probly averaging about 1 good 100% syneth every 4 syneths. So thats 75%. I play on East server random times in the day and it makes no difference. Busting rate stays consistent. Currently doing Arm BS Ach Wev and LW all at once 1 level at a time doing roughly 70 syneths a level. I have had LOTS of samples no where close to the devs. But again the percentages are off.

[Niwashi]

heres the problem I us 3 syneth abilities that require an 80% chance to syneth. 1 in 3 is not 80% then we get to the multi bust syneth busting mulitple times in 1 reciepe. So..... I am probly averaging about 1 good 100% syneth every 4 syneths. So thats 75%.

Ok, the game uses the term "synthesis" referring both to completing a recipe and to the action in a single step of that recipe advancing it's progress number, so it gets a little confusing where you're using which meaning. To avoid that confusion, I'm going to use different terms here: "completing" for when you finish a recipe with either a newly crafted item or a bust, and "progressing" for when you use an individual action like Basic Synthesis on one step of your crafting.

I think the problem is that you're looking at how often you successfully complete a recipe and expecting that to fall into line with the percentage that indicates how likely you are to successfully progress the recipe in an individual action. It won't, not if you're using three progressing steps in the process. If you're relying on three actions with an 80% success rate each, then you only have a 51.2% chance that all three will succeed, along with a 38.4% that two will succeed, 9.6% that only one will succeed, and a 0.8% chance that they'll all fail. So you should expect about half to have failures, and over ten percent to have multiple failures. That isn't going to feel like an 80% success rate, but it is.

Add in the human tendency to remember failures easier than successes, and we keep having threads like this pop up with people thinking the success rates are off. Yet whenever someone actually keeps track over a statistically large sample of individual actions, they find it actually comes out about right. (Admittedly, it has been a really long time since I've heard of anyone actually keeping track like that. But quite a while back, like maybe two or three years ago, a particularly long thread on the subject led to someone doing it. Their result was that the success percentages are right after all.)

[Frizze]

I have had LOTS of samples no where close to the devs.

Add in the human tendency to remember failures easier than successes, and we keep having threads like this pop up with people thinking the success rates are off. Yet whenever someone actually keeps track over a statistically large sample of individual actions, they find it actually comes out about right. (Admittedly, it has been a really long time since I've heard of anyone actually keeping track like that. But quite a while back, like maybe two or three years ago, a particularly long thread on the subject led to someone doing it. Their result was that the success percentages are right after all.)

Jrbaetz, you say youre leveling 5 classes from 1 to 70? Then you certainly have a decent sample size in front of you. Record your findings. Track your quality and progress actions(you can skip recording the 100% ones - we know they work right). All you need for each action is % chance to succeed and whether it did or not. This will prove to all the doubters that youre right(or more likely prove to you that the doubters are right). Because every time i read someone who says something like "I am probly averaging about 1 good 100% syneth every 4 syneths", all i can think is "human tendency to remember failures easier than successes".

[Elamys]

If I didn't know better (and I probably don't) I'd almost make a guess that when condition is Excellent, that some under-the-covers modifier gets baked into the equation.
I would venture to say that I have a <=50% success rate on Synths when Excellent. I see "Excellent" and say "oh boy.." cross my fingers...

Just my observation, YMMV. Just seems that way to me.

This is the only "conspiracy theory" related to listed success rates that I sort of believe. It seems that on Excellent procs, it's like I have a baseline lower chance to succeed. I almost never succeed an unbuffed touch on Excellent.

[Jojoya]

This is the only "conspiracy theory" related to listed success rates that I sort of believe. It seems that on Excellent procs, it's like I have a baseline lower chance to succeed. I almost never succeed an unbuffed touch on Excellent.

And yet I've had several times when I've done it successfully without Steady Hand I/II, definitely more than it's failed on me. I will say that it seems to fail more on a high level item than a lower level item but that could just be the typical perception bias.

My personal beef is with Armorer and Rapid Synth. You can't get Rapid Synth to 100% even with Steady Hand II so there is always that chance of failure. My failures still seem more or less in line with the success chance in all the jobs except ARM. With ARM, I get about 2 out of 3 failing. I mentioned the situation as a joke in FC chat and the other omni-crafter went "you too?". He had noticed the same thing - his failures were pretty much in line with what they should be except for ARM.

It makes me wonder what they use to determine the seed for the random number generator on crafting procs. Bad seeds have been known to skew results. If the seed is influenced by character/account data, it would explain why some players consistently see a result outside of what's expected while others don't - they likely ended up with a bad seed.

[Elamys]

And yet I've had several times when I've done it successfully without Steady Hand I/II, definitely more than it's failed on me. I will say that it seems to fail more on a high level item than a lower level item but that could just be the typical perception bias.

My personal beef is with Armorer and Rapid Synth. You can't get Rapid Synth to 100% even with Steady Hand II so there is always that chance of failure. My failures still seem more or less in line with the success chance in all the jobs except ARM. With ARM, I get about 2 out of 3 failing. I mentioned the situation as a joke in FC chat and the other omni-crafter went "you too?". He had noticed the same thing - his failures were pretty much in line with what they should be except for ARM.

It makes me wonder what they use to determine the seed for the random number generator on crafting procs. Bad seeds have been known to skew results. If the seed is influenced by character/account data, it would explain why some players consistently see a result outside of what's expected while others don't - they likely ended up with a bad seed.

That's really strange. I haven't noticed any differences from crafter to crafter. I might start keeping track of stuff like that though. I'm not generally one for theorycrafting and crunching numbers in the game but it would be nice to know

[jrbaetz]

Yeap broke busted a 90% syneth about 45% success rate for a level 50 master reciepe when I am 65....

[jrbaetz]

okay to clarify because I am saying "syneth" and I mean to progress. Now it says it has an 80% or 90% chance to progress... if its not 80% or 90% why is it labeled that way?

[jrbaetz]

statisically that is impossible to bust that many progression steps if it is 80 or 90 %

[Elamys]

You have a 10-20% chance to fail on each individual step. It's not "you do ten steps then two of them fail." That is a significant chance to fail on each of those actions. It's labeled that way because it is an 80-90% chance to progress.