Synthesis Percentages Flawed?

[Legato_]

I'm a hunter that uses dual sabres. My Personal machine has striking level 21. I bought slasher, the 9250 board , so that I could make the strongest dual weapons. The board has 3 tries.

At my PM's current level, the board has 62% chance of success. When I tried to synth the slashers, all three failed. I was confused. I bought another slasher board, and it failed for a fourth time.

I wanted to determine the % chance that all four boards would fail given 62% chance of success. By using a binomial distribution, setting n to 4, x to 0 p (chance of success) to .62, and q (chance of failure) to .38, you determine that there is a .02085 chance that this happens.

This means there is slightly more than a 2% chance that all four were to fail on me. If I were to do it a fifth time, which I am about to do, and it fails again, I will be almost certain that the percentages for synthing are broken. The chance for all 5 to fail in this scenario is less than 1% (.0079).

Does anyone have insight into this?

[Dhylec]

What's your luck when you tried to synth? Have you taken that [important but often forgotten] factor into your calculation?

[Nani-chan]

Did the % go down when you added a ??? property to it? I had that happen when I made armor before.

[Randomness]

It goes down with elemental slightly, and a LOT if you use el or im photons.

[SonicTMP]

ouch real bad luck man. I had 3/3 shooters work for me at liek 57% and then had my megaline synth fine. Had zero luck that day too. It's just a matter of how the dice lands for you personally.

[Randomness]

On 2006-11-04 15:50, Dhylec wrote:
What's your luck when you tried to synth? Have you taken that [important but often forgotten] factor into your calculation?



I dont forget, its why I'm waiting for a 3 luck day to make a lightbulb.

[Lyrise]

This is just PSU's RNG at work. While your math is spot on, all your calculations do is prove an ideal condition. However, as with real life, you can never expect anything to be ideal. Personally I think while it sucks for stuff like this to happen, this level of randomness is kinda fun to think about.

Don't feel too bad, I've failed synths at 82% success, at 91% rate, and am 5/12 on 8 star weapon synthesis rated at 53%. I've also succeeded in synths that were 39% and 32% as well. As long as you don't try to take the randomness of it all too seriously, you eventually have things go the way you want them to.

edit: Luck does nothing as far as your success rate. In synthesis, luck affects the result of the product (re: elemental ratio) and nothing else.

<font size=-1>[ This Message was edited by: Lyrise on 2006-11-04 15:58 ]</font>

[DrewSeleski]

2% means it's still possible though. It is in fact possible for every board with a chance to fail failing throught your entire time on PSU. Of course very unlikely... but still possible.

[KireekSr]

They need to do something about that... I mean I think that the danger in losing something is really cool but... sometimes its just flat out unreliable to go by those percentages.

I made a 55% chance Busterline (light -based!) today. I was happy about that lol.

yes, noob I know. I would have been furious if it messed up, I can't imagine how upset you must be.

better luck next time man.

[Ceresa]

You are 1 player out of thousands, someone has to hit the outliers, sucks it has to be you, but there it is.

I don't have the exact numbers, but a few weeks ago when I made some 40 haujirods, the average was damn close to the 80% my pm said.

4 is simply a bad sample, your luck will turn around eventually.

[Cause_I_Own_U]

I screwed up my 74% synths twice at the beginning of the game, but ive goten a bit luckier since then

[Legato_]

"you are 1 player out of thousands, someone has to hit the outliers, sucks it has to be you, but there it is.

I don't have the exact numbers, but a few weeks ago when I made some 40 haujirods, the average was damn close to the 80% my pm said.

4 is simply a bad sample, your luck will turn around eventually. "

The number of people playing is irrelevant to the tests I used.

Also, someone mentioned luck. The day I did this, I had maximum luck. Regardless, even if you don't have more luck, the chance per trial should be at least what the game tells you it is.
And when you run five independent, identical trials with a .62 chance of success, your chance that all will fail is .0079.

Under any organized system that uses statistics, this is enough evidence to question whether the percentages they present are accurate.

[PALRAPPYS]

Ouch. That's gotta hurt. Hope you get it good soon! http://www.pso-world.com/images/phpbb/icons/smiles/icon_smile.gif

And how do you raise synth levels? I do some synthing with my PM, but I assume you have to feed your PM foods and stuff, is that correct?

[Legato_]

Yeh. Dimates, trimates, atomizers, scape dolls add a lot to striking.

[Dusk21]

I feel ya Legato, a little bit ago I was tryin to synth me up some 5* weapons, I failed 3/3 on the rifle at 57 or 58%. A little pissed and thinking ok...with all those fails how could I fail some more, the chances can't be good. I buy the 5* dual handguns board and proceed to fail 3 straight on those at 60% (fed the PM a bit between). After that I went crazy on feeding my PM, an evolution and a ton of M later and I gave it another go on the 5* rifle, 3/3 successes (however as luck would have it I broke 2 of those 3 rifles trying to grind to +3 with 3 luck both times).

Anyways, moral of the story is...I dunno, but you'll get your item soon enough. Try levelin your PM some more, that seemingly helped me.

[Ceresa]

Did your test say there was a 0% chance of it happening? No? ok then, did you think arbitrarily running some graphs and numbers is going to effect the result of an item you make in a game?

People playing is irrelevant? People playing is everything. You think the other 20000 ppl online are just using the starter weapons? No they're synthing the same shit you are, with the same %, some success, some fail. I made 5 canes in a row at 56%, 5.5% chance of that with your beloved and meaningless distribution. Do you not understand the concept of outliers?

The only thing that can prove the synthesis % given by PM to be incorrect is if you take thousands of results from players at 62% success. If you get 3000/10000 made with 62%, then yes something is wrong, but you will find that you will be very close to 6200/10000.

The fact that you think an experiment with 4 tests is accurate is laughable at best, and a complete insult to the fundamentals of statistics that you try to preach, really did you skip the first day of class?

[Legato_]

I don't necessarily care as much about getting the item as I am interested in whether the percentage that the game presents is accurate. If you're telling me that you failed 3/3 two times in a row, with over a 50% chance of success both times, that seems to reinforce my inquiry.

Perhaps the percentages are not calculated properly for the five star weapons? Or if they are, those percentages aren't displayed properly? It certainly seems possible.



<font size=-1>[ This Message was edited by: Legato_ on 2006-11-04 17:38 ]</font>

[Legato_]

"The fact that you think an experiment with 4 tests is accurate is laughable at best, and a complete insult to the fundamentals of statistics that you try to preach, really did you skip the first day of class?"

I'm not running a t test to make inferences on the true population distribution of successes. If I were doing that, 4 trials would be insufficient.

All I'm trying to do is determine the odds that something would fail 5 times in a row under these circumstances. Getting results that are less than one percent chance, in addition to other people telling me they failed 3/3 on 5 star items, is enough to question whether the percentages for 5 star weapons are calculated correctly.

Am I making an assertion that they're wrong? No.
I'm saying there is evidence to suggest that something may be wrong with the percentages they provide, nothing more.



<font size=-1>[ This Message was edited by: Legato_ on 2006-11-04 17:40 ]</font>

[Ceresa]

No matter how small the % is, if it's greater then 0, and there's enough ppl banging their head against it, someone will get owned by the worst case scenario.

It just happened to be you this time. There's no incorrect math on your part or the game whatsoever, because there are enough ppl that fall in the 98% that don't fail 4 times in a row, they just don't post about it. Kinda like "squeaky wheel gets the grease" or something.

[etlitch]

The chance of failing 4 62% synths in a row would be ~15%, which is'nt exactly impossible. At least it's not like my PM, failing 80% synths 4 times in a row.

[Ras]

I failed my first 3 56%s. Chance of that? 8.5%. I thought I was really unlucky.
Next 7? All successes. Chance of that? 1.7%. Nice!

Lyrise and Ceresa are right.

[DummieAcount]

I've succsessfully made 4 60%> chance items in a row.

You're just really unlucky.

[Dusk21]

I think it does even out in the end to be quite honest, taking into account all your synths successes and failures. Hopefully the game will repay me for all those grinding failures though, that'd be nice (I DID get my 5* rifle up to +4 though, and it finally gave me an appreciation for rifles too. Nice http://www.pso-world.com/images/phpbb/icons/smiles/icon_smile.gif )

[excal]

Sorry, you don't understand probability. You don't use binomial dis. or any fancy crap. All you need to know that, is that everytime you do it, the chances reset. It's 62% the first time. 62% the second time. 62% the third time. Etc. Every single time, you have 38% chance of failing. It doesn't reduce the more you do it. 38% chance is quite high, I dont see why it wouldn't happen, while you might be unlucky.

[Razorback]

there's always a chance of failing

[Parn]

On 2006-11-04 17:34, Legato_ wrote:
I'm saying there is evidence to suggest that something may be wrong with the percentages they provide, nothing more.
What evidence? If I roll two dice four times and keep getting numbers lower than 7, that's hardly evidence that the dice are rigged. That's just the nature of the game. You win some, you lose some.

My partner machine is 100 armor, and I synthed three pieces of armor this evening. One was a 5 star with a 56% chance of success, the second was a 6 star with a 53% chance of success, and the third was a 3 star with a 70% chance of success. The five star was successful and Remedy is the happy owner of some really good armor. Kamin's 6 star armor failed unfortunately, and my own personal synth for the 3 star also failed despite higher odds. It sucks, but again, you win some, you lose some.

Now I just hope that my rappy and pannon figures synth successfully. If they don't, then that's that. No sense in dwelling on it.

[Legato_]

"Sorry, you don't understand probability. You don't use binomial dis. or any fancy crap. All you need to know that, is that everytime you do it, the chances reset. It's 62% the first time. 62% the second time. 62% the third time. Etc. Every single time, you have 38% chance of failing. It doesn't reduce the more you do it. 38% chance is quite high, I dont see why it wouldn't happen, while you might be unlucky. "

You're correct in saying that it's 62% the first time, 62% the second time, etc. That's why they're called independent, identical trials. However, you either have a fairly insufficient background in statistics or misinterpreted the situation.

An example of a binomial dist. problem - A hunter has been found to have 85% accuracy when shooting at a target. Given this information, what is the probability that he will hit 3 or more targets in 8 trials?

Just like with flipping a coin, or synthing, there are only two options. In this case, it is hit or miss. But to determine this probability, you need to use a binomial dist. formula to determine the answer.

I'm not saying that since I have already tried 5 times, the sixth trial will not have 62% chance of success and 38% chance of failure. What I am saying is that the probability of 6 independent, Identical (I.E same percent for each one) trials with 62% chance of failing ALL failing is below one percent.