View Single Post

finelinebob's Avatar

12.06.2012 , 01:03 AM | #9
Quote: Originally Posted by Wolvereen View Post
no, 20% aren't really 20%^^

actually the chance is MUCH lower ...unfortunately
Adding to the chorus: no, you are wrong. You are making a naive, as in "lacking enough knowledge" not "silly", mistake about what type of probability is being listed.

There is something called a Simple Probability. This is an expression of a ratio of the number of successes per number of attempts. Although this number must refer to multiple attempts, it is a measure of what level of success you should expect for any one given trial.

This also happens to be the percentage listed under the tooltip "Chance to Research a Schematic". It means exactly this: If you have one item in your inventory with a Chance to Research a Schematic of 20%, then for precisely that single item you may expect success 2 times in 10 trials. If you have a second, third, tenth, or thirtieth item of the same type in your inventory, as you RE them your chance for success remains 2 in 10, regardless of your results from any past or future attempts of the same task. Simple probabilities are calculations that reference independent trials.

Quote: Originally Posted by Khevar View Post
The last time I was worried about the RE percentage, I started keeping track in a spreadsheet.

691 greens RE'ed- 132 successful = 19.1% (tooltip says 20% chance)
316 blues RE'ed - 35 successful = 11% (tooltip says 10% chance)

Statistically, this is a pretty small sample size, but it was good enough for me. I've gone 45 attempts without a successful RE. I've also done 3 in a row successfully.

Toolip is correct.
This, on the other hand, shows a couple of other things that are useful to know if you want to understand how probability works in the game (or elsewhere). The primary thing to note is the idea of "Convergence", the idea that if you have a 20% chance of something happening, you should expect your results to approach a 20% success rate if the events are independent of one another and the number of trials is very large. "Very large" taken strictly means an infinite sample (this is a simplified version of something called the Law of Large Numbers and it is often misconstrued as the "Law of Probability" based on common popular perceptions about how probability should work.

Again, not how it does work, but how people want it to work regardless of the statistical laws that define it.

Now, the perception of the RE process being "broken" because you have 10 items with a Chance to Research a Schematic of 20% and you got any result other than 2 successes, particularly if you got 1 or 0 successes? That is mistaking the idea of Simple Probabilities, which are applied to independent trials, with Binomial Probability, which looks at series of related random events and can tell you things like (1) the expected rate of success below a certain outcome, (2) the expected rate of success for precisely a certain outcome and (3) the expected rate of success above a certain outcome. Or, to put it bluntly: (2) how lucky were you, (1) how many people had worse luck than you had, and (3) how many people had better luck than you.

Binomial probability works a lot different from simple probability because it relies on a series of related events rather than completely independent trials. Convergence in simple probability means that if you make a graph of the number of the attempt (x-axis) versus the average value of the results up to that attempt (y-axis), you would get something that jumps around a bit before flattening out to a straight line for the expected value. Take throwing one six-sided die. At the start, the numbers may jump around quite a bit and look chaotic, but within a few thousand trials that line marking the average roll over time is going to settle down on 3.5. That is what is meant by convergence, quite literally.

Binomial probabilities more closely follow something called the Central Limit Theorem. The Central Limit Theorem essentially says that as you increase the number of attempts towards infinity. then (1) a little less than half of the attempts will be worse than you expect, (2) a little less than half of the attempts will be better than you expect, (3) what is left over will be precisely what you expect, and (4) it will all come out as a normal distribution, a bell curve, with that expected value as the mean.

This is a Bad Thing™, as the Crew Skills system has been implemented. It is working as designed. It may even be working as intended, give the developers who did it did not know enough statistics to realize they were playing with fire, given people with a more naive (silly this time, because they should have known better) understanding than we deserved. It's bad because it says that for 20% RE attempts, nearly half of the players doing these REs over the life of the game will realize a success rate of better than 20% of their attempts and nearly half of the players doing these REs over the life of the game will fail more than 80% of the time.

There is no such thing as luck. In our crafting and Crew Skills systems, it is absolutely certain that there is no such thing as skill. Or knowledge. Or mitigation. Or anything else that would change that horrendous failure rate of 80% to 90% for reverse-engineering.

No, there is something. It's "the developers and producers realizing what a horrible system they have devised and how by brute force randomization they have turned what could have been a game of skill into a purely and ruthlessly skill-free exercise in probable futility ... and that this should be changed."
Jedi Guardian JonBonJovi'wan
"Wanted: Dead or Alive"
Defenders of Monkeys - Prophecy of the Five
DatacronHunter's Video Guides to Datacrons | Datacron Spreadsheet