View Single Post

finelinebob's Avatar

12.06.2012 , 04:55 PM | #11
Quote: Originally Posted by SteveGarbage View Post
...The truth is there IS a lot of luck involved, because it's a RNG system, that, yes, has been talked about to death but still not changed.
Luck is an epiphenomenon. It doesn't cause things to happen; it's something we see after the fact and only exists in our heads. It's not part of the game.

Quote: Originally Posted by SteveGarbage View Post
I don't have the game in front of my right now, but if I'm not mistaken, the chance to RE a schematic off a blue-grade item is 10 percent? Or is it only 5 percent when you get to those higher levels? ...

That means on top of a 1 in 10 chance or 1 in 20 chance to not RE anything, you're adding another 1 in 5 chance to get the one you want. As far as I'm aware, just because you already RE'd one of those last five schematics doesn't take it out of the pool. It's not like, "Oh you already have Expert, here's Vehemence instead!" It's more like, "You got Expert! Oh you already have Expert? OK then you get nothing."
On prefixed items, blues are 20% up to around level 23 combat rating, 10% from 24 on up. It may be level 32, can't recall right now. And if selection of what schematic you get occurs "without replacement" as you suggest -- with the game always using all 4 or 5 Tier 2 schematics possible and not removing the ones you already have from the mix -- then the odds of researching a new schematic would drop every time you learn one. For example, when REing a higher level item already with a Critical prefix, you initially have a 10% chance of learning a new schematic. Once you learn one of the five available on a Critical item, if things worked out the way you suggest your odds at learning one of the remaining four would be (Chance for successful RE) x (Chance for getting something other than what you know). If you know one Tier 2 schematic, your chance to learn the second would be (0.1)x(0.8) or 0.08 ... a 2% drop from the stated 10% chance. Similarly, knowing two Tier 2's would drop your chance to learn to 6%, knowing three drops your chance to 4% and finally knowing four Tier 2's would drop your chance of ever learning the final schematic to 2%. That's not what the tooltip says -- it stays at 10% no matter how many you know until you know them all, then it says No Research Available.

You're right that in independent trials, the odds of getting a specific Tier 2 schematic from a Tier 1 with a 10% chance of success -- granted that you do not know any of the Tier 2 schematics for that particular Tier 1 item -- is the product of the probability of succeeding and the probability of being assigned a particular schematic. Initially, that number is 0.1 x 0.2 or 0.02: a 2% chance of any single, independent Tier 1 RE giving you exactly what you want (if you don't already know anything). The assumption that what Tier 2 schematic you are assigned each time you succeed being independent of past success for a given Tier 1 schematic is a fallacy, tho, as I just mentioned above.

Quote: Originally Posted by SteveGarbage View Post
Again, those tries are all independent of one another. You don't gain anything by rolling multiple times. You odds will be 1 in 100 on try No. 1 and they will be the same on try No. 100. Your percentage chance to win isn't increasing with each try. You're not "due" to win eventually. That's the gambler's fallacy. So while the chance of failing 30 times in a row is whatever small percentage, the chance of you failing on each try is 99 percent, period.
I'm well aware of the Gambler's Fallacy and I'm not advocating that it is true here. I am, however, pointing out the difference between independent trials and dependent trials. As I have discussed elsewhere, a series of dependent trials relies on binomial probability, not simple probability. A core aspect of understanding large numbers of dependent trials is the Central Limit Theorem. In a way, we're kinda talking about the "quantum mechanics" of Crew Skills. Just like electron spin has a 50% chance of being up or down, once you actually measure it that spin will be up or it will be down: probability in this case talks about possible future outcomes. If you start measuring things, that act of measurement creates a dependency. Calculating a binomial probability of your series of measurements is not a calculation of how probable your particular out was but a comparison of your outcome against all possible outcomes given the same initial conditions.

This is exactly where the crafting system breaks down ... and it's not broken mechanics -- it works as designed. It is the design itself that is "broken". Consider combat for a second. Your chance to hit your opponent is also based on RNG. Would you put up with a failure rate for your attacks that was as bad as what you get for crafting? Of course you wouldn't, and you don't have to because of the design of the combat system. Every class starts off with an extremely high chance to succeed, relative to your chance to succeed at crafting. As you level up, your "skill", your accuracy, improves. Furthermore, you can choose skills in your skill tree that provide additional mitigation against your own chance to fail to hit, or mitigation that lowers an opponent's chance to hit you. You can even get gear that will improve your accuracy to the point that, expressed as a probability, your chance to hit is greater than 1.0. As a simple probability, that makes no sense. But it's there all the same, to give you a chance to out-mitigate your opponent.

What mitigation of chance is there for crafting? Very little. A few companions start with an increased chance to crit on a craft or mission. Increasing your affection also increases crafting mitigation by a small amount. At best, you have something like a 17% chance to get an augment slot on the hardest items you can craft with the right companion and the right affection level. Reverse-engineering, however, is sadistically free of any chance to mitigate your chance to fail. Companion affection does not alter it. Companion specials do not alter it. Even your SKILL level does not alter it. A level 49 Premium item, with a difficulty rating of 108, has precisely the same chance of researching a new schematic as a level 9 Premium item with a difficulty rating of 28. Does that make any sense? Does it make sense that an item with a rating of 28, something you can craft at skill level 1, has precisely the same RE failure rate when your skill level is 400? What does that extra 399 points of "skill" count for?

Because there is no way to mitigate your chance to fail at REing, the distribution of player failure rates will remain pretty much fixed in a normal distribution, as the Central Limit Theorem points out for a significantly large number of measurements. That means that almost HALF of the players in the game who have tried to RE Premium items currently have a failure rate GREATER THAN what the tooltip says it should be. Equally, almost half of the crafters out there have a failure rate LESS THAN what the tooltip says. What makes this broken by design is that the tooltip says for Premium items you have an 80% chance to fail. For combat, we start off with somewhere around a 20% chance to fail to hit your target and by the time you hit level 50, your chance to fail to hit has become negligible. For reverse-engineering, the number of crafters out there who have a history of failure that is negligible is, well, probably non-existent. Given the distribution in a normal curve, by far the majority of the "lucky" people out there who fall on the high side of the curve for RE success still do not reach that 20% chance to fail that combat level 1's have, let alone a 50% failure rate.

Quoting the simple probability of your chance to fail on your very next attempt to RE a particular item is a simple thing: look at the tooltip. Considering your HISTORY of failures in REing, on the other hand is not quite so clear. We see the tooltip and think it should be the same, but the Law of Large Numbers says "talk to me after an infinite amount of trials; maybe I'll give you an estimate after a few thousand." And what are we hoping for? A failure rate of either 80% or, for the more important gear, a failure rate of 90%. That's something to hope for? That's what our skill buys us?

If people cared as much about crafting success as they do about combat success, there would be no way we would be saddled with the failure of a system we currently have. Yes, it is working exactly as designed given the latitude RNG requires. That is exactly the problem.
Jedi Guardian JonBonJovi'wan
"Wanted: Dead or Alive"
Defenders of Monkeys - Prophecy of the Five
DatacronHunter's Video Guides to Datacrons | Datacron Spreadsheet