Live Letter XXX Discussion Thread

[Kaze3434]

There are two different (yet related) concepts that people have been discussing here, and it's important not to confuse them.

The first, which we'll call event probability, represents the chance of any one isolated action happening in a certain way. In the map dungeon scenario, this would be the chance when opening a door that you'll get the correct door. As I briefly mentioned in my original post, this may not necessarily be 50% depending on SE's exact implementation, but it's easier to assume that it is. What this means is that any time you clear a room and go to open the door to the next room, you'll always have a 50% chance of your chosen door being the correct one. That is, regardless of whether you're opening the door to the second room or to the seventh room, you still have a 50% chance of being right.

Another way of representing this is to say, "Given that I've already reached the N'th room, I will always have a 50% chance of reaching the N+1'th room."

This should not be confused with the other concept, cumulative probability, which represents the chance of a given series of events happening in a certain way. In the map dungeon scenario, this would be the chance of reaching a particular room without any pre-conditions. What this means is that if you wanted to figure out your chance of reaching, say, the third room--before you even pop the map in the first place--you need to consider it in the context of the chain of events that will lead you to that outcome. For that particular example, it's very straightforward to derive the cumulative probability using the individual event probability form discussed above:

You have a 50% chance of having the portal spawn once you beat the initial pop.
Given that you had a portal spawn, you have a 50% chance of selecting the correct door from the first room.
Given that you made it to the second room, you have a 50% chance of selecting the correct door from the second room.

Therefore, the probability of making it at least as far as the third room is the product of the individual event probabilities, or (1/2)^3, which works out to 12.5% for any given map. Of course, this doesn't mean that if you did 1000 maps, you would make it to the third room exactly 125 times, but any given sample set should be close to that proportion--assuming SE is using a reasonable PRNG algorithm in the proper way; all bets are off if that isn't the case.

thats the problem. SE RNG is stupid for the most part. small example from yesterday is when i was melding some materia for a friend. went thru 8 materia at 26% before one finally went in, and then the next chance was 10%. it went in on the second one. for me, i have stopped using the collective probability for this game, because its largely depressing, and at times, useless because rng gonna rng.

[Ariannaid]

thats the problem. SE RNG is stupid for the most part. small example from yesterday is when i was melding some materia for a friend. went thru 8 materia at 26% before one finally went in, and then the next chance was 10%. it went in on the second one. for me, i have stopped using the collective probability for this game, because its largely depressing, and at times, useless because rng gonna rng.

I know exactly what you mean (I think anyone who's done any crafting, gathering, desynthing, etc. is really familiar with the streaky RNG in this game). That doesn't necessarily mean that the algorithm itself is implemented or used incorrectly, but it does point out one of the problems I have with the game: namely, true randomness isn't very people-friendly, in my opinion.

People instinctively look for patterns in data, which causes number sequences with a lot of streakiness to appear less than random, even if they truly are randomly generated (e.g., sourced from radioisotope decay). Another problem is that the observable data sets we typically see in game as players are small enough that the effects of any streakiness are magnified. If you were able to meld a thousand materia, you probably would see something closely approximating the displayed success rate, but that doesn't make it feel any better when you're only melding 10 and they all fail (something that would happen about 6% of the time at a 25% individual success chance; considering how often you meld materia in the game, seeing something like this is almost guaranteed every now and then).

I would much rather see something like an asymptotically adjusted RNG, where every consecutive failure slightly increases your chance of success until you do succeed. Tuned properly, this could give results that wouldn't be truly random but would look much "nicer" from a person's perspective yet wouldn't deviate too much from true probabilities in the long term. Systems like this have been used in heavily-random games (see "smart RNG" in games like Destiny, mods for XCom, etc.), and from my own experience, they do feel better than truly random systems. The main problem with them is needing to track the current RNG weights for each action on a per-player basis, which in a game like FFXIV could be a pretty significant overhead.

So, TL;DR: server limitations (or developer philosophy, but it's fun to blame everything on server limitations ).

I agree, I feel RNG should be used to make events seem more alive and less scripted, but often in this game stuff is still scripted and RNG is just added for the sake of it. I don't see any reason why any of this RNG is in place (20% chance to spawn portal, 2 doors 1 = move on, 1 = cry). It seems very unnecessary.

I agree with this, in that I'd much rather have the rooms escalate in difficulty such that making it to deeper-in rooms would feel like a real accomplishment rather than just getting lucky. Maybe the Deep Dungeon coming in 3.35 will scratch that itch.