Tiering Philosophy - It's that time again

[Scarlet]

I don't think you have any idea on what you're talking about. I'm asking you why we should go quickly rather than go as fast as possible, but you've been dodging the question for multiple posts now.

You need to work on your reading comprehension, then, as apparently I'm dodging questions that either have never been asked or already been answered.

"Different playstyles and/or reliability", to make it short. The same reason Warp is banned in some tier lists, really: to make using a greater selection of units realistically possible while still allowing for discussion.

In any case, I've made my point, so I see no reason to continue this discussion.

[Chiki]

This is very true.

Okay. That said going "as fast as possible" is often less reliable. Do you include reliability Olwen?

http://serenesforest.net/forums/index.php?showtopic=40004

[Snowy_One]

How about a 'brisk pace' is determined by how long it would take the lord to kill the boss/seize the square assuming he moved to the maximum extent of his unmodified movement every turn, had no aid, and killed every foe in one round/was never in any danger?

Here's the thing. When rescue-dropping, smiting, and refreshing it's possible to get the lord to move at a VERY fast pace. So fast that it would be 'better' from a turn-count perspective to simply rescue the lord and keep refreshing the mounted unit on all but the smallest maps. This is VERY biased though since it assumes that certain units MUST be played and the like and can reward units for having no impact at all. For example, a Titania on a team with Reyson will 'do better' than a Titania on a team without Reyson even if she is statistically identical and never gets refreshed just because Reyson refreshed Marcia carrying Ike to clear the chapter faster. Meanwhile, a masterfully played team of super-strong foot-units would be inherently worse than a team of fliers even if the foot-units were hacked to have 100's in every stat except movement simply because they can't move as fast as the mounted units. Since the point of the list is to measure how good the units are and does so through a proxy measurement of turncounts, assuming the lord is receiving a boost from outside sources sort of defeats the purpose.

In otherwords, assume two teams. Both identical except that team A has a dancer named 'Michael Jackson' and team B has an Archer named 'Chuck Norris'. Norris has infinite range, no minimum range, every good skill, 100's in every stat, and so-forth (for the sake of argument assume he can't attack bosses on kill-boss chapters). Any sane person would put Norris as the best unit because he's basically invincible and destroys everything, but Michael Jackson's team will get the lower turn-count and be 'better' even if Michael Jackson is a completely average unit aside from his dancing simply because he can refresh a lord/carried unit.

[Redwall]

Okay. That said going "as fast as possible" is often less reliable.

I don't disagree.

bearclaw: You don't understand what we mean by "going as quickly as possible." I can, for example, go as fast as possible (subject to reliability constraints) using trash like Virion, Donnel, Kellam, and so on. I can also go as fast as possible using top-tier units like Sumia, Cordelia, and Frederick.

It is less arbitrary to say "subject to reliability constraints, faster is better" than it is to say "subject to reliability constraints, faster is better, except when I don't like it."

You need to work on your reading comprehension, then, as apparently I'm dodging questions that either have never been asked or already been answered.

"Different playstyles and/or reliability", to make it short. The same reason Warp is banned in some tier lists, really: to make using a greater selection of units realistically possible while still allowing for discussion.

I don't think we're on the same page. We can, for example, play as quickly as we can, subject to reliability constraints, with Donnel, Virion, Kellam, and so on. The argument is that this is less arbitrary than playing "pretty quickly," subject to the same reliability constraints, with the same set of characters.

In otherwords, assume two teams. Both identical except that team A has a dancer named 'Michael Jackson' and team B has an Archer named 'Chuck Norris'. Norris has infinite range, no minimum range, every good skill, 100's in every stat, and so-forth (for the sake of argument assume he can't attack bosses on kill-boss chapters). Any sane person would put Norris as the best unit because he's basically invincible and destroys everything, but Michael Jackson's team will get the lower turn-count and be 'better' even if Michael Jackson is a completely average unit aside from his dancing simply because he can refresh a lord/carried unit.

I've never played the Tellius games so I can't comment on that part of your post. But your claim that versatility matters is one that is already accounted for in tier lists. If Michael Jackson indeed required a very specific team to get a lower turncount than team B, then MJ would be penalized. I don't know if there are any cases quite so extreme in the games, though.

Edited June 3, 2013 by Redwall

[Espinosa]

That would be a strange restriction Snowy. For example, can you picture a tier list of FE DS H5 that assumes that we turn Marth into a ORKO machine, in a mode where you generally need several units to gang up on an enemy (or effective weaponry)? That'd be a nonsensical way of playing the game, though having a Marth that's capable of combat obviously pays back in the final map.

If you know what you need to beat the game, you needn't do any more training or, God forbid, grinding than is necessary.

[Chiki]

Anyway, I don't think the "playing as fast as possible" criterion will work simply because people don't seem really interested in it. People are more interested in banning Rescue and Galeforce to "be a gentleman" and babying Nowi and such.

SDS is such a gentleman, he must have the most one-night stands out of anyone here.

Edited June 3, 2013 by Chiki

[XeKr]

http://serenesforest.net/forums/index.php?showtopic=40004

Just curious, you wording seems to imply you see "reliable" as a discrete thing. Clears above an arbitrary cutoff of ~60% are reliable while others are not. Assuming some set turncount, do you consider a 100% clear ~= 60% clear > 59% clear ~= 0% clear? All clears above ~60% are "reliable" and equivalent? Also, are strats that transition on into 1 turn more clears equivalent to strats that are success or death?

If 100% is better than the 60% > 59 > 0, etc, why not just weigh them by reliability for a continuous metric instead of discretely "reliable" or not? If an arbitrary cutoff, why not 95% since that's more standard in terms of "scientific" cutoffs.

[Chiki]

There's no such thing as a 100% clear, first off.

I don't think you can compare this to science. It really is something that depends on a person's patience--are they willing to restart? Second, which do we prefer? The 5 turn with 60% or the 6 turn with 95%? Personally, I'm up for the 5 turn, but if people have different opinions, then that's fine too. We should, together, establish arbitrary limits for reliability.

This is another necessary arbitrary limit that we have to argue on and decide.

[Espinosa]

In all games besides FE9 (Fixed), reliability is such a phoney concept when brought up. The only 100% 'reliable' strategy involves a team of units at their base stats (because it's not reliable if you have to rely on any procs in terms of growths, which might not activate) and every enemy attack that could hit you hits and that can crit you crits. There's also the problem of your own attacks not being 100% accurate most of the time (sometimes fixable in games with forges, to a certain extent), and if you whiff, you may or may not lose that turn, and you may or may not compromise reliability on the following turn (e.g. the whiff exposes your frail healer or dancer to the opposition whom you fail to overpower).

Then there's games like FE6 where even Rutger misses bosses a lot with skill capped and you could potentially never move forward because all of your attacks miss.

Of course, degrees of reliability should be taken seriously.

[Chiki]

We can use the probability of a character like Rutger having a certain amount of strength in a chapter to measure the reliability for a clear of that chapter.

[Espinosa]

We can use the probability of a character like Rutger having a certain amount of strength in a chapter to measure the reliability for a clear of that chapter.

Yeah, but even a Rutger who had only perfect level-ups is often very likely to repeatedly whiff the boss.

[Chiki]

I don't see the problem. We only consider the overall reliable alternative (60% or so).

[Hawk King]

Uh Chiki, there are such things as 100% reliable clears. However rare they may be, they do exist.

[XeKr]

There's no such thing as a 100% clear, first off.

I don't think you can compare this to science. It really is something that depends on a person's patience--are they willing to restart? Second, which do we prefer? The 5 turn with 60% or the 6 turn with 95%? Personally, I'm up for the 5 turn, but if people have different opinions, then that's fine too. We should, together, establish arbitrary limits for reliability.

This is another necessary arbitrary limit that we have to argue on and decide.

Well there are, but they're very, very, very rare. Some of FE9 fixed could be. Some bosskills in this game are probably 100% though, or can be made so (Basilio vs. flier with Rallies). But in general you're very right, especially with growths (which the community generally doesn't consider the statistics of... <_<).

How about a 80% vs. a 60% clear then? Or in general, are all clears above a certain cutoff reliable?

Instead of a cutoff, above which is reliable and below which is not, why not treat reliability more continuously. Things are relative: they are more reliable or less reliable instead of "reliable" or not.

[dondon151]

even i can't be bothered to dredge up 4-year old posts of mine, so what does it prove? that i used to have an opinion? surely even you can figure out that it has no meaning in this context.

[Redwall]

I prefer a different definition of reliability: one that is based on chance of death.

If we have 0% chance of death for a strategy that is 40% likely to clear in four turns and 60% likely to clear in five, then we simply compute the expected turncount and separately note the chance of death, which in this example is 4.6 turns with 0% chance of death.

Edited June 3, 2013 by Redwall

[Chiki]

Uh Chiki, there are such things as 100% reliable clears. However rare they may be, they do exist.

Well, you have to consider what I mean by 100% reliable. I also consider the fact that a unit will have certain stats at that point of the game. It's a very complicated matter, and I doubt each and every contributing factor to a clear has a 100% chance.

I guess there's FE9 Prologue which has a 100% reliable clear, though.

Well there are, but they're very, very, very rare. Some of FE9 fixed could be. Some bosskills in this game are probably 100% though, or can be made so (Basilio vs. flier with Rallies). But in general you're very right, especially with growths (which the community generally doesn't consider the statistics of... <_<).

How about a 80% vs. a 60% clear then? Or in general, are all clears above a certain cutoff reliable?

Instead of a cutoff, above which is reliable and below which is not, why not treat reliability more continuously. Things are relative: they are more reliable or less reliable instead of "reliable" or not.

That's the point. We have to decide where the cutoff is. I personally go with 60%, but we pretty much have to vote and it and see what people want.

I prefer a different definition of reliability: one that is based on chance of death.

If we have 0% chance of death for a strategy that is 40% likely to clear in four turns and 60% likely to clear in five, then we simply compute the expected turncount and separately note the chance of death, which in this example is 4.6 turns with 0% chance of death.

But there's also some clears that require rigging a crit (like in my LTC run, Morgan getting 2 crits on Grima with a forged Killer Axe). Why not consider them? Crits and skill activations and such can save a lot of turns.

[XeKr]

I prefer a different definition of reliability: one that is based on chance of death.

If we have 0% chance of death for a strategy that is 40% likely to clear in four turns and 60% likely to clear in five, then we simply compute the expected turncount and separately note the chance of death, which in this example is 4.6 turns with 0% chance of death.

That's where I was going. I've favored expected turn count for a long time. >_>

@Chiki: Crits/skills are considered in the weighted probabilities.

Edited June 3, 2013 by XeKr

[Redwall]

But there's also some clears that require rigging a crit (like in my LTC run, Morgan getting 2 crits on Grima with a forged Killer Axe). Why not consider them? Crits and skill activations and such can save a lot of turns.

My proposal doesn't neglect those possibilities. Looking at your example, we can compute things like a) Morgan getting the requisite number of +Str level-ups, b) the possibility of Morgan procuring the two required crits, and c) the possibility that, if Morgan doesn't get two crits, he will get the two crits on the next turn.

[Chiki]

Could you give an example of that? I'm not sure I understand it fully.

[Redwall]

OK, let's say Morgan has a 40% chance of procuring two crits (thereby defeating Grima) on any given turn. Suppose for simplicity that Morgan cannot die, and that Grima gets healed back to full health during the enemy phase. Also assume that Morgan can't attack on the enemy phase.

The probability of a one-turn is 0.4

The probability of a two-turn is 0.6*0.4

The probability of a three-turn is 0.6*0.6*0.4

...and so on for higher-order possibilities.

The expected turncount would wind up being a series

<TC> = 0.4*1 + 0.4*0.6*2 + 0.4*0.6*0.6*3 + ...

I think this particular series simply converges to 1/0.4 = 2.5. So I would say that the clear has a 0% chance of death, and can be expected to clear in 2.5 turns on average.

[Chiki]

Why not value failure as a whole in general, though?

[dondon151]

that's too theoretical. what is more often the case is that if something doesn't work the first time, then you're dead.

[Chiki]

There's nothing wrong with giving metaphysical examples to see the limits of his definition--so we don't have to change our criteria simply with the advent of a new game in which it does happen. And it does happen in practice. Jill in 3-13 will never die against Ike but she needs to proc Stun to kill him.

[Snowy_One]

That would be a strange restriction Snowy. For example, can you picture a tier list of FE DS H5 that assumes that we turn Marth into a ORKO machine, in a mode where you generally need several units to gang up on an enemy (or effective weaponry)? That'd be a nonsensical way of playing the game, though having a Marth that's capable of combat obviously pays back in the final map.

If you know what you need to beat the game, you needn't do any more training or, God forbid, grinding than is necessary.

Not really. The reason I'm assuming '1RKO' is because this is basically what would happen if the lord was moving as fast as possible unassisted and was not in any real danger. A simpler way to state it would be 'brisk is basically what happens when the lord moves as fast as possible to clear the game without any enemy obstructions or assistance from other units'.