How should we handle a long-term Lunatic tier list?

[XeKr]

RE: philosophy: I'll let others more seriously debate this, but I'll just briefly say (to address a few points and to hopefully elucidate the position of some of us in this topic.): I don't think anyone wants to reiterate years of "efficiency" and turn count arguments. As far as I know, most people tend to agree that faster is better, more reliable is better. Expected turn counts combines both into a continuous metric, therefore it's one that resolves differences between characters using criteria we value. It's an arguably cleaner way to resolve turncount vs. reliability arguments which tend to crop up because they have somewhat a direct relationship. As turncount decreases, as does reliability, and vice versa. Statistics are useful (rather, that's their use), to summarize a wide range of potential occurrences (subject to constraints like fast and reliable play). There is already precedent for using statistics for average stats.

The issue some have with "brisk" play has been stated ad nasauem already. Unproductive arguments arise like “for this chapter, we go this fast, but not that fast” or “the strat must be this reliable, not that reliable” or “noooo u, that’s not efficient/brisk because I don't think it is. This is" There is a tendency to use turncount arguments when convenient for their position but not otherwise.

Redwall has already shown that Lunatic (including a few paralogues) can be completed in sub 120 turns with fairly reliable strategies (I think <40 chance of death in most cases, except C20). It's apparent that the character placement is heavily inspired by such a pace. He (and others) would probably consider that brisk. This is not an indictment of your playstyle/playlog (I apologize if it sounds aggressive/elitist), but you were at Chapter 10 then. If we consider multiple playstyles as equally important and both as "brisk", character placement can vary significantly (see: Sully). I realize some people like debating this kind of vaguer stuff, but generally I think more mathematically inclined people do not.

There are numerous other objections to be made with these present criteria as well. Including but not limited to, the criteria currently favor 70% success or death strats over 70% strats that complete with 100% success in 1 more turn. In addition, the ranking can be roughly summarized as "optimal units" > "units that act as utility/support for optimal units" > "non optimal units". This would be more troublesome for intuition if the "non optimal units" were still good fighters or w/e, but in this case they are legitimately pretty bad (HM Donnel is an apt comparison). We have this issues which may be unintuitive but they are imo less arbitrary and still preferable than trying to subjectively place each character by alternate criteria.

That's my 2 cents.

Re: Lon'qu, He barely avoids the ohko, which is important since other Non-Avatar/Frederick characters are probably 2-3hko'd. Lon'qu has significantly better avoid (overall durability) due to how weapon triangle calculations work (the earlygame is primarily an axefest) and his skills/stats. For example, I believe Lon'qu faces around 20% hit or 8% true hit when he gets a C support. He has a ~3.4% chance of dying if faced with 4 enemies and is 2hko'd. That's far better than most other characters can claim (I believe so at least, I'm willing to accept other evidence regarding this). Lon'qu also has the base Spd/growth to double and C Swords, which not many characters can actually claim. The result is his offense is among the best of the team.

EDIT: This was a lot longer than I intended. >_>

Edited June 8, 2013 by XeKr

[Chiki]

I don't think anyone wants to reiterate years of "efficiency" and turn count arguments.

Actually, no one's complained at all about the vagueness of efficiency since this year. So this is very new.

But I agree with your post.

[Redwall]

I'll type up a proper response later, but how the hell do I make new polls? If you notice the current poll, I can't figure out how to clear the votes from the other poll.

[Chiki]

There's a red "x" at the top right corner of the poll popup.

Also, you should say "somewhat brisk" for the description of SDS's tier list.

Edited June 8, 2013 by Chiki

[Titamon]

Re: Lon'qu, He barely avoids the ohko, which is important since other Non-Avatar/Frederick characters are probably 2-3hko'd. Lon'qu has significantly better avoid (overall durability) due to how weapon triangle calculations work (the earlygame is primarily an axefest) and his skills/stats. For example, I believe Lon'qu faces around 20% hit or 8% true hit when he gets a C support. He has a ~3.4% chance of dying if faced with 4 enemies and is 2hko'd. That's far better than most other characters can claim (I believe so at least, I'm willing to accept other evidence regarding this). Lon'qu also has the base Spd/growth to double and C Swords, which not many characters can actually claim. The result is his offense is among the best of the team.

You've pretty much nailed it. Aside from Fredrick and Avatar, everyone is 2-rounded/1-rounded in the 1st part of the game. Lon'qu can use his naturally high avoid to seek EP exposure when the others can't.

[Redwall]

slightly unrelated to the discussion at hand but:

HEY GAIZ I GOT IT

I think I've figured out a self-consistent method of accounting for the chance of a Game Over: we simply add the turns spent in that unsuccessful clear to the turncount of a successful clear.

Let 0 <= x <= 1 be the chance of a clear without a Game Over.

Let 0 <= 1-x <= 1 be the chance of a clear with a Game Over.

Let t be the conditional mean turncount, given no Game Over occurs (probability x).

Let 0 < k <= t be the turn on which a Game Over occurs (probability 1-x).

<true TC> = xt + (1-x)k +(1-x)*[xt + (1-x)k] + (1-x)^2 * [xt + (1-x)k] + ...

= [xt + (1-x)k] +(1-x)*[xt + (1-x)k] + (1-x)^2 * [xt + (1-x)k] + ...

= [xt + (1-x)k] * {1 + (1-x) + (1-x)^2 + ...}

= [xt + (1-x)k]/x

= t + k(1-x)/x

Man, I feel silly for not thinking of this sooner. The reader is encouraged to think about the physical meaning of this so-called <true TC> when x = 0, 0.5, and 1.

Again, I'll post a proper response to Red Fox later.

Edited June 8, 2013 by Redwall

[Florete]

The issue some have with "brisk" play has been stated ad nasauem already. Unproductive arguments arise like “for this chapter, we go this fast, but not that fast” or “the strat must be this reliable, not that reliable” or “noooo u, that’s not efficient/brisk because I don't think it is. This is" There is a tendency to use turncount arguments when convenient for their position but not otherwise.

Thing is, we operated fine under our system for years. The RD HM list got almost 10,000 posts on it. Clearly we all understood approximately how fast we were going, so we never bothered to think of 'exactly' how fast we were going. Then people came in and wanted concrete details, even down to specific turn counts, and we neither had an answer nor understood why we needed one.

However, I don't think there was ever an upper limit for reliability...

Redwall has already shown that Lunatic (including a few paralogues) can be completed in sub 120 turns with fairly reliable strategies (I think <40 chance of death in most cases, except C20). It's apparent that the character placement is heavily inspired by such a pace. He (and others) would probably consider that brisk. This is not an indictment of your playstyle/playlog (I apologize if it sounds aggressive/elitist), but you were at Chapter 10 then. If we consider multiple playstyles as equally important and both as "brisk", character placement can vary significantly (see: Sully). I realize some people like debating this kind of vaguer stuff, but generally I think more mathematically inclined people do not.

One thing I notice is that his playlog seems to rely on pretty specific characters and placement. This is great for absolute LTC and drafts, but it's not what we like in our tier lists because it doesn't allow a whole lot of flexibility in who the player can use. Basically, we don't like making a tier list assuming we always use the same basic team with the same strategies allowing for little variation to put others in.

You're not sounding elitist, but I will note that I already stated my run could/should have been faster in the early maps, specifically Prologue, 2, 3, 5, and Paralogue 1, the last of which should just be thrown out because <_<. Those can all probably be done 3-5 turns faster in our tier setting with better planning, and it's after those that I stopped needing to be so careful, so my turns improved. I also only used Rescue on the maps that are annoying and a lot easier to get away with little planning because I wanted to demonstrate not using Rescue-skips as much as possible.

Re: Lon'qu, He barely avoids the ohko, which is important since other Non-Avatar/Frederick characters are probably 2-3hko'd. Lon'qu has significantly better avoid (overall durability) due to how weapon triangle calculations work (the earlygame is primarily an axefest) and his skills/stats. For example, I believe Lon'qu faces around 20% hit or 8% true hit when he gets a C support. He has a ~3.4% chance of dying if faced with 4 enemies and is 2hko'd. That's far better than most other characters can claim (I believe so at least, I'm willing to accept other evidence regarding this). Lon'qu also has the base Spd/growth to double and C Swords, which not many characters can actually claim. The result is his offense is among the best of the team.

When he's fighting Axe users, fine, but I usually find Axe users are never alone. In Ch 5, there are Myrmidons and Mages wherever Fighters are. Ch 6 is a bit nicer depending on where you put him, but he's still at risk of getting knocked if he kills something unexpectedly. And, I don't remember the enemy atk values in Ch 5+, but it's already been pointed out that Sully can be OHKOd in Ch 2; Lon'qu has the same defensive bases three maps later. Who is his support partner such that this doesn't happen?

EDIT: I fixed the poll for you. I think...I can vote now (it was telling me I needed to pick an option for all questions before), but it still says over 20 members have cast votes.

Edited June 8, 2013 by Red Fox of Fire

[Titamon]

Axemen are the only enemies that OHKO Lon'qu in Ch. 5, but standing on a forest tile gives him enough defense to be 2HKO'd. Alternatively you can give him a defense tonic. Regardless, 1 HP proc is enough stop this from happening. The magic users in Ch. 5 are primarily Dark Mages with bad hit, and some Elfire!Mages up the cliff, that harass everybody. Myrm's 2-shot him, but surprisingly their hit against him is kinda shaky (67ish hit on a forest tile If I recall).

[Blademaster!]

Personally, I would put Anna above Libra and maybe Oliva because of her usefulness with Levin Swords, since I haven't really ever found Libra to be as useful both offensively and defensively as well.

Edited June 8, 2013 by Blademaster!

[XeKr]

slightly unrelated to the discussion at hand but:

HEY GAIZ I GOT IT

I think I've figured out a self-consistent method of accounting for the chance of a Game Over: we simply add the turns spent in that unsuccessful clear to the turncount of a successful clear.

Let 0 <= x <= 1 be the chance of a clear without a Game Over.

Let 0 <= 1-x <= 1 be the chance of a clear with a Game Over.

Let t be the conditional mean turncount, given no Game Over occurs (probability x).

Let 0 < k <= t be the turn on which a Game Over occurs (probability 1-x).

<true TC> = xt + (1-x)k +(1-x)*[xt + (1-x)k] + (1-x)^2 * [xt + (1-x)k] + ...

= [xt + (1-x)k] +(1-x)*[xt + (1-x)k] + (1-x)^2 * [xt + (1-x)k] + ...

= [xt + (1-x)k] * {1 + (1-x) + (1-x)^2 + ...}

= [xt + (1-x)k]/x

= t + k(1-x)/x

Man, I feel silly for not thinking of this sooner. The reader is encouraged to think about the physical meaning of this so-called <true TC> when x = 0, 0.5, and 1.

Again, I'll post a proper response to Red Fox later.

I require Matlab or R plots for verification

@Red Fox of Fire: I'm getting the impression that we don't really disagree that much. Just that for me, stricter standards are more theorectically satisfying (there's a certain elegance to optimized strats) than lenient ones, when available. For example, if you could complete earlygame chapters 3-5 turns faster, why not 4-6 turns faster? Alternatively, why 3-5 turns faster and not 2-4 turns faster? etcetc

I'll likely address your points in more detail later.

Edited June 9, 2013 by XeKr

[Redwall]

I initially would not restrict Rescue at all, but I don't think this has as large an effect as made out to be. From my experience, the only maps significantly sped up by Rescue are 19, 22, and 25. Potentially 18 as well, but in my run (which didn't have Galeforce) I found it very difficult to find a 2-turn clear and would not rely on it in a tier setting. It can speed up plenty of maps, but most others have too much enemy density to so easily warp your way through.

So I assume you'd restrict Galeforce? That would make sense, since the presence of Galeforcer would force every non-Galeforcer to deal with table scraps for EXP. As for Rescue, it depends on what you mean by large.

C16: With Rescue-spamming, you don't need Galeforce for a one-turn clear; then again, you're likely to clear the level quickly regardless because of the reinforcements.

C17: With Rescue-spamming, you still don't need Galeforce for a one-turn clear. Related question: in your conception of a briskly played Lunatic run, would we stick around for the Boots? In his tier list, SDS has been rather vague about whether or not we should skip the C11 Speedwings.

C21: Even without Galeforce, Rescue allows you to end the level reasonably quickly (3-5 turns).

I'd expect that, in these three levels, a non-Rescue clear with a "pretty balanced but not necessarily super-optimized" team would probably take around three or four-ish extra turns on average compared to the fastest clear with Rescue. As I will argue later, those three or four turns can make a difference; enemy-phase combat during these turns will be plentiful, and can alter the rankings of specific characters.

We are never going to have a concrete answer for that question. However, in my logged run, my aim was for the kind of speed I see a "brisk pace" tier list being, potentially a bit faster in the early maps since I was being pretty careful and not allowing anyone to die (which would make Ch 2 faster).

However, I don't think concrete criteria should necessarily be our aim.

Thing is, we operated fine under our system for years. The RD HM list got almost 10,000 posts on it. Clearly we all understood approximately how fast we were going, so we never bothered to think of 'exactly' how fast we were going. Then people came in and wanted concrete details, even down to specific turn counts, and we neither had an answer nor understood why we needed one.

However, I don't think there was ever an upper limit for reliability...

While your approach avoids many empirical problems (for example, no one disputes that Marcus is top tier in FE7), there are cases at the margins where rankings can change sharply in response to ostensibly small changes in turncounts. For example, Nowi is entirely dependent on an EXP dump during Paralogue 4; if we're LTCing or even just playing "pretty quickly" with Sumia, Cordelia, and the Avatar, we can complete the level before Nowi finishes her lunch, which has the consequence of making her too slow to double the C12 enemies. On the other hand, if we slow down and take six turns (instead of two or three), Nowi becomes a badass capable of lasting into the endgame. For this reason, I think it is preferable to establish clear definitions when possible.

I suspect that making a stringent reliability limit can have the effect of making more characters competitive long-term units since the player will be forced to slow down.

Another reason I don't like the extreme turn-centric approach is because of the point in time we're at. Lunatic hasn't exactly been "mastered," so to speak, by the community. Many people still don't play it much and plenty even it consider it too much to tier. I disagree with that last part, but I do think the list should start out a bit more lax while allowing people to test out multiple teams and units, find strategies, etc.

I have pointed out that, for example, the player can simply eschew the Avatar and, from there, optimize efficiency under this constraint. No tier list I have seen restricts the player to only the most optimal team, and this one is no exception.

Lastly, regardless of where the list goes, I still think it's completely ridiculous to have Chrom so far above Sully given their statistical similarities. This could just mean Chrom needs to go down as he's never really that great.

Under the definition of efficiency I've adopted, I simply cannot do that. If you want Chrom to go down, it will have to wait for this brisk-play tier list you are planning.

EDIT: I also just plain don't get Lon'qu. Has anyone noticed that he has the same defensive bases as Sully, only he comes four chapters later? I've had trouble using this guy in Hard mode...

A Defense Tonic allows him to just barely survive the P1 and C5 Steel Axe guys. In C5, you can also hide him in a forest tile and have him go nuts during the enemy phase. In C6, he can tank Thieves (lol) with a Kellam Pair Up (lol).

When he's fighting Axe users, fine, but I usually find Axe users are never alone. In Ch 5, there are Myrmidons and Mages wherever Fighters are. Ch 6 is a bit nicer depending on where you put him, but he's still at risk of getting knocked if he kills something unexpectedly. And, I don't remember the enemy atk values in Ch 5+, but it's already been pointed out that Sully can be OHKOd in Ch 2; Lon'qu has the same defensive bases three maps later. Who is his support partner such that this doesn't happen?

Chrom should be his support partner IMO to allow him to double, and like I said, Def Tonic saves Lon'qu from a OHKO. This allows him to become an offensive focal point with no investment

Remember that with a Str Tonic and Killing Edge, Sully and Lon'qu only need either a single crit or Dual Attack to ORKO Barbarians. What are the odds of this happening? Assume Sully has gained five levels (I am being way generous here), giving her a 68.3% chance of doubling when given a Lon'qu Pair Up (no C support just yet); that both Sully and Lon'qu have a 30% chance to crit on any given KE attack; and that the Dual Attack likelihood on any given hit is 35% with a Chrom support and 25% with a Lon'qu support.

Lv 7 Sully (Lon'qu Pair Up, no supports): 0.683*(1 - [0.7*0.75]^2) + 0.317*0.3 = 63.67%

base Lon'qu (Chrom Pair Up): 1 - (0.7*0.65)^2 = 79.29%

Even with five levels under her belt, Sully will ORKO Barbarians at a lower rate than will Lon'qu.

EDIT: I fixed the poll for you. I think...I can vote now (it was telling me I needed to pick an option for all questions before), but it still says over 20 members have cast votes.

Thanks. I think there were three votes before you for this specific question (and not the one from the other day): two people voted yes, and one person voted for other. Is there any way to recover those votes?

Edited June 8, 2013 by Redwall

[Florete]

I accidentally closed the tab while writing a response, fuck. Gonna cut unnecessary stuff and try to go fast.

@Red Fox of Fire: I'm getting the impression that we don't really disagree that much. Just that for me, stricter standards are more theorectically satisfying (there's a certain mathematical elegance to optimized strats) than lenient ones, when available. For example, if you could complete earlygame chapters 3-5 turns faster, why not 4-6 turns faster? Alternatively, why 3-5 turns faster and not 2-4 turns faster? etcetc

There's nothing wrong with it. There was nothing special about 3-5 turns specifically.

So I assume you'd restrict Galeforce? That would make sense, since the presence of Galeforcer would force every non-Galeforcer to deal with table scraps for EXP. As for Rescue, it depends on what you mean by large.

C16: With Rescue-spamming, you don't need Galeforce for a one-turn clear; then again, you're likely to clear the level quickly regardless because of the reinforcements.
C17: With Rescue-spamming, you still don't need Galeforce for a one-turn clear. Related question: in your conception of a briskly played Lunatic run, would we stick around for the Boots? In his tier list, SDS has been rather vague about whether or not we should skip the C11 Speedwings.
C21: Even without Galeforce, Rescue allows you to end the level reasonably quickly (3-5 turns).

I'd expect that, in these three levels, a non-Rescue clear with a "pretty balanced but not necessarily super-optimized" team would probably take around three or four-ish extra turns on average compared to the fastest clear with Rescue. As I will argue later, those three or four turns can make a difference; enemy-phase combat during these turns will be plentiful, and can alter the rankings of specific characters.

Yes, perhaps restricting Rescue, and maybe Galeforce, would be a good idea.

While your approach avoids many empirical problems (for example, no one disputes that Marcus is top tier in FE7), there are cases at the margins where rankings can change sharply in response to ostensibly small changes in turncounts. For example, Nowi is entirely dependent on an EXP dump during Paralogue 4; if we're LTCing or even just playing "pretty quickly" with Sumia, Cordelia, and the Avatar, we can complete the level before Nowi finishes her lunch, which has the consequence of making her too slow to double the C12 enemies. On the other hand, if we slow down and take six turns (instead of two or three), Nowi becomes a badass capable of lasting into the endgame. For this reason, I think it is preferable to establish clear definitions when possible.

I suspect that making a stringent reliability limit can have the effect of making more characters competitive long-term units since the player will be forced to slow down.

Yes. Units like Donnel will still be bottom of the barrel, but units like Sully and Nowi who just need a turn or two over the bare minimum will be able to shine.

I don't know about Sumia, though, since she kinda sucked in my run, Galeforce or no, as she didn't reach --/15 for a while and Galeforce relies on killing stuff, something she just isn't good at.

I have pointed out that, for example, the player can simply eschew the Avatar and, from there, optimize efficiency under this constraint. No tier list I have seen restricts the player to only the most optimal team, and this one is no exception.

It does seem restrictive in a sense, though. In top and S tier, only one unit is a long-term first gen reliable combat unit. A tier has 1-2 depending on how you look at it, B is a mix of utility and tanking, and C is everyone else. The inference is that the list relies on pretty specific strategies because most units are where they are in the higher areas for a very specific role.

Under the definition of efficiency I've adopted, I simply cannot do that. If you want Chrom to go down, it will have to wait for this brisk-play tier list you are planning.

I must ask: what is it that puts Chrom so high for you at all? His combat isn't that special and he has no special utility aside from a decent Pair Up bonus. To start, how is he above Cordelia?

I do not think I am "planning" any list.

Remember that with a Str Tonic and Killing Edge, Sully and Lon'qu only need either a single crit or Dual Attack to ORKO Barbarians. What are the odds of this happening? Assume Sully has gained five levels (I am being way generous here), giving her a 68.3% chance of doubling when given a Lon'qu Pair Up (no C support just yet); that both Sully and Lon'qu have a 30% chance to crit on any given KE attack; and that the Dual Attack likelihood on any given hit is 35% with a Chrom support and 25% with a Lon'qu support.

Lv 7 Sully (Lon'qu Pair Up, no supports): 0.683*(1 - [0.7*0.75]^2) + 0.317*0.3 = 63.67%

base Lon'qu (Chrom Pair Up): 1 - (0.7*0.65)^2 = 79.29%

I don't know what the relevance of all this. It doesn't surprise me that Lon'qu has a higher chance to crit kill enemies upon joining. I was talking about durability.

Thanks. I think there were three votes before you for this specific question (and not the one from the other day): two people voted yes, and one person voted for other. Is there any way to recover those votes?

They'll just have to re-vote.

[-Cynthia-]

I believe Chrom is as high as he is because he significantly improves the turns/reliability of the final chapter. His early Pair Up bonuses also seem to have a variety of uses. He's not as high as he is because his actual combat is all that great in most situations.

[Scarlet]

Just popping in to say that I think too much credit is given to characters for just the pair-up boni they give. If we're at a point where Sumia is in S Tier merely for her pair-up boni and flying, we're really tiering the classes more than the characters themselves. If Sumia were, say, a cavalier with the exact same bases and growths instead of a pegasus knight, she'd probably be in C Tier, wouldn't she?

Edited June 9, 2013 by Scarlet

[Redwall]

Yes. Units like Donnel will still be bottom of the barrel, but units like Sully and Nowi who just need a turn or two over the bare minimum will be able to shine.

My mistake, I didn't specify my example was drawn from HM. I would be surprised if anyone got Sully or Nowi going with a mere turn or two of investment in each Lunatic level.

It does seem restrictive in a sense, though. In top and S tier, only one unit is a long-term first gen reliable combat unit. A tier has 1-2 depending on how you look at it, B is a mix of utility and tanking, and C is everyone else. The inference is that the list relies on pretty specific strategies because most units are where they are in the higher areas for a very specific role.

Here's how I see it: the B-tier units do their jobs (ferrying, thieving, etc.) whether you're training the Avatar or whether you've eschewed the Avatar and have instead chosen to train Stahl; getting money helps cut expected turns, as does ferrying staffbots to good squares. You can't say the same about most of the C-tier units, who won't contribute much without receiving EXP. Again, I do make an effort to give due credit to the C-tier guys for being useful when fewer good units are available; notice the high positions of Nowi and Sully, for example.

I must ask: what is it that puts Chrom so high for you at all? His combat isn't that special and he has no special utility aside from a decent Pair Up bonus. To start, how is he above Cordelia?

Cynthia already answered it. Chrom isn't strictly necessary to defeat Grima, which is why I haven't put him in some sort of special tier, but he is extremely helpful in doing so.

I do not think I am "planning" any list.

Sorry, that was a bit presumptuous of me. I'd personally like to see one, either stewarded by you or someone else, even if it abides by criteria I don't personally like, since arguing about characters is still kind of fun at the end of the day; but obviously it's totally up to you if you want to manage one or not, and I guess also up to the community if they want to take part in it.

Should I change the topic title to something like "What should we do about a long-term Lunatic tier list?" It doesn't seem like many people have voted.

I don't know what the relevance of all this. It doesn't surprise me that Lon'qu has a higher chance to crit kill enemies upon joining. I was talking about durability.

I was thinking of your claim that Lon'qu wasn't entitled to the KE when I wrote that.

Just popping in to say that I think too much credit is given to characters for just the pair-up boni they give. If we're at a point where Sumia is in S Tier merely for her pair-up boni and flying, we're really tiering the classes more than the characters themselves.

Classes are partly what define the usefulness of a character. Also note that combat by itself has nothing to do with tier lists.

If Sumia were, say, a cavalier with the exact same bases and growths instead of a pegasus knight, she'd probably be in C Tier, wouldn't she?

I don't see a problem with that. If you abide by the current criteria, Sumia needs to be ranked highly because she shaves a lot of turns.

Edited June 9, 2013 by Redwall

[XeKr]

Thing is, we operated fine under our system for years. The RD HM list got almost 10,000 posts on it. Clearly we all understood approximately how fast we were going, so we never bothered to think of 'exactly' how fast we were going. Then people came in and wanted concrete details, even down to specific turn counts, and we neither had an answer nor understood why we needed one.

However, I don't think there was ever an upper limit for reliability...

It seems unsatisfying to not be able to provide those details, if requested. It seems fair for people to expect those answers, if desired. Maybe that's just me.

I was referring to a statement in the vein of "The strat should be >80% reliable to be considered reliable, not >50% reliable”. Where an arbitrary cutoff is can drastically influence analyses.

One thing I notice is that his playlog seems to rely on pretty specific characters and placement. This is great for absolute LTC and drafts, but it's not what we like in our tier lists because it doesn't allow a whole lot of flexibility in who the player can use. Basically, we don't like making a tier list assuming we always use the same basic team with the same strategies allowing for little variation to put others in.

I think this a consequence of the metric. Trying to go super fast in Lunatic is inevitably restrictive. This is intended, in order to resolve differences between characters. There are the optimal characters, then slightly less optimal ones, etcetc in a gradient until the characters are all ranked. I already noted that in this case it makes the ranking “optimal” > “utility” > “non-optimal in a gradient”, which some may see as unfortunate. This is a matter of tiering philosophy and the current criteria. It's something hard to resolve cleanly, but this way is at least self-consistent.

Going slower results in more variable conclusions. Nowi is probably the case study for this. Why can we slow down enough for Nowi to shine but not slow enough for Donnel to shine? Or anyone, because everyone in this game is good when trained.

I didn’t mean to imply that playlog is the only evidence we should consider. I just think it’s fairly compelling compared to other evidence we currently have for determining distinctions in Lunatic.

There's nothing wrong with it. There was nothing special about 3-5 turns specifically.

It seems like it’s intuitive to go “somewhat briskly” and not whatever-who-cares pace. I suppose my point is then, why not just tend toward the minima of expected turn count (given some team composition)? I don’t think anyone anticipates calculating the exact values here all the time either. It merely sets up more rigorous framework for those tricky-to-judge cases that intuition cannot easily resolve.

In Lunatic, I do not think a single turn (or 10) salvages Sully and Nowi.

Re: Lon’qu: I think the point is, Lon’qu barely avoids the ohko. Other characters are still solidly 2hko’d or such. Axe users are the most common and hit the hardest. Avoiding all damage from them with high probability is a huge boon earlygame. I’d welcome evidence that Sully or someone can replicate his role.

Re: Chrom. I originally thought he was too high, but his Spd Pair up is actually amazing for Frederick before Sumia, and then Lon’qu or Avatar or someone. It is extremely hard to double without +Spd. He apparently improves Endgame by a bunch, turncount and reliability wise (I’m theorizing Lucina can do it better, but we’ll ignore that for now) .

I wouldn’t consider it the same way as Premonition because it is surprisingly complicated to get Chrom strong enough. It requires substantial planning throughout the game. It’s not really a interactive cutscene like Premonition. And there are alternate ways such as Braves+Ignis/Luna (requires more exp though).

@Scarlet: If Seth was an armor knight, would FE8 be known as The Sacred Seth?

Edited June 9, 2013 by XeKr

[Scarlet]

Classes are partly what define the usefulness of a character. Also note that combat by itself has nothing to do with tier lists.

I don't see a problem with that. If you abide by the current criteria, Sumia needs to be ranked highly because she shaves a lot of turns.

I'm aware that this abides by the current criteria of the list, it just... ...I dunno, strikes me as undesirable.

Especially considering that pretty much all of her turn-shaving entirely relies on making Frederick better - if Frederick isn't used as a serious combatant past the stuff he's absolutely required for, she's not very useful at all.

I'm not actually trying to make a serious argument out of this, just... ...I don't even really know myself. I just had to say it. Don't mind me.

@XeKr: Probably not, I know. Don't mind me.

[Florete]

My mistake, I didn't specify my example was drawn from HM. I would be surprised if anyone got Sully or Nowi going with a mere turn or two of investment in each Lunatic level.

My idea wasn't that you would randomly take a turn or two extra, but that the clear strategy in general would take a bit longer because you aren't doing the fastest strategies. Having looked at your log, if Sully were used and every map took 2 more turns, I couldn't see using her being a significant problem. But for now, that is really neither here nor there.

Cynthia already answered it. Chrom isn't strictly necessary to defeat Grima, which is why I haven't put him in some sort of special tier, but he is extremely helpful in doing so.

Well, I've already explained my dislike of the strict focus on turns, which is where this comes from. This is the same logic that would auto-top Edward and Black Knight in RD. I'm pretty sure even Aeine doesn't submit to this logic because of the complexity factor. I would sooner judge Chrom based on how he performs throughout the game as a whole, which on this list would probably put him in B tier.

Sorry, that was a bit presumptuous of me. I'd personally like to see one, either stewarded by you or someone else, even if it abides by criteria I don't personally like, since arguing about characters is still kind of fun at the end of the day; but obviously it's totally up to you if you want to manage one or not, and I guess also up to the community if they want to take part in it.

I guess I could start it and see what happens. I just am not sure if I like the idea of six tier lists running around.

Should I change the topic title to something like "What should we do about a long-term Lunatic tier list?" It doesn't seem like many people have voted.

That could help.

It seems unsatisfying to not be able to provide those details, if requested. It seems fair for people to expect those answers, if desired. Maybe that's just me.

I understand. I didn't mean to say we shouldn't have to provide those details, just that the way things have gone, it's a bit difficult for us to simply explain. This might not be a great example, but it's the first thing that comes to mind: when I was learning to drive, I kept failing the test while only my parents were teaching me. I finally passed when my sister helped me. We came to the conclusion that my parents had just done it for so long that what they did was ingrained in them and they didn't actually know what it took to pass the test.

Going slower results in more variable conclusions. Nowi is probably the case study for this. Why can we slow down enough for Nowi to shine but not slow enough for Donnel to shine? Or anyone, because everyone in this game is good when trained.

We can slow down for Donnel. It's just that the amount of slowing down required isn't worth it for the payoff.

It seems like it’s intuitive to go “somewhat briskly” and not whatever-who-cares pace. I suppose my point is then, why not just tend toward the minima of expected turn count (given some team composition)? I don’t think anyone anticipates calculating the exact values here all the time either. It merely sets up more rigorous framework for those tricky-to-judge cases that intuition cannot easily resolve.

If I'm reading this right: the idea still is to tend toward lower turns. Units who can cut more turns still get ranked higher, it's just not so strict in the sense that certain units can't get any experience or that these turn cuts always happen.

[Redwall]

OK, I fixed the poll. Somehow I couldn't vote in yesterday's version.

[Espinosa]

You need to delete your old vote first if the polls have changed.

[Florete]

You need to delete your old vote first if the polls have changed.

If only that were all. I'm once again being told I need to choose an option from each question.

Edited June 9, 2013 by Red Fox of Fire

[Redwall]

If only that were all. I'm once again being told I need to choose an option from each question.

Lovely . Should I just make a new topic with the same poll?

[Integrity]

I got the voting issue and decided to mess with the poll a little bit and I'm not getting the voting issue anymore. Somebody not-me test if it works now?

EDIT: Update that to now: I think I found the issue with the poll and fixed it.

EDIT2: Judging by the votes, it seems to be ok now.

Edited June 9, 2013 by Integrity

[Florete]

I got the voting issue and decided to mess with the poll a little bit and I'm not getting the voting issue anymore. Somebody not-me test if it works now?

EDIT: Update that to now: I think I found the issue with the poll and fixed it.

EDIT2: Judging by the votes, it seems to be ok now.

Enlighten me on how you managed to pull this off.

[Integrity]

Enlighten me on how you managed to pull this off.

The first question in the poll was Question 2, which implied a vestigial first question that sort of existed and sort of didn't anymore, so I nuked the entire poll and saved it blank and then re-edited the post to include the poll again so that it would be Question 1.