Serenes Tiering

[MrSmokestack]

1 hour ago, MaskedAmpharos said:

Or are we going to look at each unit in a neutral natured and un-merged (ignoring the possibility of future merges) state to compare all units on a level playing field?

Unmerged with neutral natures. That way, free units like Lloyd and Sharena can't be rated lower simply because they can't be optimized like paid units. Trust me, even if that argument doesn't come into play, we would still be rating units based on different metrics, which isn't fair.

Also, there is no guarantee as to what units a player can obtain. The units being rated should be able to work at a baseline level--a la average stats for units in other games; for example, if Setsuna *needs* +Atk to function, but another archer with neutral stats can run her set better, then she would be rated lower.

[XRay]

On 6/19/2017 at 9:45 AM, MrSmokestack said:

Tier lists are best constructed using a metric--usage stats--that do not currently exist in an accessible form in Heroes. Involving the community in building a tier list is perfect fuel for flame wars and endless arguing that will divide more than it brings together.

 

On 6/19/2017 at 9:56 AM, Xaos Steel Wing said:

If anything, wouldn't win rates would be the statistic you want?

What about only getting data from First Dragons who survived Tier 20 for at least one week? Additionally, we can place further restrictions such as getting data only from teams that can earn a deathless run by themselves. The sample size is certainly very small right now, but once T20 has more players, our sample size will be more reliable and accurate. Additionally, another benefit is that First Dragons who survive multiple weeks are also more likely to be whales, and whales are less affected by the bias against characters locked at five stars.

Flame wars cannot happen if we remove subjectivity. Either your team won a deathless run or it did not. There is no way to check if a player is telling the truth though besides making them record their run, but I think that maybe far more hassle than it is worth and make our already small sample size smaller.

We can also do a similar list for defense teams. I think it would be even easier since we only care about the player being a First Dragon since we do not care about what their team is, but what the opponent defense teams are.

[MrSmokestack]

4 minutes ago, XRay said:

What about only getting data from First Dragons who survived Tier 20 for at least one week? Additionally, we can place further restrictions such as getting data only from teams that can earn a deathless run by themselves. The sample size is certainly very small right now, but once T20 has more players, our sample size will be more reliable and accurate. Additionally, another benefit is that First Dragons who survive multiple weeks are also more likely to be whales, and whales are less affected by the bias against characters locked at five stars.

Flame wars cannot happen if we remove subjectivity. Either your team won a deathless run or it did not. There is no way to check if a player is telling the truth though besides making them record their run, but I think that maybe far more hassle than it is worth and make our already small sample size smaller.

We can also do a similar list for defense teams. I think it would be even easier since we only care about the player being a First Dragon since we do not care about what their team is, but what the opponent defense teams are.

That doesn't resolve the issue of usage stats being difficult of access. What you are suggesting is precisely the definition of self-reporting, which should be avoided for reasons you and I have both mentioned: small sample size and unreliable data collection. How else do you propose we collect this information? The Fire Emblem fanbase is by no means Serenes-centric.

Flaming and trolling aren't likely to happen at this stage since the discussion thus far has been pretty civil and we're still talking in terms of hypotheticals and what ifs. However, if (and this is a big "if") this project continues the chances of these events occuring will continue to rise.

Your scoring range affects your rewards tier as well as the teams you face, but your tier is independent of the other two variables. Because of this, teams you face in Tier 20 are not necessarily going to be strong or consistent; Galeforce and Aether are objectively poor skill choices though they offer more points for scoring because of their SP. In addition, there is no way currently to check the legitimacy of a defense win since you don't know anything about the team yours won against, whether they played poorly or surrendered out of pity. In fact, you could even argue that the Arena Defense meta has no real value since your team only needs to win once in a given week to recieve the maximum reward, unless you like racking up wins to grow your e-peen. It also constantly changes with the bonus unit every season.

[Ice Dragon]

4 minutes ago, MrSmokestack said:

Unmerged with neutral natures. That way, free units like Lloyd and Sharena can't be rated lower simply because they can't be optimized like paid units. Trust me, even if that argument doesn't come into play, we would still be rating units based on different metrics, which isn't fair.

Also, there is no guarantee as to what units a player can obtain. The units being rated should be able to work at a baseline level--a la average stats for units in other games; for example, if Setsuna *needs* +Atk to function, but another archer with neutral stats can run her set better, then she would be rated lower.

I disagree.

Your argument for why non-neutral natures should not be taking into consideration is no different than asking for availability to be taken into consideration for one aspect of a character's viability. Why should natures be treated any differently than, say, skill inheritance sources?

In a tier list, a rating should represent a performance ceiling for that character under a specific set of circumstances. If you're not going to take into account non-neutral natures, that's a circumstance you have set and you should not take them into account at all. If Setsuna needs +Atk to function and you've taken that into account in your tiering, you're no longer judging Setsuna on the same criteria as everyone else. This situation is rectified by either re-judging Setsuna using only neutral stats or re-judging everyone else with while accounting for non-neutral natures.

That brings me up to my next point. Non-neutral natures are virtually unavoidable. A full 95% of all summoned units have a non-neutral nature. The set of circumstances you have set for this tier list do not apply at all to 95% of summoned units. If by chance the player pulls a unit with an optimal nature, the tier list is meaningless to them. Near-optimal natures account for about 14% of all summoned units (for most units, one of [+Atk, =Spd] or [+Spd, =Atk] is near-optimal, which corresponds to 3 out of 21 natures each), which is by no means a low probability. What this means is that the performance ceiling you're using for the tier list is easily breached, which makes the tier list not useful.

I'll make the argument that the first sufficiently solid ceiling is with optimal natures and no merges. Optimal natures is rigid because there's no possibility of a player pulling a character with a nature better than the optimal nature. Sub-optimal natures are easily understood that they fall somewhere below the presented ceiling and never above the ceiling. It's better to have a soft floor and a hard ceiling than to have both a soft floor and a soft ceiling. The former sets one rigid and reliable bound; the latter says nothing.

Merges have two very specific properties that make them a good fit for being excluded on the grounds of availability and why the ceiling remains solid despite accounting for availability in this one case. The first is that the player has full control over whether or not a unit is merged. The player knows full damned well that upon merging the unit, it will be just a bit stronger than the tier list expects. The second is that the difference in stats that a single merge makes is small and is on the same scale of difference as (1) changing a sacred seal as well as (2) an opponent having a difference of one merge.

[bethany81707]

4 minutes ago, Ice Dragon said:

I'll make the argument that the first sufficiently solid ceiling is with optimal natures and no merges. Optimal natures is rigid because there's no possibility of a player pulling a character with a nature better than the optimal nature. Sub-optimal natures are easily understood that they fall somewhere below the presented ceiling and never above the ceiling. It's better to have a soft floor and a hard ceiling than to have both a soft floor and a soft ceiling. The former sets one rigid and reliable bound; the latter says nothing.

...You know, this makes quite a bit of sense.

[XRay]

1 minute ago, MrSmokestack said:

That doesn't resolve the issue of usage stats being difficult of access. What you are suggesting is precisely the definition of self-reporting, which should be avoided for reasons you and I have both mentioned: small sample size and unreliable data collection. How else do you propose we collect this information? The Fire Emblem fanbase is by no means Serenes-centric.

Flaming and trolling aren't likely to happen at this stage since the discussion thus far has been pretty civil and we're still talking in terms of hypotheticals and what ifs. However, if (and this is a big "if") this project continues the chances of these events occuring will continue to rise.

Your scoring range affects your rewards tier as well as the teams you face, but your tier is independent of the other two variables. Because of this, teams you face in Tier 20 are not necessarily going to be strong or consistent; Galeforce and Aether are objectively poor skill choices though they offer more points for scoring because of their SP. In addition, there is no way currently to check the legitimacy of a defense win since you don't know anything about the team yours won against, whether they played poorly or surrendered out of pity. In fact, you could even argue that the Arena Defense meta has no real value since your team only needs to win once in a given week to recieve the maximum reward, unless you like racking up wins to grow your e-peen. It also constantly changes with the bonus unit every season.

So, basically, we do not have the manpower nor resources to create a proper tier list. Will we ever have enough though?

Would another community, say Reddit or Gamefaqs, have enough manpower and resources to do it?

[Ice Dragon]

8 minutes ago, XRay said:

So, basically, we do not have the manpower nor resources to create a proper tier list. Will we ever have enough though?

Would another community, say Reddit or Gamefaqs, have enough manpower and resources to do it?

No, we don't have enough resources to determine usage statistics. That's relevant, but not required for creating a tier list.

[MrSmokestack]

@Ice Dragon

14% of units with optimal natures versus 86% of units with neutral or suboptimal natures. Ok.

Out of all the possible stat spreads available, neutral natures are still the only ones that are guaranteed, even if only for a few units; no other stat variance has been attached to a unit as a reward for clearing a quest. The odds of non-neutral natures are still substantially higher, yes, but there are natures besides neutral that function about the same in practice; Spd and Atk variances, moreso the former, have the most influence over any given matchup spread, and not every unit is going to have a nature that affects either or both of those.

The +Atk Setsuna wasn't intended as an actual argument. +Atk or no, Setsuna still performs objectively worse compared to other units that run the Quad set competently. Non-neutral natures are still not being taken into account here.

+Def -Res on your typical player phase offense unit, like Reinhardt for instance, is a nature that is the same as neutral for all intents and purposes--his Atk is untouched. His enemy phase is irrelevant because the only Reinhardts that get hit are the ones controlled by bad players or the AI.

If nothing else, allowing optimal natures to be taken into consideration causes the free starter and hero battle units to bottom out, because when everyone else is able to function at their full potential and they can't, they no doubt perform worse--which isn't at all an accurate representation of their effectiveness as a unit. This is the most important reason to not consider optimal natures, by far.

[XRay]

7 minutes ago, MrSmokestack said:

If nothing else, allowing optimal natures to be taken into consideration causes the free starter and hero battle units to bottom out, because when everyone else is able to function at their full potential and they can't, they no doubt perform worse--which isn't at all an accurate representation of their effectiveness as a unit. This is the most important reason to not consider optimal natures, by far.

I think having optimized natures makes sense though, because in reality the GHB heroes and especially the Askr Trio do bottom out. Tier lists should accurately reflect that.

[Ice Dragon]

@MrSmokestack

The biggest problem with using neutral natures as the benchmark for rating is something you still haven't addressed yet, and that's the issue of a soft ceiling. What relevant knowledge does someone gain from a tier list that tells you the midpoint of a character's viability without at all indicating how far away that is from a sufficiently hard ceiling? A hard ceiling at least says, "No matter how much you try, you can't make this character better than this."

Second, there are very real cases where the ceiling is significantly higher with a particular nature than with a neutral nature. Users of Brave weapons are the most notable, but there's one very concrete case I can present: Robin (M) in the game's first arena meta. Robin was a counter to Takumi if and only if he was +Spd. Neutral Spd was double attacked by Takumi [+Spd], making Robin [=Spd] at best a check and not a counter. With Hone Spd 3 available only from Laslow, Matthew, and Ryoma at the time, it simply wasn't a feasible to expect to be able to use buffs to make up the difference.

 

30 minutes ago, MrSmokestack said:

If nothing else, allowing optimal natures to be taken into consideration causes the free starter and hero battle units to bottom out, because when everyone else is able to function at their full potential and they can't, they no doubt perform worse--which isn't at all an accurate representation of their effectiveness as a unit. This is the most important reason to not consider optimal natures, by far.

No, they don't bottom out as long as merges aren't used as a judging criteria. Why? I can say with 100% confidence that Alfonse's optimal nature is [=].

There is literally nothing functionally different between "Zephiel is a 55/35/16/38/24 unit that is locked to a neutral nature" and "Zephiel is a 55/32/19/38/24 unit that you can only get a pretty damned sweet [+Atk, -Spd] version of".

Worded another way, saying that Alfonse would really, really, really like to be [+Atk, -Res] is no different than saying that Reinhardt would really, really, really like to be [+Atk, +Atk, +Atk, +Atk]. Neither is any more possible than the other and therefore should not even be in the realm of consideration. A stat spread that is impossible for a character cannot be optimal.

[Refa]

48 minutes ago, XRay said:

I think having optimized natures makes sense though, because in reality the GHB heroes and especially the Askr Trio do bottom out. Tier lists should accurately reflect that.

I don't get how this is true considering several of the GHB units are ridiculously good, and the rest are generally fairly decent.

[XRay]

1 minute ago, Refa said:

I don't get how this is true considering several of the GHB units are ridiculously good, and the rest are generally fairly decent.

Besides Xander, Camus, and Ursula, the rest are good, but not great. I like Legion's stat spread a lot, but I think ranged glass cannons are better than melee glass cannons.

[Refa]

Just now, XRay said:

Besides Xander, Camus, and Ursula, the rest are good, but not great. I like Legion's stat spread a lot, but I think ranged glass cannons are better than melee glass cannons.

I agree, but that doesn't really contradict what I said anyways.  For the record, I'd put Lloyd, Legion, Female Robin, and Michalis under pretty good.  Narcian is the only one that is underwhelming in my experience.

[XRay]

Just now, Refa said:

I agree, but that doesn't really contradict what I said anyways.  For the record, I'd put Lloyd, Legion, Female Robin, and Michalis under pretty good.  Narcian is the only one that is underwhelming in my experience.

I guess I will rephrase my statement. The Askr Trio and most GHB characters can be replaced by someone better.

[Beddlam]

Looking at infantry mages, I think Soren has to be the same tier as Celica as he has almost the same and even probably slightly better stats and in my experience (admittedly not at the very top tier) green is at least as good a color (if not better) for a mage. Celica has access to good exclusive tome, but -blade tome is still optimal. As has already been said: Celica, Tharja, and Katarina are essentially identical. Soren is clearly below Nino however IMO, because the three points of speed separating them actually make a huge difference in terms of doubling commonly seen units and builds. So it seems to me that the 3 red mages should be knocked down a tier to be the same tier as Soren, Bride Caeda, and Spring Lucina, Leaving Nino and Linde alone at the top of infantry mages.

Related, to players that are top tier: is there a significant difference in the usefulness of different colored blade mages? Where I am, green seems the best because you can run with Olivia and Ephriam or Ninian/Azura and Eirika and they can take on Reinhardt and Hector. I also find that their biggest weakness, red mages, are less common than the counters to the other mages.

I also think that separating into 1 range vs. 2 range is more useful than separating by move-type because it seems like comparing infantry mages to mounted mages or even archers and dagger-users is a more direct comparison then comparing them to sword/axe/lance or even dragon infantry.

[MaskedAmpharos]

Alright so 

I think part of the problem with Ice Dragon and Smokestack's disagreements is a difference in what you want the tier list to do. 

@Ice Dragon seems to see the tier list from the perspective of someone who has (or can get) a bunch of units and wants to know which ones are the best to work on. He views the list as a resource for more experienced players who have more resources at their disposal.

@MrSmokestack, on the other hand, seem to view the list from a more beginner-friendly perspective. He doesn't know if people will be able to get certain units or natures and so he just wants to give newer players a rough idea of where certain units stand in comparison to each other on average. 

Both have their merits, but they're two different types of tier lists. And from what phineas said when the thread was made, this tier list is meant to be more of a guide for newer players who don't have a lot of units or resources at their disposal, so I'm going to side with Smoke on this one.

[Rosetta]

I'm on the side of considering average-case natures, or approximately, neutral natures, as it's uncommon to get optimal natures. Most people have most units with "not bad" natures, such as boon/banes in HP, def and res. I guess it's them who need to refer to tier lists, instead of the hard-core players who have many nature variations of one character. If one has pulled a unit with optimal nature, he or she can just expect the unit to be slightly better than what the tier list shows.

28 minutes ago, Beddlam said:

Related, to players that are top tier: is there a significant difference in the usefulness of different colored blade mages? Where I am, green seems the best because you can run with Olivia and Ephriam or Ninian/Azura and Eirika and they can take on Reinhardt and Hector. I also find that their biggest weakness, red mages, are less common than the counters to the other mages.

In my impression, B tome breaker is more common than G tome breakers, and R tome breakers is rarely seen. Does this count?

[Nathair]

I'm on the side of considering neutral natures for reasons already mentioned above, but only for 5 star only units. Those are so rare that the average player will only pull one or two copies of the same character. So the chances for optimal natures are pretty slim. The picture is a different with units that are available at 3 star rarity though. Those are pulled in such a amount that average players can get a near optimal nature in the long run. 

[Ice Dragon]

3 hours ago, MaskedAmpharos said:

MrSmokestack, on the other hand, seem to view the list from a more beginner-friendly perspective. He doesn't know if people will be able to get certain units or natures and so he just wants to give newer players a rough idea of where certain units stand in comparison to each other on average. 

What value is there in knowing how well a character when it is exactly average? How is knowing the floor* for a character more useful than knowing where the ceiling is? If a player happens upon pulling a second copy of a character, they're not going to merge into the more-average character, they're going to merge up.

Furthermore, if availability of natures is the argument for not taking non-neutral natures into account, why shouldn't it be taken into account for rare skills?

Pulling a Lucina [+Spd, =Atk] from the Tempest Trial banner (Lucina is appeared in several banners at this point) at 0.37% is 1.5 times more likely than pulling an off-banner Hector (the only banner Hector has appeared on is the Golden Week banner, and I highly doubt he'll ever appear on any other banner because a Golden Week banner almost by definition is meant to be a blatant cash grab) at 0.25%, even before considering the fact that the typical player isn't going to want to sacrifice their only Hector for Distant Counter.

Finally, pulling a new merge base is virtually no different than pulling a character to supply a rare skill. Both move the character up, and an actual ceiling is useful to know because merging into a new base or teaching a character a new skill will always move the unit closer to the ceiling and never move the unit past the ceiling (noting again that merge levels don't push the ceiling up much if at all).

*Noting how often the question, "It it worth using [character] with [bad nature]?" comes up in the questions thread and how often the answer is "no", neutral is basically a floor, even if it's not a particularly solid one.

 

18 hours ago, MrSmokestack said:

14% of units with optimal natures versus 86% of units with neutral or suboptimal natures. Ok.

One thing that I forgot to comment on last night, so I'll do it now.

You missed the point of me mentioning that number. The point is that any player with a half-decently sized roster (50-100 units) will have a decent number of units with optimal natures (not to mention 3/21 is a conservative estimate). They might not all be on characters that the player wants to use, but the player will have them.

Edited June 21, 2017 by Ice Dragon