Author Topic: Theorycrafting! (Because I like competitive metagames too much)  (Read 110279 times)

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4373
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #750 on: June 18, 2017, 10:36:11 PM »
And...update to the dip list too.  Moving Ninja and Mediator down as a dip as I posted about several days ago.  I mean, you can use them to enhance another build, but it's pretty reasonable to just...not.

Moving Monk up a few slots on the dip list.  HP restore is pretty legit, I gained respect for it this playthrough, bad autopotion is alright, something you can get from spillover, and also in the build path of some builds (like Samurai) and not far out of the way for others (anything that needs to unlock Geo, so Ninja, Dancer, Geomancer).  I think Monk is in competition for carrying Jump (along with Ninja) cause it has good PA and ok speed.  Sometimes Ninja wants martial arts too.

Knight (for weapon guard, sometimes Equip Shield) and Mediator (for faith raising) are notable for being things that higher tier (read: mage) builds use, so there's an argument that they should be higher, but they're pretty optional/minor.  In the case of Knight...you probably won't actually job change to Knight, so rely on someone going Knight (ew) or Samurai (eh, I guess).

Oh also, moving Summoner above Wizard on the dip list.  So many builds work because "just dip Summoner for Ramuh".  A lot of builds, when they are done learning everything, want to move to Wizard and set MAU, sure.  But by the time you are doing that you probably don't need the help.  Actually, fuck it, moving Summoner above Squire too.  Notably a Priest with Summon and no Gained JP Up, managed to be my overall most consistently strong PC this past game.

new (tentative) dip list

1   su
2   sq
3   wi
4   ch
5   tm
6   or
7   pr
8   th
9   ar
10   mo
11   ni
12   me
13   kn
14   ge
15   ca
16   sa
17   la
18   ba
19   da
20   mi

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4373
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #751 on: June 18, 2017, 10:56:38 PM »
Whew, ok, going back to respond to people now.

- How much JP will one earn over the course of a game?  Not counting spillover JP here; there's no guarantee it will be in the classes you want. 

It's...nonlinear, but about 5000?

400 or so by the end of Chapter 1.

Cumulative 1350 or so near the end of Chapter 2.

Cumulative 2400 or so near the end of Chapter 3.

Cumulative 5000 or so near the end of Chapter 4.

This is just based off rough recollections in this last playthrough of when I got to Level Jump 8, when I got to Vertical Jump 8, when I actually finished learning things in Samurai.

But the JP formula adds your class level (so sticking Lancer is not the best model).  The JP formula adds your actual level too (so class level matters less in Chapter 4 if you want a second specialization).  Also I might be overestimating Chapter 3 JP values, as obviously my Lancer was high job level through most of the chapter.

Quote
- What about JP investments to unlock classes?  Angel Song and Nameless Dance are both useful tools and only cost 100 JP but unlocking their classes takes investing over 1000 JP (after factoring in the 100-199 JP every unit starts with in every class).  1100-1200 JP is a more likely minimum commitment since the investment is unlikely to be perfectly distributed.

Yep, those are pretty real.  My rough calculation for these is that you should expect to overshoot the JP requirements a bit, taking more actions than you need.

Quote
- Crystals: is there in impact?  Uncooperative teammates are a thing, I guess.

Generally teammates will let you get appropriate crystals.  Like "oh, you're a Summoner, you should get that Summoner crystal".  Partly cause there's not much to be gained by stealing crystals (if you don't have the job unlocked you can't even get the skill).

Quote
- How many classes and/or setups can use a skill well?  Ramuh is awesome and 200 JP isn't too far out of the way though needs good MA and enough MP to make use of it so mainly the five main mage classes and maybe Geomancer will get mileage from it.  Without high PA, Punch Art come out like a cherry tap and some will argue Martial Arts is needed as well.  At the other end, Item with Potion and Phoenix Down can contribute to sandbagging at worst and Time Magic for Haste and Slow can be viably wielded by many classes.  These are just examples.

Yeah, it's definitely a consideration, for sure.  Like...you don't dip for Punch Art--if you're going Monk, you do Monk focus, and then can't even take the skillset very many places (I ended up just going back to Monk).

Quote
- Golem seems like a handy Summoner skill to have access to.  Or is it?

Oh, that's true.  Expensive (500 JP, 40 MP) but good.  Not usually a dip.

Quote
- Shield using setups like Abondon sometimes, even if it only shines in chapter 4.

Sure, yep.  That said, Abandon falls into a category of "has much steeper unlock requirements than Auto Potion, and is not as strong"--a lot like Blade Grasp.  It's something you get if you're a focus Ninja, and maybe you use a mantle or Equip Shield.  (You can still attack twice with a shield equipped as long as you put a shield in your first hand, and your second hand unarmed; you get two punches).

Quote
- Magic DefendUp is handy at Velius and a few other select places.  I guess offensive supports are more favored?

I think MDU for Velus is reasonable.  You'd probably only set it if you have a focus Priest, and managed to get it from spillover, but yeah, it's fine.

Quote
- Everyone benefits from Maintenance in the three fights with Mighty Sword using opponents.  Though Math Skillers can escape without it at Rofel 2 due to AI manipulation.

True.  250 Chemist JP, though; my Chemist JP tends to be precious (like...if I'm one of the many builds that sets item, I'd probably grab X-Potion before Maintenance).

Quote
- Pray Faith is one of those skills I never seem to want to spend JP on yet it has its uses with mage allies.

It's really good on Math Skill, since you just hit the entire battlefield with Faith.  Outside of that, though...it's a 400 JP ability in Oracle which has lots and lots and lots of good abilities.
« Last Edit: June 18, 2017, 11:07:16 PM by metroid composite »

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4373
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #752 on: June 19, 2017, 06:59:01 AM »
OK, for the focus list, the Priest/Wizard/Ninja/Chemist group...I decided to start a new run with these; 3 PCs cause it's a higher tier team.  Decided I didn't want Chemist cause they would give spillover Autopotion and that would break any semblance of challenge (indicating that yeah, Chemist is probably the best of the bunch, and maybe I should reevaluate Chemist vs Oracle).

Chapter 1 went about as you'd expect.  Wizard Good.  Priest also good.  Thief (on the way to Ninja) bad.

Chapter 2 Thief damage remained stagnant, and then Move+2 was obtained.  Move+2 combined with classes that have melee physical damage (Knight/Monk/Geo) is pretty solid.  I believe all of them were hitting close to 60 (60 for Monk 56 for Geo, I probably hit 49 with Knight cause I hadn't bought a Coral Sword yet, but with a Coral Sword would be 56 as well).

Ninja unlocked around Bariaus Valley, with Move+2 and Concentrate...and I look at my daggers (4WP and 5WP) and I unequip them since fists deal more even with Green Beret...and then I set Martial Arts instead of Concentrate.  60x2 = 120 damage.  Worth noting, however, Equip Sword would be better than Martial Arts here dealing 64x2 = 128 damage.  I'll probably try to sneak out of Ninja on easy fights to squeeze in the JP for the various competing support abilities (Equip Sword and Attack Up).  But then again, I also want to sneak out of Ninja to learn X-Potion on Chemist.

But uh yeah, 64 long range damage with balls, and 60x2 = 120 melee damage is...now a solid unit, but still third at the moment.  Wizard deals 200 with Bolt 2.  Priest is like...140 or so with Ramuh.  (Although they both have charge times and are not 8 speed units).  That being said, with all three party members having damage, Priest is actually using white magic a bunch, since someone needs to heal, and the Priest is better at it than the rest of the party.  Still busting out the Ramuh if enemies are kind enough to cluster ofc.

Playing Priest a bit more "normal" this time (using Gained JP Up, will switch out of Priest at some point and set WM secondary).

DragonKnight Zero

  • Denizen
  • *
  • Posts: 502
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #753 on: July 17, 2017, 01:36:05 AM »
More random thoughtdump because I enjoy FFT discussion that much:

Good catch with Abandon: I totally forgot about the class unlock requirements.  That was a brain fart on my part.  If one already got Move +2 and Concentrate, then sure, it's not too far off but otherwise takes a large investment.  Come to think of it, most of my Chapter 4 specials won't have Ninja or any of the other multi-class prerequisite classes unlocked in a normal, casual "anything goes" game by the end of Morund.

- If a 7 Speed unit Moves without Acting on it's first turn or vice versa, even if the 6 speed enemy moves without acting, Ramuh will still go off before its second turn.  Fun times.  My easiest Golgarand battle experience came from Priest Ramza with Magic Attack up nuking things with offensive magic.  (There was also a lucky first turn snatching of Gaf's Blood Sword, hush)

- Squire flat out sucks.  If hitting things with swords is the goal, then Geomancer does it better.  Only unique thing Squire has over Geo is Scorpion Tail jumping which, haha. (oh, and knockback but that comes into play rather rarely)

- Geomancer I feel need Rune Blade to be notable as a magic user.  So for most of the game, I find them mediocre mages.  I have run one with Time Magic secondary (and Gold Hairpin for MP) at Velius since I wanted a durable Slow and Demi caster but that's a setup for that fight specifically.  Rune Blade and Aegis Shield do outmagic Priest at certain level ranges, heh.

- As much as I value Secret Hunt normally, I suppose in this context, the main worthwhile benefits are Holy Lances (if a spear user is around) and killing undeads without giving them a chance to revive in the 3-4 battles they appear in.  Poor Thief; great speed and mobility but poor offensive stats to channel those attributes through.

- Knight gets the short straw outside of Chapter 1.  Geos are better at whacking things with swords come Chapter 3 and its abundance of PA boosting armor.  The breaks are used better by gun or bow users or Ninjas.  Being subject to Counter and getting shut down by evasion just makes things worse.

Dark Holy Elf

  • DL
  • Denizen
  • *
  • Posts: 8135
  • Well-behaved women seldom make history
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #754 on: July 17, 2017, 01:49:00 AM »
Quote
Geomancer I feel need Rune Blade to be notable as a magic user.  So for most of the game, I find them mediocre mages.  I have run one with Time Magic secondary (and Gold Hairpin for MP) at Velius since I wanted a durable Slow and Demi caster but that's a setup for that fight specifically.  Rune Blade and Aegis Shield do outmagic Priest at certain level ranges, heh.

With Rune Blade they should outmagic Priest at any level, their multiplier is just 5% worse (= 1 point at most, often 0) and Rune Blade + Aegis Shield is 3 magic to Wizard Staff's 1. Post-Rune Blade/Aegis Shield they actually hold up well compared to any non-Wizard job for MA. I agree that they're inferior at magic until then, though sometimes you do want that bulk and/or move; heck I've run Knights with magic for that reason and legit consider it an advantage they have over e.g. Monk/Archer/Thief that their magic stats are 80/80 (as opposed to something even worse) and that they can use robes.

Erwin Schrödinger will kill you like a cat in a box.
Maybe.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4373
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #755 on: July 21, 2017, 03:20:10 PM »
Yep, for the period when Rune Blade is available (last half of Chapter 4) Geos are solid users of magic.  Then again, this is also the period when Excalibur is available, if you don't have someone else using it, so sometimes Knight is a very solid option at this point too.  Like...I can certainly remember zodiac fights where I went Excalibur Knight with Life Drain, which straight up was the strongest option (better than mage classes, better than Geo).

Quote
Post-Rune Blade/Aegis Shield they actually hold up well compared to any non-Wizard job for MA.

Usually equal or sometimes 1 less MA than Summoner/Oracle/Time Mage, yeah.  They also will have something like 14-18 less MP at level 30-40, which can be pretty relevant for some specific builds.  (If your build is Short Charge Meteor, you might be out of MP after the second Meteor or even not have MP for the second Meteor so...).

(As a quick side note, if you're female and you've been going through stat growth on the mage side of the job tree, and you're equipping for magic, don't expect your physical with Rune Blade to be that exciting.  It'll deal about 84.  Which...sure, is higher than Wizard Staff which deals around 60.  Just...the main thing a mage gets out of Geo is HP, Ageis Shield evade, and move+1).

Quote
- Squire flat out sucks.  If hitting things with swords is the goal, then Geomancer does it better.  Only unique thing Squire has over Geo is Scorpion Tail jumping which, haha. (oh, and knockback but that comes into play rather rarely)

You know, I remember Dark Holy Elf did a spreadsheet of how good each class was at physical attacks at every store checkpoint, and Squire came out overall 5th.  "Worse geo" isn't exactly a dumpter fire.  It's more of a "but why?  Why would you spend time in Squire?  There's no point!"

Quote
- Knight gets the short straw outside of Chapter 1.  Geos are better at whacking things with swords come Chapter 3 and its abundance of PA boosting armor.  The breaks are used better by gun or bow users or Ninjas.  Being subject to Counter and getting shut down by evasion just makes things worse.

Knight is niche, but still manages to come up for me more often than I expect.

If I'm a mage, I want to use a shield for one fight (Balk 1 or Balk 2 probably) and I don't have the 350 Monk JP to unlock Geo (why would I?)   Guess what? Knight gets dusted off.  (Or, if I've spent a bit of time as Knight already, I might pick up Equip Shield).  Also, Reflect Mail has some neat (very niche) applications in theory.  And obviously Excalibur is good.

It's just....very very very much a dip class, and a terrible focus class.  You switch to it for one or two fights; you don't spend a whole chapter as Knight (except maybe in Chapter 1, but even there, I feel like by the end of the chapter you shouldn't have a JP incentive to be in Knight).

Dark Holy Elf

  • DL
  • Denizen
  • *
  • Posts: 8135
  • Well-behaved women seldom make history
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #756 on: July 22, 2017, 02:14:48 AM »
Quote
You know, I remember Dark Holy Elf did a spreadsheet of how good each class was at physical attacks at every store checkpoint, and Squire came out overall 5th.

*digs up the spreadsheet in question*

Seventh, actually; behind Ninja, Geomancer, Knight, Oracle, Monk, and Lancer. That said your broad point is basically correct; in particular there is a significant dropoff from #7 to #8 (which is... Mime, technically, but Mime is relatively better in chapter 1 lol so the list overrates them; otherwise Archer). And they have 4 move. They're not an AWFUL carrier just... totally outclassed by many others thanks to unimpressive stats and a terrible primary; their only niche is being a 4-move job available to units unwilling to spend any time in Monk (at which point the only other option is Thief which is a lateral move at best).


I'm always a bit leery about Excalibur hype for Knight as it often seems based on assumptions that we're playing with 5 generics (challenge runs etc.). In reality, Ramza is always in play and makes super-great use of it (its synergy with Scream is very strong). Like... if you have a build that is "mage with Excalibur" you should really just make Ramza that and get better equips/a better skillset/better MA and MP. To be sure, there are times where yes, going Knight on another build to get Excalibur is valid, but it feels very corner-case. ("Well Ramza has another build entirely AND I'm not using Orlandu or Agrias AND then Knight with Excalibur is the best setup for this particular mage/etc. in this fight.") It can happen but is worth extremely little consideration.

Erwin Schrödinger will kill you like a cat in a box.
Maybe.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4373
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #757 on: July 22, 2017, 11:20:49 AM »
So...stat values of FE Heroes explained.

Know how in some cases +Atk on say Olwen is +4 attack?  Or -HP on Robin is -4 HP?

Well, there is a stat growth table:

http://feheroes.gamepedia.com/Stat_Growth#Level_40_Stats

GP   1Icon Rarity 1.png   2Icon Rarity 2.png   3Icon Rarity 3.png   4Icon Rarity 4.png   5Icon Rarity 5.png
0   6   7   7   8   8
1   8   8   9   10   10
2   9   10   11   12   13
3   11   12   13   14   15
4   13   14   15   16   17
5   14   15   17   18   19
6   16   17   19   20   22
7   18   19   21   22   24
8   19   21   23   24   26
9   21   23   25   26   28
10   23   25   27   28   30
11   24   26   29   31   33

Basically, all characters have a stat growth value from...1-10 at neutral IVs; 0-11 with IVs.  These then supposedly use a lookup table to determine how much to increase between level 1 and level 40.

Now, if you look closely, you'll notice that 3 star is always gaps of 2 for the growth.  (IVs will change starting stats by 1 level, and growths by 1 level).  Why do 5 star units have these gaps?  Specifically the gap from 5 growth to 6 growth that a lot of people hit?

It's a multiply and round, that's why!

Formula is ROUNDUP((3starStats) x (StarMult))

Star mults are...

5 star StarMult = 10/9 = +11%
4 star StarMult = 29/28 = +3.6%
3 star StarMult = 1
2 star StarMult = 15/17 = -12%
1 star StarMult = 9/11 = -18%

4 star mult could also be 28/27.  This is ambiguous.

2 star StarMult...I honestly have to assume that there is a typo in the wiki table here, seeing as there are no IVs on 2 star units, and the 10 growth characters aren't available below 3 star.  The way the table is set up, the small growth jumps are at 1, 5, and 11.  The gap between 1 and 5 is 4.  The gap between 5 and 11 is 6.  No way to square that circle; you could alternate 4 and 5, or 5 and 6, but not 4 and 6; that is not a pattern that can be produced via multiplication and rounding.  That said, since there's no way to test the 10 growth at 2 star, I honestly suspect a typo is more likely.

You might be looking at these StarMults and thinking "wow, 3 star to 4 star is a small jump; 2 star to 3 star is a huge jump" and...that's not exactly true.  It's a ceiling function, so the jump from 3 to 4 star could be 1%, and it would still give you +1 to all stats.  (And that's just from the growths; going from 3 to 4 stars will also increase two of your starting stats--your highest two non HP stats).

I am 99% sure the calculation uses CEILING and not FLOOR or ROUND_TO_NEAREST just because of the way that 4 star growth table looks.  Although sure, obviously you can simulate any of these three functions using any of the other three by adding values before the multiply and subtracting them afterwards.  Just...I doubt it.

In practice, people care about the 5* growth point jumps, which happen at 2, 6, and 11.  The jump at 11 is pretty irrelevant (no, +HP Faye isn't a good IV.  I don't care if +HP is +4).  So...the key growth point jumps to get +4 instead of +3 are 2, and 6.  Conveniently there's a table with all growth points for all characters:

http://feheroes.gamepedia.com/Growth_Point_Table

Worth noting--Trainees have 6 more growth points, but start with lower stats (8 lower).  "Veterans" have 6 less growth points, but start with more stats (8 more).  Which means...yeah, that's right, Jagen and Gunther are actually Jagens.  They'll have 5 less BST at level 40 on average, while having 8 more BST at level 1.

If you're thinking about BST, and are looking for characters that have a 6 in every stat, look at trainees (Y.Tiki, Tobin, Nowi, Faye--the archer).

SnowFire

  • DL
  • Denizen
  • *
  • Posts: 4937
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #758 on: October 01, 2017, 02:07:17 AM »
So…  MC's old FFT drafts for hypothetical single-player-class-limited playthroughs vaguely came up again in IRC recently, and plus, in LoL, it's Worlds season, so catching up on some ESPORTS drafts there.  I did a little mini WAXF draft on the side, so…  a completely hypothetical Bravely Default draft!  (No, BD is a long enough game that hell no to ACTUALLY PLAYING one of these.)  BD does have the nice property that there's some notable synergy / anti-synergy in classes, so it's not quite "go to your in-game rankings thread and draft top to bottom."

Completely arbitrary rules:
* Everybody gets JL1-10 of Freelancer for free.  (No dopey worries about getting locked out entirely in C0-C1 from not enough early picks there, or about missing general quality-of-life passives like being reminded of how many treasure chests are around.)  You have to draft Freelancer if you want the JL11-13 stuff though, i.e. C5-8 Freelancer.
* No bans, but the last two picks go undrafted, so you aren't entirely guaranteed your pick of the chaff.  Everybody will get 11 classes this way.
* I'm assuming that you're allowed to turn off encounters for boss runs.  (Some of the MP-slurping classes get much worse if you don't, and classes like Valkyrie shoot up in value.)  I'm still assuming that a team that has major random woes is less impressive, though.


1A: Okay.  For first pick, there are basically four reasonable options IMO:
* White Mage: Early class.  Invaluable MT healing for any boss fight that goes slow (which, against some HP sponge enemies like Braev, matters), better revival than Phoenix Downs.  Even has some okay-if-MP-inefficient offense.  Resist Water for Rusalka.
* Dark Knight: The strongest class in the game?  But it's C5 & beyond only, and C0-C4 are no joke.
* Spiritmaster: Makes WM much better, but more importantly, elemental cheezing outright wins some tough battles for free (the Dragons), and makes a number of battles considerably more sane, especially for enemies that give you an elemental weakness (Bugzilla, Victor & Victoria / Ominas refights).
* Black Mage: Not at the same power tier as the above three, but it is an early class, and it's extremely good vs. Orthos (EDIT: and Heinkel).  It also unlocks C5+ mage builds with Pierce M. Def.

Let's not kid ourselves, though.  The pick is gonna be White Mage.
A: White Mage

1B: Well, looking at the above set of 4 classes, we get 2 and hand over the other.
Black Mage & Spiritmaster: Sure, go ahead and have the two best classes (DK & WM)!
Black Mage & Dark Knight: Sure, go ahead and have the WM / Spiritmaster combo!
Spiritmaster & Dark Knight: Sure, go ahead and snipe Knight / Black Mage, and subject me to some sort of horrible Monk/Thief/Freelancer challenge in C0 / C1!

None of these options are attractive.  That said, I think the DK / BM combo is the best; mass BMs can blitz out early bosses and is good for dealing with Orthos sans white magic, and DK gives the late-game kick required.

A: White Mage
B: Dark Knight, Black Mage

--
2A: Well…  Spiritmaster, obviously.  A lot of options otherwise, but Knight & Pirate come to mind.  Knight is offense early and a super-tank late that will synergize well with healing.  Pirate is better, but also C3.  Since we're missing out on magic offense, I'm thinking the Pirate debuff train is gonna be more important than usual.  So let's do that.

A: White Mage, Spiritmaster, Pirate
B: Dark Knight, Black Mage

2B: B has to start thinking about potential hatedrafts of good revival / stalling.  B also wants some more C1-C2 classes since all Black Mages tends to be very binary for bosses; either they blitz them out with offense or die horribly, and Rusalka is gonna be REALLY INTENSE with no WM.  BMs also have MP woes, so it's gonna make randoms annoying.  Knight is early and is also really good for Team A, so take that, I think.

A: White Mage, Spiritmaster, Pirate
B: Dark Knight, Black Mage, Knight, Salve-Maker

--
3A: I think A needs to grudgingly take Monk since Knight got sniped so that team A has some semblance of offense in C0/C1.  There's some potential hatedrafts A can do since B is stalling on mages & Spell Fencer, but…  eh, going back to things this team wants, Performer fits well the "buff/debuff/heal/smash" strat.  So take that.

A: White Mage, Spiritmaster, Pirate, Monk, Performer
B: Dark Knight, Black Mage, Knight, Salve-Maker

3B: Okay, B has stalled long enough at grabbing Spell Fencer.  Late game DKs are notably worse without Drain Sword, especially if B ends up running a bunch since they lack other healing, so pick that up.  B doesn't really need much more lategame classes, so…  maybe Red Mage?  Use that for healing / revival in C3 at least and pretend to have a more normal run there, maybe even C4, before switching to Medication DKs & magic.

A: White Mage, Spiritmaster, Pirate, Monk, Performer
B: Dark Knight, Black Mage, Knight, Salve-Maker, Spell Fencer, Red Mage

--
4A: Well, White Mages healing Monks is very one-dimensional for the early game, but the other options left are just…  uninspiring.  Thief has some nice Speed boosting passives?  Ranger has Targeting for Braev / other heavy Default users?  There's also sniping Time Mage, which while not great lategame without BMs, might still be useful for Gaia Gear earth magic.  Vampire offers a lategame carry and redundant Pirate support, and Templar has Default Guard and can be a psuedo-Knight.  I'm thinking the TM snipe (hey, magic damage that's better than WM's crappy wind in C2-C3) and Templar are the way to go here (Default Guard is useful on stally teams).

4B: Well losing TM is too bad, but eh, that just means going more all-in on lategame Dark Knights rather than Pierce M. Def Meteors.  I'm thinking that the worse healing means that dopey Ninja immunity stratz are a bit better, as well as maybe Kairi onto Knights?  Vampire also looks tempting with no Pirate for a lategame source of debuffs, and super level 4 magic better than Dark if you get the right wandering Norende beasties to go with Pierce M. Def.

A: White Mage, Spiritmaster, Pirate, Monk, Performer, Time Mage, Templar
B: Dark Knight, Black Mage, Knight, Salve-Maker, Spell Fencer, Red Mage, Ninja, Vampire

--
5A: Getting to cleanup now.  Sure take Thief.  Maybe Valkyrie as well?  More random clearing in late C2/C3.  This makes me feel bad about missing out on Vampire, since it's strictly better than Valk later on, but oh well.

5B: Well if we have Ninja, might as well take Swordmaster as well for that silliness.  Also pick up Arcanist for Fenrir statusing bosses hype since we have Black Mage, although by the time this really comes online, we could also be running some sort of 3x Dark Knight party.

A: White Mage, Spiritmaster, Pirate, Monk, Performer, Time Mage, Templar, Thief, Valkyrie
B: Dark Knight, Black Mage, Knight, Salve-Maker, Spell Fencer, Red Mage, Ninja, Vampire, Swordmaster, Arcanist

--
6A: I guess dopey Freelancer Mimic tricks can do some cool stuff with this set (Time Mage Slow World passives, etc.).  And sure, Merchant for some efficient money grinding and maybe Low Leverage if we get really desperate.

6B: Well none of Ranger, Summoner, or Conjurer are that important.  I could actually be convinced by CONJURER on Team A, with healing and potential MP woes, but not team B where we have Salve-Maker.  Summoner I guess, it's good for refights of Rusalka!  Ranger isn't even a bad class, it just doesn't have tons of synergy and the only boss fight I'd hype their skillset on is Braev really.  They're just consistently okay, while something that was usually terrible but occasionally great would be more useful for this kind of challenge.

A: White Mage, Spiritmaster, Pirate, Monk, Performer, Time Mage, Templar, Thief, Valkyrie, Freelancer (Greater), Merchant
B: Dark Knight, Black Mage, Knight, Salve-Maker, Spell Fencer, Red Mage, Ninja, Vampire, Swordmaster, Arcanist, Summoner
Unused: Ranger, Conjurer

--
I think A "wins" here with a very reliable stallfest featuring buffs, debuffs, healing, revival, and a bunch of good defensive passives (Angelic Ward, Default Guard, Speed +20%, etc.).  Orthos will be exciting, but the early game isn't THAT rocky, and they can sustain through a lot of fights.

Team B was kinda stuck by not being able to get WM, so they do make a good show of having hyper-offense teams that can just kill quickly and maybe only use the occasional choice revival, or Widen Area potion.  The gang of Drain Swording BK's with Rise from the Dead and See You In Hell is still a solid strat, though it might run into trouble for some of C7/C8's more brutal Asterisk team-up fights.  Early game will be exciting but if you don't have WM, having BM is the next best option IMO for maybe not needing it.  And if you're really bored, some bosses can just get plain Utsusemi cheeze'd out, so there's always that for avoiding the need for good MT healing.
« Last Edit: October 01, 2017, 07:19:51 AM by SnowFire »

Dark Holy Elf

  • DL
  • Denizen
  • *
  • Posts: 8135
  • Well-behaved women seldom make history
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #759 on: October 01, 2017, 05:36:02 AM »
Ooh, interesting. I'll probably look over this more with time.

First thought: as always with any of these drafts, there's a bit of a question of how we what the goal of each team is. Less resets is always desirable, but beyond that... fewest overall turns? Fewer random encounters (similar to the previous but without rewarding faster completion of boss fights)? I assume the goal is to defeat Ouroburos, too, though that should probably be stated (game's much faster if we can get the other ending, and jobs that only exist lategame would suffer in that case); I notice you mention C7-8 asterisk boss fights and I'm sitting here going "why on earth would a draft fight those" so if there are any secondary objectives those should probably be stated too.

Second thought: Heinkel is an asshole, and you probably never noticed because you used Black Mage. But seriously that fight is hell without Black Mage (he has two allies who he will cover from any physical attacks, so you can't kill them without magic); I know Magey tried to beat him with just White Mage and gave up; I did beat him with Monks but it took a little grinding and a few resets. So... for that reason alone (though also making sure he or she has non-shit offence as soon as possible, I really think player A must grab Knight as one of his or her second picks. Probably instead of Pirate, they're too similar, and Pirate's niche can be filled by other jobs easily enough (e.g. Performer, Vampire, Time Mage; there's also a lategame katana that mimics one of their best skills). This obviously has a trickle-down effect on the rest of the draft which I'm not sure about yet.

Erwin Schrödinger will kill you like a cat in a box.
Maybe.

SnowFire

  • DL
  • Denizen
  • *
  • Posts: 4937
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #760 on: October 01, 2017, 07:18:28 AM »
Same as your FF5 job rankings, really.  "Don't reset", "don't have a high RNG factor that leads to sometimes resets", "require less grinding" (bonus credit for being able to LLG dash to the boss and still win with no encounters in a game like BD!  Penalty if a boss is super-hard and requires lots o' grinding above and beyond normal play!), and so on.  And yeah, assumption is "get the final post-C8 ending."  I'd certainly consider a team that was able to take all the asterisk fights in C6-C8 better than one that can't, for all that the rewards are low enough that it'd be optional flavor....  those are among the hardest battles in the game so they're a good sanity check of the max power of a team.  (i.e. my goal is more "be awesome and powerful and able to take whatever the game can throw at you" than "speedrun to the final" IMO, which would render these matches irrelevant.  YMMV.)

I agree that Heinkel is gonna be hard, but how would drafting Knight help?  You get it after the fight, not before!  (For all that I agree that Knight is a perfectly reasonable option for A's third pick for other reasons.)

Dark Holy Elf

  • DL
  • Denizen
  • *
  • Posts: 8135
  • Well-behaved women seldom make history
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #761 on: October 01, 2017, 08:01:30 AM »
Oh yeah, I derped there. Man, that makes Team A want Monk a fair bit, which is funny given that the job usually isn't rated too highly. Probably not enough to go with Monk as a second pick though... I guess having a Freelancer/WM mix isn't THAT much worse than a Monk/WM mix. Funny how it is worth thinking about though.

Erwin Schrödinger will kill you like a cat in a box.
Maybe.

Dark Holy Elf

  • DL
  • Denizen
  • *
  • Posts: 8135
  • Well-behaved women seldom make history
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #762 on: October 01, 2017, 08:15:44 AM »
Quote
Late game DKs are notably worse without Drain Sword, especially if B ends up running a bunch since they lack other healing, so pick that up.

Wellll... keep in mind Drain Sword can be itemcast through the Blood Sword once you do Vampire Castle. Of course that's behind a six-dragon sidequest chain so it depends how we feel about that. Drain Sword does put in good work midgame regardless though, and the elementals are solid too, so I don't think Spell Fencer is a bad pick or anything.

Vampire probably should go later. Like... there's a case to not take DK in the first three picks as you noted, and it's all due to the availability. Vampire is that, but both later (effectively) and not nearly as good. Debuffs are nice, but... again, remember one of the key debuffs (attack) is covered by a universally available sword which you get around the same time as Vampire, give or take, and a second (speed) isn't available to Vampire at all. B should take Valkyrie over it solely for Thundaga Sword Crescent Moon being super-potent against Rusalka (plus random-sweeping pre-DK). Team A has a little more use for it since no DK gives Vampire's physical MT more of a niche, but still shouldn't take it with its first seven picks or anything.

Erwin Schrödinger will kill you like a cat in a box.
Maybe.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4373
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #763 on: February 15, 2018, 09:42:32 AM »
So..I've been thinking about 5e a bunch recently.  Most of this has been relatively straightforward stuff, but one angle I've been thinking along is full on thought experiment.

So...the concept goes like this: you're making a build, and must take at least one level in each class (each PHB class).  It's a subject I've seen discussed before, but there's a new sourcebook out now (Xanathar's Guide to Everything).

The core of it has not changed.  You need five stats at 13 or above to meet multiclassing requirements, which means there's two obvious options for stat arrays.  (Vanilla human seems to be the way to go).

14/14/14/13/13/13  (14s probably going to CON/DEX/WIS, although the last spot is flexible)

16/14/13/13/13/10 (where the 10 has to go in CON)

There are 12 classes, which means 12 of your levels are 100% locked.  This gives you about 8 levels to play with.  Your final HP (before adding in CON modifiers and other equipment) will be in the range of 97-124 (97 being taking level 1 in Wizard, and all your extra levels in Wizard.  124 being taking level 1 in Barbarian, and all your extra levels in Barbarian).

You will have at least 5 levels of full spellcaster progression (up to level 3 spell slots) and at most 13 levels of full spellcaster progression (up to level 7 spell slots) but your highest level spell slot will always be at least 2 levels higher than your highest level spell known.

In other words, it's good to have some kind of plan for what to upcast or do with higher level spell slots.  That said, finding decent up-cast options isn't too hard.  Warlock spells offer some solid upcast options, since both Armor of Agathys and Hellish Rebuke are designed to work well with upcasting.  On top of this, Magic Missile is pretty good to upcast if you have the Hexblade Warlock's curse on the target (76 damage with no save or prevention mechanism.  108 if you also manage to cast the Hex spell on the target, although this takes two bonus actions and requires learning Hex when you probably want both Armor of Agathys and Hellish Rebuke and only know two Warlock spells at level 1).  Anyway, point is, you'll have some solid uses for spell slots even if you don't plan ahead.

Cantrips are obviously going to be good if singletarget damage is a goal, being something that scales with player level and not class level.  But not all cantrips are created equal.  Booming Blade/Green Flame Blade where you get the full damage will be around 40 damage.  Eldritch Blast from 20 CHA with Agonizing Blast if the beams hit will be around 40 damage.  A vanilla cantrip like Firebolt will be around 20 damage.  It's also worth noting that multi-hits have value.  The Hexblade's Curse adds 24 damage to Eldritch Blast (assuming all beams hit).  The Hex Spell adds 14 damage.  Booming Blade having fewer hits is a real downside to it due to comboing off less.

Another thing that's quite good with Cantrips is quicken metamagic.  Lets you cast a spell and a cantrip in the same turn (or two cantrips).

Now...magic items, we do need to factor these in.  We are assuming level 20 characters.  Looking in the DMG I see tables to roll on for loot.  Based on DMG stuff I'm going to say the following.  Yeah, I know there's no spell shops etc, but if you want an uncommon item, you can probably get it.  You're a level 20 character, the loot tables for treasure hoards involve rolling multiple times on table F (uncommon items) starting at challenge level 0, uncommon items take like...4 days to craft according to the DMG.  Like...yeah, if a level 20 character wants an uncommon item they can probably get it.  On top of this...if there's lots of different items that do similar things (like +1 +2 +3 versions) I'm going to assume you won't have +3 across the board, but on average your stuff will probably be +2.  So...assume middle of the road rare options are almost guaranteed to be obtainable when there's lots of similar equipment.

What does this mean for character building?  Shields are good (+2 shield is +4 AC, plan accordingly).  Attunement is a real issue, and a real advantage of physical builds (+2 weapon does not require attunement, +2 rod of the Pact Keeper or Wand of the War Mage does).  Instrument of the Bards makes the skill check on Bard skills harder, so that's a real advantage for Bard.  Since there's lots of strength setting stuff, the rare option is available (Belt of Hill Giant Strength, set to 21 strength).  Headband of Intellect is uncommon, so that's available (set to 19 INT).  The CON version...no...it shows up as rare it has a single entry in table G, and there's no variants.  There's a couple CON+2 items, though, so...Belt of Dwarvenkind if you'd like.

And then a smattering of other stuff that you might want to attune.  Probably some spare rare wands lying around if you want them (wand of fireballs, wand of paralysis, wand of lightning bolts etc).  Probably some rare attunement weapons if you want those (Flametongue, Sun Blade).  And various defensive utility things you might want to use attunement on (Ring of Protection for +1 AC and all saves; Mantle of Spell Resistance for advantage vs magic; winged boots to fly; etc).  You also probably want to save at least one attunement slot for mystery super special legendary items (don't plan your build around it, but assume you're basically working with 1-2 attunement slots if your build really needs uncommon/rare items attuned).

(In other news, medium vs heavy armour is really not a decision worth putting much thought into as long as DEX is 14.  Heavy Armour can be 1 extra AC if you have 15 STR, but you lose the ability to rage, but raging prooobably wasn't important to your gameplan anyway with cantrips being so good).

Anyway...builds...I kind of assume you're going to want to go deep in a small number of classes...and the really funny thing is that this tends to start looking like 10th level builds.  Going 8 in one class, 2 in another, or 6 in one class, 4 in another are good ways to get 2 ASIs (or 3 ASIs with Figther).  This means you can only reach 18 in one stat if you start with 14 CON, or 20 in one stat if you start with 10 CON.  Going from 14 to 10 CON is like a drop from 140 HP to 100 HP, it's pretty big.  If you were actually a god of damage, maybe that would be a reasonable trade.  (Spoiler alert: you won't be a god of damage).

Full glass cannon build would be like...focusing on Eldritch Blast...20 CHA, 10 CON, Shadow Sorcerer 8, Warlock 2.  Shadow Sorc lets you cast Darkness and see through your own darkness, so hey, that's nice, no need to waste an invocation on it.  Attune Rod of the Pact Keeper (+2) and Wand of the War Mage (+2) cause technically they stack for +4 to hit!

Needless to say going full glass cannon is probably not good.  Fighter 8, Warlock 2 or Fighter 6, Warlock 4 lets you get 20 CHA, 14 CON.  The Eldritch Knight 7th level ability is interesting here; let's you make a weapon attack (probably with a hand crossbow) after you Eldritch Blast.  But I...think shields for +4 AC is probably better.  I think there's a real argument for Samurai here--the focal feature of Samurai doesn't do much, but proficiency in Wisdom saves is a big big deal.

But among Eldritch Blast builds, I think there's some merit to being Paladin 8, Warlock 2, 18 CHA, 14 CON.  Follow your party around and give them Paladin Auras.  Having a ranged move in Eldritch Blast is good when it lets you stick close to your party.

Cleric Builds...I think are a pretty standard 18 WIS, 14 CON, and just leaning on the fact that Clerics get a lot of spells that are worth up-casting (Spiritual Weapon, Spirit Guardians).  And other spells that are a few spell levels earlier than other classes (Revivify) so are easier to squeeze into the build.  Tempest specifically stands out as having a good setup for upcasting (maximize spells scale very well for upcasting, even when they are like...shatter).  Cantrip wise it's Booming Blade, but they do get to add d8 to their weapon attacks, so that's something.  Granted, a Divine Soul Sorcerer can pick up all the good Cleric spells while building CHA, but misses out on the channel divinity which works well with upcasting.

Wizard Builds...so this gets weird.  The existence of Headband of Intellect at uncommon pushes you towards CON builds, ending up with 18 CON, and 19 INT from the weapon.  Or 16 CON and a feat (like Inspiring Leader to give your whole party +21 temp HP after every short rest).  Not entirely sure on school (Necromancer, War Mage, Diviner, and Transmuter (CON save!) are probably the competition for a human with 6th level traits).  So...you'll be pretty durable cause of raising CON, and a pretty high variety toolbox cause Wizard.  You probably do also need Belt of Hill Giant Strength if you want to actually hit enemies with Booming Blade, which fills out our attunement plans.

Rogue.  Sneak Attack with Booming Blade?  (Sneak Attack adds about 17.5 at level 9 Rogue, which...exists I guess.  Although if going this route I think you might want to take Paladin 2 for Smites, which means 14 damage damage sneak attacks).

Bards.  Bards have the instruments of give me advantage on my stuff.  Pretty sure Lore Bard is going to outclass other options here (at 6th level can take the best scaling spells from any spell list).  Being Cha means they can pick up Eldritch Blast too.  Can only go Bard 9, though--can't get magical secrets at 10.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4373
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #764 on: July 02, 2018, 07:05:01 AM »
I'm going to go a little bit weird and stretching the definition of meta here, but it links back to DND, so it still kind-of applies.

So...lately I've been on an ancient history binge ever since I saw bill wurst's video.  One aspect I started looking into was how 50,000 years ago, there were several different species of human living on the planet.  Such as...

Homo Floresiensis, which has been nicknamed the Hobbit.  About 3 ft tall, lived on a small island in indonesia.  Becoming smaller is a trait common in animals that change habitat to a small island, so it's not too surprising it happened.  But...wow, so like hobbits are real.  Cool.

And this got me thinking a little bit about D&D, and how last D&D game I was in, everyone shunned the stereotypical tolkienesque races like dwarves and elves and halflings, and instead went for Catpeople and Ratfolk and Aasimars and Kitsune.  Why would you want to be a very minor variation on humans?

Well, turns out minor variations on humans that are still the same species are...pretty historically accurate.

Homo Neanderthalensis is the well-known one to famously live alongside humans, of course.  Similar brains, berried their dead.  But they had a different habitat (Homo Sapiens lived in Africa at first, Neanderthals lived in Europe and northern central Asia).  And due to their different habitat they were adapted for the cold.  (Shorter, thicker, more hairy).  Yeah, know what D&D race this sounds a lot like?  Dwarves.  They even have wide flat noses commonly associated with pop culture dwarves.

Denisovan So...I don't want to say too much about Denisovans, they've only found a single finger bone.  But one thing they (and Neanderthal) both confirm is the half-types.  In D&D Half-Elf or Half-Orc.  Denisovans interbred with both Neanderthals and Homo Sapiens.  Can't say too much about what D&D race they might correspond to, other than how their genome manifests in modern humans (seems to provide a better immune system to polynesians, and help the Tibetans survive at high altitudes with little oxygen).  Better disease immunity and not needing as much oxygen doesn't really have a D&D equivalent.  Although thematically D&D elves have vaguely similar themes (a bunch of immunities, and don't need to sleep).

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4373
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #765 on: August 15, 2018, 05:17:02 AM »
So...hearthstone has a deck referred to as APM priest, so named because it takes a lot of actions to do the combo, and Hearthstone has about a 1 minute turn limit.  It's hard enough that someone made a simulator to practice it, and shared it on reddit:

https://patashu.github.io/Combo-Priest-Simulator/

I started playing around with this, and also talked to people on Reddit about the theoretical limits of the deck (if there was no timer, could you do X?)  This eventually led me to making a puzzle, and then seeing some new tricks I hadn't thought of in the first puzzle, making a second puzzle:

https://www.reddit.com/r/hearthstone/comments/978y7z/ridiculous_impractical_norope_apm_priest_lethal/

So...Hearthstone has multiple constraints.  10 cards in hand (any extra will be destroyed) and 7 creatures on board (any extra can't be summoned) and this puzzle is tuned very specifically to these numbers.  But it also uses several tricks that weren't showing up in the normal combo.

Trick #1: double layer recursion.  Ordinarily the combo would only ever have 1-2 of the creature "Test Subject" around at a time.  You'd make a copy, kill the copy, and then cast a bunch of spells, and repeat.  This can't go on forever, because every time it returns more spells, and eventually it will return so many spells that it doesn't return the spell you need to continue the combo (topsy turvy).  One of the big innovations is to, instead of keeping 1-2 test subjects, keep 3 test subjects.  So...cast a spell copying one test subject, lets say it and its copy would now return 4 spells.  Next, cast a spell on the copy, now there's one test subject with 4 spells it returns, and two test subjects with 5 spells they return.  Pop one of the higher number ones, and then later copy one of the higher number ones again, until you hit the cap.

This way, instead of getting back 1+2+3+4+5+6+7 spells, you get back (2+3+4+5+6+7)+(3+4+5+6+7)+(4+5+6+7)+(5+6+7)+(6+7)+(7) spells.

Why not do triple or quadruple layer recursion?  Well, cause you don't have the board space for that.

Trick #2, divine spirit less, and get rid of divine spirit early to clear up space.  So...this combo is known for using a card called divine spirit, which doubles your health (and then flipping health into attack).  But you just don't need to cast this very many times.  Player health starts at 30 in Hearthstone, so if you cast Divine Spirit 5 times, that gets you from 1 to 32.  All you need.

You definitely don't need it on every test subject, so don't use it on one of the main-line test subjects.

But you also need to get it out of your hand before you start hitting the hand limit, so you might as well use it on the first big branch of test subjects.

Trick #3, saving 1/1 boars

So...the enemys blocking the way are a pretty specific unit--Void Lords, which are 3/9 taunts, that split into three 1/3 taunts BUT, BUT your enemy is also restricted by only having 7 minions on board.  So if you kill a 3/9 when their board is full, they only get one 1/3 instead of three 3/3s.

Now, one thing in this puzzle that I did was make your number of power/toughness flips extremely limited (only one), so once you flip your boar from 1 attack 16 health to 16 health 1 attack, you can't flip anything else, which means you can't get new spells.  This means the absolute maximum you can have in hand is...well the puzzle starts you with two dead cards, so 8 total spells, one of which is Topsy Turvy, so 7 spells that can copy the 16/1 charge minion (once we count the original minion that's 8 attacks of 16 power).

But we have to kill the 3/9s first right?  Otherwise they'll spawn three minions with the extra board space.  And once we kill all of them then we still have seven minions we have to attack.  So we are forced to use our 16 attack boars on 1/3 taunts, and not the 3/9 taunts.  Gosh, that sucks.

Or...are we forced to do this?

So...the cool innovation that I didn't even think of (which is why I ended up posting a second puzzle) is that at that point, you're at the very very end, you've wiped off all the test subjects from the board, which means you have two whole more board spaces opened up.  Wow, two.  What can we do with those?  Save some 1/1 boars in those spots.  Long before any of this happens, use some other 1/1 boars to lower some existing void walkers from 1/3 to 1/1.  Then we use the 16/1s to attack two of the 3/9s, and ~~after those are gone~~ use the 1/1s to kill the now 1/1 void walkers.

Trick #4, the extra divine spirit is a red herring

I needed a tiny bit extra for this verison of the puzzle to work, so I added an extra divine spirit, which I knew would trip people up.  If you're used to this combo, you see a divine spirit you put it on a test subject, this is intuitive.  But hand space is so tight that test subjects returning two divine spirits means that you would get 2+2 divine spirits back from test subjects, which is 4.  But if you instead only cast one divine spirit on the test subject and the other one on the boar, then you free up hand space to get an extra divine spirit back from the test subject (3 DSs from test subjects) and you also cast one DS on the boar.  So...1+3 = 2+2 = 4.

Honestly, kind of a dirty cheap trick, not even that deep, but it got someone who fully understood the other three tricks.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4373
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #766 on: February 25, 2019, 06:54:01 PM »
D&D 5e

So a few years ago in this topic, Elfboy and I had a debate over taking Great Weapon Master as the first ASI instead of taking +2 strength; (Great Weapon Master being the ability that lets you take -5 to hit for +10 damage).

Me arguing the average damage was higher, Elfboy arguing that added consistency was worth slightly lower damage.

And...I was not sure how to handle consistency in D&D.  Like...everything in D&D is random.  Everyone can miss.  The damage you deal is ramdom.

Well I’m taking a stab at measuring consistency with an enormous ugly spreadsheet.  Calculations are in progress but early results are here:

https://www.reddit.com/r/3d6/comments/aun56y/reanalyzing_gwm/

(Spoilers: Early results suggest Elfboy has a point).

NotMiki

  • Denizen
  • *
  • Posts: 4476
  • Social Justice McNinja
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #767 on: February 25, 2019, 11:29:34 PM »
It's fascinating trying to get at the assumptions that underpin the utility of this sort of calculation.  How sure of enemy AC can you be?  How sure of enemy HP can you be?  Are the situations where GWM is the better choice common scenarios?  How does GWM fare when you make certain assumptions about the rest of your party - like for example that you already have a consistent source of less damage?
Rocky: you do know what an A-bomb is, right?
Bullwinkle: A-bomb is what some people call our show!
Rocky: I don't think that's very funny...
Bullwinkle: Neither do they, apparently!

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4373
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #768 on: February 26, 2019, 12:46:07 AM »
I’ve also genuinely got a problem that I’m not calculating all the things I want to calculate.

Like...basline if you can imagine two attacks, one that deals 20 damage with 50% accuracy, and one that deals 10 damage with 100% accuracy, obviously they both deal the same average damage, but also obviously the 100% to hit one is better.

This calculation captures some of why it’s better.  If you have an enemy with 30 HP, you’d rather have the 100% attack and kill in three turns on average.

But I would argue it’s also better against the 40 HP enemy.  In some games it wouldn’t be, in like in a sufficiently hard roguelike you might prefer a high risk effect.  But in D&D where a party death is a permanent loss of character, you want the more reliable result.

But I’m not sure where to measure that.  I already have numbers for how often each build kills within X attacks.  Intuitively I’m now eyeballing 80% chance to kill as maybe a good measuring stick (if you can imagine a 20 HP enemy, the random character would have a 75% chance to kill by turn 2, so...should be higher than 75% I think).

But my issue is that 80% odds of killing by turn X is just a number I pulled out of my ass.  I feel like it should be maybe one standard deviation...but that’s 68% so...not enough.  Or two standard deviations, but that’s...actually pretty extreme.  Like...if you have GWM and want to have your 5th percentile performance of killing be as good as possible you should basically not activate your -5 to hit for +10 damage at all, which also feels weird and probably not correct either.

80% feels about right, but it’s literally just a made up bullshit number.

EDIT or...is it.  68% for one standard deviation cuts off both ends.  But we’re looking for only the bad performing end not the high performing end.  Since we’re only interested in the poor end, we can look at 84% chance to kill (one standard deviation where only the bad end of the standard deviation is a serious concern).
« Last Edit: February 26, 2019, 12:51:55 AM by metroid composite »

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4373
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #769 on: January 07, 2020, 08:04:57 AM »
Alright, so Laggy and I played a playthrough of Vanilla FFT with the following restrictions:

(EDIT: made a bunch of edits this morning, Jan 7 2020).

1. Each class had three banned skills (supposed to be the three strongest skills)
2. (also just outright banned Ramza's special squire class).


We thought some builds might be too prevalent (like black magic in chapter 1), so we had an extra rule that no two characters could have the same secondary, and no two characters could have the same primary.  (So e.g. you could have two people at most with black magic, one wizard, and one person with black secondary).

We also did this fairly low level (like lower than level 18 before riovanes).


Here's the ban list we used on our first attempt, and then some commentary on the list:


- Basic Skill: Accumulate, Gained JP Up, Move+1
- Item: Potion, Phoenix Down, Auto Potion (This was probably a mistake, should be X-Potion instead of Potion banned)
- Battle Skill: Speed Break, Weapon Guard, Equip Shield (This was probably a mistake.  Should ban Equip Sword over Speed Break)
- Charge: Concentrate, Charge+2, Charge+3 (This was probably a mistake.  Should probably ban arrow guard and Jump+1 and not touch Charge)
- White Magic: Raise, Magic DefenseUP, Holy
- Black Magic: Bolt 2, Flare, Magic AttackUP
- Punch Art: Earth Slash, HP Restore, Martial Arts
- Steal: Steal Heart, Secret Hunt, Move+2 (This was probably a mistake.  Should ban Jump+2 and unban Steal Heart).
- Yin-Yang Magic: Life Drain, Silence Song, Defense UP
- Time Magic: MP Switch, Short Charge, Teleport
- Jump: Level4 Jump, Level5 Jump, Level8 Jump
- Elemental: Hell Ivy, Kamaitachi, Attack UP (We definitely messed up here--should have banned Counter Flood)
- Talk Skill: Invite, Praise, Preach
- Summon Magic: Shiva, Ramuh, Ifrit
- Throw: Abandon, Sunken State, Two Swords
- Draw Out: Kiyomori, Kikuchimoji, Blade Grasp
- Math Skill: CT, Level, Height
- Sing: Angel Song, Nameless Song, Move+3
- Dance: Wiznaibus, Slow Dance, Nameless Dance

Commentary on these:

Basic Skill: GJPU and M+1 are obvious.  Accumulate probably isn't the third strongest move, but Laggy was just annoyed at the people who sit in a corner accumulating and swear by that as the fastest way to grind.  (Counter Tackle is probably the third most relevant ban here, but Counter and hammedo are unbanned so whatever).
Item: PDown+AutoPoption are obvious.  Potion is a weird choice, but potion as a secondary is still a secondary that lots of people would just slap in chapter 1, and item is much more of an earlygame skillset now without PDown.  After some debate, we're thinking X-Potion is probably the right hit here, though.  With all other support abilities nuked, Throw Item is basically a free slot.  And despite thinking Item was dead, we did have someone running Item with X-Potions at endgame.  Item with just X-Potion was pretty relevant on Ninja, which was on the strong side.  Item is still a reasonable pickup for mages who want things to do when they run out of MP.  In Chapter 1 regular Potion is strong buuuut Black Magic is stronger, so X-Potion may actually be the right ban.
Battle Skill: Weapon Guard and Equip Shield absolutely belong here; they'd be used on lots of characters if legal.  Speed Break probably doesn't; it's capable of doing some silly things with zodiac bosses, but you could also just...kill the bosses.  Equip Sword or Equip Armour might belong in this slot--Equip Sword is a decent damage boost to Ninjas who use bracers (like about as good as Attack Up).  Equip Armor is a big HP boost to monks.  At the moment Ninjas are looking like maybe even still a problem class and Monks are sketchy so Equip Sword to target Ninjas is probably correct.
Charge: Concentrate is obvious.  On further debate we're thinking it was incorrect to not ban Arrow Guard (which we allowed the first go through) as it's looking like a super premium reaction compared to other reactions available, and pretty accessible even to mages.  So the debate is between Jump+1 and hitting the charge skillset, and Jump+1 came up on a bunch of characters whereas the charge skillset was always a temporary thing and not the long term skillset.
White Magic: Holy is obious.  Raise...we know from LFT that if PDown is banned, everyone will just jam raise, so it was an obvious target.  MDefUp...the fear was that if it was allowed, everyone would run it lategame (and our lategame was like 4x maintenance even in fights without equipment breaking, so...yeah, 100% it would be run).
Black Magic: MAU is obvious.  There was a desire to cripple the early game power of black magic, but without banning all three of fire/ice/bolt that wasn't really happening.  So...instead Bolt 2 (for being 10 MP, and also weakens thunder rod setups in general).  And then Flare--it just had the biggest impact of a ban; mathskillable, the only non-elemental damage in the skillset, maybe the best assassination spell now that all the better options were banned.
Punch Art: Earth Slash is obvious.  Martial Arts is good even on non Punch Art users, but removing it also just cripples punch art setups cause you can't really leave monk.  We were just not that worried about dedicated monks after those cuts, so HP Restore as their best (budget) reaction was the third cut.
Steal: Move+2 is obvious.  Secret Hunt officially bans chantage shenanigans (instead of just being informally banned) so sure, why not.  Steal Heart is what we banned, but that was almost certainly incorrect as several characters grinded Jump+2 from unlocking things, whereas Steal Heart would be temporary.  Jump+2 is probably the correct ban.
Yin Yang: Life Drain is obvious.  Defence Up is here for the same reason as MDU: all the good support abilites are banned so everyone would use it.  Silence Song...this is the slot that's fairly debatable.  Situational, but very strong when it's good.  Whereas the rest of the options are...ok.  Sleep could cripple people who invest deep into YYM, but banning Life Drain already does that.  Move MP Up is a possibility, but I only ended up getting it on one of my mages.
Time Magic: Teleport is obvious.  Short Charge...banning has a huge impact, yes, so sure.  MP Switch would probably go on every physical setup if it was allowed (especially if Move MP up wasn't banned).  This does leave Haste legal, and it's pretty good, but it still takes a skillset slot.
Jump: Yeah, three highest jumps, now capped at jump 3.  Fairly obvious.
Elemental: Attack Up is obvious.  We really should have set Counter Flood here, as with all the other reactions banned it kind-of became the go-to for a lot of characters and was quite good.
Talk Skill: Invitation is obvious.  Praise makes sense to me; just...brave is good on everyone, and technically with no way to raise brave you disable some busted strategies.  There was some debate about Solution vs Preach.  Eh, not sure which is right still; I expected lategame magic to be worth doing, but it did fall off kind-of hard due to lack of short charge, but high faith is still desirable for support skills like haste and raise 2.  Also math skill.  Yeah, Preach ban is probably fine.
Summon: Shiva/Ramuh/Ifrit Obvious.  So...Titan can kind-of stand-in for Shiva/Ramuh/Ifrit, but it's also earth damage, so can't boost it with ice rods or thunder rods, and also means the multitude of chapter 4 enemies that float or have earth clothes really puts a damper on Titan.  And then the lack of short charge limits the rest of the skillset.
Throw: Yeah, you can't really impact the throwing all that bad; ball would be annoying I guess, but you'd substitute shurikens.  So all the relevant RSM instead (Abandon, Sunken State, Two Swords).
Draw Out: Kikuichimoji is obvious.  We quickly ended up agreeing on Blade Grasp since all the other strong reactions were out.  Which leaves Kiyomori vs Muramasa.  You could substitute Heaven's Cloud for Muramasa,  though, unlike Kiyomori which was unique and good.
Math Skill: So...with 4 bans you could blank the skillset.  Only 3 bans, so no.  EXP just seemed like the weakest thing to let through (Much harder to control).  Ended up feeling pretty random--when the EXPs linked up it felt great, but sometimes you could only like...haste a single party member or status a single enemy.
Sing: Move+3 and Angel Song seem obvious.  Nameless Song was our best guess for third best.  I guess Angel Song could be debated with MP Switch being banned, but it is the best MP restoration in the game for MP heavy parties.
Dance: The three obvious ones (Nameless, Wiznaibus, and Slow).



The characters we did:

Agrias with nothing from Holy Knight banned but all other restrictions in place.  (Laggy wanted to see how weak or strong she was relative to nerfed generics).  She was strong and carried the party, threw balance a little out of wack.

Ninja performed well when it got to Ninja.  Unlocking is always a pain of course.  Really worth noting that all the move skills are banned (Move+1, Move+2, and Move+3) Concentrate is banned, it's a lot worse than normal.  These all hurt.  But the class still seemed solid.

Samurai Ramza.  No Kikuichimoji and the massive lack of movement abilities mean that movement was a problem.  Although was still strong and one of the few ways to bypass evasion.

Summoner, that detoured for Move MP-Up:  Carried through the first half of the game.  Fell off super hard in chapter 4.  I got Odin, but 9 ctr no short charge just isn't good enough, and so many enemies in Chapter 4 immune or absorb earth.  Ended up going to Time Mage as carrier over Wizard so we could keep haste in the party, and ended up just using a lot more Time Mage moves than summons towards the end of the game in general (haste, slow, demi).

Calculator: underperformed just due to JP shortages.  Didn't unlock calculator until the start of chapter 4.  (I typically hit end of chapter 2 caulculator with Gained JP up).  And then needed to bail before getting all my skills cause the game was going to end.  (Got EXP, 5, and 4, did not get 3, Prime, or Damage Split).  Kind-of a product of not having Gained JP Up and doing no propositions or randoms.  In a group that grinded more this would not be an issue.  Overall math skill was...pretty random.  Sometimes the exps would line up perfectly, and it was like fairly god mode.  And sometimes I'd just like...haste two of my characters and one enemy ninja.

The calculator also started off as Priest and rushed Raise 2 as its big expenditure from the 550 JP in Priest required to unlock Calc.  It felt...worthwhile to have one raiser, but with no holy in the skillset, and so much of the JP invested in the skillset sunk into Raise 2, it wasn't always correct to even set white magic.  It got slotted for longer grindfest maps where an early death could otherwise cause a wipe, and black magic got used otherwise.

Black Magic falling off in Chapter 4 isn't news to anyone (gets evaded, now also lacks flare) but it still was pretty consistently not bad, often a fallback move for characters with other skillsets (the summoner using level 1 spells when Titan's 5 CTR was too slow.  The Samurai using Ice 2 when out of range).  Was still all around very good in Chapters 1-2 and most of chapter 3.  (Chapter 3 does have some enemies with Ageis Shields or White Robes which kind-of brick black magic, but I was often running white magic in those fights, so that weakness didn't really come up).

Other thoughts:

RSM was very very weak.  Jump+2, Maintenance, Counter Tackle.  These were all set by multiple people at some point.  We ended up deciding Counter Flood was way too good given everything else that we had banned.  Bypassing evasion is definitely a premium here.

This had knock-on effects on what classes we cared about.  Geomancer being one of the only 4 move classes was a go-to class as a carrier.  At one point (because we weren't allowing ourselves two of any class such as Geomancer) we actually had Agrias as an Equip Sword Thief just to have two characters with 4 move (Thief ended up being not durable enough, but that's how desperate we were for move).  Equip Shield being gone also made the classes with shields matter a little more.

Builds not tested:

Lancer: Initially, the thought was that it would be too hurt, with level Jump 3 as the max.  But really, it's probably actually fine.  One of the few ways to bust through evasion that isn't banned with Concentrate being gone.

Monk: We're kind-of up in the air on this one.  No Earth Slash is pretty bad, no martial arts is pretty unfortunate if you want to go anywhere.  But you do get Revive and Chakra.  And Wave Fist gives you range.  Will struggle with breaking evasion lategame.  You could also (and would) use Equip Armour for a helmet, with basically no opportunity cost because there's almost no other relevant support abilities left.  So...Monk is probably an ok tank with revival options.

Oracle: I'm not entirely sure--they're known for being swiss army knives, but Life Drain was also kind of the solution to a lot of problems like enemies without evade, needing to heal themselves, killing zodiacs.  Take away Life Drain and they are hurting.  (Dedicated oracles are fine with losing silence, they have other backups like berserk, but losing life drain really opens up holes in their skillset).

Mediator: Talk Skill exists, Mimic Daravon and Solution are both out there.  Didn't mess with any of those on this run.
« Last Edit: January 07, 2020, 05:52:23 PM by metroid composite »

Dark Holy Elf

  • DL
  • Denizen
  • *
  • Posts: 8135
  • Well-behaved women seldom make history
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #770 on: January 08, 2020, 01:24:09 AM »
I'm a bit surprised you didn't ban Equip Gun over Preach/Solution, neither of which feels remotely broken to me if you allow the player to farm generics at the start of the game, 40 faith is "low enough" for many purposes and 70 faith is already pretty good offensively (80 is better, but taking ~20 actions to get there is bleh). Being able to hand out solid ITE ranged damage seems pretty potent especially with most other support abilities gone, and especially if you steal from Balk and/or the Goland chemist.

Magic Defence Up I was initially skeptical about but your reasoning makes sense, with pretty much every premium support being banned it would start looking really good.

Erwin Schrödinger will kill you like a cat in a box.
Maybe.

Laggy

  • ReDux'd
  • DL
  • Denizen
  • *
  • Posts: 1145
  • Generations of suffering & all I got was a stick
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #771 on: January 08, 2020, 02:26:26 AM »
It's a fair argument - I spent time in Chemist getting Item for the Ninja and viewed it very favorably, especially using Mythril Gun right as it became available. Definitely felt strong. But it's also 750 JP in Mediator, a job that feels pretty bad to hang out in for that long. Still, you might be right that it's about as much effort as Faith raising; depends on how much you would place value on long term Math Skill performance in a theoretical DD postgame setup. I'm a bit dismissive of bothering to steal spellguns as a kneejerk so I didn't consider that angle, but yeah it's there.
<Eph> When Laggy was there to fuel my desire to open crates, my life was happy.  Now I'm stuck playing a shitty moba and playing Anime RPGs.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4373
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #772 on: January 08, 2020, 05:14:43 AM »
Hm, yeah, I was wondering if we were underselling Equip Gun.

It's a big investment (750 JP for Equip Gun, and for elemental guns 600 JP for Steal Weapon, not necessarily on the same character).  But that just means it needs to be part of a planned build.

Among the builds we did...

Ninja: lolno

Agrias: lolno

Summoner (Time Mage with Summon secondary): TBH, getting Odin was kind-of a failure, came too late and enemies were too fast so...sure: maybe skip Odin and get Equip Gun instead.  Loses 1 MA from swapping out Wizard Staff, but I think Equip Gun could make the character better overall; often she just couldn't even cast Titan due to 5 ctr being too slow.  (Even a 64 damage Mithril Gun would probably make the setup stronger).  Although...as far as "skip Odin and get X instead" options, there was also Half of MP as an option--maybe Equip Gun was stronger than Half of MP though.

Samurai (Wizard with draw out): I mean...probably wouldn't give up Wizard Rod for just a Mithril Gun, but an elemental gun hypothetically?  yeah, probably.  Draw Out Wizards need range help, and they can afford to sac a little melee power to get better range moves.  That said...Samurai is pretty JP hungry, I don't really think the build had time to get 750 Mediator JP.  With more grinding though?  Sure, could do.

Calculator: (Priest with Math Skill): Mmmm...this character was so JP starved.  Never learned 3 or prime number.  Never learned some pretty relevant stuff like Reraise, Cure 3, Pray Faith, Reflect, Demi 2, Ice 3.  I can see the value of Equip Gun--sometimes EXP-only mathskill was godly, but sometimes it would blank hard and hit like...one friendly character and nobody else.  Having a reliable option like a gun sounds decent.  But it would come at the cost of speed (H-Bag) and 750 JP.

Additionally, while we didn't play this one, one other theorycraft I had was Monk with Equip Gun.  Lets them brick Archers with Hammedo.  Gives them actual range.  Gives them a no-evasion move.  They can still wave fist for slightly less than an unarmed punch.  Obviously there's a tradeoff here, though (not using Equip Armour in the support slot).

Another theorycraft: if you steal an elemental gun, you can then use Jump with it, and deal better damage than the best storebought spear.  (Like...jump for 300ish, shoot for 150ish from a Ninja with Bracer, Thief Hat, Power Sleeve).

---

I'm...undecided TBH.  Like...I see potential strength there but Equip Gun tends to come with some loss from the weapon it replaces (typically losing Wizard Rod/H-Bag, or in the case of Monk losing 100 HP, or in the case of Ninja with Equip Gun and Jump, losing the Ninja auto-attack).  If it was cheap, I'd just grab it and switch between guns and other weapons depending on the fight, but it's also a pretty heavy JP investment.

Whereas like...setting aside weird debades about super-lategame setups, raising faith even a small amount (from 70 to 73) is enough to make Haste 100% on yourself instead of 92%.  That's...something I certainly care about.  77 faith on two neutral zodiac people gets Raise 2 to 100% instead of 83% at 70 faith.  I kinda care about that too.

That might just be a psychological thing on my end, where it really stings every time a 92% haste misses, and 100% is really not that much of an improvement, but...still, it does stick out in my mind.

If we can come up with a really clear power build for Equip Gun where it's like "if people knew about this build people would use it", then yeah, it's probably the ban choice.  But right now the examples I can think of are kind-of muddy like "yeah, I guess you could jam Equip Gun into this build, but it's kind-of a sidegrade over the weapons we were using without learning Equip Gun."
« Last Edit: January 08, 2020, 07:05:08 AM by metroid composite »

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4373
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #773 on: January 08, 2020, 09:01:11 PM »
Alright, Laggy and I have been hashing this out over chat.

Spellguns could be crippled a little by banning Steal Weapon over Jump+2.  This doesn’t prevent getting spellguns but it makes them slower and grindier to get.  We decided spellguns weren’t so big of a deal to make this change; you get them really late, they requires effort to even get, whereas every Ninja would get Jump+2 for free.

We did actually come up with a sensible build that really benefits from Equip Gun.  Grab the bare minimum from Summoner to crush the early chapters (Titan, Moogle).  Then GTFO to a lategame mage (Oracle or Time Mage).  When Titan becomes trash (late chapter 3 most of chapter 4) go Mediator with Time/YYM secondary, learn Equip Gun, then exit to Priest with Time/YYM and get Raise 2, Cures, Shell, protect.

The thing Equip Gun does here compared to being a Mediator with Time/YYM is that you get reviving (and misc healing).  And compared to doing the same setup without Equip Gun it gives you offence (basically zero offence spells in all three skillsets now).

So...yeah, I’m on board with the ban Equip Gun train.  Just debating what to unban.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4373
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #774 on: January 08, 2020, 10:40:06 PM »
Options to unban: Invitation, Praise, Preach

Praise value is lowered in a world with a lot less reactions.  But it’s not a world with zero reactions.  Damage Split, Dragon Spirit, Counter, Hammedo, shenanigans with Quick and Countermagic are potentials.  There’s also just Ninja unarmed punches (roughly attack up level boost if you have 97 brave).

On the flip side, small amounts of brave raising (3-6 say) does little.

Faith you can definitely feel the benefits with small investments; 100% self haste.  But it’s more niche, and makes you take slightly more damage.

Invitation mostly boosts builds that use Mediator as a carrier.  (Status mages like oracles and time mages that want to have guns in chapter 4 when attack magic becomes trash, basically).  They would go to mediator for robes+guns, and get invitation for free as a strong panic button to turn around fights that are going poorly.

I’m...wondering if Invitation might be the right hit here?  I definitely see the value in it as a boost to the Mediator Carrier, I’m just less scared of the things that can be done with a Mediator using Time Magic than I am of a Mathskiller with Preach boosting or a Ninja with praise boosting.