Register

Author Topic: Theorycrafting! (Because I like competitive metagames too much)  (Read 133632 times)

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4378
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #600 on: October 26, 2015, 07:05:12 AM »
Just based on your rough descriptions I would assume they are something closer to 19-25, but more information would be helpful.

19-25 definitely sounds low.  These are creatures that I'm fairly sure could see a bullet and step out of the way in response (not that the bullet would hurt them).

Quote
You're probably overrating the strength somewhat too. Leafing through the 5e monster manual... a kraken (gargantuan creature, can lift ships let alone 2000 kg objects) "only" has a strength of 30, so it seems unlikely that Twilight vampires are better than that. The stat appears to be strongly logarithmic past a certain point.

Size plays a role in carrying capacity, at least in Pathfinder.  Not sure about 5e, but in Pathfinder Gargantuan creatures get a x8 modifier on their carrying capacity.  Which means a heavy load of a 30 strength character gets into the 13,000 lbs range.  (Depends what you mean by ships I guess).


Quote
(Actually, in general, D&D stats tend to work a bit differently at certain ranges. 3-18 is supposed to cover the range of human experience with a distribution you'd expect from rolling 3d6. 1 and 2 cover "worse than any human" which can be a wide range in some cases, e.g. animal Int stats. Then higher numbers cover everything from the barely super-human to godlike creatures... but for the most part, don't go much past 30 if at all, and are scaled to cover this range.)

Ah yeah, Pathfinder was definitely describing 4-5 as "worse than any human, or any human that's not seriously crippled".

Quote
DC 45 Tumble check allows you to treat a fall as if it were 30 feet shorter. If you want to add extra credit, DC 60 allows you to treat a fall as 40 feet shorter. And a DC 100 check allows you to ignore fall damage completely.

I will also note that DC 120 Balance checks allow you to stand on clouds.

Mmmm...well standing on clouds is definitely a no, but ignore fall damage is probably a yes.  It's not 100% clear, though, whether fall damage is a joke to vampires because of acrobatics, or because terminal velocity is too low for it to actually be a threat.

Quote
B) Bella is 18 Charisma.  Not because the author is remotely convincing at making her authentically charismatic (a la, say, Kefka or Balthier), but because everybody just wants to be friends with her on meeting her or just generally thinks she's Important and should be Paid Attention To despite being a random mortal high schooler (e.g. for the villains).  That's a side-effect of 18 Cha; you carry a reality-warping field of "everybody looks up when you enter the room."  In all seriousness, this is one of the nice things about roll-play rather than role-play; sometimes a shy player wants to try their hand at a super-charismatic character, so sure, even if you can't charm them off their feet yourself, go ahead and roll a Diplomacy check...  okay, yes, Bella knows the right things to say, and her brown hair is shimmering at just the right angle from the sunlight, etc. etc. you're fine and your character just naturally knows what to do even if you don't.

Eh, if anything I would give Bella a Charisma bonus when interacting with vampires.  (Which is even explained in the books--damn near every vampire wants to drink her blood because it smells unusually delicious.  Also, the whole immunity to mind magic thing can make a room full of vampires stand up and take notice).  Although, honestly, MOST of the interest she draws from other vampires is the fact that there are vampires keeping a human as a pet and not killing her, which is really fucking weird in the Twilight universe, not to mention illegal (breaking the Masquerade laws).

I really don't think she's 18 charisma when interacting with humans, though.  Above 10, but not 18.  She doesn't smell unusual to them.  None of them have mind magic.  None of them know about vampires.

So maybe something like, 13 cha, with a large bonus to charisma checks with vampires.

SnowFire

  • DL
  • Denizen
  • *
  • Posts: 4964
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #601 on: October 26, 2015, 07:54:02 AM »
Disclaimer: I've only watched the movies...  with Rifftrax...  and while I'm familiar with the outline of the books, I haven't read them.

Bella as a transfer student instantly gets a network of (mortal) friends at the high school who are for the most part pretty decent.  She *whines* occasionally but nobody actually is mean to her or the like in a way that matters.  Additionally, it's not just vamps who think she's the bees knees- Jacob & co. all think Bella is totally awesome and should be protected blah blah blah, for the some-mortal-friend / some-supernatural crowd.

Bella's delicious blood / weird aura / whatever can obviously be interpreted a lot of ways, and I doubt it's meant to be taken too scientifically, but...  there's definitely an alternate universe where vegetarian vampires find someone with delicious blood, and say "yup, now we should never ever be around this person, because that's asking for trouble."  Anyway, even if it partially is the "aura" / delicious blood's reason for why supernatural critters just think hanging around her is awesome, I agree that it should be seen as equivalent to higher Charisma.  Base Bella is Cha 13 but she's actually 18 for all intents and purposes due to weird magical boost, perhaps.  But she is portrayed as having that 18 Cha aura in effect, I think!

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4378
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #602 on: October 27, 2015, 12:46:30 AM »
Bella as a transfer student instantly gets a network of (mortal) friends at the high school who are for the most part pretty decent.  She *whines* occasionally but nobody actually is mean to her or the like in a way that matters.

Mmm...she gets attention initially due to being the only new girl in a small town in ages (the school has like 200 students).  There are definitely humans who don't like her (Lauren jumps to mind, Jessica once Bella does something stupid; also Jessica attached herself initially because of the extra attention drawn by the "new girl in a small town" effect, but never particularly liked Bella--she was just using her).  The number of human kids who sit with her at lunch by book 3 (once the dust has settled) is...three.  Those humans being...Angela, who is also smart, considerate, and introverted; fairly natural friendship.  Ben, who is dating Angela.  And...Mike Newton, who has a crush on Bella.  (And then the rest of the people sitting with her are vampires).

Admittedly, I haven't watched the movies, but I wouldn't be surprised if the human characters didn't get much screen time or were mostly replaced with extras.  Gotta cut something to make a 2 hour movie, and the humans are a pretty obvious place to start.

Quote
Additionally, it's not just vamps who think she's the bees knees- Jacob & co. all think Bella is totally awesome and should be protected blah blah blah, for the some-mortal-friend / some-supernatural crowd.

Almost strictly Jacob, tbh.  I mean, to some degree Werewolves have the whole hive mind thing going on.  So you do have, for example, Leah Clearwater bitching about how Jacob thinking about Bella so much caused her to dream about kissing Bella.  But the reverse is also true--Leah pining after Sam had a lot of male werewolves pining after Sam.  Discounting the hive mind, though...pretty much only Jacob actually cares about Bella.

Like...let's see...opening up Breaking Dawn to the page where the werewolves are talking about killing Bella to destroy her unborn halfbreed child...Jacob objects, because he likes Bella.  Seth objects, because he likes Edward.  The rest of the werewolves are pretty much on board with murdering Bella as collateral damage to get to their intended target.  She isn't seen as important.

Supporting cast to the werewolves in La Push...Billy likes Bella, but Billy's been good friends with her father for years, and babysat Bella for years.  The rest don't really interact with her.

So...no, I don't really think the werewolves as a whole consider her important.  Jacob does, 'cause he's in love with her.  The rest not so much.
« Last Edit: October 27, 2015, 12:50:54 AM by metroid composite »

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4378
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #603 on: October 30, 2015, 09:53:20 PM »
Sand Rat Cellar: Level 3, (20 actions)
Meeting at lionel: Level 10, (70 actions)
Orbonne: Level 17 (120 actions)
Bethla: Level 27 (170 actions)
Before final dungeon: Level 36 (220 actions)

Alright, let's talk about these.

First of all, Gained JP Up in Chapter 1.  Let's say you start with 150 JP in Squire; how quickly will you get it with spillover?

It'll be 2 JP per action while Delita and Algus are JL1, but very quickly 3 JP per action once they're JL2.  If you ignore GJPU and get it strictly from spillover, you'll have it sometime around Sand Rat Cellar.  No wait, there's two of them, so earlier than that.  6 JP per action; you'll get it after two fights of five actions.

Is this worth it to ignore assuming you're going the magic route?  With your first two fights you could get about 50 JP in Chemist, 50 JP in Wizard.  Or you could get 50 JP in Squire, and then 75 JP in Chemist.  Mmm...give me that Wizard JP most of the time, I think.

OK, so assuming class hopping, it's looking like gaining about 250 JP total.  If you drill into one single class, taking advantage of job levels, you get 305 JP total.

Best setups:

1. Wizard with Summon (Linen Robe + Red Hood gives +18 MP, base MP is 21, total 39.  MA is...still 7 not 8.  HP is 25 + 34 from equipment = 59, Thunder Rod available) -- Ramuh damage 80.  Bolt 1 damage 47.
2. Wizard with Item (See above mostly)
3. Wizard with Auto Potion (technically you can afford it, but you'd need to be a Chemist with only Potion and Black Magic secondary for four straight fights, so it's not the most practical).

Other classes:

Squire: if you are female, you have 3 PA.  18 damage with Iron Sword.  Or you can use a Mythril Knife and deal 15 damage.  For males, it's 24 with Iron Sword, 20 with Mythril Knife.  Chemists are 3 PA regardless of gender.  Mages with staves will generally be dealing 18, unless they're Priests.  Thief's actually a little less crappy than most of these, with 20 damage across both genders.  Lancer/Geomancer you...won't have unlocked.  Archer is 20 Longbow damage across both genders.

So...alright, the somewhat serious one,

Knight
36 damage (30 female).  43 base HP + 60 HP from equipment = 103 HP.  Buckler is 13% evade.

Head and shoulders above the other physical classes.  I should also mention, outside of literally going all-in on Auto Potion, Weapon Guard is probably your best reaction at this point (20% evade on most mages).

And for a slightly less serious entry...

Monk
36 damage (21 female).  47 base HP + 18 from equipment = 55 HP.  (So...in Chapter 1 hats give most of the HP.  So...enjoy that less-than mage HP total...).


Classes that stand out as good.

1. Wizard
2. Chemist
3. Squire
4. Summoner
5/6. Knight/Time Mage (I haven't mentioned it, but Haste is pretty rad).

Classes that stand out as bad:

None particularly.  Most other classes have nearly identical physicals.  Obviously Oracle/Priest/Time Mage are preferable due to using black magic reasonably well.  And Archer physical attacks have range, so that's nice.  But you're in these classes cause you're training for some future setup, usually.
« Last Edit: October 30, 2015, 10:13:49 PM by metroid composite »

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4378
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #604 on: October 31, 2015, 01:08:52 AM »
Chapter 2

OK, so the level is 10.  Mages can expect 33 MP.  Triangle Hat + Wizard Robe give 34 MP.  Which means no: Meteor is not happening.

Is Mathskill happening?

Job level 4 still takes about 21 actions.  Job level 3 takes about 12 actions.  The equivalent of about 12 actions happened in Wi/Pr/TM/Or before the chapter 1 checkpoint, so subtract those off.  That's...enough JP to be a calculator, just barely, but not enough JP to have Mathskill up and running.

Is Ninja happening?  That's one JL4, two JL3s, two JL2s, 21 + 12*2 + 4*2 = 53, but a few actions in Chapter 1 are enough to make this a yes, yes Ninja is happening, as is Samurai with just the basic Koutetsu learned.

Is Bahamut happening?  Um...actually yes, 1200 JP all in one job (or more realistically 1400 JP all in one job because you're not skipping Ramuh) is helped a lot by the job level system.  It's just barely enough JP and just barely enough MP but yep, it's happening.  That said, honestly, there's no JP wiggle room with that.  Might be better to go 300-350 JP cheaper with Leviathan or Odin, and then get MAU.


Alright, setups:

Summon
Wizard MAU Odin: 268 damage (50 MP, 9 ctr, 4v3)
Wizard Bahamut: 231 damage (60 MP, 10 ctr, 4v3)
Wizard MAU Ramuh: 201 damage (24 MP, 4 ctr 3v2)

And the moral of today's story is...why bother, when Thunder Rods are the only equipment your Wizards are going to be using anyway?

(Granted, if you're not in Wizard then Ramuh damage is comparably smaller.  In like...Priest it goes down as low as 100!)

Alright, how good is Draw Out?

Wizard Koutetsu: 144 (132 with Battle Boots)
Samurai Koutetsu: 96 (84 with Battle Boots)

It's not Summon, but it might fight for the #2 slot.

How good is Jump?  Lancer PA is 7, I'll assume Power Wrist for this.  You're looking at like....around 1300 JP for this, so LJump2, VJump3, LJump8?  Something like that.  So...8 range is not unreasonable at this point.  It's no like everything else has been tabled for it.

Lancer Jump: 108 damage (8 ctr)

Is Dance happening?  21+21+12+12+4+4 = 74.  Nope!  Not enough JP.

(Bard probably could happen, but why?)

Ninja:

They have 7 speed, 7 PA.  Green Beret and Power Wrist are available to boost damage (as well as headgear)

Martial Arts + Headgear + Power Wrist: 156 (only 7 speed)
Martial Arts + Green Beret + Power Wrist: 120 (8 speed)
Equip Sword + Green Beret + Power Wrist: 128 (8 speed)
Platina Daggers + Green Beret + Power wrist: 80 (8 speed) (100 with Attack Up)

Throw Ball: 64 ranged damage (56 if you don't wear Green Beret)

I'll be blunt--wow, Attack Up and Concentrate are kinda mediocre at this point.  Concentrate gets points for boosting Throw, at least, so there's a reason to consider it.  But otherwise...Martial Arts; Equip Sword is even better if you have it.  Although maybe flails...?  Nope, not out in stores yet!

Oracle physicals:

77 damage, 2 range

Time Mage physicals:

55 damage

Not amazing, but similar to what Knight or Geomancer can do, so...worth mentioning.

Monk:

Earth Slash: 48 damage (30 if you don't use Power Wrist)

It's not terrible, but you could be a Chemist with a gun dealing 36, and that takes no training, so it's pretty underwhelming.

Hell, a male monk could set Summon Magic secondary, and deal like...40ish damage with Ramuh?  Admitedly, only enough MP for one Ramuh.  But the point is this Earth Slash plan isn't really paying off yet.  It's not painfully bad; it's not Thief or Calculator.  But it's not good either.  You're training in Monk now for a payoff later.

Geomancer:

Males have...6 PA, 4 MA.  Use MA+4 equipment to get the MA+2 = PA sweet spot and you have...

32 Elemental damage

Going for higher PA instead cause that also boosts your physical attack...

30 Elemental damage (But 64 melee damage instead of 48)

Mmmm...well Chapter 2 is the "mage chapter", I was thinking maybe Elemental would benefit from all the free MA, but...ehh...not enough.  Like...Ninja's Throw outdamages Geo's everything, while having more speed.


-----------------------


Overall best archetypes:

1. Any mage with Summon.  Like...even when training in Oracle/Time Mage, you're still going to outdo any physical setup.

2. Black Magic.  Like...Bolt 2 without MAU is still doing 113 damage at range with AoE.  If you had any real incentive to min-max Black Magic well...at least consider MAU for 150 damage here....

3. Lancer.  I think range 8 with 108 damage is more impressive than Ninja melee damage where they need to use a Power Wrist over Battle Boots to get decent numbers.

4. Ninja.  The numbers really are quite high, especially considering they come with lots of speed.

5. Samurai. Samurai are kind of in this awkward "just recently unlocked" phaze, where if you rush to Samurai, you won't have Teleport, you won't have Magic Attack Up, you probably won't have Move+2.  You won't have a good secondary to stick on Samurai.  Still lots of damage.  Similar melee instant damage to Ninjas, but with good AoE...except unlike Ninjas they don't get into range easy or have high speed, and they need to stop training Samurai to get the good damage numbers.

6. Holy (white magic).  Nuking a single enemy and then running out of MP immediately...there aren't really any assassination missions in Chapter 2, so this is underwhelming.  But like...still an option I would take seriously.  (The next best "blitz them down" offence option is...something non-serious like...quite possibly Oracles with Sticks or some shit).


Healing/buffing/R/S/M/Utility:

1/2/3: Chemist, Time Mage, Squire

* Chemist: it's not entirely weird if half your characters are still using pheonix down.  Also, Auto-Potion is an option for people who don't need buckets of JP, like Summoners
* Time Mage: Haste is really good.  Teleport is near Auto Potion levels of busted, but in a class that Summon types are happy to train in.
* Squire: if you want to get anywhere in Ninja/Lancer/Samurai, you need Squire pretty badly.  For summoners, it's not as high a priority now.  Also most of your characters probably have Move+1 now.

4/5/6/7/8: Wizard, Thief, Monk, Knight, Oracle:

* Wizard: MAU is good; you're probably still training and only setting it if you hit a hard fight, though.
* Thief: Ninjas and only Ninjas get Move+2, but they love it.
* Monk: Martial Arts looks like the premium Ninja support in this Chapter, unless you can get Equip Sword.  Also, some MP restoration.  Also sometimes a reaction ability like Counter.
* Knight: Equip Sword technically best Ninja support, and Weapon Guard is still very much appreciated by mages
* Oracle: Lots of nice tricks that you don't -need- to invest in, but you have the spare JP to do so now, like Silence Song and Life Drain and Spell Absorb.  Move-MP Up is also fine, although I think it's most attractive for Calculators in training (who might want the MP Switch combo and aren't rushing for anything in particular in Oracle).

9/10/11: Mediator, Bard, Archer

* Mediator: Invitation?  Only need it for one fight, and it's basically free.  Maybe Faith raising if you're in the mood.  Neither of the benefits of these will really be significant till Chapter 3, though.
* Bard: Angel Song is pretty good, although it's fairly out of the way, and only Ramza can go for it, which is a lot of investment for just a utility move.
* Archer: Concentrate still has decent arguments for it on a Ninja, unlike Attack Up.  Charge can be nice on a Chemist.


Actively Painful classes to be in at this point

* Calculator
* Bard
* Thief
* Monk without a skillset (Unless you equip something other than Battle Boots, which...you're pure melee, so that has problems too, you're usually looking at 40 damage melee only character; maybe 50 with some brave raising.  And...HP only slightly higher than Triangle Hat Wizard Robe mages...and lower than Ninjas).
* Knight without a skillset (But if it has Summon or Black Magic it does fine TBH.  It can use Wizard Robes, and has a decent MA stat)

Hmm....how is Archer at this point?  Lightning Bow is 6 WP.  You can attack for about 42 with that.  Eh, not quite bad enough that I'd file it under "painful".  Slightly more damage than guns even!  But it can miss and has less range usually.  Still not something to file under "painful".
« Last Edit: October 31, 2015, 01:19:27 AM by metroid composite »

Dark Holy Elf

  • DL
  • Denizen
  • *
  • Posts: 8161
  • Well-behaved women seldom make history
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #605 on: October 31, 2015, 01:29:09 AM »
Looks right so far.

Random thoughts for C1:
-Priest is obviously better than many of the random leftover classes, and may even compete for that 5-6 slot. They do only slightly less damage than other fighters, have decent enough MA/MP for magic secondaries, and Cure alone is a cooler skillset than most others at this point (it's great economy if you can AoE it effectively, in particular). I'd certainly value it above Haste, though TM's +1 MA may make them slightly better anyway... still, it's competitive.
-I see Weapon Guard get brought up as a mage reaction pretty often, but I feel it's a touch overrated? A modest evade push just doesn't feel worth spending time in Knight to me; all the mage jobs are pretty big JP sinks. If you get it through spillover from someone else going into Knight, though, it's a nice bonus, sure, so I suppose this is still making a small case to send someone to Knight, anyway.

For C2:
-Well Wave Fist should really be mentioned for Monk. 60 ranged damage probably keeps them off the "painful" list, at least as much as it does Archer (who has less damage but more range/move).
-Yeah, Attack Up/Concentrate Ninja hype really relies on them having weapons better than pathetic 5-WP knives.
-Holy is mainly notable as one of the better ways to wreck Queklain. Obviously Ramuh still works pretty well (slower summons not so much, dat 9 speed)... but one Wizard with Holy can all but solo the fight even if the rest of the team gets wrecked by Bio/Nightmare or something. (Heck, I've never specifically set up for this, but I bet good-or-best zodiac MAU Holy actually pulls a OHKO there.) It also one-shots Gafgarion. Otherwise yeah, lack of assassination battles makes it questionable before then.

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

Grefter

  • Villain.
  • DL
  • Denizen
  • *
  • Posts: 10386
  • True and Honest. Smarter. More aggressive.
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #606 on: October 31, 2015, 01:50:24 AM »
I think Weapon Guard is almost exclusively pushed from an expectation of spillover JP?  I kind of assume it comes about from a mix of a character or two climbing the physical branch and Delita after JP Gained Up is acquired?
NO MORE POKEMON - Meeplelard.
The king perfect of the DL is and always will be Excal. - Superaielman
Don't worry, just jam it in anyway. - SirAlex
Gravellers are like, G-Unit - Trancey.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4378
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #607 on: October 31, 2015, 07:11:45 AM »
Yes, Weapon Guard primarily through spillover is the idea.

You can even pretty easily get guest spillover if you already have Move+1 and Gained JP Up on everyone who wants it.  Gafgarion and Agrias are a-ok with being Knights.  (Chemist is the other obvious Guest Spillover choice if you're done with Squire spillover, but you can't really get Gaf/Agrias guns without taking guns away from Mustadio, so that's a big power drop.  Rafa, on the other hand, I'll make Rafa a Chemist all day).

-Well Wave Fist should really be mentioned for Monk. 60 ranged damage probably keeps them off the "painful" list, at least as much as it does Archer (who has less damage but more range/move).

Oh, trained Monk is fine.  Not like...super good or anything, but passable.  Trained monk is like...slightly better than untrained Chemist!  Maybe!

Partially trained Monk (with just Wave Fist but not Earth Slash) is still fine.

Wave Fist is a 300 JP ability, though.  Like...if you're a Ninja, and you're grabbing Martial Arts on your way through, you will never actually learn Wave Fist.

If you're a Samurai, and you're not interested in HP Restore, then sure, you'll have Wave Fist for about half the time you're getting 550 Monk JP.

Quote
-Priest is obviously better than many of the random leftover classes, and may even compete for that 5-6 slot. They do only slightly less damage than other fighters, have decent enough MA/MP for magic secondaries, and Cure alone is a cooler skillset than most others at this point (it's great economy if you can AoE it effectively, in particular). I'd certainly value it above Haste, though TM's +1 MA may make them slightly better anyway... still, it's competitive.

Hm, interesting...

I mean...Haste, if you hit 4-5 people with it, and there's a lot of flat ground to do so...

On clocktick 19, with 14 CT, your speed becomes 9.  10 clockticks later, on clocktick 29, you get your second turn.  Enemies who move-acted get their second turn on clocktick 34 (you can bolt 2 them a second time).  Enemies who move-waited get their second turn on clocktick 30 (ok, yeah, that's bad).  If you can manage to drop a spell on the enemies without moving, and also be in the haste, though, then you're....34 CT when haste starts on clocktick 19, which gives you your second turn on clocktick 27...NOPE, that's not fast enough to cast spells on move-wait enemies either.

Certainly if you're a physical setup though (Knight, Thief, whatever) getting hasted round 1 lets you attack twice before any opponent gets their second round.  This might mean not bunching up as much for haste, though.

Hm, yeah, maybe I'm overestimating haste.

That said, Priest numbers aren't great.

15 physical damage is...not much different from any other class at this point, but nevertheless still game-lowest.

34 healing with Cure (assuming your party has 70 faith across the board) is nice AoE healing.  I still wouldn't set it as a secondary over Potion+Phoenix Down on a Wizard, though--Wizards run out of MP.  (And on all other mage classes, Black Magic).

29 damage with Bolt.  Yeah, not bad.  Less than all the other mage classes (47 on Wizard, 41 on Oracle and Summoner, 35 on Time Mage).



I guess my biggest issue is that I'm not entirely sure the reason why you'd be in Priest in the first place.    Knight you might be on the way to Samurai or Ninja, you might be picking up Equip Sword.  You might just be chilling long enough in Knight to get all your mages Weapon Guard.  You might be Delita.  Time Mage you might be on the way to Summoner, and Time Magic secondary (with just haste learned) is an acceptable skillset on physical classes like Archer, Knight, Thief, etc, cause it mostly ignores MA and Haste is really good if you aren't worried about lining up CTs.  You might also be in Time Mage 'cause you want Teleport or will eventually want Short Charge.

But like...what's the motivation to be in Priest in the first place?  There's no rush to get Holy for chapter 2...or really any Priest ability at all.  I guess if you're en-route to Calc, then yeah, might as well do your Priest levels in Chapter 1, cause Priest gets worse in Chapter 2.

But sure, you can and do multiclass between all the mages fairly easily, cause they all have robes, decent stats,  and you only need 50-200 JP to have a decent skillset for multiclassing.  Priest has the worst stats in Chapter 1 (but not too bad), catastrophically terrible stats in Chapter 2, and actually pretty good stats in Chapter 3 where the extra point of speed is pretty nice.

Don't get me wrong, if Auto Potion was a Priest ability I would learn it way more often.  Priest is a lot nicer to be in than Chemist.  But like...if we're going there, Oracle can use Thunder Rods and has more MA, and thus is probably nicer to be in than Priest at this point (and certainly nicer in Chapter 2).

I'm trying to figure out what I'm actually trying to say here hmm....  Well...I guess mages are pretty spoiled.  All of them have both "pretty good stats" and "abilities I want to learn".  Priests having both the worst earlygame stats and the fewest abilites I want to train towards for Chapter 2 makes them...frequently get passed over for me I guess?  At this point in the game, anyhow.  (They pick up again in Chapter 3/4).

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4378
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #608 on: October 31, 2015, 09:01:38 AM »
Alright, so Chapter 3.  All that stuff I said wasn't quite within the cutoff for Chapter 2?  It is now!

Calculator

Wizard MAU CT5 Holy with Chameleon Robe: Deal 378 damage to all your enemies, and heal 441 HP on all your allies

CT5 Raise, CT5 Raise 2: Infinite range instant multitarget revival at half/full health if a bunch of people die early.  Notably, the hit-rate on Raise 2 isn't terrible (MA+160) just the ctr, which Mathskill ignores.

Time

Short Charged Meteor from a Wizard: 403 damage 4v3 area 7 ctr

So...here's the part that I actually really really like about sending at least one person down the Short Charged Meteor route.  Time Mage spillover JP is really good.  Fighters can use Teleport and MP Switch, and they're both probably better than whatever they had before (possibly except for Move+2 Ninja).  Summoners want Short Charge or Teleport.  Everyone might consider Demi for Zodiacs (it's only 250 JP).  Short Charge Meteor is theoretically not worth the slightly higher JP investment over Short Charge + large Summon.  But in practice?  It's good enough, and literally your entire party will appreciate Time Mage Spillover.

Summon

108 Gems MAU Leviathan from a Wizard: 351 damage 4v3 area, 9 ctr
MAU Magic Gauntlet Bahamut from a Wizard: 405 damage 4v3 area, 10 ctr

(Ramuh deals about 221 damage; that hasn't changed much.  3v2 area, 4 ctr)

Alright, so let's think about 10 ctr.

Speed 7 enemies get their second turn on clocktick 29.  Unless they move-wait, then they get their second turn on clocktick 26.

If you're speed 6, then you can expect to get your first turn on clocktick 17.  10 ctr is actually too slow to hit a move-wait opponent.  9 ctr is fast enough, though.

What about speed 8 opponents?  They get their second turn on clocktick 25.  This is too fast for either Leviathan or Bahamut (but Meteor will still nail them).  Unless they Move-Wait, of course, then they get their second turn on cloctick 23 (Meteor hits on clocktick 24).  This only applies to a few Thieves and whatnot (and possibly the Ninjas?  Can't remember if they're speed 8 or 9).

Whatever, point is, Bahamut is a bit of a red herring, and Summon builds should go Leviathan unless they're grabbing Short Charge.

White Magic

Holy is relevant!  All these assassinations!  Um...462 damage if cast from White Magic directly (with MAU 108 gems Wizard)

Ninja

Oh one sec, now I have to compute a stat that isn't Wizard MA.

9 PA
8 speed

Twist Headband, Power Sleeve Martial Arts: 342 damage (can boost this to 378 with Power Wrist, 420 with a little brave raising)
Green Beret, Power Sleeve, Concentrate: 200 damage (With Twist Headband and Power Wrist, can boost this to 220).
Equip Sword, Twist Headband, Power Sleeve: 260 damage (Diamond Sword?  Really?  In a full chapter swords only went up by 2 WP?)

It should be noted, Martial Arts is only really great if you don't want to use the Green Beret.  With the Green Beret it drops to 224 damage (288 with Diamond Armlet), which really isn't better than Attack Up.

Equip Sword definitely isn't better than Attack Up this time (Attack Up easily paired with Green Beret and the same damage).

Lancer

Alright, so you were pretty deep into Lancer before, and wanted to pick up Vertical Jump 7 or 8.  Given your job level how long does this take you...?  28 actions for Vertical jump 7 and Equip Spear.  (Why does Equip Spear cost 400?)  So...about 20 actions left...is this actually enough JP to unlock like...Ninja?  I'm...not sure it is.  (Do we even want to unlock Ninja?  8 speed in Chapter 3 is bad for Jump TBH).

Alright, so the spear is Partisan (11 WP).  Equip Axe is honestly a pretty good substitute for Equip Spear, and you might get Equip Axe for free (16 WP means pretty similar jump damage.  Weaker regular attack, granted).

PA in Lancer is 8.  Jump damage from Lancer is 143 assuming Diamond Armlet.
Jump damage from a Geomancer or Archer with Equip Spear, Power Sleeve, Twist Headband, Diamond Armlet: 198 (192 if you go cheap and use Equip Axe...not that you'd need to set Equip Axe on a Geomancer!)

Samurai

Well...it's kind of like last time, except now Samurai get Heaven's Cloud, have time to learn Teleport and Magic Attack Up, and Blade Grasp is on the table for sure.  Ok, just how much shit is this, and can I really afford it all?

1460 Samurai JP, 400 Wizard JP, 600 Time Mage JP

Ha, yeah, no, that Samurai JP alone is like...nearly 50 actions, and we had only just unlocked Samurai last time.  Well, time to prioritize.  Teleport > Blade Grasp.  And...that's probably the only really necessary concession, because Wizard and Time Mage spillover will help.

Anyway, MAU Wizard numbers:

Koutetsu: 252 damage
Heaven's Cloud: 294 damage

(I'm not adverse to skipping one or both of Murasame or Heaven's Cloud either.  Not the end of the world to delay MAU either since most fights will be in Samurai with Gained JP Up for the forseeable future).

Damage numbers in Samurai with Gained JP Up are definitely starting to be kinda mediocre however.

Koutetsu: 108
Heaven's Cloud: 126

(This assumes Magic Gauntlet, although Red Shoes are available from Alma and not too big of a damage drop)

100 is...not trash, but it does relegate Samurai to being a pretty clear training class now (whereas near 100 damage in Chapter 2 was pretty standout).

Punch Art

Let's see...Ninja stats....

9 PA.  Can get +5 from equipment.  Sure, make it so.

Earth Slash damage: 147

Yeah, this is actually the real deal as far as Ninja secondaries right now.  It's long range damage that goes on a Ninja with Martial Arts (which is still really good compared to other supports).  It's better than Throw at this point by a reasonable amount.  It's better than Jump on a Ninja right now due to Ninjas being too fast to land jumps right now.  It doesn't miss, which is something Ninjas with Martial Arts lack.


(Ok, I need to sleep; continue this later...).

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4378
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #609 on: October 31, 2015, 09:10:46 PM »
Ninja

I miscalculated their PA in the previous post.

8 PA
8 speed

Twist Headband, Power Sleeve Martial Arts: 288 damage (can boost this to 342 with Power Wrist)
Green Beret, Power Sleeve, Concentrate: 180 damage (With Twist Headband or Power Wrist, can boost this to 200).
Equip Sword, Twist Headband, Power Sleeve: 240 damage (Diamond Sword?  Really?  In a full chapter swords only went up by 2 WP?)

It should be noted, Martial Arts is only really great if you don't want to use the Green Beret.  With the Green Beret it drops to 210 damage (224 with Diamond Armlet), which really isn't better than Attack Up.

Equip Sword definitely isn't better than Attack Up this time (Attack Up easily paired with Green Beret for the same damage).

Punch Art

Recalculating this given the bad Ninja stats...

8 PA.  Can get +4 from equipment.  Sure, make it so.

Earth Slash damage: 108 (114 with Diamond Armlet)

Well...this is less impressive now that I've recalculated for my miscalculation earlier.  This is still probably the best secondary for Ninja right now, given that Jump just doesn't work at all (you don't want to be 8 speed with Jump in Chapter 3; you'll never hit anything).  And Martial Arts is probably still the support of choice for Ninja.  And if you're using Martial Arts, Earth Slash gives you a 100% hit move.  But yeah, if you're willing to shell out cash, Throw deals 128 damage (144 with Green Beret).  Without shelling out cash, you're throwing balls, so it's 64 (72) damage.  So...Earth Slash is nice to have, but not a massive increase to the quality of the build at this point.

Monk

Of course, a more reasonable path might be to rush to Ninja first, and then return to Monk in Chapter 3 when they have Power Sleeves.  How good/bad is Monk at this point?

8 PA
7 speed

Wait...8 PA?  There's no way it should be lower than Ninja?  Let me check...yeah, must have miscalculated with Ninja; time to go back and redo that.

Power Sleeve brings that to 10.  (Diamond Armlet to 11, but that's not actually very useful).

75 Earth Slash Damage
90 Wave Fist Damage

You no longer have a big physical attack either (105 damage physical attack).

I...honestly expected more.  They have power sleeve now!  But their punch is less than Koutetsu from a Samurai with Gained JP Up.

Dancer

5 PA. (female Ninjas will have 6 PA)

PA+5 equipment, no reason not to.

Wiznaibus deals 17 damage...to all enemies, 2-3 times per round.  (Ignores all damage modifiers).  Assuming 6 enemies this is 204-306 damage per round.

It has multitarget curse, of course, where leaving all your enemies alive is actually not as good as focus firing down...any target really.  But yo, that's actually pretty solid damage.  Better than Lancer.  Worse than a Ninja rushing into melee range.  But you know, infinite range and all that.  It's kind of like having a Romanda Gun and shooting every enemy once per turn from a speed 9 character and infinite range.

Know what doesn't have multitarget curse, though?  Nameless Dance.  Nameless Dance hits a status 50% of the time, and a relatively fatal status 25% of the time.  With 6 targets, on average 1.5 are taken out of the fight every 10 clockticks.  The downside is that it's random.  But on the flip side, this is technically more targets per clocktick than Ninja incapacitates.  10ctr wait before the first Nameless Dance, though, so realistically Ninja gets a head start, and Dancer probably doesn't catch up.

Dance is....actually pretty good TBH.

How about the Dancing Samurai plan?  Where you use Dance at long range to train as Samurai, and clean up at close range with Draw Out?  Is there enough JP to have both Samurai and Dancer unlocked at this point?  Hmm...yeah, looks like it, with about 20 actions to spare.  (Enough to get Teleport if you want it, although you probably want to catch up on Samurai JP instead).

What about Dancing Calculator?  Can you have both Dancer and Calculator unlocked at this point...?  Nope!  Not enough JP!

(ok switching computers, will continue)

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4378
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #610 on: November 01, 2015, 01:57:00 AM »
Guess I haven't computed

Geomancer

7 PA, 5 MA.

Twist Headband, Power Sleeve, Diamond Armlet

120 damage physical (well...I believe that's better than Monk)

42 damage Elemental.  Can make it 48 damage with Wizard Robe.

Geomancer, I'm glad you finally joined the Chapter 2 club.

R/S/M/Healing/Buffing

Mathskill, of course, surpasses all other healing/buffing skillsets at this point.

Other than that, not a whole lot has changed.  Squire/Chemist/Time Mage are still the premium. 

Dancer now exists, but I'm filing it under offence, cause it mostly is.

Wizard is becoming more and more important for MAU as well. 

New noteworthy stuff: Oracle/Time Mage/Summoner pop up a bit for Demi/Life Drain/Lich

I think Move-MP Up (Oracle) is increasing in value, as lots of spellcasters learn nuke spells that take almost all their MP

Mediator is probably now in the tier 2 level, as faith modification kicks into gear a little bit

Punch Art is now a viable secondary for Ninja, although it's not that big of an upgrade.

Geomancer is now kinda real as a carrier, specifically it's the best class to take Jump at the moment (and the best class to take Agrias whom I haven't been calculating).


Overall Offence plans/skillsets/whatever

1. Math Skill

2. Time Magic (Yeah, I think it's a little bit better than Summon at this point.  More damage for less ctr.  Investing in Time Mage gives very valuable spillover JP to your whole team.  It's still very close and only narrowly ahead of Summon even at this point).

3. Summon (While Summon and Time could be swapped probably, there's no question that they're #2 and #3).

4. Ninja (Next highest damage per clocktick.  Yeah, Jump has 8 range, and about 60% of the damage, but it also has only 75% of the speed.  The damage gap has widened, despite the fact that to even maintain this damage gap, Lancer needs to go to a carrier like Geo or Archer, which means you can't have a good secondary like Item.  Ninja is still flexible enough to work with lots of secondaries, while having lotss of damage).

5. Draw Out (unfortunately Samurai need to still spend most of their time training, which means most of their time in the Samurai class.  Fortunately it's...still pretty good now that there's time to get Teleport and maybe something like Blade Grasp or Auto Potion to negate the dangers of teleporting into the enemy team.  If you catch two people in a Koutetsu from a Samurai, you deal more damage than Jump.  If you catch two people in a Heaven's Cloud, you deal similar damage to the maximum output of Wiznaibus.  You can actually boost the Samurai damage more than I listed, using C-Bags.  And catching more than two people is not unreasonable.  The real kicker, though, is that if you hit a hard battle, you can just switch to mage and throw on MAU and be better than Ninja.  But yeah, still mostly training, because Kiyomori, Muramasa, and Kikuichimoji cost 1740 JP, and you want them all).

6. Dance (you...outdamage Jump if you have 6 targets, and you do so with more range.  You incapacitate targets at a faster rate than jump.  Dance is only really bad in assassinations.  But like Ninja it's a lot more flexible than Jump--you can get most of the power of Dance on nearly any class.  Dance is probably slightly better than Draw Out from Samurai stats, but Samurai have the panic button of "spend one fight as a mage".  Dance might be somewhat competitive with Ninja in some fights, but Ninja is sooooo much better at assassinations).

7. Jump (almost 200 damage that can't be dodged at long range is still quite good.  The bad is that you are limited in what classes can use it, and you don't get much of a skillset if you want PA.  Your options are Geomancer/Archer (pump to 12 PA), Monk/Thief/Priest/Dancer/Squire (pump to 11 PA), Chemist/Mediator (pump to 10 PA).  Huh...o...kay, I guess it really just needs a class that can wear Twist Headband and Power Sleeve and doesn't have mage level PA.  At 11 PA you deal 176 damage.  At 10 PA it's 165.  Compared to 198 at 12 PA.  So yes, there's potential there.  You won't have a skillset in any of these classes yet, though.  Lancer is a JP hog, and only just finished learning Equip Spear.  Also, if you're using Equip Spear, and you need to in any class that doesn't equip axes if you want damage, then you won't learn your skillset fast.  Still, at some point you'll have jump + a skillset like Dance/Punch Art/White Magic).

8. Yin Yang Magic (YYM is like...a collection of situational spells.  But learn them all and you're pretty good in every fight.  There's enough JP floating around to justify that now). 

9. Holy (You...actually can't get enough mana for two holies easily, so that sucks.  45 MP from your stats + 54 MP from your equipment is still less than 112 MP.  Nearly 500 damage from a Wizard, but only singletarget and only once.  Fine in assassinations, but not great otherwise).

10. Black Magic (Bolt 2 with MAU is...166 range AoE damage.  But enemies dodge now; there's Aegis Shields in this chapter--50% magic evade.  And Elf Mantles are not unheard of--25% magic evade.  And then Diamond Shield with 15% magic evade--the highest shield evade before this chapter was 5%.  The problem with evade and charge times is that your allies don't even know whether you're going to kill your target; do they need to attack it or not?  This would be fine if the damage makes up for it, but the numbers are similar to jump unless you really tried to force BM to work and picked up Flare or something.  It's still something you'll fall back on when you're using Wizard as a carrier and run out of MP for your big bad summon/Meteor).


Anything below this and I'd say the character is definitely in training.  (which is fine.  Still two more checkpoints.

DragonKnight Zero

  • Denizen
  • *
  • Posts: 518
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #611 on: November 03, 2015, 02:36:33 AM »
Random scattered thoughts

Priest, Time Mage, and Oracle I find underwhelming until after Zirekile.  Oracle can use Paralyze to shut off boss' Reactions in Chapter 1 but offers little else other than Silence Song.  Yes, one can give a Thunder Rod to an Oracle though might as well be a Wizard unless the Oracle JP is part of a longer term plan.

Monk really hits its low point in Chapter 2 due to a lack of PA boosters and better weapons for other fighters.  Still very much a mage's playground due to equipment availability.

Short Charge + Meteor already?  Guess so with JP up available.  in a magic users challenge run, I found that keeping a unit in the same class for the whole game netted them about 4000 JP.  Didn't have Short Charge + Meteor until the Morund sequence.  Wasn't using JP up and had spent JP on Haste, Slow, Don't Move, Demi, and Teleport so the Time Mage could contribute throughout.

Speaking of mages, Haste worked out nicely later in chapter 2 when their physicals are at their relative peak.  Easiest Baraius Hill battle I've ever had.  Not enough time for another attack magic sure but bashing things on the second turn with sticks and staffs worked well enough.

Tide

  • Malice Tears
  • Mod Board Access
  • Denizen
  • *
  • Posts: 1373
  • Cacophony of Sorrow
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #612 on: November 06, 2015, 02:08:07 AM »
Quote
Yeah...I don't count JP scroll as legal.  Some skillsets are JP scrollable.  Some are not, and this means that some jobs that aren't scrollable are just arbitrarily garbage.

Funny story, there actually IS  a No Math Glitchless route. It is however not as optimized as the other categories because no one really has an interest in it since without JP scroll glitch, battles tend to be more random, generally slower and sometimes more dangerous. Not touching the question of legality of the JP scroll glitch since that's not what this is about.

Rifling through the notes, here's a few notable changes I pick up on immediately:
- Compatibility matters. A lot. You are however allowed to take 2 Taurus instead of all Capricorn, although 1 generic must be Capricorn.
- Gained JP is never required.
- Mustadio and Agrias are both recruited. Agrias does not change jobs as it much slower and she cannot be sent on any Props, which your generics can.
- Holy is never used.
- Flare is never used.
- Physical classes, despite having potentially faster animations are never used outside of Squire. Holy Swordsman and Holy Knight are used but that is because they are they base class for Agrias and Orlandu. This would primarily be due to JP mapping and you do not have time to swap from Mage classes to Physical classes. Mage classes are obviously superior.
- Ramza never enters Samurai due to the above, even discounting the JP costs on Draw Outs.
- Quick is used relatively little if at all. Most likely to save proposition time as well as JP efficiency. Without Draw Out, Quick is also less efficient since you lose no charge time damage. Ramza never has time for it as he MUST learn Demi, Teleport and Haste.
- The route primarily uses 3 Female Generics with 64 Faith minimum + Ramza, Orlandu and Agrias. Most of the time, you sacrifice initiative so you can sync your CT to go immediately after and drop Ramuhs and Titans.


tl;dr relevance to topic:
- Summon Magic is most important with Time Magic as second for utility for real time purposes.

I would actually be very interested in what the DL thinks of the overarching game plan and if there are better ways of time saves. I'm sure other FFT runners would appreciate it as well, although this is not a popular category at all. Think of it as theorycrafting (like the topic title)!

Jobs Paths look like:

Ramza
Squire (Move+1), Chemist (Potion/P.Down), Wizard (Bolt, MAU), Time Mage (Demi, Teleport, Haste), Summoner (Ramuh, Titan)

Generics A+B+C
Squire (Move+1), Chemist (Potion/P.Down), Priest, Wizard (Bolt, MAU), Time Mage, Oracle (Life Drain), Summoner (Ramuh, Titan)
« Last Edit: November 06, 2015, 02:35:13 AM by Tide »
<napalmman> In Suikoden I, In Chinchirorin, what is it called when you roll three of the same number?
<@Claude> yahtzee

<Dreamboum> Everyone is learning new speedgames!
<Dreamboum> A bright future awaits us gentlemens
<Pitted> I'm learning league of legends
<Dreamboum> go fuck yourself

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4378
    • View Profile

ParaSC2

  • New User
  • Posts: 5
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #614 on: January 13, 2016, 07:29:39 AM »
It seems like I don't have the ability to make new threads so I'll just post this here.

I recently returned to SC2 (again) and got interested in a new arcade game called StarJeweled. It's basically Bejeweled with SC2 units added in. I don't really enjoy playing the game itself though... matching groups of gems repeatedly gets boring and I'm not that great at the game (I average around 1200 EPM). But I found that I really enjoy making a robot to play the game for me.

Currently my bot (written in Python) averages around 3500 EPM, give or take 500 energy depending on luck. And it seems like I've hit a plateau at 3500. I've spent a good amount of time trying to come up with a better algorithm to improve the bot's performance to the mid 4000s range, with no luck. The current algorithms I've come up with involve thinking multiple moves ahead, but they're way too slow and actually end up cutting the performance in half (1700 EPM, yikes) instead of improving it.

I'm curious anyone on these forums have made a StarJeweled bot before that has broken the 4000+ EPM barrier or is interested in helping to come up with a better algorithm.

ParaSC2

  • New User
  • Posts: 5
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #615 on: January 13, 2016, 07:37:23 AM »
Bot is written in Python and uses the following modules: pyscreenshot, autopy, itertools
All testing was done in a 13" Macbook Air with 1.7 GHz i7 (Haswell) and 8GB RAM, running at 1440x900 resolution and medium graphics

Algorithm:

1. Take a screenshot of the 480x480 gem grid using pyscreenshot and store it in memory
2. Go over each gem (60x60 square pixels) and calculate the average RGB value of 20% of the pixels (spread evenly)
3. If the RGB values fall within a given tolerance, predict the color of the gem (right now, this is almost 100% reliable if no gems are falling or text is on the board)
4. Return a 2D list representing the colors on the board
5. Given this 2D list, find all pairs (eg: two gems together vertically or horizontally, or two gems with a gap in between)
6. For each pair, find a gem that can be moved next to it to make a group of at least 3. If it exists, then add it to the move list
7. Go through the move list and calculate a score for each move (Score = Points Gained + Moves Created * 5)
8. Order the moves based on score, from best move to worst move
9. Process all these moves in order by moving the mouse and clicking


Currently some flaws with the algorithm are that some moves can conflict with each other and they are not removed from the move set (however, conflicts don't seem to be too costly in terms of EPM performance), and we don't wait for chains of gems (eg: swapping two gems and then having a match after the gems fall), so this lowers EPM.

ParaSC2

  • New User
  • Posts: 5
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #616 on: January 13, 2016, 07:44:04 AM »
Actually I'll just post the bot code here in case anyone's interested in taking a look. It's open source on Github.


[Admin note -- link deleted ~mc]
« Last Edit: January 15, 2016, 06:38:51 AM by metroid composite »

SnowFire

  • DL
  • Denizen
  • *
  • Posts: 4964
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #617 on: January 13, 2016, 08:16:12 AM »
Well, you came to the right place, as metroid was like master tier at StarJeweled...  I dunno about calling it a "new" game, though.  It came out with SC2 arcade, which was....  uh....  checking my achievement history....  September 2011?  So pre-Heart of the Swarm, still Wings of Liberty days.

Your algorithm sounds solid as a first pass, although from looking at the code, maybe I'm missing something, but the point calculation looks off - where are you adding in the chain reaction bonus?  StarJeweled hugely rewards follow-on matches off chain level & all, so maybe I'm just missing it, but that seems super-important for getting a greedy algorithm right.  I also think you might be overrating creating new match possibilities a bit - pressing the "refresh board" button is not THAT bad if you somehow lock yourself in.  I don't see logic about pressing the button, although I guess it's safe to ignore if you're always opening new possibilities, sure.

I'd be curious to see the more detailed logging about where you're spending your time, since spending time efficiently is pretty important here.  Notably, how fast is our take-a-picture & OCR the board cycle?  If that's really fast we can be a lot lazier about rankings and just take a picture of our new pieces more often, if it's really slow we need to be more careful.

As an irrelevant nit, if you want to avoid duplicates in your potential-move-list from the start, you can use a dictionary instead.  i.e.
maybeMoves[str((x + 1, y - 1, Down))] = (x + 1, y - 1, Down);
if you're using the string representation still.  (then iterate on the values.)

ParaSC2

  • New User
  • Posts: 5
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #618 on: January 13, 2016, 09:42:03 PM »
The points from the chain reaction are calculated using the _get_points() function from the Match class. It is equal to OriginalPoints * (2^Chain). Right now my algorithm delays 250 milliseconds for every chain level, so
"Good Job" results in 1/4 second delay, "Fantastic" 1/2 second, etc. The appropriate delay time would vary based on the lag. My algorithm does overcreate match possibilities. Out of all the possible moves it generates, only about half of them can be made concurrently without conflicts since making some moves will invalidate others. Currently the bot does not try to reset the board.

Right now each "cycle" (which involves screen grab -> OCR -> decision -> make move) takes about 200 - 250 milliseconds. Breakdown is as follows:
Screen grab -- 150 ms
OCR -- 20 ms
Decision -- 30 ms
Make moves -- 10 ms


So screen grab takes the most of the time. 7 screenshots a minute isn't really that fast, I'd rather prefer 20 but it seems unlikely. I'm currently using pyscreenshot module for this which is the fastest library I could find. I tried other ones but they were 2-3x slower.. I think I could definitely speed up each cycle to 100-150ms if I either rewrote the bot in a faster language like Java, or ran it on Windows with a faster computer (I'm currently using a 1.7 GHZ ultra-low-voltage processor which obviously don't help).

Here's a video of the bot in action: https://dl.dropboxusercontent.com/u/32170031/StarJeweled.mp4



metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4378
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #619 on: January 15, 2016, 06:45:50 AM »
I'm kind-of not comfortable posting the github link on these forums (since this topic is a known location people come for non-bot starjeweled advice).  To put things into perspective, I actually try to limit how much I play Starjeweled these days, because I know when I play a lot, new players get frustrated with losing all the time and quit.  And I'm not a bot, I'm just a somewhat fast human.

5k has already been done with decision trees, although the bot was never released to the public (it was just someone doing a programming project for fun).

Once you can match that with decision trees, though, you should be able to pass it by adding in some of the known glitches.

DWIR is when you swap two gems almost simultaneously in a way that would cause a match if they could be swapped at the same time...and then spam reset.  Should be very easy for a computer to do.  Described here (it's actually not something I do myself when playing, so use this description not mine):

http://starjeweled.tumblr.com/

And there's also DWIM--I'm not sure exactly how it works, but it involves swapping two blocks that don't match, having the swap fail and that somehow keeps your combo counter.  Again, not something I personally use in my own play, so my description is going to be a bit crappy.
« Last Edit: January 15, 2016, 06:48:29 AM by metroid composite »

ParaSC2

  • New User
  • Posts: 5
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #620 on: January 18, 2016, 06:01:42 AM »
Interesting bugs... never knew they existed.


Grefter

  • Villain.
  • DL
  • Denizen
  • *
  • Posts: 10386
  • True and Honest. Smarter. More aggressive.
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #622 on: February 28, 2016, 11:26:10 AM »
I like that you mathematically come to the solution to the problem where as the best answer to using point but to me is to avoid drama and just for the sake of expediency.
NO MORE POKEMON - Meeplelard.
The king perfect of the DL is and always will be Excal. - Superaielman
Don't worry, just jam it in anyway. - SirAlex
Gravellers are like, G-Unit - Trancey.

Dark Holy Elf

  • DL
  • Denizen
  • *
  • Posts: 8161
  • Well-behaved women seldom make history
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #623 on: February 28, 2016, 11:14:34 PM »
Nice.

As a DM I favour pointbuy just because it's fair. You avoid situations where players may be jealous of others who rolled better stats, or worse where some characters are just legitimately much stronger than others and it makes the players of the weaker characters feel bad.

It's nice to see the math, for sure, but in practice the concerns of the playstyle and group will take precedence over the slight differences over which is optimum in a vacuum, almost every time.

I should also point out that calling 5e a "three stat game" can be a little misleading, and certainly isn't always true. (e.g. some builds such as non-arcane rogue pretty much only care about two stats for combat; not everyone uses Dex for AC; some characters needs durability far more than others in general depending on role) And everyone is different, but personally a 15/15/15/8/8/8 stat build makes me twitch: that's going to be a lot of skills you're hideously bad at, and anyway no race is well-positioned to use 15/15/15 besides human (as you'd need a feat to improve the third 15 to an even number, and the only other race that boosts three stats, half-elf, boosts one of them by +2 anyway).

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

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4378
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #624 on: February 29, 2016, 01:28:33 AM »
I should also point out that calling 5e a "three stat game" can be a little misleading, and certainly isn't always true. (e.g. some builds such as non-arcane rogue pretty much only care about two stats for combat; not everyone uses Dex for AC; some characters needs durability far more than others in general depending on role) And everyone is different, but personally a 15/15/15/8/8/8 stat build makes me twitch: that's going to be a lot of skills you're hideously bad at, and anyway no race is well-positioned to use 15/15/15 besides human (as you'd need a feat to improve the third 15 to an even number, and the only other race that boosts three stats, half-elf, boosts one of them by +2 anyway).

Mmm...well I'm joining a campaign that's above level 4, so I can use an ASI pretty much regardless of race.  I.e. if I wanted to run 15/15/15, then I would put +2 to one of those stats, +1 to another stat, which at level 1 would be 17/16/15, buuuuuut then at level 4 would become 18/16/16 by using the "+1 to two different stats" option.  (I don't plan on running 15/15/15 though).

Quote
e.g. some builds such as non-arcane rogue pretty much only care about two stats for combat

Oh, well, yeah.  If you only care about two stats in combat, then you can go 15/15 or 15/14 depending on your class bonuses, and still get like...13/12/10/8 for the rest of your stats.  Rolling instead of pointbuying will give you somewhere in between 15/15 and 15/14 on your key two stats on average.
« Last Edit: February 29, 2016, 01:32:32 AM by metroid composite »