Theorycrafting! (Because I like competitive metagames too much)

[metroid composite]

Messing around with Evolution Chamber now (the genetic algorithm for zerg build orders...).

Fastest route to...

5 Roaches:
8 Pool

10 Roaches:
11-Overpool, 14-Gas

20 Roaches:
11-Overpool, 11-Gas

40 Roaches:
10-Overpool, 14 Hatch, 14-Extractor, 30-Extractor (I...what?  You confuse me, genetic algorithm)

20 Roaches + Lair + Roach Speed:
13 Extractor, 15 Spawning Pool, 14 Extractor (Does it really take that much gas to get lair and one upgrade...?)

Hm, interesting; doesn't feel that informative, though...

[metroid composite]

Plants vs Zombies (Survival Endless)

So...it's bugged me a bit that the map of Survival Endless is fixed to the Pool map, notably because the pool lanes are very weak.  A lot of Survival: Endless players rave about how Glooms are the best thing ever, when the Pool really inflates glooms.  Allow me to elaborate...

Strategies often end up looking something like this:

http://www.youtube.com/watch?v=GJ2QQZzci7I

The pool lanes dedicated to stuff that will help the land lanes; the six pool Glooms, for instance, use AoE on the outside lanes.  The four Cob Cannons blast the outside lanes.  The pool sunflowers give you sun.  The pool Winter Melons are there for the AoE freeze+damage on the inside land lane.  Now stop and think--what else are you going to use in the forward pool slots?

Cob cannons?  The pool lanes may be super-easy, but they're not quite that easy.  (For all that yes, there are crazy Chinese setups with 16 Cob Cannons that actually do use these spots for Cobs, and just nuke dolphins constantly).  Twin sunflowers and sleeping freezies are also options--and yeah, you might do those, but then you need to place your offence elsewhere.

Which got me thinking--I know there's some hack that lets you play survival endless on other maps; how does that metagame shake out?

A brief google search turns up some stuff from Steam forums:

http://img246.imageshack.us/img246/7200/anotherday.png

The basic idea being to keep the middle three rows bombarded with Cob Cannons, and then kill the outside two rows with Gloom Shrooms.  But this is just some random guy on Steam who never managed to get the build particularly sustainable.  The most advanced Plants vs Zombies players tend to come from China, where....

http://tieba.baidu.com/f?kz=673175483
http://tieba.baidu.com/f?kz=671412511

8 cobs seems to be the answer (8 cobs going 1000+ flags, it looks like).  Notably this also covers a big weakness of the 3 cob build above (Gargantuars can throw their imps onto the 4th row, so cobs need to be placed on row 5/6, not 4/5).  It also has a clever way of sneaking cobs into the back rows.

The back two rows are in danger of being killed by digger zombies; the "standard" solution is to throw two Glooms in row 2 (which help kill imps as well).  And just do a little bit of repairing when needed.  Specifically, plants have a random wake-up time when they could start firing (something like 0-2 seconds), digger zombies spend a certain amount of time dizzy after they're done digging, which is shorter than the max wake-up time.  With two Glooms, they only get a rare nibble, easily repaired.  With three Glooms they'll never even get a nibble.

Now, with three cobs in back, it's not actually possible to cover them all with three Glooms.  The middle one is covered (four Glooms, in fact).  But the sides are not.  ...Which is where the Torchwood and 2x Split Peas come in.  If you watch him playing, he doesn't even pay attention to the digger zombies--they just auto-die.  (I have some concerns about Imps blocking the Split Pea shots a small percentage of the time, but it seems to be rare enough).

Now, those of you who've played some endless have probably spotted a gaping hole in this plan: where's the Umbrella Leaf?  There isn't one--the player just substitutes excellent APM and timing instead--constantly nuking the basketball zombies.  (Insert joke about Asians being genetically gifted with APM).

"But mc", you say, "if this is all about proving that Gloom isn't all that, then why does this player have 8 glooms?"  Well...6 of them are rear defence; glooms being one of three plants that shoot backwards, it's pretty natural in this role (and I don't think any sane person would argue Starfruit or Split Pea are better than Gloom in Endless).  The two in the front though...

The two in the front are interesting.  Here's why I suspect they're there.  Gigas have 225 HP.  The player can cob once on fast zombies, and twice on slow zombies (like Gigas) before they reach the front line.  That leaves an extra 25 damage to deal.  I think Melon Pults will actually have more DPS in this role, but they lack consistency--if everything in the Melon Pult's row is dead, it won't continue to blindfire and splash the surrounding rows.  Gloom, however, will continue to fire.  The only other plants that will continue to fire in that situation are...Threepeater and Starfruit, neither of which are really serious options (neither gets AoE).  So...sure, since you only need to deal 25 damage (and probably less than that because there are Winter Pults which should usually deal 20 or so; so like...8 damage) then yes, Gloom for the slightly better consistency.

But wait, China provides another fascinating build:

http://hiphotos.baidu.com/%C6%BD%D4%A8/pic/item/c6e28307e5a1b15503088134.jpg

The top two rows can take care of themselves, more or less (Zambonis will probably give fits to the firepeas, granted).  What's interesting is that there's no digger defence.  I haven't seen a video for this one, but I would guess that perfect regular timing on the Cob Cannons on the bottom three rows kills the digger Zombies while they're digging and still moving with the rest of the pack.  I know it can be done, I just wasn't aware that 5 cobs per 3 squares was enough to keep the Diggers constantly killed.

And there's also this one:

http://hiphotos.baidu.com/%C6%BD%D4%A8/pic/item/9ad2be42c5bd0a3f9213c60e.jpg

Flawlessly protected from diggers and basketball zombies.  But it's only 6-cob and needs to nuke both sides.  My guess would be that the extra walking space allows for three nukes.  Either that, or the extra walking space allows the Winter Pults to keep up.  It's also clever in that since the Cob Cannons nuke a 3x3 area, they're double-covering the middle three rows, which means that the middle row can squeeze in some sunflowers.  What I don't know is how this deals with Imps thrown by Gargantuars onto the fourth row (eating the Cob Cannon).  Maybe the player is just careful not to deal too much damage to Gargantuars so that they always walk far enough forward that lobbed imps hit the third row?

Anyhow...Night and Day are basically the same map.  Fog is basically Pool except annoying.  Roof is...

http://tieba.baidu.com/f?kz=669766721

Apparently like Night/Day except that there are no digger zombies ever, and Imps land on row 4/5 instead of row 3/4.  (Which...looks easier, other than that whole "if you want more than five cobs, you need cobs in row 6+7")  Which...on the whole sounds a lot more straightforward in terms of "here is the ideal setup" --although less forgiving with all the front row pumpkinless plants.

[SnowFire]

Night is not quite the same map as Day, since you will have gravestones forcibly upending your expensive plants, and will probably have to reserve a plant slot for the gravestone eater.  So you'd need more Sunflowers in your setup to be able to afford the constant replacement, and a much greater chance of randomly getting screwed over entirely.  It'd probably be the toughest endless map, actually.

That said, more seriously, I am childlishly amused at the Google Translate translation of the linked webpages.

* For boutique operations have objections to it on their own study of the Friends of the video, above, I Tuisan the
* Finally, you are not a shameless grab a couch, playfully down
* Meow a microphone harmony you 2 you O death
* You can separate the four guns hair, I will rosy, Keke this gun is already a mature 8 to moldy. . . . . .
* As long as there is love, What kind of roof wind

Good stuff.  Chinese is hard for computers!

[metroid composite]

Um, from the PvZ wiki:

"In Survival: Night and Survival: Night (Hard) if the whole lawn is filled the plant will be removed, and a grave placed there instead. However, if there is even one empty space on the right half of the lawn, the grave will not remove the plant."

So...doesn't sound like it should be an issue.  You weren't going to fill the whole lawn anyway--if you do you'll just be flattened by Zombonis/Gargantuars.

The sun drain, though, yes, is absolutely an issue.  When you only have two twin sunflowers, a decent percentage of your sun comes from the day.  And yeah, grave busting isn't free, and probably is still something you want to do (since Zombies spawn from graves, and you need the space).

As for Google translate, I'm guessing it has trouble with text that's full of gamerspeak shorthands.  Like...if I were to say "I six pooled and Snow went proxy two gate", most English speaking people would have no idea what I just said (even though a majority of DLers probably understood that sentence just fine).

[metroid composite]

Hmm...a bit of semi-sorta stat topicing for Plants vs Zombies:



"Fast" recharge plants (Sunflowers) take about 7.5 seconds to reload.

"Slow" recharge plants (Potato Mines) take about 30 seconds to reload.

"Very Slow" recharge plants (upgrades, Cherry bombs) take about 50 seconds to reload.

Cob Cannons take 40 seconds to reload.

Potato Mines take 15 seconds to arm.

Sunflowers produce sun right away, and then once every 24 seconds after that.

The normal attack cycle for peas, etc is once every 1.5 seconds.  The normal attack cycle for catapult type plans is once every 3 seconds.

A normal zombie crosses a square in 7.5 seconds (which is to say, they get hit five times by a pea in the time it takes them to walk one square).

Sun Shrooms grow large after 120 seconds (which means you spend 25 less initially, but get 50 less sun back over time, assuming a night level, which means Sunflower has a 25 more sun return on investment overall).

Sun Shrooms when they go big, and twin sunflowers when they are upgraded maintain the same old sun production schedule.

Butter seems to freeze zombies for about four seconds.

Kernel-pults fire butter approximately 25% of the time.  (I'm not entirely sure on this number, just statistical testing; it's also possible they've been buffed slightly, as they were slightly higher in-game than they were in youtube video I checked).

Which brings me to stats.

Peashooter:
Damage per tile: 5 (10 with Torchwood)
Self-defence tiles affected: up to 9
Other tiles: none
AoE: None (1x1 with Torchwood, dealing 5 damage)

Repeater--same as above but
Damage per tile: 10 (20 with Torchwood)
AoE: None (1x1 with Torchwood, dealing 10 damage)

Gatling pea--same as above but
Damage per tile: 20 (40 with Torchwood)
AoE: None (1x1 with Torchwood, dealing 20 damage)

Threepeater (Peashooter but hits three rows)

Cabbage-Pult (Peashooter with minor differences; ignores screen door shields, hits scuba zombies, but doesn't combo with torchwood)

Kernel-Pult
Hmm...ok, so one Kernel-Pult on its own is not interesting at all.  It has a DPS that's 62.5% of a Peashooter.  When being shot at by a Kernel-Pult you'll spend about 33% of your time buttered.  All of this comes together to make it about...94% of the damage per tile of a pea shooter.  That's...actually quite a bit better than I had guessed, anyhow.  With two Kernel Pults, it's now...

(1 - (0.75^2)) + ((((0.75^2) * 1) / 3) * (1 - (0.75^2))) = 0.51953125

So 52% of time that you spend buttered.  Now the damage per tile is 130% of a pea shooter.  With three kernel-pults it becomes...

(1 - (0.75^3)) + ((((0.75^3) * 1) / 3) * (1 - (0.75^2))) = 0.639648438

So 64% of the time that zombies spend with butter on their head.  Now the damage per tile is 174% of a pea shooter.

That said, there is a fundamental flaw with singletarget snare effects when combined with low DPS.  If there is, say, a single bucket-head zombie advancing on three plants, then yes, triple Kernel-Pult will perform better than triple pea shooter.  However, if there is a bucket head zombie spawning once every 30 seconds, three pea shooters will handle that much better than three kernel pults, because they have the DPS to handle that much HP spawning on a regular basis.

Melon-Pult
Damage per tile: 10
Self-defence tiles affected: up to 9
Other tiles: up to 18
AoE: 3x3 (dealing 5 damage per tile)

Melon-Pult is kind-of nuts.  Singletarget it's a Repeater that costs more; not remarkable.  Multitarget it's got a massive AoE.  It's kind-of mind boggling how this costs less sun than the Three-Peater: it also hits three lanes, defends itself better (even with torchwood the damage per tile is the same, the AoE damage is the same but area is larger, and anti-pea plants don't apply).

Winter-Pult (largely same as melon-pult, but also does MT freeze)
Damage per tile: 20 (factoring in snare)

Yeah, know how I said singletarget snare was kinda questionable because it only slows the frontmost enemy while the back enemies catch up?  Multitarget is really good.

Puff Shroom
Damage per tile: 5
Self-defence tiles affected: 3
Other tiles: none
AoE: None

Puff Shroom is kinda broken, because it costs 0.  Although yes, it's otherwise pretty sad.

Fume Shroom
Damage per tile: 5
Self-defence tiles affected: 4
Other tiles: none
AoE: 1x4

Somewhat evangelized by the Survival endless types (and hey, I use them in Survival Endless as well) these are...honestly unremarkable.  One of these can't hold off a conehead zombie on its own (28 HP vs a maximum of 20 damage dealt over four tiles).  You can handle conehead zombies with like...four rows of Fume Shrooms, but such a setup still won't stop a single bucket head zombie from walking up and eating the first Fume Shroom.

I mean, Fumes are handy when your entire plan revolves around AoE and you want to AoE forward, but past that....

Scaredy Shrooms (Peashooter minus, but at least it's cheaper...on some maps)

Gloom Shrooms
...are a little different.  Instead of firing every 1.5 seconds or every 3 seconds like every other plant in the game, they fire every 2 seconds.  The end result is...
Damage per tile: 15
Self-defence tiles affected: 1
Other tiles: 7
AoE: 3x3

Gloom Shrooms are funny; it's a lot of damage, but it's hard to use because they need to be near the zombies, but they also don't defend themselves (a single gloom is slightly better at self-defence than a single Puff Shroom--in that Zombies with more than 15 HP can walk up to either one just fine, but at least the Gloom will kill the Zombie faster once the Zombie is already eating the Gloom).  If you can make enemies ignore the Gloom's row and walk alongside it instead, though (garlic, for instance) then Glooms are great; 45 damage to a normal-speed zombie as they walk past those three tiles.  (By comparison, Fumes are like 20 over all their tiles, Torchwood-Gatling...if it always hit the same spot instead of moving with the enemy pack would be 20, and similarly Watermelon would be 10 to one row...well technically 15 if it constantly blindfired on one spot).  So yes, if you can get enemies to go right past it without actually attacking it, then it's amazing.

Cobb Cannon
Damage per tile: 19 (although Cob Cannons are basically two plants, so I guess I should cut that in half to 9)
Self-defence tiles affected: up to 9
Other tiles: ALL OF THEM (45, I guess?)
AoE: 3x3

Cob cannons are good, because you pretty much always take advantage of the full 3x3 AoE, so that's like...27 AoE damage per lane if you constantly bombard one area.

Chomper
On paper, if they eat a solid diet of full health football zombies, they're doing good damage (2 DPS; compare to repeater's 1.3 DPS).  Realistically, it usually doesn't work that way.

Cattail
Damage per tile: 10
Self-defence tiles affected: up to 9
Other tiles: ALL OF THEM (45, I guess?)
AoE: no

Cattails are more of an earlygame thing, but they're revoltingly good in the earlygame role; cover every row?  Sure.

Starfruit
Damage per tile: 5
Self-defence tiles affected: 0.2 (where it does deal 10 damage per tile, to be fair)
Other tiles: behind, diagonally up/down, and to the sides.

Starfruit is deceptively a lot worse than it looks.  A lot of people see them and think "ok, so it shoots in odd directions, but I'll just use masses of them and they'll shoot to cover each other".  And yeah, this works, but it isn't actually good; if you pay close attention when doing setups like this, you'll notice most plants are only hitting one target, especially if they're towards the top or bottom of the screen, or towards the left.  At that point...may as well use the Pea Shooter (or Split Pea if you need to deal with Digger Zombies).  More cost-efficient.

[metroid composite]

SC2 tests

Laggy had been saying that Hydras are cost effective against Marines if the Marines don't have any upgrades.  The argument being that Hydras have lots of damage and range and Marines lack HP.  So...I decided to test this.

Hydra (w Grooved Spine) vs Marine

Hydra x 10, Marine x 30:
Marines win with 17 Marines remaining.

Well...maybe that's not enough units; there weren't so many units that there could be two rows firing.  Let's try more....

Hydra x 20, Marine x 60:
Marines win with 26 Marines remaining.

Hydra x 30, Marine x 90:
Marines win with 26 Marines remaining--This is with zero micro, however; if the marines step forward so all of them can shoot at once....

Hydra x 30, Marine x 90, with Marine Micro:
Marines win with 52 Marines remaining


Ok, what about cost disadvantaged situations for the Marines?  After all, we could go by something other than mineral cost + gas cost.  Maybe we should go by...supply!

Hydra x 5, Marine x10:
Marines win with one (heavily damaged) Marine remaining.

Hydra x 10, Marine x 20:
Hydras win with two units remaining.

Hydra x 20, Marine x 40:
Hydras win with 7 units remaining.

Hydra x 40, Marine x 80
Hydras win with 17 units remaining.  Although...again this was without micro.  With micro...

Hydra x 40, Marine x 80, with Marine stutter step forward micro....
Marines win with 26 units remaining.

I tried to do the same micro the other way around, but really, nothing Hydras have is better than A-move.  For some reason when I try to stutter step with them, more than half of the marines live; it's weird.



Oh, and while I'm at it:

20 Hydras vs 20 Stalkers:
Hydras win with 3 units remaining

30 Hydras vs 30 Stalkers:
Stalkers win with 5 units remaining.

Granted, yes, one Stalker does cost slightly more than one Hydra.

35 Hydras vs 30 Stalkers:
Hydras win with 13 units remaining.

Hm, weird, pretty large swing there; re-test.

30 Hydras vs 30 stalkers:
Hydras win with 11 units remaining.
30 Hydras vs 30 stalkers:
Hydras win with 3 units remaining.
30 Hydras vs 30 stalkers:
Hydras win with 10 units remaining.
30 Hydras vs 30 stalkers:
Hydras win with 2 units remaining.

60 Hydras vs 60 Stalkers
Stalkers win with 2 units remaining.

1 Hydra vs 1 Stalker:
Mutual destruction (they literally both die!)

100 Hydra vs 100 Stalker:
Hydras win with 21 units remaining.  (retest: Hydras win with 10 units remaining)

Hm, interesting.  On the theory that Hydras pack more densely--that doesn't seem to be the case.  Observing how they line up, it's about 14 or 15 on each side in the unit tester map, so they're very similar size; Stalkers seem to have a harder time pathfinding to open spots in spite of their higher movement, though.


Other units...

Marauders: are mildly cost effective against Hydras, even without stim.  (6 Marauders for every 5 Hydras).  Notably, equal numbers of Hydras beat equal number of un-stimmed Marauders, but if you have the 6:5 ratio it usually tips towards the Marauders.  Even when it was 60 vs 50, which you wouldn't expect because Hydras are packing more densely.

Immortals win in equal-supply fights, and lose in equal-cost fights.  That said, Immortals upgrade better, and do win if both sides are fully upgraded.

One queen beats one hydra (unmicroed).  Multiple hydras beat multiple queens, at least off-creep.

Reapers (with no upgrades) are cost efficient against Hydras.

Unmicroed red flame Hellions are cost efficient against Hydras.

Hydras are cost-efficient against unsieged tanks!

Hm, a comparison that laggy requested but went offline...

80 Zerglings vs 20 Zealots, unmicroed (with all upgrades except Protoss Shields):
Zerglings win with 29 Lings left over

80 Zerglings vs 20 Zealots, unmicroed (all upgrades):
Zerglings win with 27 Lings left over

80 Zerglings vs 20 Zealots, unmicroed (no upgrades except speed and charge):
Zealots win with 7 left over
Second run, Zealots win with 4 left over

So...it depends on the crackling upgrade, really.

Back to Hydras!

Hydras are...cost effective against Banshees?  Whuh?

30 Hydras vs 18 Banshees:
8 Hydras survive

Umm...let's run that one again

30 Hydras vs 18 Banshees:
4 Banshees survive

10 Hydras vs 6 Banshees
5 Hydras survive

Moving on....

Brood Lords are weird.  In a small enough fight (18 Hydras vs 5 Brood Lords) Hydras can win, but only if microed properly (in particular, it seems to work best to alternate between killing a BL, and cleaning up broodlings).

Battlecruisers seems...oddly slightly Hydra favoured, as long as they shoot at their targets (so...stutter step forward if there's enough BCs so that the Hydras in the back can shoot).

Carriers...seems to be probably Hydra favoured (without micro they just kill all the interceptors--although granted, not if the interceptors are set on auto-build).


Hmm...shorter ranged units.

40 Hydras beat 12 Ultralisks
(With like...2 Hydras left).

40 Hyras lost to 60 Chargelots
(With 34 Zealots left over)

40 Hydras lost to 240 Cracklings
(with 127 Zerglings left over)

40 Hydras lost to 240 Speedlings
(with 113 Zerglings left over)

40 Hydras lost to 240 Slowlings
(with 86 Zerglings left over)

40 Hydras beat 15 Archons
(with 13 Hydras left over)

While I'm at it...
10 Archons vs 160 Cracklings
Cracklings....win!  (0/0 upgrades on both, mind you; Archons probably upgrade better?)

10 Archons vs 40 Chargelots
(Zealots win with like...23 Zealots alive).

10 Colossus vs 200 Cracklings
Cracklings...win! (With like...5 left)

10 Colossus vs 50 Chargelots
Zealots win with 26 units alive

The moral of this story is: Surface Area?  What's that?  (Unless you're up against Blue Flame Hellions, in which case, yes, 20 Hellions without micro beat Zealots/Zerglings just fine).

EDIT:

5 Hydras vs 9 Infested Terrans (the number you can spawn when you have a freshly hatched Pathogen Glands Infestor group of equivalent cost):
Infested Terrans win with 2 remaining.  (They lose if they need to hatch in front of the Hydras, though).

[metroid composite]

And more testing:

Hellions, in sufficient number, are cost effective against just about everything on the ground through simple "run up to their face and AoE them" micro.

With a few exceptions.  The exceptions are...

Thors
Collosi
Ultralisks
Archons
Roaches
(And sort-of Sieged Tanks; Hellions beat Sieged Tanks in low numbers; it's the reverse of how they normally work).

Marines are similarly ridiculous.  Stim, Combat Shield Marines, in sufficient number are cost effective against everything but...

Banelings
Hellions (including red flame Hellions)
(Sort-of Siege Tanks; only when there are a lot of Siege Tanks).
(Sort-of Colossus; only when there are a lot of Colossus).

And...hmm, let's test Colossus for fun.  Colossus in large numbers (10) are cost effective against everything but...

Marauders
Thors
Zerglings
Ultralisks
Zealots
Immortals
Sieged Tanks

Hmm...a mmuch longer list than I imagined.  OK, let's match stuff against 20 Tanks...

20 Sieged Tanks are cost effective against everything but...

Zealots
Ultralisks

While we're at it...

10 Thors are cost effective against everything but...

Marines
Siege Tanks
Zerglings
Roaches
Ultralisks
Immortals
Zealots

The interesting exclusion from this list being Marauders, which get just barely 2HKOed if they stim (and are lame without stim).

[Monkeyfinger]

Air unit vs ground unit calculations in SC2 are pretty tough because of the fact that the air units can ball up and all fire, or focus fire even, without impediment. Ground units:
- Struggle to get into a good firing position if terrain features or buildings are near the fight
- Struggle similarly if the unit counts are big, as they trip over each other

With banshees in particular the building thing is a big problem. What cost efficiency hydralisks and marines normally have vs. banshees is easily ruined as they march single file through a mineral line or a row of barracks into a formation of banshees that's firing from range 6 in perfect unison. (Stalkers have blink and warp in so it's a lil easier for them)

This is why queens are much worse practical AA than the raw stats suggest. They're fat as hell and they're slow even on creep so that means positioning and microing them is impossible if the guy controlling the air units knows where to put them and hard even if he doesn't. Compounding this problem is their durability-over-damage setup which means that a hurt banshee/mutalisk/void ray can just pull back a few feet and queens go fucking spastic trying to do anything about it.

[metroid composite]

Air unit vs ground unit calculations in SC2 are pretty tough because of the fact that the air units can ball up and all fire, or focus fire even, without impediment

Well right--this was true in SC1 as well.  What I meant was more...in SC2, Banshees are cost effective against some ground units; like...in equal cost quantities, Banshees beat Stalkers.  This is without cliff abuse, without the fact that they ball better than land units ever could.



Anyhow, my weird revelation of the day is: Ultralisks are really weird.
They apparently have a "frenzied" ability that lets them move while fungal growthed, move while being hit by strike cannons, be immune to mind control.  I'd say something about countering infestors, except they probably can't handle it if the infestors summon a bunch of infested Terrans.
In the mean time, something that hadn't really registered with me--they have the third biggest gap between damage to armoured and damage to light.  Beaten only by Immortals and Blue Flame Hellions (and they will no longer be more extreme than Hellions post-patch due to the blue flame nerf).
Oh but wait, there's more.  Thors deal more singletarget DPS than Ultralisks, even against armored (so much more that Thors win the 1v1).  You might not expect this because Thors have 7 range and hit air, and don't absurdly specialize in one damage type.  But uhh...yeah.  Thors.  The only thing Ultras have going for them when compared to Thors is movement speed and splash damage.

[metroid composite]

How units have changed since SC1

For this, I will assume that the marine un-stimmed firing rate is the constant.

Marine:
HP: 37% increase
Stim HP: 50% increase
DPS: ---
Stim DPS: used to be 33% better
Cost: ---

Firebat/Reaper:
HP: --- (although they used to have 1 armor)
DPS to Light: 29% increase (but loses splash and stim)
DPS to Armored: 129% increase (but loses splash and stim)
Cost: used to be 33% better
Range: Increased from 2 to 4.5

Missile Turret:
HP: Increased by 25% AND gains 1 armor
DPS to Armored: Increased by 20%
DPS to Light: Increased by 140%
Cost: used to be 33% better

SCV:
HP: used to be 33% better
DPS: used to be 74% better

Siege Tank:
HP: 7% better
Sieged DPS to Armor: 3% better
Sieged DPS to Light: 43% better
Unsieged DPS to Armor: 70% better
Unsieged DPS to Light: 104% better
cost: used to be 10% better.  Also, supply used to be 50% better

Drone:
DPS: used to be 19% better
Range: used to be 2 instead of 1

Zergling:
DPS: used to be 52% better
DPS after Adrenal Glands: used to be 70% better

Mutalisk:
DPS: 13% better

Hydralisk:
DPS to armored: 24% better
DPS to light: 149% better
Range: +1
Cost: used to be 50% better.  Also supplu used to be 100% better

Spine Crawler/Sunken Colony:
HP/Armor: ---
DPS to Armored: used to be 34% better
DPS to Light: 24% better
Cost: 17% better

Ultralisk:
HP: 25% better
DPS to Armored: 75% better (and gets splash)
DPS to Light: used to be 33% better (but gets splash)
Cost: used to be 25% better.  Also, supply cost used to be 50% better


IMPORTANT NOTE for Protoss units--shields counted as both Light and Armored (they took full damage to both concussive and explosive).  This makes all SC1 Protoss units less durable when faced with typed damage.

Dragoon/Stalker:
HP: used to be 13% better
DPS to Armored: used to be 20% better
DPS to Light: 20% better

Zealot:
HP: used to be 7% better
DPS: 5% better

Photon Cannon:
HP: 50% better
DPS: 1% better

Archon:
DPS Biological: used to be 31% better
DPS Mechanical: used to be 83% better

Dark Templar:
DPS: 14% better
Cost: used to be 11% better

Carrier:
You know, I can't actually find the cooldown of SC1 interceptors on the internet.  Laggy's stat topic says that the cooldown is 30, though, so I'll go with that.
DPS: used to be 5% better
Armor: used to be 4 (instead of 2)

[metroid composite]

Star Jewelled:

I'd really like to stat topic this up, get stats for all the units and good stuff like that.  For now, I'll at least document the in-game UI which tells the "costs" and the "counters"--which I actually have not found documented elsewhere on the internet.

Zealot 50:
Strong: Roach Hydra Immortal Ultralisk
Weak: Mutalisk Banshee Colossus
(Dies to storm)

Roach 75:
Strong: Ghost, Hydralisk, Colossus
Weak: Zealot, Mutalisk, Banshee, Ultralisk

Hydralisk 100:
Strong: Mutalisk, Banshee, Immortal
Weak: Zealot, Roach, Colossus
(Survives one storm at low HP)

Ghost 100:
Strong: Mutalisk, Banshee, Hydralisk
Weak: Roach, Siege Tank, Immortal, Colossus
(Dies to Storm)

Mutalisk 200:
Strong: Zealot, Roach, Siege Tank, Immortal, Colossus
Weak: Hydralisk, Ghost
(Dies to Storm, melts to cannons)
(Apparently Marines either don't shoot Mutalisks, or deal no damage to Mutalisks; weird...)
(I've read that they're supposed to have Glaive bounce, so work better than Banshees against groups, but I haven't noticed it if so).

Banshee 250:
Strong: Zealot, Roach, Siege Tank, Immortal, Colossus
Weak: Hydralisk, Ghost, Mutalisk
(survives one storm)
(Deal a lot more damage than mutas against single large targets like Colo/Ultra)

Siege Tank 300:
Strong Against: Structures
(Siege Tank dies to one storm if it's sitting still for the whole storm, not if it's moving)

Immortal 300:
Strong Against: Structures
Immortal Hardened Shields blocks two attacks

Colossus 500:
Strong: Zealot, Hydralisk, Ghost
Weak: Roach, Siege Tank, Immortal, Ultralisk

Ultralisk 500:
Strong: Roach, Siege Tank, Immortal, Colossus
Weak: Zealot mutalisk, Banshee
(Ultralisks deal ridiculous amounts of damage to other Ultralisks.  If your opponent is bad and doesn't cast spells, but spams ultras, the best counter may well be Ultra+Warp Cell)

Warp Cell: 150 (Disables target for 15 seconds)
Time Bomb: 200 (AoE 50% speed within bubble, lasts 12 seconds)
Heal Wave: 250 (Heals 300 HP, bounces a MAX of 4 times)
Storm: 300 (deals 100 damage, lasts 4 seconds)

[SnowFire]

I stopped playing StarJeweled when the ultimate strategy became clear: Stockpile energy, then release 3 Ultras at once, then stick everything into Storms to support them.  (Or Warp Cells if your opponent goes Ultras as well).  Storm is just stupidly overpowered; there seems to be a lot more strategy if Storm costs double.  Take banshees for example; interesting unit, has counters, and counters Ultras...  except...    a single Storm pretty much ruins a Banshee.  Catch some nearby marines and you have parity.  If you ever release 2 Banshees or more, they tend to stack up, so you're now just donating points to the coming Storm.  Zealot / Roach / Hydra / Ghost / Muta are all nearly worthless except as a way to use spare change thanks to Storm, too.

So, yeah.  1 Banshee to force a Storm and maybe scare Ultras, 1 Siege Tank or Immortal if you have the advantage and want to blow a cannon, and then Ultra spam.  Although the Ultras only work if you do the stockpile energy + release simultaneously approach, since otherwise you die to Warp Cell.  Make Warp Cell cost 250 and Storm cost 600, and I'd totally be interested in StarJeweled again.

[metroid composite]

Ok, tried to get a real read on unit HPs.

Marine- 2 cannon hits
Banshee- 8 cannon hits
Mutalisk- 6 cannon hits
Zealot: 4 cannon hits
Hydralisk: 6 cannon hits?
Roach- 9 cannon hits
Ghost- 5 cannon hits
Tank- 5 cannon hits
Immortal- 15 cannon hits
Colossus - 25ish cannon hits?
Ultralisk - 35ish cannon hits?

I'm fairly certain that Cannon damage is unchanged--which is to say, one cannon hit is 20 damage.  This seems to line up with Storm dealing 100 damage.  So...this puts my best estimates for HP at...

Marine: 30
Zealot: 80
Roach: 180
Hydralisk: 120
Ghost: 100
Mutalisk: 120
Banshee: 160
Tank: 100
Immortal: 300
Colossus: 500
Ultralisk: 700 (doesn't line up with Heal Wave's description, since Heal Wave heals more than half an Ultra's HP.  Possibly there's armor in the mix here?  Possibly my HP estimates are a little high?  Possibly it's due to zerg units having some passive regen--I suspect Mutalisk is actually 100 HP, and just regenerates the 1 HP required to survive a storm/five canon hits)

And yes, Hydralisks, Mutalisks, and Banshees all survive one storm.

[metroid composite]

I stopped playing StarJeweled when the ultimate strategy became clear: Stockpile energy, then release 3 Ultras at once, then stick everything into Storms to support them.  (Or Warp Cells if your opponent goes Ultras as well).  Storm is just stupidly overpowered; there seems to be a lot more strategy if Storm costs double.  Take banshees for example; interesting unit, has counters, and counters Ultras...  except...    a single Storm pretty much ruins a Banshee.  Catch some nearby marines and you have parity.  If you ever release 2 Banshees or more, they tend to stack up, so you're now just donating points to the coming Storm.  Zealot / Roach / Hydra / Ghost / Muta are all nearly worthless except as a way to use spare change thanks to Storm, too.

So, yeah.  1 Banshee to force a Storm and maybe scare Ultras, 1 Siege Tank or Immortal if you have the advantage and want to blow a cannon, and then Ultra spam.  Although the Ultras only work if you do the stockpile energy + release simultaneously approach, since otherwise you die to Warp Cell.  Make Warp Cell cost 250 and Storm cost 600, and I'd totally be interested in StarJeweled again.

It's funny--talking to Laggy he claims that Storm is overcosted.

As for Ultras, they're good, but they mostly force a warp-cell from the opponent, and then act as a meat shield.  If your main goal is meat shielding, then Roaches exist.  You can buy 6-7 roaches for the cost of one Ultra, and the roaches are certainly more HP to chew through.  I dunno--I usually open with a pair of Ultras to test the waters, but if someone on the other team is competent, they'll get warp celled, probably a Banshee or two will get made, and I'll usually give up on supporting them and just switch to a harder-to-counter strategy.

[metroid composite]

I need to do full starjeweled damage tests (read: I would like a volunteer to test with me because this is really hard to test against the computer), but Ultras seem very typed in their damage.

Like...I think they're somewhere in the range of 15 damage to light, and somewhere around 75 damage to armored.

[metroid composite]

Posting these here in case I decide to continue calculating logarithms by hand for fun.

the_rowan posted...
You can't really find a decimal value without a calculator or memorizing tables.

You can; I mean logs have been around since the 1600s. Obviously someone had to make the log tables.

Here, let me start grinding some stuff out by hand.

2^10 = 1024, which is really close to 10^3

therefore
10 * log 2 ~= 3

therefore
log 2 ~= 3/10 = 0.30
(actual value: 0.3010...)

3^2 = 9, which is close to 10.

Therefore
2 * log 3 ~= 1
log 3 ~= 0.5
(actual value: 0.4771...)

Or, and admitedly, I didn't calculate this by hand:
3^21 = 10,460,353,203

Therefore
21 * log 3 ~= 10
log 3 ~= 10/21 = 4.76 (and that I did calculate by hand)

Log 4 = 2 * log 2 ~= 0.60
(actual value: 0.6020)

log 5 = log 10 - log 2 ~= 0.70
(actual value: .6990)

log 6...well I could just add the log of 2 and 3 together, but I cheated a little on log 3, so let's see...

6^1 = 6
6^2 = 36
6^3 = 216
6^4 = 1296
6^5 = 1300*6 - 4*6 = 7800 - 24 = 7776
6^6 = 7777*6 - 6 = 46662 - 6 = 46656

Ok, I don't want to compute much further by hand; the one that looks the most useful at a glance is 6^3 = 216

3 * log 6 ~= log(200) = 2 + log 2 ~= 2.3

log 6 ~= 2.3/3 = 2/3 + 0.1 = 0.76666...
(actual value: 0.7781)

Log 7...

Right away, 7^2 = 49 ~= 50 jumps to mind.

2 * log 7 ~= log(50) = 1 + log 5 ~= 1.70
log 7 ~= 0.85
(actual value: .8451)

Log 8...

Log 8 = 2 * log 2 ~= 0.90
(actual value 0.9031)

Log 9...

Well...I did cheat a bit on the log 3 stuff, so let me go back and use the log 6 result to re-calculate log 3.

log 3 = log 6 - log 2 ~= 0.77 - 0.30 = 0.47
(Actual value: 0.4771)

Which means...

Log 9 = 2 * Log 3 ~= 0.94
(Actual value: .9542)

So there you go: all the one-digit logarithms hand calculated to *almost* 2 digits of accuracy (I was off by 0.1 in a couple of places--3, 6, and 9. Damn you multiples of 3 *shakes fist*).

Actually, let's see if I can get a better estimate on log 3/6/9. I know 12^2 is 144. I know the square root of 2 is 1.41.... So...without even calculating it, I know that 12^4 is going to be pretty close to 20000 (closer than 216 is to 200, anyhow).

4 * log 12 ~= 4 + log 2

log 12 ~= 1 + 0.30/4 ~= 1.08
(actual value 1.0791)

Which gives us...
Log 3 = Log 12 - Log 4 ~= 1.08 - 0.60 = 0.48
(actual value: 0.4771)

Log 6 = Log 12 - Log 2 ~= 1.08 - 0.30 = 0.78
(actual value: 0.7781)

Log 9 = 2 * Log 3 ~= 0.96
(actual value: 0.9542)

Arg, nooo, I'm still off.

Wait, 81, is pretty close to 80.

2 * log 9 ~= 1 + log 8 ~= 1.90
log 9 ~= 0.95
(actual value: 0.9542)

There--now I've gotten estimates for every one-digit logarithm that's accurate up to two digits. All calculated by hand. Yaaay.

So, an inquiring reader might ask, how would I go further? How would I get, say, three digits of accuracy? And how, without looking up on the internet, would I know that I had three digits of accuracy?

Ok, let's cover accuracy first.

You'll notice that when we were very close to powers of ten (2^10 = 1024) that we were quite accurate. Notably, 1024 is only 2.4% away from being 10^3, and 0.30 is actually even more accurate--0.34% away from being log(2). We can predict this with a good degree of accuracy using Calculus.

The derivative of Log( x ) at any one point is...well Log( x ) = ln(x) / ln(10), so the derivative is 1/x * 1/ln(10). I don't even need to know ln(10) here, just that it's somewhere between 2 and 3 (cal it 2.5).

So, this means that at x = 1000, the slope is roughly 1/1000 * 1/(2.5) = 1/2500. Going 24 away from this, we'd expect the answer to be off by 24/2500, or ~0.01. And indeed, log(1000) = 3. log(1024) = 3.0103. (Actually...we'd actually be freakishly accurate here if we had a better estimate on ln(10) than 2.5. My initial plan had just been to show that we could get a pretty good guess on the error, but this might be worth pursuing...).

Man, I was going to go calculate the square root of ten, cube root of ten, etc by hand, but now I'm tempted to screw that and just get a really good estimate on log(e), and trusting in the power of calculus. Well...square root of ten is still going to come in handy, and is honestly probably easier to calculate by hand than e.

Ok, square root of 10, let's do this!

Pass 1:
3*3 = 9
4*4 = 16
First digit is 3, second digit is small.

Pass 2:
3.1*3.1 = 9.61
3.2*3.2 = 10.24
Second digit is 3, third digit is largish.

Pass 3:
3.17*3.17 = 3.1*3.1 + 2*(3.1*0.07) + (0.07)^2 = 9.61 + 2*(0.217) + 0.0049 > 9.61 + 0.43 + 0 = 10.04
3.16*3.16 = 3.1*3.1 + 2*(3.1*0.06) + (0.06)^2 = 9.61 + 2*(0.186) + 0.0036 = 9.61 + 0.372 + 0.0036 = 9.9856
Third digit is 6, fourth digit is smallish.

Hmm...well, let me state up front that I'm not going to calculate the fifth digit because this is becoming a pain, so that lets me cut a couple of corners here
3.163*3.163 ~= 9.9856 + 2*(3.16*0.003) ~= 9.9856 + 2*(0.009) ---- (Nope--too big. That's adding 0.018. We're looking for 0.0144)
3.162*3.162 ~= blah blah + 2*(3*0.002) ~= blah blah + 0.012 ---- (Yep, that's better--closer to the 0.0144)
(Fourth digit is 2, fifth digit looks small)

So to four digits of accuracy, sqrt(10) = 3.162

Ok, how much of a pain is it going to be to compute e?

e = sum( 1/n! ) from n = 0 to infinity
e = 1 + 1 + 1/2 + 1/6 + 1/24 + 1/120 + ...
~= 2.5 + (20 + 5 + 1)/120 = 2.5 + 13/60 = 2.5 + 1.3/6 = 2.5 + 1/6 + 0.3/6 = 2.5 + 1/6 + 0.05 = 2.55 + 1/6 ~= 2.55 + 0.16667 = 2.7167

The next term is 1/720, which we can totally ballpark as roughly 1/700 ~= 0.0014. So...that brings us up to 2.7181

The next term is 1/5040, which we can totally ballpark as roughly 1/5000 ~= 0.0002. So...that brings us up to 2.7183

And the term after that is going to be too small to register at 5 digits of accuracy. Bam, 5-digits of accuracy on e. That was remarkably painless. Man, four digits on the square root of 10 hurt more than that.

Ok, just a ballpark, though, I'm going to take it back down to 2.7.

e^2 ~= 4 + 2*(2*0.7) + 0.49 = 4.49 + 2.8 = 7.29 ~= 7.3
With three digits..
e^2 ~= 7.29 + 2*(0.02*2.7) + something_tiny = 7.29 + (0.108) ~= 7.4

Well, at very least this gives us a very rough bead. We know to a good degree of accuracy, that log(8) = 0.90, and log(7) = 0.85, so let's say that log(7.4) = 0.87. That means log(e) is pretty close to 0.435, which in turn means that ln(10) is pretty close to 1/0.435.

And...I don't really feel like calculating 1/0.435 right now.  Stupid long division....

[metroid composite]

Back to Starjeweled for a bit.

Things I see in the metagame:

Ultra Ghost
Ultra Hydra
Heavy Banshee
Heavy Muta
Mass Colossus
Mass Hydra
Mass Ultra
Siege Tank Ghost
Mass Zealot
Mass Roach
Immortal+whatever

Some of these are reactionary (Zealots, Banshees, Mutas, honestly Roaches too).  So...let's cover the plan A.

Siege Tank Ghost is just awful.  all your units die to Storm; I've beaten teams that had 50% more energy production and went this combo.  I've lost because my teammate kept trying to force this combo when we were slightly outproducing the opponent.

Mass Hydra is interesting--they take low damage to Ultras due to being light, they tend to spread out and thus not get all caught in the same storm, and don't die to one storm anyway.  Colossi are easily frozen.  The best response I've found is Roach+Storm.

Mass Colossi is weird because it loses to an equal number of Ultras.  That said, if it does get a decent energy advantage, it becomes hard to break.  Sending a wave of two Ultras just doesn't do much to five Colossi.  Air is probably the best counter if you get behind, but it doesn't work as well as it does against Ultras, because the air needs to run forward and expose itself to the full army, and sometimes defensive cannons.  Otherwise, equal Ultras, gogo.

Mass Ultra: counter with Banshees; be ready to storm Ghosts/Hydras, or heal-wave if they storm the Banshees.  You might also want to counter with Muta Banshee, just because it spreads things out.  Also acceptable to counter with mass zealots, and freeze the ultra just as the Zealots arrive (yeah, Ultras deal low damage to Zealots, but they still have no HP and don't like AoE)--this is counterable with Storm, though, and the zealots will die fast if they have Ghosts.

So hmm...that's my intuitive feel of the metagame right now.

Notably, yes, Ultras are good, but there is a rock-paper scissors still; I'm kind-of down on Roaches as a rule--the point of them is usually to tank, but they tend to get stormed, they're not good against mass colossi, and they take ridiculous damage from Ultras.  Ultras tank better, even if they have less HP for the energy, just because they're not vulnerable to storm.  I'm definitely feeling pretty negative towards Siege Tanks and Immortals.  Immortals do take three storms to kill, but they also stand with the rest of your army that is often getting stormed anyway.  Siege Tanks die to one storm, and are just all-around awful if they're getting hit.  Ultras do the job of "I threaten to kill your canons quickly so you'd better have a response" anyway.

Just my current gut reactions.

[metroid composite]

A fun trick someone just quoted at me.

If you are given a percentage growth per year, say 7% per year, you can get an approximation of the doubling time by doing 70/percentage, so in the case of 7% per year 70/7 = 10 years to double.  (Actual value: 10.2 years to double).

So...why does this equation work, and where does it break?  What we have is...

(1+p)^n = 2

Taking the log of both sides:

n * ln(1+p) = ln(2)

And here's where the approximation comes in--using a linear approximation for ln(1+p), we get ln(1+p) ~= 0 + p

n * p ~= ln(2)

n = ln(2)/p

ln(2) = 0.693 ~= 0.70.

And where does this break?  Since we're just doing a linear approximation around ln(1), it breaks far from ln(1).  So, for example, increasing by 70% per year is well short of doubling every year (more like doubling every 1.3 years).  It's also noteworthy that 69.3 is higher than 70.  For example, the approximation would tell us that increasing by 1% per year gets you to double in 70 years, but the actual time to double is actually 69.6 years--less than the estimate (whereas when the percentage is low, we slightly overestimate).

Which leads me to wonder, what's the tipping point when they are exactly equal?  So...

0.7/x = ln(2)/ln(1+x)

1+x = e^(ln(2)/0.7*x)
1+x = 2^(10/7*x)

Umm...yeah, as far as I know there isn't an algebraic solution to something like this.  Fortunately, there are numerical solvers on the internet, like Wolfram Alpha.

x ~= 0.019838

Which is to say, the 70/oercentage equation is closest to perfectly accurate around 2%.  Let's say we want to change that--let's say that our most-common use case is more like 5%, and we want to use something like 71/percentage, or 72/percentage; what's the exact value then?

Q/0.05 = ln(2)/(ln(1.05))

Which is to say, the correct value to use would be 0.710334954.  Or rather, that multiplied by 100 for percentage purposes, so 71.

[metroid composite]

Ok, you know, I'd really like to drill down and finish the FFT ban list.  Last time you joined us, things looked something like this:

Ban List:

1. Calculator
2. Summoner
3. Wizard
4. Chemist
5. Squire
6. Ninja
7. Time Mage
8. Samurai
9. Lancer
10. Priest
11. Oracle
12. Monk
13. Geomancer

Remaining classes are Thief, Knight, Archer, Mediator, Dancer, Bard, Mime

Let's drill down and get some damage numbers for assassinations, as those proved important in the past.

Chapter 1:

Knight (36 or so)

Chapter 2:

Knight (56 or so)  Although Equip Sword Archer is similar, and Equip Sword Thief with Power Wrist instead of Battle Boots is also similar.  I want to say Thief is probably the best of these due to having the option of 5 move, and unlike Priest, 7 speed is usually fine and acceptable on a melee fighter.  Depends if you want shields, though.

Chapter 3:

Archer/Thief with Equip Sword setups (about 110).

Chapter 4:

Archer/Thief with Equip Sword setups (168, although Knight is ~160 with Ice Brand).


Breaking this down a little further...

Chapter 1:

Bows deal ~20, so swords are a 75% improvement.  (Charge+3 is a 50% improvement).

Chapter 2:

Guns deal 36 so swords are a 55% improvement.  (Charge+3 is a 42% improvement).

Chapter 3:

Guns deal 64, Bows deal ~72, so swords are a 52% improvement (Charge+3 is a 27% improvement)

Chapter 4:

Crossbows deal 120, so swords are a 40% improvement (Charge+3 is a 21% improvement)


So generally, the gap between guns and not guns is bigger than the gap between Charge and not Charge.  Although there's a couple of caveats: with both Guns and Bows you can raise your speed without dropping your damage, effectively upping your damage per clocktick.  Additionally, against enemies with evasion, obviously guns ignore that (whereas you can't use both Concentrate and Equip Sword on the same unit.  Concentrate Knights are possible, though, and only significantly less damage in chapter 3).

Also in the favour of charge is that it's usually best to use a mix of melee and range, however (3-2 split or so) as it's often not possible to hit a boss from 4 sides.

So...for boss smashing, swords and Charge are the important abilities.  All four of Thief/Mediator/Archer/Knight are worth considering for stats/equips.

Oh yes, I said I would stick Chocobos somewhere on this list due to Boco (presumably somewhere below Mediator for obvious reasons).

Chapter 2:

Choco Meteor is 44

Chapter 3:

Choco Meteor is 68

Chapter 4:

Choco Meteor is 100

...Yeah, given that they don't charge...non-Tiamat monsters still suck.  Red Chocobos get pulled out against anti-physical bosses like Elmdor and Rofel, but that's it.  They can't even play the Geomancer game of having both ranged chipping and melee smashing, because Choco Meteor outdamages their melee until like...over level 30.  (And black chocobos aren't a credible option.  They have 33% more physical damage than reds...but reds have 39% more speed, and therefore more physical offence anyway.  Combine this with how reds prefer magic offence, and have 70% more HP, and...yeeeah).


Onto mook smash.

Dancer wins, so very, very hard.  All the other Mook smash got worse--Attack Up Mythril Gun lost Attack Up.  Elemental and then clean up with melee lost Elemental.  But...Dancer also is stuck as female, and stuck going through a lot of dead classes now, and not on the radar for boss smash, so it's going to be hard to be overly centralizing.

Ok, so for the non-females, what are the setups?  I'm not so hot on pure-melee setups against mooks--melee often can't get in range turn 1, and often doesn't want to rush into range even if it can--that's asking for an early death timer.  So basically, past about Chapter 1, all anti-mook setups must have range.  Most of the time, this is going to mean you're either using Talk Skill as your secondary, or you're equipping a gun or bow.  This is...actually quite Mediator-dominant, given that whole "guns are better than bows" and "the best Archer support for most of the game is Equip Gun, not Concentrate".

There are some alternatives.  Steal for Steal Heart (although...Steal Heart is mostly useful on females, and females will have Dance, so...yeah.  Even if it could hit either gender, Mimic Daravon is arguably better anyway).  And there's...Sing.  Cheer Song tends to be better than waiting on spot.  Life Song is some minimal healing.  I'm still really skeptical about Bard, though.  I think Cheer Song is an upgrade over Mimic Daravon, but it's a marginal one for such a long detour.

That said, there are a couple of notable exceptions to "Don't use pure melee for mook smash".  As Elfboy noted in his playthrough, Knight is the best class in Chapter 1 (there are no guns, and the damage gap is larger).  In Chapter 4, Ice Heal gives you something great to do when out-of-range, and more to the point all you have to do is sit around and ice heal while the Dancers win you the fight.  Add to this, Chapter 4 is when guns drop off significantly in power (less than half of Knight melee damage).


So...to summarize.

Assassination missions: Mostly Knight/Archer for the majority of the game.  (With guest appearances from Thief, Mediator, and on rare occasions Red Chocobo).
Mooks: Knights in Chapter 1. Mediators in Chapter 2.  Dancer Mediator in Chapter 3.  Dancer Knight in Chapter 4.  (With guest appearances from Archer and Thief)

Combine this with Elfboy's musings on whether Knight should be banned before Geomancer just because Chapter 1 was the hardest chapter for him...yeah.  Knights are looking really dominating here.  Even the mook-killing Mediator setups in the middle of the game don't really escape Knight--a lot of them are either Equip Shield Mediators (because Equip Shield is basically the only support option Mediators have), or some variety of sword user with Talk Skill.

On poaching, just taking the "You have a 33% chance of Uribo" argument at each relevant fight, ok, so you have a 33% chance of murdering the second half of the game, and a 56% chance of murdering the last quarter of the game.  Once you add that up, on average poaching murders 22% of the game, as opposed to the 80% of the game that Knights are relevant in the event that poaching doesn't come up.

Ban List:

1. Calculator
2. Summoner
3. Wizard
4. Chemist
5. Squire
6. Ninja
7. Time Mage
8. Samurai
9. Lancer
10. Priest
11. Oracle
12. Monk
13. Geomancer
14. Knight

[Grefter]

Yayyyy more stuff I can comprehend instead of just understanding the math porn aspect. <3 this stuff as always.

[metroid composite]

So...with the banning of Knight...

Thief goes from having "Nice stats for Equip Sword, and having an important movement ability" to being totally irrelevant.  Almost everything left is like...range 8 or more, so Move+2 is pretty questionable.

Dancer, Mediator, and Archer are still highly relevant of course.

Your target probably includes being an Archer with Equip Gun.  Equip Shield is banned now, so Archer is the only way you're getting a shield.  Equip Gun is the only way you're actually using that shield (because LOLcrossbows).

The rest of your setup...your reaction is probably Arrow Guard, although since you have a shield there's a decent argument for Caution (!!!).  Movement is...probably empty most of the game, maybe filled by Move+2 eventually.  You probably want to lower your faith on some characters, because between Arrow Guard and low faith you won't fear ranged attackers.  (And stutter stepping deals with melee attackers).

Your secondary...actually, is your secondary just basically Dance with all characters?  Guns don't use PA.  Males and females deal the same damage with longbows in Chapter 1.  Females have a slightly less painful time unlocking Mediator.  So...the logical path is...in chapter 1 go Archer and get some charges; next go Mediator until you have Equip Gun, and then run around with Charge and Equip Gun in the dead physical tree (most of whose weapons are banned now, like swords, but Equip Gun handles the no-weapon issue, and a lot of these classes give you shields so you're not really sacrificing anything by not being in Archer).  And...really, what's going to destroy mooks harder?  Arrow Guard?  Or freaking Nameless Dance?

Hmm...is there a reason to go males?  There's the...three battles in Chapter 1 where you don't have Longbows, so females will be dealing 12 damage with Bow Gun compared to males' 15 damage with Bow Gun.  Maybe you want to go for an unarmed melee attack in order to blitz a boss.  Hmm...using the same numbers as before...

Unarmed damage
Chapter 1: 15 damage
Chapter 2: 35 damage
Chapter 3: 77 damage
Chapter 4: 96 damage

...Ehh...mostly lower or similar to the equivalent ranged damage.  You could raise your brave, but if we start talking about brave raising, I'm pretty sure "reset until Uribo" is much less time consuming.  Ok, so screw unarmed damage.

Then there's...Gastrifitis damage in the later half of Chapter 4.  It's somewhere between a 50% to 100% damage improvement over Mythril Gun, and still lets you use shields; this is a big deal, and probably worth the range drop.  Granted, if you're using Gastrifitis, you're using a minimum of Power Sleeve and Charge, so males are like...only 10% better damage with it.

Also...we're talking later half of Chapter 4 here.  56% chance of Chantage (assuming it's not worth it to just force a Uribo at Zigolis Swamp).  Nonzero chance of Hydra/Tiamat (216 damage per triple flame hit).  Nonzero chance of Dark Behemoth (240 damage from 60 brave).  Elemental Guns require a detour, but are similar.

So hm, the big argument for males seems to be Chapter 1.  Thing is, 12 damage and 15 damage are both so incredibly awful.  Could you maybe just rely on Delita and Algus and play more of a supportive role instead?  Hmm...actually yeah, maybe.  You probably won't unlock anything like Mimic Daravon before Longbows show up, but Steal Heart is only 150 JP so you'll often be able to learn it without being a Thief at all.  And...hey wait!  MYTHRIL KNIVES!  Thief with Mythril Knife deals 20 damage, from either gender.  (I had been ignoring Knives because they spend 90% of the game with less WP than bows).  So...yeah, there are alternatives to attacking with crossbows for those three battles where they're relevant.

Which brings me back to thinking that you do run all females; mostly due to "why not?"


So...the game breaks down something like this:

Before Guns: all about Archer, with a tiny bit of Steal Heart and Knives

After Guns: all about Mediator/Equip Gun, with guest appearances from the Archer Charge skillset, but Equip Gun helps you so much more than Charge on the route to Dancer.

After Dancer is unlocked: Primarily about Dance (vs mooks) and Charge (vs assassinations) and sometimes debilitation like Solution/Threaten/Slow Dance/Persuade (vs zodiac assassinations).  Guns are still cool, although you're probably back chilling in the Archer class, and might even have Gastrifitis, so they're not helping you as much as the other two classes.

After the game is broken via Chantage or invited Tiamats or Elemental Guns or whatever: Mediator.


Alternatively, if you go males, there's...less of a focus on Equip Gun (although you still want it for your Archer) and more of a focus on Solution+Arrow Guard.  And...Talk Skill probably becomes your secondary of choice, although you might grind for Sing (which could be argued as a slight upgrade over Talk Skill).  It probably plays something like a Mediator SCC, except with Shields, Charge, Arrow Guard, the ability to use invited monsters, and breed/poach.

Hm, yeah, this is awkward.  I'm pretty sure the best party is four females, and yet Dancer STILL doesn't feel dominant; dominates one half of one section of the game (out of four), and is relevant in some assassinations.

Archer vs Mediator is like...
Guns > Bows (For most of the game)
Charge > Talk Skill (in most fights, not counting out-of-battle effects)
Arrow Guard ~ Solution (Not sure which of these makes the bigger impact, and I'm not sure it matters when "get some Dancers instead" is probably a better option in most fights where these are relevant)
Shields are good, but it takes Equip Gun to make effective use of them usually

This feels kind-of balanced.  Well...maybe Guns > Bows doesn't hold for "most" of the game, when it's false in Chapter 1 and half of Chapter 2, and half of Chapter 4 (that's technically half the game).  Although...bows don't completely dominate Chapter 4--Guns still have 4 more range, and the damage gap is a lot smaller if you're focusing on speed/durability over PA, and guns can be used on any class (whereas Equip Crossbow means not using Concentrate, and accuracy closes the damage gap).

This feels...kind-of-sort-of balanced; maybe slightly Archer-leaning.  Except...yeah, throw in the spice of "Mediators can randomly break the game with chantages.  Or intentionally break the game if you expect more than 2 resets in Chapter 3/4".  I...think that seals it?

Ban List:

1. Calculator
2. Summoner
3. Wizard
4. Chemist
5. Squire
6. Ninja
7. Time Mage
8. Samurai
9. Lancer
10. Priest
11. Oracle
12. Monk
13. Geomancer
14. Knight
15. Mediator

[metroid composite]

Next up...same as before, except with Archer/Dancer as the only really relevant classes.  Maybe a little Bard.

So...there's the before-Dancer time, which is Archer dominated, with some Steal Heart action.  And there's the after-Dancer time which is...

mooks (50% of fights) dominated by Dance
assassinations (50% of fights) mostly you want to be using Charge, Concentrate, and bows, although some fights you do want to Dance (Balk 2, Altima).  Although...speaking of special exceptions like Balk 2, you also want Shields for that fight, which only Archer offers.  There's also a couple of special exceptions that pull out red Chocobos (Elmdor, Rofel).

So...Dance dominates somewhere in the range of 55% of remaining fights...probably used from the Archer class because Archer offers the most durability, and the only credible offence for cleaning up a fight quickly before the Nameless Dance statuses wear off, and is tied with Dancer for the best PA multiplier if you're thinking of using Wiznaibus.  Oh, and you're probably using both Concentrate and Arrow Guard.

Add to this, being male matters again; not a whole lot, but it does.  And there isn't a hilariously easy way to get to Dancer through all the dead classes like Equip Gun+Charge.  You could use Equip Crossbow, but that's not even remotely effective until like...Hunting Bow (Orbonne Monestary checkpoint in Chapter 3).  Before then, Equip Crossbow doesn't even necessarily increase your damage above "unarmed female punch".  Which is to say, it's going to be tempting to not even start unlocking Dancer until half-way through Chapter 3--just leaning on Archer with a bit of Steal Heart.  ...Yeah.

Ban List:

1. Calculator
2. Summoner
3. Wizard
4. Chemist
5. Squire
6. Ninja
7. Time Mage
8. Samurai
9. Lancer
10. Priest
11. Oracle
12. Monk
13. Geomancer
14. Knight
15. Mediator
16. Archer

[metroid composite]

Ok, so...the classes remaining are Thief, Bard, Dancer, Mime.

You could chill out in Thief, and suck forever.  Except for that one glorious moment when you steal Meliadoul's Chantage with your 500 JP Steal Accessory ability.  Or maybe when you poach some Ryozan Silks (15 WP on a class with good stats from a common poach?  OMG!!!) at the start of Chapter 4.

You could head to Bard as a low-faith male mage, sucking hard while you got there, sucking while you're in there, but coming out with an almost decent skillset.  Maybe you can get a party of Battle Song Thieves that raise their PA and then punch people!

You could head to Mime, and suck for 90% of the game, but be...actually pretty awesome for the remaining 10% since you probably have at least two allies with Sing/Dance, and Mime HP doesn't look as atrocious when your options are Bard, Dancer, and Thief.

Or you could go to Dancer and not suck.  (The best skillset of all these classes.  And...I think also the best physical attack of all these classes).


Thief is doing a decent amount to help here, between poaching and stealing, but Dancer feels like the only real power in this group.

Ban List:

1. Calculator
2. Summoner
3. Wizard
4. Chemist
5. Squire
6. Ninja
7. Time Mage
8. Samurai
9. Lancer
10. Priest
11. Oracle
12. Monk
13. Geomancer
14. Knight
15. Mediator
16. Archer
17. Dancer

[metroid composite]

In all of this, I've forgotten about Chocobos.

Do Chocobos come before Dancers?  Hmm...I doubt it.  You may want to use a bunch of red chocobos for most assassination missions, but you can deal similar damage (at shorter range with a chance of missing, granted) when using humans.  And humans gain JP.  And there are certainly points in the game when Wiznaibus is similar singletarget damage to Choco Meteor.  They're good, but they don't crowd out the actual use of humans (and Dancer in particular).

When we're just down to Thief/Bard/Mime, though?  I uhh...well this is probably a pretty accurate Thief damage curve:

Unarmed damage
Chapter 1: 15 damage
Chapter 2: 35 damage
Chapter 3: 77 damage
Chapter 4: 96 damage

Daggers are something like...
Chapter 1: 20
Chapter 2: 30
Chapter 3: 56
Chapter 4: 90

By comparison...

Chapter 2:
Choco Meteor is 44
Chapter 3:
Choco Meteor is 68
Chapter 4:
Choco Meteor is 100

So...Chocobos are generally winning damage.  Speed for Red Chocobos is...generally acceptable (they reach 7 speed at level 3, 8 speed at level 16, 9 speed by level 28).  Which means they're usually parity or one speed less than Thieves using Green Beret/Flash Hat/Thief Hat.  Their HP is...

Level 10 (Chapter 2): 93 HP (less than Thieves--Green Beret+Brigandine alone beat that)
Level 18 (Chapter 3): 146 HP (less than Thieves--Twist Headband+Power Sleeve is 126 HP, so at level 18 their base HP of 67 would add up to 193 HP)
Level 30 (Chapter 4): 225 HP (less than Thieves--Thief Hat+Power Sleeve is 170 HP, so at level 30 their base HP of 95 brings that to 265 HP)

Ok, how much more HP is this on average?

Chapter 2: Thieves have 45% more HP
Chapter 3: Thieves have 32% more HP
Chapter 4: Thieves have 17% more HP

There are other intangibles, of course, like Thieves can wear status blocking, Thieves can wear Angel Rings and Feather Mantles.  Thieves can have Sing and have something that almost counts as a skillset.  Thieves can get more movement (although not Ignore Height, which is probably more important once we're quibbling about 7 move vs 6 move).

Hmm...here's the thing.  Red Chocobos seem better in assassination missions.  But Red Chocobos also seem better in most mook-smash situations.  They can stutter step, they can cliff abuse (without spending an extra 1200 JP in Bard for Fly--and they have a better height-ignoring attack to use with cliff abuse--Choco Meteor > Harp attack).  Lots of mooks have evasion, which they ignore.

There are some fights where I think you still want Humans.  Zalera (because you probably want something with an angel ring.  Then again, Red Chocobos might be able to just blitz Zalera).  Maybe Hashmallum (they'd need to be pretty high-level not to get insta-cast by Quake...although maybe you can bait him into charging something on Ramza?  Although they're decent at spreading once they do get a turn).  Maybe Altima (doing this without Angel Rings sounds nasty).

These aren't a lot of battles, but they're important ones--important enough that you might not want to just let your humans lag behind in levels while you take a flock of Chocobos into every fight.  Hmm...yeah, I might need to think about this a bit more....

[Dark Holy Elf]

You know this analysis has turned awesome when there is in-depth numeric comparisons on Thieves vs. Red Chocobos. *grabs popcorn*