Author Topic: Theorycrafting! (Because I like competitive metagames too much)  (Read 133484 times)

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #100 on: January 10, 2011, 11:14:21 PM »
Oh I see: you're saying movement is higher in SC2.  That makes more sense.  (I thought you were saying damage was higher relative to HP, which...it really isn't.  Stim is only a 50% attack rate boost not 100%.  Zerglings attack slower.  Stalkers deal less damage than Dragoons.  One Reaver shell hurts much worse than one Colossus laser, and you can't stick two Colossi in a shuttle like Reavers.  Pretty sure it's easier to get out from under Storm now; not sure if Storm deals damage slower or movement is faster.  Old Hydras basically became Roaches, with more HP and less DPS.  New Hydras are admittedly glass cannons.  Sentries serve to make fights longer.  Photon Cannons have more HP.  Archons deal less damage.  Both Robo units have way more HP than Reavers.  Three Goliaths don't have collectively as much HP as one Thor.  Thor has more ground damage, but less air damage than three Goliaths...although splash....  Infestor has more HP than Defiler, but less armour.  Terran Bio has more HP.  Spider Mines are gone.  Protoss shields no longer take full damage from everything.  Oh, they also regenerate faster.  Hellions deal less damage to light than Vultures, but they splash.  Hellions also have more HP...but probably less relative to their cost.  Siege Tank damage is down, and HP is up.  Ghost has double the HP and damage...but costs way more too).

Airforce probably deals more, but a lot has been done to encourage mixing air units into your army, so this is no surprise.

So...yeah, I was scratching my head about the blink of an eye comment, but more mobility I'd buy.  Especially when there's nothing as dumb as Dragoons, and very little that walks slower than a Marine (Thor, High Templar, off-creep Queen/Spore/Spine).

Laggy

  • ReDux'd
  • DL
  • Denizen
  • *
  • Posts: 1147
  • Generations of suffering & all I got was a stick
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #101 on: January 10, 2011, 11:19:43 PM »
I wasn't really thinking about numbers when I made that statement, and more the general sense that in practically every game of SC2 I play, battles end really fast, whereas in SC1 they didn't. Things like missing a stim or a forcefield in a handful of seconds literally makes engagements go sour for one side.

I THINK it's partly the mobility and partly the much superior grouping and pathfinding that means everyone fights each other basically right away, which makes them conclude that much more quickly. Also, ladder maps are smaller as well, so these clashes happen pretty easily (and look at the number of base trades that happens relative to SC1, and how fast they go down...)
<Eph> When Laggy was there to fuel my desire to open crates, my life was happy.  Now I'm stuck playing a shitty moba and playing Anime RPGs.

SnowFire

  • DL
  • Denizen
  • *
  • Posts: 4964
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #102 on: January 11, 2011, 02:22:11 AM »
Sentries serve to make fights longer and win them.

Just adding that in.  (And saying it as somebody who initially disdained sentries for anything but ramp-blocking or Marine-spoiling.)  Which is not to say my micro with sentries is perfect yet, but I've seen enough times where well-placed FFs cut an army and half and the front half dies horribly unsupported.  Part of the whole "things die faster now" bit I guess.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #103 on: January 11, 2011, 03:38:15 AM »
Sentries are obviously good.  Like...Guardian Shield: who doesn't want +2 armour?  Hallucination also offers some of the best scouting.  Force Field is probably their best use, but even without it they're obviously something you might want.

And...I'm still not entirely convinced on battles happening faster.  I just have memories  of losing 7 units to a spider mine I somehow missed.  Not to mention Reavers.  They tell you to make Hydras against reavers, as units that can't hit air can't stop the shuttle micro.  All fine and good, but if I mess up my micro even once...then half my army dies to one scarab.  And Mutas, which have barely changed, felt fragile as hell in SC1, like you could lose your entire group in one misclick.  Less so in SC2, with the stim nerf, the lack of irradiate, storm seeming easier to dodge, Archons nerfed, and Corsairs being gone.

The lack of base trades in SC2 is noteworthy, though.  I'm sure the maps are partially to blame (the idea that your armies won't cross each other in an SC2 sized map is silly).  Drops seem less popular except possibly for Terran; Protoss strangely ignoring the new shuttle, and Zerg usually getting a Nydus worm instead, which allows for retreat and defend very quickly.  Kinda like Muta harrass.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #104 on: January 17, 2011, 02:55:45 PM »
So...one thing that's caught my attention recently: completely ridiculous Mario World hacks.  They've taken a step beyond Kaizo (the "lol, look how hard I am", which honestly isn't very interesting to me) and stepped more into an exploration of what's possible--what the limits are of the game system.

A leading example being Item Abuse:

http://www.youtube.com/watch?v=QvY1MNF26-w

Which seems to be as much about showing off glitches as anything else.  (There's also an Item Abuse 2, which has one or two glitches that weren't done in IA1, but it's mostly more of the same).


Then there's Cool or Cruel.  I don't know why, but I just find the game design of this game cute to look at:

http://www.youtube.com/watch?v=uOH5jJAvxDg
http://www.youtube.com/watch?v=x7jrzcYj9oA
http://www.youtube.com/watch?v=cqg2Gkx4ckQ

One thing I do like about what this game does (other than obvious style) is that it seems to actually show outlines on the invisible blocks.  I'll accept the existence of invisible blocks to make some interesting jumps, but there's no reason they should be invisible.  The IWBTG style "LOL, I screwed you" is funny to see once, but especially when I'm watching someone else's successful run (which is the only way I'm likely to experience Cool or Cruel) it's nice to see what they're actually dealing with.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #105 on: March 17, 2011, 06:29:27 PM »
So...a few things I've noticed on SC2.

First, for 2v2, Marauder Zealot is a very good earlygame combo.  Concussive shells means you can't kite/run from the Zealots.  (By extension, Marauder Baneling is probably superb too).

Second, Colossus are really good lategame units.  I'm sure you're thinking "well duh", but stop and think: what are desireable lategame qualities?
AoE (check, perhaps best in game)
Resistane to AoE (check: large size, and top five HP for a ground unit across all races).
Range (check, can fight for #2 here)
Mobility (Decent.  Probably about average.  Walks a little slow, but has cliffwalking).
Mana (Nope).

Now compare that to other options.  High Templar and Siege Tanks and Ghosts have serious mobility issues and are much less resistant to AoE.  Archons, and Ultralisks have decent mobility and AoE resistance, but no range.  Hellions and Banelings need to be suicidal to use their AoE fully, and don't resist AoE themselves (that said, hellions have mobility.  They lack stats, but large groups of hellions will beat, say, large groups of even armoured units like Stalkers provided they can get close to AoE).

Anyhow, Colossi are just everything you could want in a lategame unit.  You could nerf them pretty hard (drop their damage a bunch) to the point that they were no longer viable in the midgame, and I suspect we'd still see them pulled out in 30 minute games.  (As is, Blizzard seems happy to just provide air units with bonus damage to massive to keep Colossi in check).
« Last Edit: March 17, 2011, 06:31:30 PM by metroid composite »

SnowFire

  • DL
  • Denizen
  • *
  • Posts: 4964
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #106 on: March 20, 2011, 08:15:16 AM »
Yup.  In a 2v2 push, I'll generally pump more Zealots than usual if I'm playing Protoss.  They tank well for terran bio support, too, which when stimmed is notably more deadly than stalkers.  And Colossi are great, yes.  My default 2v2 tech vs. good players is Robo, since you want Obs anyway, and Colossi work against their hordes.  (vs. worse players when part of teams in the lower leagues, just go Council->Dark Shrine and warp in 4 DTs, then send them to the back of the opposing mains + expos.  Always hilarious, but this is countered by them pushing earlier, which is hard to count on against competent opposition.)

Banelings require fairly precise control in 2v2, as you're going up against a ball that is potentially twice the size it normally would be early, and thus can mow down unsupported banelings before they blow easily.  It's even more imperative than usual to hit from multiple angles simultaneously with 'em.

I'd say that mass Hellion, even microed well to get MT damage (which does not happen if you A-move!), should only beat Stalkers under weird circumstances.  Stalkers outrange them and tank their damage magnificently.  They may also have Blink, too.  You are almost always better off just retreating your Hellions and building Marauders in such circumstances, even without the MT, unless the Stalkers are literally in your base and Hellions are all you have. 

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #107 on: March 21, 2011, 05:12:28 PM »
Direct 1v1 no-micro matchups...

My methods were as follows: calculate the time needed for one unit to lower the other unit to 0 HP.  Compare the two times; this is the combat ratio.  No attack/defence upgrades (usually they don't change much).  No abilities used.  Yes passive upgrades (combat shield for Marines).  Notice that this does not directly translate into a cost ratio--generally if you outnumber n:1, you can take on something which outstats you by a ratio of n*(n+1)/2:1.  Cost ratio is just "add their mineral and gas counts; take ratio".  With about 30 units, there's about 435 matchups, so I'll stick to the close ones.

Marine vs Sentry:
Combat Ratio: 1.3:1 (Sentry wins)
Cost Ratio: 1:3
(Marines are cost efficient)

Landviking vs Roach:
Combat Ratio: 1.1:1 (Viking wins)
Cost Ratio: 1:2.25
(Roaches are cost efficient)

Landviking vs Hydralisk:
Combat Ratio: 1.4:1 (Viking wins)
Cost Ratio: 1:1.5
(Hydralisks are cost efficient)

Roach vs Hydralisk:
Combat Ratio: 1.3:1 (Roach wins)
Cost Ratio: 1.5:1

Colossus vs Archon
Combat Ratio: 1.3:1 (Colossus wins)
Cost Ratio: 1:1.25 (or 1:1 if made from DTs)
(Technically, if you can spread them flawlessly so that they never get splashed while still focus firing, then 5 HT-morphed archons beat 4 colossi, making them mildly cost-efficient.  Good luck pulling that off.)

Stalker vs Hydralisk
Combat Ratio: 1.1:1 (Stalker wins)
Cost Ratio: 1:1.17
(Hydralisks are cost efficient)

Stalker vs Roach
Combat Ratio: 1.3:1 (Stalker wins)
Cost Ratio: 1:1.75
(Roaches are cost efficient)

Marauder vs Hydralisk
Combat Ratio: 1.15:1 (Hydralisk wins)
Cost Ratio: 1:1.2
(Marauders are cost efficient.  ...Without stim.)

Marauder vs Stalker
Combat Ratio: 1.1:1 (Marauder wins, but it's closer than I thought; though again, this is without stim)
Cost Ratio: 1.4:1

Marauder vs Landviking
Combat Ratio: 1.2:1 (Marauder wins)
Cost Ratio: 1.8:1

Stalker vs Landviking
Combat Ratio: 1.1:1 (Stalker wins)
Cost Ratio: 1.5:1

Thor vs Ultralisk
Combat Ratio: 1.2:1 (Ultralisk wins...but not by much)
Cost Ratio: 1:1

Archon vs Ultralisk
Combat Ratio: 1.3:1 (Ultralisk wins)
Cost Ratio: 1:1.25


And for some fliers.....

Viking vs Phoenix
Combat Ratio: 1.3:1 (Phoenix wins)
Cost Ratio: 1:1.11
(Phoenixes are cost efficient)

Viking vs Corruptor
Combat Ratio: 1.1:1 (Corruptor wins)
Cost Ratio: 1:1.11
(Vikings are cost efficient)

Viking vs Mutalisk
Combat Ratio: 1.4:1 (Viking wins)
Cost Ratio: 1:1.12
(Vikings are cost-efficient)

(uncharged)Void Ray vs Phoenix
Combat Ratio: 1.6:1 (Void Ray wins)
Cost Ratio: 1:1.6
(Phoenixes are cost efficient...although they're pretty much tied if the VR is fully charged)

Mutalisk vs Void Ray
Combat Ratio: 2.7:1 (Void Ray wins)
cost ratio: 1:2
(Mutas are only barely cost efficient, and not if the VR is fully charged).

Viking vs Void Ray
Combat Ratio: 2.4:1 (Void Ray wins)
cost ratio: 1:1.78
(Almost tied for cost efficiency...until fully charged when VR's damage spikes against armored)

Corruptor vs Void Ray
Combat Ratio: 2.3:1 (Void Ray wins)
cost ratio: 1:1.6
(VR's are flat out cost efficient here, even not fully charged).

That's all for now; wtf Void Rays.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #108 on: March 28, 2011, 04:25:22 PM »
Hm, so since it's not feasible to do every single comparison, I went and did an approximation.

Roughly speaking, this first list is durability*DPS, but it also accounts for stuff like burst damage (i.e. in Starcraft 1 it doesn't matter that Siege Tanks shoot once per minute, they've already OHKOed you with the first shot).  It accounts for stuff like armor (with some best-guesses for stuff like Immortals).  You'll also notice that there's multiple versions of some units, like Immortal A and Immortal (the "A" stands for their damage against armored units).  Brood Lords I just assumed that people focused down the Brood Lords and took full damage from the Broodlings.

Planetary_Fortress:41.91
Ultralisk A:23.35
Battlecruiser land:19.42
Thor ground:18.12
Battlecruiser air:13.98
Immortal A:13.39
Carrier:9.01
Mothership:8.56
Ultralisk:8.24
Archon Bi:6.52
Spore_Crawler:5.82
Missile_Turret:5.56
Void Ray A Charged:5.32
Spine Crawler A:5.27
Thor L air:5.17
Colossus:5.07
Brood_Lord:4.31
Archon:4.23
Spine_Crawler:4.18
Photon_Cannon:3.97
Immortal:3.96
Siege Tank A TM:3.42
Dark_Templar:3.15
Void Ray A:3.1
Siege Tank A SM:3.06
Void Ray Chrarged:2.41
Thor air:2.35
Phoenix L:2.22
Corruptor Ma:2.04
Siege Tank TM:1.86
Banshee:1.86
Siege Tank SM:1.83
Void_Ray:1.73
Stim Marauder A:1.72
Zealot:1.42
Marauder A:1.39
Corruptor:1.35
Viking A air:1.23
Queen Air:1.19
Stalker A:1.18
Auto-Turret:1.1
Viking land:1.03
Phoenix:1
Ghost L:0.97
Roach:0.94
Queen Land:0.93
Viking air:0.83
Stalker:0.8
Stim Marauder:0.77
Hydralisk:0.75
Hellion L:0.65
Marauder:0.62
Reaper Bu:0.59
Mutalisk:0.59
Reaper L:0.47
Ghost:0.44
Sentry:0.3
Infested_Terran:0.26
Stim Marine:0.25
Marine:0.21
Reaper:0.18
Hellion:0.18
Zergling (attack rate upgrade):0.14
Zergling:0.12
Broodling:0.08
SCV:0.08
Drone:0.07
Probe:0.07

Next, the efficiency list.  Basically, I took the gas and mineral cost, added those together, then took the above number and divided it by (cost+100)*cost.  This is by no means a perfect formula (it somewhat underrates the five units that cost less than 100 minerals, namely Marines, Zerglings, and workers--four Zerglings totally beat one Roach, for an equal cost example.  Stuff that costs 100+ should be reasonable, though).  Oh yeah, buildings; I included the cost of the drone/CC in the cost of the crawlers and PF.

Missile_Turret:2.78
Spore_Crawler:2.07
Spine Crawler A:1.4
Spine_Crawler:1.12
Photon_Cannon:1.06
Immortal A:0.85
Ultralisk A:0.78
Planetary_Fortress:0.75
Zealot:0.71
Stim Marauder A:0.61
Thor ground:0.6
Marauder A:0.49
Roach:0.47
Zergling (attack rate upgrade):0.45
Zergling:0.38
Dark_Templar:0.36
Battlecruiser land:0.35
Siege Tank A TM:0.33
Stim Marine:0.33
Archon Bi:0.33
Hellion L:0.32
Queen Air:0.32
Siege Tank A SM:0.3
Reaper Bu:0.3
Marine:0.28
Stim Marauder:0.28
Ultralisk:0.27
Void Ray A Charged:0.27
Phoenix L:0.25
Immortal:0.25
Battlecruiser air:0.25
Queen Land:0.25
Stalker A:0.24
Reaper L:0.23
Corruptor Ma:0.23
Marauder:0.22
Banshee:0.21
Archon:0.21
Hydralisk:0.2
Siege Tank TM:0.18
Siege Tank SM:0.18
Thor L air:0.17
Colossus:0.17
Viking A air:0.17
Stalker:0.17
Carrier:0.16
Void Ray A:0.16
Corruptor:0.15
Viking land:0.14
Brood_Lord:0.12
Void Ray Chrarged:0.12
Mothership:0.12
Phoenix:0.11
Viking air:0.11
SCV:0.1
Mutalisk:0.1
Reaper:0.09
Hellion:0.09
Drone:0.09
Probe:0.09
Void_Ray:0.09
Ghost L:0.08
Sentry:0.08
Thor air:0.08
Ghost:0.04

A few surprises.  Carriers are pretty bad compared to Battlecruisers.  They deal more damage, but 550 HP instead of 450 HP, and all of that has 3 armor instead of some 0 armor, some 2 armor.  Thors suck against air; I had assumed they were better because they're supposed to be the new Goliath.  (Although sure: they splash).  Vikings are not as good in the air as I thought (really low firing rate; they need to kite against a lot of stuff) but better on the ground than I thought.  Ignoring stuff like range and movement and spellcasting, Queens are overall better in combat stats than Hydralisks.  175 HP and 1 armor = unholy tank.  I had been mentally comparing Roaches to Zealots, and thinking "I know Roaches have a bit more range, but Zealots just outstat"--as it happens, being near-Zealot level makes Roaches pretty damn efficient.  Thor is the best unit to be neither melee nor specializing in a specific damage type.  I'm a little surprised the Siege Tank Armored is so much higher than the Colossus--might just be a result of the burst damage calculation (which assumes an armored unit with about 140 HP, but Colo has 350 which can soak up the burst).

(Interestingly, Corruptors and Vikings have a pretty close match.  Phoenixes and Vikings have a pretty close match.  Corruptors crush Phoenixes, but go fairly even with uncharged Void Rays.  Void Rays and Mutas are pretty even.  Vikings and uncharged Void Rays are pretty even.  Phoenixes and Void Rays are pretty even.  Phoenixes crush mutas.  Carriers and Vikings are pretty even.  It feels like they went out of their way to make air-to-air fairly even).
« Last Edit: August 07, 2011, 03:03:01 PM by metroid composite »

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #109 on: April 19, 2011, 11:17:37 PM »
Desktop Dungeons

Yeah, it was only a matter of time....

Spell efficiencies!  Assuming a normal run--fighting the boss around level 8-9, the most boringest class/race ever (so like...Goblin-Fighter) and a default boss (i.e. one of the ones that deals 75 damage) how efficient are all of the spells?

BURNDAYRAZ: 0.67 damage per mana per level
HALPMEH: 0.72 damage per mana per level
CYDSTEPP: 0.65 damage per mana per level
BYSSEPS: 0.75 damage per mana per level

Halpmeh varies considerably on the damage output of the boss, of course (much better against low-damage bosses provided it doesn't mana burn...).

What about for a Berserker, though, whose usual math is screwed because they add +2 MP to everything?  (Still assuming Goblin and ignoring deity).  I will also assume that the enemy is higher level and thus Berserker triggers that extra bonus.

BURNDAYRAZ: 0.5 damage per mana per level
HALPMEH: 0.63 damage per mana per level
HALPMEH vs mage boss: 1.37 damage per mana per level
CYDSTEPP: 0.79 damage per mana per level
BYSSEPS: 0.37

As you might expect, adding 2 to all mana costs makes less expensive glyphs trashier, and more expensive glyphs less trashy.  Having high physical attack emphasizes glyphs that push attacking (to the point that cydstepp is more mana-efficient on a Goblin-Berserker than a Goblin-Human).  And naturally, high damage resistance makes healing more efficient.

The next obvious question is...what happens in the other direction, for the Wizard where all spells cost one less?  (But attack power is lower).

BURNDAYRAZ: 0.8 damage per mana per level
HALPMEH: 0.87 damage per mana per level
CYDSTEPP: 0.58 damage per mana per level
BYSSEPS: 1.5 damage per mana per level

This is perhaps an unintuitive result--on, say, an Elf-Wizard, you're better ditching your attack magic, to go slap the opponent around with healing and strength buffs.


The relative value of halpmeh vs cydstepp in general depends entirely on enemy damage--it doesn't matter if you're human and are buffing your strength.  Obviously the relative value of these glyphs compared to attack magic does depend pretty heavily on your strength.  Bysseps is actually quite independent.  Outside of little buffs like Fine Sword, it's always going to add 1.5 damage per level (much the same way Burndayraz always deals 4 damage per level).  The key problem with bysseps is that it doesn't self-stack.  If you want to spend all your mana on Bysseps, you'd better have some plan to get a lot of physical attacks (Halfling Priest!)  (In practice, I find that if I'm human it's usually better to just convert Bysseps for the damage buff than to actually keep the glyph...once I have another damage dealing glyph that I can sink all my mana into, of course).


Hmm...more to come later.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #110 on: April 20, 2011, 03:08:51 AM »
Desktop Dungeons

How much damage can you deal to a boss without mana?  (Ceil(PlayerHP/BossDmg) - 1)*PlayerDmg

Where PlayerHP is the total HP available including through potions.  And obviously this changes a bit if the boss damage is more than player max HP (or even if it's close to the max HP as there will be a tendency to waste potions).


So...let's look at these at level 9 against a typical 75 HP physical boss, assuming conversion of all but one glyph, and again this is only touching physical damage, no magic damage or spellcasting noted here at all.

Standard vanilla (goblin Fighter): 4x58 = 232
Attack vanilla (Human Fighter): 4x76 = 304
Defence vanilla (Dwarf Fighter): 5x58 = 290
Potion vanilla (Halfling Fighter): 6x58 = 348
Max out power (Human Berserker): 4x103 = 412
Balanced Berserker (Dwarf Berserker): 5x85 = 425
Potion Berserker (Halfling Berserker): 6x85 = 510
Standard Monk (Human Monk): 8x54 = 432
Potion Monk (Halfling Monk): 13x36 = 468
Max defence stats Monk (Dwarf Monk): 10x36 = 360
Potion Thief (Halfling Thief): 11x63 = 693
Attack Thief (Human Thief): 8x85 = 680
HP Thief (Dwarf Thief): 11x63 = 693
Priest...is a weird case as their potions full-heal.  If they could somehow use 100% of that healing, and still take Halfling, then 16*58 = 928.  Realistically...can't take two boss hits, so one hit between each potion makes for more like 9x58 = 522.  Dwarf actually does marginally better at these very specific numbers, only because it can survive two hits (so gets 10x58 = 580).
Standard Rogue (Dwarf Rogue): ignoring dodge and first strike: 3x81 = 243.  With first strike: 324.  With one dodge (60% chance!): 405
(All other Rogues actually get OHKOed by the standard boss at level 9...).
I said physical boss, right? (Halfling Paladin): 9x58 = 522
Let's just compare... (Human Paladin): 5x81 = 405
And for completeness (Dwarf Paladin) 7x58 = 406
(Of course, once you allow for the extra attacks granted by the healing spell, Human is slightly better than Halfling for Paladin...and Dwarf is not worth mentioning).

So...yeah.  Dwarf is...solid for balancing out a character that emphasizes offences over defence (like Berserker and Rogue), but is otherwise quite possibly underpowered (only boosts end-game potion use, not healing spells or exploration regen, and is often worse than Human even in the end game!)  Halfling is, as it should be, often the best boss-killer, because it certainly doesn't do anything before the boss.  In fact, I was surprised how often Halfling was the best (Monk, for instance, I assumed was just auto Human; apparently not).  But hey, if you're a level lower when you fight the boss because you picked Halfling?  That would put the class behind.  This also doesn't account for mid-fight level-ups, which are basically free heals (and free heals make the race of extra healing...less special).

Should do magic next...
« Last Edit: April 20, 2011, 03:11:35 AM by metroid composite »

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #111 on: April 20, 2011, 03:03:12 PM »
Magic!

In general, the most relevant stat here is how much mana you have available.  There's not really any such thing as magic attack (although some physical characters in combination with healing or reraise effects deal more damage per mana).

Mana also varies immensely, since deity choice can more than double it.

Same system--assume the conversion of all-but-one glyph, level 9, using all mana potions, and...go:

Typical Berserker (physical Berserker): 33 mana (144 damage with fireballs)
Generic (goblin Fighter): 33 mana (180 damage with fireballs)
Generic elf (Elf Fighter): 53 mana (288 damage with fireballs)
6 Glyph elf (Elf Warlord): 59 mana (324 damage with fireballs)
Generic Gnome (Gnome Fighter): 53 mana (288 damage with fireballs)
6 Glyph Gnome (Gnome Warlord): 58 mana (324 damage with fireballs)
Pactmaker'd Generic Elf (Elf Fighter w/ PM): 79 mana (468 damage with fireballs)
Pactmaker'd Generic Gnome (Gnome Fighter w/ PM): 95 mana (540 damage with fireballs)
Typical Wizard (Elf Wizard): 59 mana (396 damage with fireballs)
Typical Wizard (Gnome Wizard): 58 mana (396 damage with fireballs)
Ideal Wizard (Gnome Wizard w/ Pactmaker): 104 mana (720 damage with fireballs)
Typical Sorc (Gnome Sorceror): 74 mana (432 damage with fireballs)
Mana Sorc (Elf Sorceror): 66 mana (396 damage with fireballs)
Pactmaker'd Sorc (Gnome Sorceror w/ PM): 116 mana (684 damage with fireballs)
Typical Thief (Halfling Thief): 89 mana (504 damage with fireballs)
MP Thief (Elf Thief): 114 mana (684 damage with fireballs)
Pactmaker'd Thief (Halfling Thief): 159 mana (936 damage with fireballs)
Pactmaker'd MP Thief (Elf Thief): 164 mana (972 damage with fireballs)
Bloodmage...see the caveats with Priest, but also that you somehow need to heal yourself from all the damage the mana potions deal.  In-practice...
MP Bloodmage (Elf Bloodmage): 115 mana...some of which can't be used (540 damage with fireballs)
Your Funeral Bloodmage (Gnome Bloodmage): 130 mana...some of which can't be used (720 damage with fireballs)
Pactmaker'd MP Bloodmage (Elf Bloodmage w/ PM): 165...some of which can't be used (900 damage with fireballs)
Pactmaker'd your funeral (Gnome Bloodmage w/ PM): 230...some of which can't be used (1080 with fireballs)

I'm a little surprised the top-end Wizard outpaces Sorceror, but hmm...extra glyph helps, and eventually with enough buffs layered on, the lower mana cost will matter.  (Top end Wizard ought to be better in-principle, as Sorc has stupid good combat stats).  The range for mages without the assistance of a religion is similar to the physical range.  (About 200 damage to boss on the low end, to about 600 damage to boss on the high end...although this is just with fireballs, which aren't always the most efficient spell especially for physical characters...but they're also the only spell you're guaranteed to have).  Just...religions can pump up mana a lot.

Unlike with the physical, Elf is often outright better than Gnome here, in spite of the fact that Gnome will not help you until the endgame.  Possibly these were balanced with religions in mind?  (With religions, the balance between Elf and Gnome becomes similar to the balance between Halfling and Dwarf).

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #112 on: April 20, 2011, 04:57:33 PM »
Desktop Dungeons overall

I'm curious about hybrid--if you have the healing glyph, how do you do?  Typical level 9, 75 damage boss, but HALPMEH glyph...

Completely generic (Goblin Fighter): 7x58 = 406
HP (Dwarf Fighter): 9x58 = 522
MP (Elf Fighter): 10x58 = 580
Potion (Halfling Fighter): 11x58 = 638 (You can actually convert the Halpmeh glyph at the end; final glyph conversions aren't useful on the mana side as there's nothing to spend the mana on)
Mana pot (Gnome Fighter): 10x58 = 580
Atk (Human Fighter): 7x76 = 532

The low value of attack is a little surprising to me here.  When you're using something like a healing glyph, both the health reserves and the mana reserves get a benefit (whereas all the other stats only buff one or the other reserve).  Granted, slight roundoff issues hurting attack here, and high attack allows for better regen games; more on those later....

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #113 on: April 25, 2011, 11:45:02 PM »
Desktop Dungeons

Regen Fighting strategies

Often the strategy is to hit the enemy, regen, and then hit the enemy again, and hopefully your regen outpaces the enemy regen.  Specifically, you want the following:

PCDmg/EnemyDmg > EnemyRegen/PCRegen

So...regen rates:

Default: 1 HP/Level (all enemies)
Monk: 2 HP/Level
Default with HALPMEH: 2HP/Level
Transmuter (if you blow up walls): 2HP/Level
Monk with HALPMEH: 3HP/Level
Wizard with HALPMEH: 2.5 HP/Level
Wizard with HALPMEH and BLUDTUPOWA: 3 HP/Level
Transmuter with BLUDTOPOWA: 4 HP/Level

It's a little bit more complicated when we mix in spellcasting.  Well...let's do pure spellcasting first.

FireballDmg/6 > EnemyRegen
or with BLUDTUPOWA you can get...
FireballDmg/3 > EnemyRegen

But as we know the regen and fireball damage formulae, we can write this properly as

PCLvl*4/6 > EnemyLvl
(w/ BLUDTUPOWA...)
PCLvl*4/3 > EnemyLvl

And now combining the formulae, a mixed fireball and attacking strategy for an entirely vanilla character would be...

PCPhysicalDmgPerTile + PCMagicalDmgPerTile > EnemyRegen
PCDmg*PCRegen/EnemyDmg + FireballDmg/6 > EnemyRegen
(PCDmg/EnemyDmg)*PCLvl + PCLvl*4/6 > EnemyLvl
(PCDmg/EnemyDmg + 4/6) > EnemyLvl/PCLvl

Notably, fireballs are more efficient than healing if PCDmg/EnemyDmg < 4/6.  As the technique is most commonly used against higher level enemies, this will rarely be the case, as we expect EnemyLvL/PCLvl > 1.  If EnemyLvl/PCLvl >= 4/3, then, in order for this to be profitable at all...

(PCDmg/EnemyDmg + 4/6) > 4/3
PCDmg/EnemyDmg > 2/3
PCDmg/EnemyDmg > 4/6

So...EnemyLvl/PCLvl < 4/3 is something that can come up if you're level 4+ and fighting a monster one level higher or less, or you're level 7+ and fitghting a monster two levels higher or less.  (Of course, if you're in a situation where the fireballs are more efficient than the regen, and you have BLUDTUPOWA, obviously turn it on).

Next thing to note, you may not want to enter an engagement where it's greater than...but only barely.  Situations where you're reducing HP by 1 every 6 panels can easily make you explore 90 panels to finish off an enemy; yeah, maybe you'll do it anyway, but on some level that's also a lot of resources consumed.


Ways to maximize your regen fighting:

1. Buff the two relevant physical stats: your damage, and your defence/magic defence.  Human's really the only race that does this.  HP is irrelevant, of course (although it can determine whether you can engage at all).
2. Buff your level.  Dracul's on-worship effect, Orc, Binlor's Heroics, Pactmaker and Taurog exp buffs.  This isn't as useful for fighting higher level enemies (where the point is to get a large level gap to abuse the quadratic bonus experience) but it can be useful for fighting bosses--particularly low-damage bosses.
3. Increase the regen stat more directly.  Notably Monk jumps to mind here.  On the magic side, Wizard for having 5-mana fireballs (which means 4/5 and 4/2.5 formulae).



Poison makes everything easy, of course.  Granted, the poison glyph itself takes resources, so the general strategy is to do some damage, do a bunch of regeneration until you're near-capped, and then do some more damage (that way you minimize the percentage of your resources being spent on poison).

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #114 on: April 26, 2011, 11:48:28 PM »
Desktop Dungeons

So...borrowing this bit of information: in a standard "Normal" run there will be the following distribution of enemies:

Code: [Select]
Monster Level 1 2 3 4 5 6 7 8 9 10 Total
Number spawned x10 x5 x4 x4 x4 x3 x3 x3 x2 x1 39

(I'm not sure if the themed dungeons follow these values outside of bosses; they...might?)

Which means, before fighting the boss, if you never get any bonus experience from fighting higher-level monsters, you'll get 149 exp.  Reaching level 8 takes 5*8*(8-1)/2 = 5*28 = 140 exp.  (Never mind that killing level 9 enemies when you're level 8 would have given you some bonus exp; maybe you cheated a little).  This means that level 9 requires 31 bonus exp, and level 10 requires 76 bonus exp.  31 bonus isn't too hard--of the 38 non-boss enemies who spawn in the dungeon, you need to kill 16 of them when they're one level above you (or 6 of them when they're two levels above you).  Adjust as appropriate if there's some high level enemies you just don't plan to kill.

On exp bonuses...Pactmaker's Training if played right can mean...probably up to 30 bonus exp if you get it after killing only 8 enemies, but probably 20 is a safer bet.  Pactmaker's bonus exp is 10.  Orc will typically get about 20 bonus exp from glyph conversion.  Taurog offers 5 bonus exp.  So...most of the exp bonuses floating around award...about half a level or less (at level 8-9).
« Last Edit: April 28, 2011, 07:58:57 PM by metroid composite »

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #115 on: April 28, 2011, 07:58:21 PM »
Desktop Dungeons

APHELSIK+CYDSTEP: what's the best race: elf of human?

The primary advantage of elf being that more of your mana is being spent on CYDSTEP, the primary advantage of human being you deal more damage per CYDSTEP trigger.  Frequently it will be as follows:

For 15 mana Human: 1.6x base damage

vs

For 25 mana Elf, 2*1.3x = 2.6x base damage

(The actual mana values tend to be 13 and 19, but CYDSTEP effectively increases your mana pool by 10 as you can pre-activate it and then refill your mana).

So...human is very fractionally better and the gap grows with stuff like converting an extra glyph, using Pendant of Mana.


Next comparison: Human vs Elf transmuter with APHELSIK and all other glyphs but ENDISWALL converted, presuming that six uses of ENDISWALL provides a second attack:

Again, pre-heal, which gives an effective +6 mana capacity

Human: for 17 mana, 3.4x damage
Elf: for 23 mana, 3.9x damage

Human is actually quite a bit more efficient in this case (about 17% more damage per tile explored, and fewer walls destroyed of course).


Hmm...yeah, this last calculation specifically applies to my current Gauntlet runs, which are doing well with Elf Transmuter, but Human Transmuter may be better.  (And certainly is better for non-poison attrition).  Of course, Elf Transmuter is more flexible, and can do fireball spam if that's more appropriate.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #116 on: April 29, 2011, 07:57:13 PM »
Desktop Dungeons

So the level curves are interesting; the enemy level curve is quadratic--in particular it looks like...

HP: L*(L+6) - 1
Atk: L*(L+5)/2

Whereas the player level curve is technically linear; sure, you collect gear and buffs from the dungeon floor, but if you get the key ones early, you're set.  Linear vs Quadratic has interesting implications--in general the quadratic curve can start out above the linear curve, dip below it for a while, and then jump back above it again.  DD has extra complications that you're often not fighting on the same level--a lot of important fights are actually against higher level enemies.

0. Fighting an equal level monster will be harder the higher your level.
1. Fighting an enemy one level higher is optimally done at level 2-3.  Level 1 and Level 6 are about equally tricky to pull off on paper (and 7-9 are the trickiest).
2. Fighting an enemy two levels higher is optimally done at level 4.  (Level 1 is the worst level to try to swing this, even if you do get all the good pickups).
3. Fighting an enemy three levels higher is optimally done at level 5.  (Level 1 is the worst place to try this, obviously).
4-9. For all higher level gaps, the most efficient level is always the highest.

Generally, for enemies two levels higher (or less) if you have all the attack power pickups, then healing > fireballs for mana efficiency.  Generally for enemies three or more levels higher, if you have all attack powerups, fireballs will outperform healing.  This assumes generic enemy and generic class, of course.

The other fun statistic:
*non-Dwarf Rogues need to be level 10 to survive a hit from a boss.
*To survive a hit when one level lower than an enemy, they need all three Health+ pickups, or two Health+ pickups and be in the range of level 2-5.
*To survive a hit when two levels lower than an enemy...they can't.

(By comparison, generics can almost always survive a hit when three levels lower, except at level 1, and can survive a hit at four levels lower provided they are level 5-6 versus an enemy of level 9-10).

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #117 on: April 29, 2011, 11:05:56 PM »
Desktop Dungeons

Store items!  So...grabbing straight from the wiki, we have this fascinating mechanic to contend with:

Quote
Item selection is not equally distributed within the Item Rank limit. Only the final shop created will be able to pick randomly from the full limit. The rest pick from a subset of the list based on what items have been unlocked and how many shops are to be created. If 9 shops were created, then the first shop would only be able to choose from the bottom 9th of the list; the second shop would then be able to choose from the bottom 2/9ths of the list, so long as it doesn't pick whatever the first shop has. This continues until all shops have selected the item to be sold.

Which brings up the interesting question of whether it might actually be a bad thing to raise the item rank if, say, the most important item to your strategy had a low or mid item rank.  The issue being that there's two variables here: you might have fewer shops that can spawn your item, but they're also selecting from a smaller number of items.  Let's math out!

These are also going to be a bit unusual, because the values are quantized--there's a big jump whenever you add a new shop.

Item Rank, best three Item Rank Limits (in order)
47: 47
46: 46, 47
45: 45, 46, 47
44: 44, 45, 46
43: 43, 44, 45
42: 42, 43, 44
41: 47, 41, 42
40: 45, 46, 47
39: 44, 45, 46
38: 43, 44, 45
37: 42, 43, 44
36: 47, 41, 42
35: 45, 46, 47
34: 44, 45, 46
33: 43, 44, 45
32: 42, 43, 44
31: 47, 40, 41
30: 45, 46, 47
29: 44, 45, 46
28: 42, 43, 44
27: 41, 42, 43
26: 47, 39, 40
25: 45, 46, 47
24: 44, 45, 46
23: 42, 43, 44
22: 40, 41, 33
21: 38, 39, 40
20: 45, 46, 36
19: 43, 35, 44
18: 41, 33, 42
17: 31, 39, 32
16: 29, 36, 37
15: 27, 34, 28
14: 21, 32, 26
13: 20, 24, 25
12: 18, 22, 19
11: 20, 17, 15
10: 13, 15, five-way-tie between 18, 12, 10, 16, and 14
9: 9, tie between 12, 11
8: Tie between 8/9/10 (all 100%)
n where n=1..7: Tie between n/n+1/.../10 (all 100%)

Some pretty interesting results here; high numbers are better in-general until about the bottom third of the table.  20 is still "45, 46, 36, 37, 47" for top five.  By the time we get to 15, lower numbers are better (but we're talking 30% chance to spawn with an Item Rank Limit of 27, versus 27% chance to spawn with an Item Rank Limit of 47; so...better, but not a huge difference).  And then things really blow up at around 10.

Speaking of stuff blowing up at 10, holy crap what is up with that five-way tie at 10???  10, 12, 14, 16, and 18 all have exactly 50% spawn rates for the item with IR-10, and I'm sure there's a clever mathematical reason why, but from a distance it looks like total coincidence.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #118 on: April 30, 2011, 01:21:44 AM »
The mystery of 10!  Beware this is basically math stream-of-consciousness (and includes a couple of "nope; did that wrong--redo") including an outright algebra error or two.

I'm going to go ahead and abbreviate IR for Item Rank, and IRL for Item Rank Limit.  (Yes, I know, IRL usually means something else.  Whatever).  Here's what's going on in the IR 10 case for 9 shops.


10 IRL
1-8 are fixed.  The 9th shop can either be IR 9, or IR 10.
1/2 = 50%

12 IRL
The 8th shop has a 1/3 chance of being IR 10.  The 9th shop has a 1/4 chance of being IR 10 provided the 8th shop is not IR 10.  The chance of neither of them spawning an IR 10 is thus...
2/3 * 3/4 = 50%

14 IRL
floor(7/9*14) = 10
floor(8/9*14) = 12
floor(9/9*14) = 14
The 7th shop has a 1/4 chance of being IR 10.  If it's not, the 8th shop has a 1/5 chance of being IR 10.  If neither of them are, the 9th shop has a 1/6 chance of being IR 10.  The chance of none of them being IR 10 is...
3/4 * 4/5 * 5/6 = 50%

16 IRL
floor(6/9*16) = 10
floor(7/9*16) = 12
floor(8/9*16) = 14
floor(9/9*16) = 16
The 6th shop has a 1/5 chance of being IR 10.  If it's not, the 7th shop is 1/6 chance.  If not, 8th shop 1/7 chance.  If not, 9th shop 1/8 chance.  The chance of none of them being IR 10 is...
4/5 * 5/6 * 6/7 * 7/8 = 50%

18 IRL
No crazy roundoff error coincidences this time--18 is divisible by 9, so the individual rank limits of the shops are 10, 12, 14, 16, and 18.
5th shop has a 1/6 chance of being IR 10.  If not, 6th shop has a 1/7 chance.  If not, 7th shop has a 1/8 chance.  If not, 8th shop has a 1/9 chance.  If not, 9th shop has a 1/10 chance.  The chance of none of them being IR 10 is...
5/6 * 6/7 * 7/8 * 8/9 * 9/10

And things break at 20, because floor(4/9*20) = 8, not 10.


The pattern is pretty obvious, but it's still feeling somewhat like a flimsy coincidence.  If all of those floors hadn't rounded off perfectly, wouldn't we see chaos?  Does this even work for any number of shops besides 9?  Let's find out!

Starting with 3 shops, trying to spawn IR 4.

4 IRL:
3rd shop can only spawn two items, one of which is IR 4.
1/2 = 50%.

6 IRL
floor(2/3*6) = 4
floor(3/3*6) = 6
2nd shop has a 1/3 chance of spawning IR 4.  If it doesn't, then 3rd shop has a 1/4 chance of spawning IR 4.
2/3 * 3/4 = 50%.

Ok...let's move on to 4 shops, attempting to spawn an item with IR 5.

5 IRL
1/2 = 50%

7 IRL
floor(3/4*7) = 5
floor(4/4*7) = 7
2/3 * 3/4 = 50%

9 IRL
floor (2/4*9) = 4 --ok so this one fails, but that's not too surprising (we're in the bottom half of shops--all the others failed at this point too).


Alright, can we generalize this?  n shops, trying to spawn IR n+1.

n+1+2X IRL
What we want to prove is, for any 0 <= D < X:
floor(((n-D)/n)*(n+1+2X)) = n+1+2(X-D)
(Or at least determine under which conditions this holds)

Obviously this is true if D = 0.  This is also obviously true if (n+1+2X)/n = 2.  (I.e. the magical half-way point; IRL 18 for 9 shops; IRL 6 for 3 shops).  Obviously it breaks if (n+1+2X)/n > 2.  So...what are the risks if (n+1+2X)/n < 2?  Well...chief among them is that it might round up.  In other words, we're trying to prove the following:

((n-D)/n)*(n+1+2X) < n+2+2(X-D)

Doing a bit of manipulation on this...

(n-D)*(n+1+2X)/n < (n^2 + 2n + 2n(X-D))/n

(n^2 + n + 2nX - nD - D - 2DX)/n < (n^2 + 2n + 2nX - 2nD)/n

(-n + nD - D - DX) < 0

(-n + D(n-1-X)) < 0

D < n/(n-1-X)

But since (n+1+2X)/n < 2, we get -n+1+2X < 0.  X < (n-1-X)

D < n/(n-1-X) < n/X

Hmm, actually, that's not getting us anywhere.  Back up.

(n-1-X)*D < n

We know D is at most X-1.

(n-1-X)*D < (n-1-X)*(X-1)

Since (n+1+2X)/n < 2, we get -n+1+2X < 0, so 2X < n-1

(n-1-X)*(X-1) < (X)*((n-1)/2-1) < 1/4(n-1)*(n-3)

Interesting; in the case of 9...

1/4*8*6 = 12

Well...no, ok, obviously I made too many approximations, because it works for 9.  Let's back up to here:

(-n + D(n-1-X)) < 0

Let's double-check the validity of this equation with 9.

(-9 + 3*(9-1-4)) <= 0, 4 <= 0.

Whoops.  Back up.  Ok, starting equation validation:

((n-D)/n)*(n+1+2X) < n+2+2(X-D)

((9-3)/9)*(9+1+2*4) < 9+2+2(4-3)
(2/3)*(18) < 13
12 < 13

Ok, good.  (Although this does point out an incorrect statement above: should be 0<=D<=X).  Oh, damn, I've caught the error, I dropped a 2.  Should be:

(-n + nD - D - 2DX) < 0

(-n + D(n-1-2X)) < 0

D(n-1-2X) < n

But we know (n+1+2X)/n < 2, which means 1+2X < n.  And...hmm...actually, I'm pretty sure this isn't true for n much larger than 9.  At 9 the worst case scenarios are...

3*(9-1-3*2) = 3*2 < 9
2*(9-1-2*2) = 2*4 < 9

Climbing up one...

3*(10-1-3*2) = 3*3 < 10
2*(10-1-2*2) = 2*5 = 10 (FAIL)

Gotcha.  So the first point where this fails is:

10 shops, want IR 11.

At IRL 15
floor(8/10*15) = 12 (FAIL)
floor(9/10*15) = 13
floor(10/10*15) = 15
Which gives us a 1/5 chance on the first attempt.  If that fails, a 1/5 chance on the second attempt.  And if those fail, a 1/6 chance on the third attempt.  Alltogether:
4/5*4/5*5/6 = 53% chance to fail.  (FAIL)

Interestingly, still works at IRL 17:
floor(7/10) = 11
floor(8/10) = 13
floor(9/10) = 15
floor(10/10) = 17
Which gives
4/5*5/6*6/7*7/8 = 50%

The oddity comes from this: it's a quadratic formula:

D(n-1-2X) < n

which means that extreme values of X are going to be low (usually lower than n) but values half-way in between may not be.  Ye olde optimization for finding the highest point of X(n-1-2X) where n is constant.

And now you know.

TigerKnee

  • Denizen
  • *
  • Posts: 41
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #119 on: May 01, 2011, 09:38:06 AM »
So what do you think are the best classes to do Gauntlet Mode in? I'm at level 10 right now and it's getting kinda painful.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #120 on: May 02, 2011, 03:41:51 PM »
So what do you think are the best classes to do Gauntlet Mode in? I'm at level 10 right now and it's getting kinda painful.

I went about 1-14 as Human Warlord on the theory that "I want to be powerful right away".  (And guaranteed Cydstepp means you can get three physical attacks off on anything that doesn't OHKO or mana drain you).  Generally going for Pactmaker, because OMG +10 mana is a lot.  Dracul was also solid for +5 mana, +20% damage, and the HP drop doesn't matter when you're planning on being a Cydstepp whore.  (And since these only take 30 piety, you can often worship Dracul right at the end and get a nice benefit).

I went about 15-17 as a mix of Gnome Warlord and Gnome Sorceror (again, generally going for Pactmaker; although Gnome Sorceror can also go for Mystera Annur).  Sorceror is ridiculously strong early on (at level 1, any spell is better healing than HALPMEH).  In retrospect, Human was the probably the wrong Warlord race; I should have been using Gnome or Elf.  (But the early Gauntlet levels aren't too hard so it didn't matter too much).

By the high end of that range, I would often start killing the boss at like...level 5, and finish killing the boss around level 6 or on rare occasions level 7.   The issue being that non-bosses just used so many resources that killing them would use up all your regeneration space--often not worth-it, especially if you have poison.

Those are the fair, but powerful strategies.  


...and...then there's just breaking the game:

I think I did one round of Human Dracul Platemail Monk.  Monk is actually just good in general--the double regeneration is extremely good when most fights are regen-based.  But the idea of this particular combo is to get 90% physical defence (and then you just win).  Or, in the event of a mage boss, you need Mageplate instead of Platemail.  The problem with the invincibility combo is that the odds are pretty low--you're hoping for a deity that spawns 50% of the time, and an item that spawns less than 20% of the time, so maybe 10% of your runs are successful.

18-27 was Elf Binlor Transmuter.  As previously mentioned, I was often fighting the boss at like...level 5, which has obvious downsides.  Binlor has Heroics, which puts you at level 10.  Transmuter is good at gaining piety with Binlor.  Just about any class at level 10 would make fighting bosses easy again.

28-37 has been Dwarf Binlor Transmuter.  I got a few bad losses to bosses who OHKOed me, so I went Dwarf to avoid the OHKO.  This definitely drops my versatility (anti-physical bosses became a lot harder) but I'd say my win rate is probably hovering somewhere in the range of 40%-70% (well...20%-35% because Binlor isn't on the map half the time, but "no Binlor" resets usually don't take too long).

I'm aiming to get to 47; that's the point when all enemies are at 335%, which means that most level 1 enemies deal 10 damage, OHKOing the typical level 1 hero.  Not that this will stop any of the gamebreaking strategies, it would just be hilarious to show to newbies.  At some point it's probably going to make sense to stop using Dwarf, of course, when enemies start OHKOing me even with Dwarf durability.  Probably back to Elf at that point, I guess?

Not sure if I'll switch class/deity any time soon.  On paper, other classes might be better than Transmuter, especially when I can't go physical anymore.  In practice...I don't know if my copy's glitched, but Binlor doesn't spawn ENDISWALL on-worship like the wiki says it does; so...any other class will have to deal with a 50% chance of Binlor, and 50% chance of ENDISWALL, so 75% chance of failure right away.  Oh, and you probably need extra mana regen, too (BLUDTOPOWA) since otherwise you need 40x8 = 320 tiles of mana regen, which is most of the map.  So...87.5% chance of failure (well...let's say 85%, since Crystal Ball and Bludtopowa are buyable).  So...not really enthusiastic about switching classes.

I have considered Gorgon, mind you.  Fireball boss twice and then poison with a physical attack.  Starts with Endiswall.  Can gain mana by attacking low-level monsters.  Death Gaze cuts the durability of most enemies in half.  But Gorgon is like...only 8% more durable than Dwarf Transmuter so....  (Although thinking on it, Gorgon can potentially win long after every other combo just can't.  Orb of Zot plus find two more glyphs to get 50% Death Gaze--one bought, one from Earthmother, I guess?  Given the 2.5% spawn rate on Orb of Zot, though, I'm never going to aim for that intentionally).

EDIT: wow, thinking on it, Gorgon can theoretically do any of three gamebreaking strategies; 90%-100% defence with Dracul, level 10 with Binlor, Orb of Zot with Earthmother.
« Last Edit: May 02, 2011, 04:33:23 PM by metroid composite »

TigerKnee

  • Denizen
  • *
  • Posts: 41
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #121 on: May 03, 2011, 02:52:53 PM »
I don't know if my copy's glitched, but Binlor doesn't spawn ENDISWALL on-worship like the wiki says it does;

I think he only spawns it if it doesn't already exist. If it does, even if it's ferreted away behind 3 level 9s or something, he won't spawn it at all, which I always thought was dumb (he should at least transport the glyph over, same problem I have with Annur)

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #122 on: May 03, 2011, 08:30:03 PM »
I don't know if my copy's glitched, but Binlor doesn't spawn ENDISWALL on-worship like the wiki says it does;

I think he only spawns it if it doesn't already exist. If it does, even if it's ferreted away behind 3 level 9s or something, he won't spawn it at all, which I always thought was dumb (he should at least transport the glyph over, same problem I have with Annur)

No, no.

That's the way Earthmother works for me, absolutely; I'm familiar with the mechanic.  But I've never seen Binlor spawn the glyph (in 20+ attempts with other classes), and there were definitely occasions where I cleared out the entire map and still couldn't find that glyph.

I'm wondering if maybe Binlor used to spawn a glyph in an earlier or later version (hence why the wiki is wrong) but I can say with confidence that Binlor never seems to spawn a glyph in the version I have.

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #123 on: May 04, 2011, 10:59:32 PM »
Desktop Dungeons:

The map is 20x20.  The specific dimensions aren't that important, but what is important is that there are 400 tiles on the map.  This has several implications:

The most piety you can get from Pactmaker's on-worship effect is 40.  You won't hit 40.  Chances are you will be aiming for one of 5, 20, or 25 piety, though--we can now say that these numbers imply 1/8, 1/2, or 5/8 of the map uncovered.

More important for me right now, though, is attrition fights against bosses.  For instance, if you're level 10, have BLUDTOPOWA active, and are burning the boss, taking a step, burning the boss, repeat, then...for every 3 tiles you uncover, you gain 6 mana, which means 40 damage to the boss, while simultaneously the boss gains 30 HP.  So...for every 3 tiles uncovered you deal a net 10 damage.  If you can pull this off for, say, 210 tiles (just over half the map) then that's 700 damage from regen tactics alone.

I can't possibly note all damage-per-tile-explored values, but here are a few:

Lvl 10, BLUDTOPOWA, fireballing: 3.33 d/ti
Lvl 10, APHEELSIK, fireballing w/ 13 MP: 3.64 d/ti
Lvl 10, APHEELSIK, fireballing w/ 23 MP: 5.22 d/ti
Lvl 10, BLUDTOPOWA, CYDSTEPP, w/ 13 MP and 65 atk: 3 d/ti
Lvl 10, APHEELSIK, CYDSTEPP, w/ 13 MP and 65 atk: 4.33 d/ti
Lvl 10, APHEELSIK, CYDSTEPP, w/ 15-23 MP and 65 atk: 5.20 d/ti

CYDSTEP, of course, allows you to act like you have more mana than you really have (by doing one activation pre-combat, then regening mana)--hence why the 13 mana APHEELSIK case goes a lot better under CYDSTEPP.

I don't have healing listed.  There's a lot of variables involved in healing, but it works a lot like CYDSTEPP in that you can have one full-heal already full before engaging, and then sneak in a second before you poison.  Mostly, healing is cool if it costs less than CYDSTEPP's 10 mana to get an extra attack in (as healing gets much, much higher d/ti values if it costs significantly under 10 mana--or does cost close to that amount but you're taking advantage of both health regen and mana regen--or costs quite a bit, but you can almost survive two hits from the boss so you're effectively storing even more mana with APHEELSIK).

metroid composite

  • m_ACac
  • Administrator
  • Denizen
  • *
  • Posts: 4375
    • View Profile
Re: Theorycrafting! (Because I like competitive metagames too much)
« Reply #124 on: May 28, 2011, 01:57:32 AM »
Magic the Gathering

So...they went and broke it again.

And...not necessarily in the way the DL expected from stuff I've seen posted here--in particular, not because they printed big scary creatures (nobody's playing the stuff I just linked anymore anyway).  But rather because of some interesting miscalculations.  Let's start with the most harmless-looking.

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

Squadron Hawk

Remember Skyshroud Sentinel?  That card was bad, eh?  You got to pay 3 mana for a 1/1 (by all accounts awful on its own) just for the right to get three more cards...but you know they're going to be terrible cards so who cares?  The best use for them was when you had other cards in your deck with lines of text like "discard a card: something awesome happens".

Now, when Squadron Hawk was first revealed, people were like "Ok, 1 less mana, and you get flying; maybe that's enough of an upgrade to be useful.  Well...there'll be some deck that uses it.  There's discard costs, there's decks that want to play two creatures in a turn, and there's some effects that put cards from your hand back into your library, so it'll probably squeeze into a niche deck."  And yeah, Squadron Hawk totally does get used in those decks, but the thing is, it's much, much better than that--to the point that it's used in just about every deck; to the point that slow white-blue control decks stopped playing big angel creatures and started playing 2 mana 1/1 birds.  Let's analyze.

Sliver Queen and her modern cousin Ant Queen are good, not because they've got big bodies (which is actually hardly used) but because "2: make a 1/1" is just a powerful effect.  For that matter, Sacred Mesa was a control staple; recently, and generally involves spending a fair bit more than 2 mana for each 1/1 flyer you get from the deal.  On the one hand, these work at instant speed, but on the other hand, they're easier to stop--you blow up the source, the flood of 1/1s ends; there's nothing to blow up in the case of Squadron Hawks.  Granted, infinite uses vs 4 uses, but 4 is actually quite a lot; given the pace of games these days, the game will often be over before you're left wanting a fifth.

The other way of looking at it: Skyshroud Sentinel was kinda 3 mana to draw 3 cards...which would be a good deal except it's 3 absolutely awful cards.  Squadron Hawk is like 2 mana to draw 3 cards...except 1/1 flying for 2 isn't irrelevant, so kinda like 1 mana to draw 3 cards...which would be broken beyond belief...except it's 3 sub-par cards (which is still quite good).

Of course, it isn't helping that 1/1 flier is better than it normally would be right now--we're in another artifact block, which means lots of powerful equipment trotting around, and 1/1 fliers love equipment.  Speaking of equipment, that leads us nicely into...

Stoneforge Mystic

Let's pause for a minute and look at Elvish Visionary (or for you old-schoolers in the audience, Wall of Blossoms).  Elvish Visionary is played in...just about every format; mostly in Elf decks, sure, but it's pretty solid.  We're back to the "2 mana for a 1/1 that doesn't cost you a card = good".

Stoneforge Mystic is 1/2 instead of 1/1.  It has an extra ability (which I'll get into later).  And instead of drawing a card it searches for a very specific kind of card (Equipment).  Now, initially this seemed balanced enough--since (at the time Stoneforge Mystic had been printed) they had been pretty careful with the equipment they were printing, search your library for an Equipment was at the time probably worse than draw a card, just due to selection.  And indeed, there's evidence to back that up--when Stoneforge Mystic first came out, a columnist I have a great deal of respect for (Craig Wescoe) got top 4 with a white weenie deck that maxed out on Stoneforge Mystic, running such uninspiring equipment as Trusty Machete.

So...that was then; what's changed since then?  For starters, decks with lots of creatures were common then; one of the downsides of equipment in general is that if you have nothing to attach it to, it's a dead card--but Stoneforge Mystic herself can always be equipped; problem solved!  You can now gain all the benefits of powerful equipment in decks filled with counterspells and card drawing.  So...that's one extra aspect that's handy.  Next up: Stoneforge Mystic has a second ability: "2 mana, tap, put equipment from your hand into play".  Now, when Stoneforge Mystic first came out, the equipment worth considering at all generally had casting cost 1 or 2--no real saving, although maybe you'd use it to avoid a counterspell or keep mana open until your opponent's end of turn.  As the artifact block came out, this became "good equipment costs 3 mana" (1 mana savings) and then "good equipment costs 5 mana" (3 mana savings).  The instant speed of this effect gained in value, with the introduction of equipment that comes with a creature token attached (so you can drop an instant-speed blocker).  Finally, the third way Stoneforge Mystic became better is just that the artifact block introduced more powerful equipment--better stuff to search for.

So...overall we went from the starting point of "this is a pretty good creature" to the ending point of "this is a scary win-con all by itself--arguably scarier than the ridiculous big creatures they print at twice the mana cost".

Jace the Mind Sculptor

You know, unlike the other two, this one isn't remotely subtle.  The other two snuck up on me a little--they initially seemed "that looks decent, I think people might even play that" not "that's going to warp the entire format", and I had to go back and think "oh, yeah, actually that does make sense when you do the math and add a few new combos."  Jace 2.0 on the other hand, was hit-you-over-the-head obvious.

All you need to know is that Jace Beleren is a good card that was already being played in most tournament blue decks.  -1 loyalty for draw a card is a pretty good deal.  Jace the Mind Sculptor admittedly costs an extra mana, but has -0 loyalty for an effect considerably better than draw a card.  Even if he had no other abilities, blue decks would likely still run him just for that effect.

But as an added bonus, if you find yourself in the lead, where you don't desperately need to draw cards, he can be your win condition.  And if you find yourself in danger, he can bounce the dangerous creature.  He just happens to be a swiss army knife in addition to being ridiculously good card draw.

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

These three cards form the backbone of the most dominating deck in Standard since Affinity.  (Backed up by hey: counterspells and card draw don't actually suck right now).  Oh but hey wait, here comes a new challenger.

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

Deceiver Exarch

Deceiver on it's own is an ok card; possibly tournament-worthy but hardly gamebreaking--just neat; certainly Pestermite was before it.  Consider: you can untap one of your lands, or tap one of your opponent's lands, so right away it's like 2 mana for a 1/4 (assuming you can't think of anything cooler to tap/untap than a land).  Not bad, kinda fun to play; pretty versatile in what it can do, but hardly terrifying.  Enter Splinter Twin:

Splinter Twin

Enchant Deceiver.  Tap it to make a copy; when that copy enters the battlefield, untap the original.  Repeat a billion times.  Now attack with your billion hasty creatures.  Yep, turn 4 infinite damage with a 2-card combo.  They had to know this was coming--older formats were dealing with Splinter Twin + Pestermite.  For combo purposes, Deceiver Exarch is actually better than Pestermite anyway: 4 toughness, so it doesn't die to Lightning Bolt.  I'm well...a little bit surprised; historically they've said that they try not to print 2-card combos in Standard, only 3 card combos.

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

Not that I'm expecting them to run around and ban a bunch of stuff now; it's 4 months till all the crazy rotates out of Standard.

And until then there should be a metagame; blue-red combo beats stupid white-blue deck (probably).  White-blue-black control with lots of combo-killing cards beats blue-red combo.  Stupid white-blue deck beats white-blue-black combo killer.  And combo is probably not going to be only one deck anyway (given that it only needs two cards; the rest of your deck can be pretty much whatever you want).  And hey, every once in a while, if these decks focus too many of their cards on killing each other, a "how bout I just kill you" deck might surface briefly.  So...strangely this might be a metagame with some diversity.

It's more that...the power of the best decks is kinda wtf-worthy.  And there's the unhealthy factor that a lot of emphasis is being put on a small number of cards, which adds more luck into the equation (did you draw key card X?)