Inspired by salien's post here, I wanted to go more in-depth on the comparison between the Bag o' Tricks and Ice Sickle versus the Haiku Katana and Pilgrim Shield. The latter is the standard weapon/offhand combination in SC speed runs, and the former is the only sensible replacement. The problem is that comparing the damage done as well as the MP returns is kind of tricky. I'm going to assume a moxie class so that myst and muscle are about the same, and I'll just call them M. Bander weight will be B. I'm also going to assume 70% combat rate and a 2-day 550-turn run, as well as a perfect spell-slinging setup. All right, so let's do this.
First, the tl;dr: The BoT with an Ice Sickle seems optimal under the conditions I assume, but it remains unproven.
Stats: 3.75, plus 0-300/day from the BoT's ML. Assuming you use it 5 times (saving one +item from worms/rats, which is doable but somewhat tricky) that's an extra 350 stats from the BoT (more if you time it well, since you lose 30% to noncombats here).
MP: Total gained = 11*550 = 6050.
Additionally, the BoT gets you 15% extra meat/turn. A while ago Darth estimated meat gains per ascension, and an extra 15% comes to 6467 on his assumptions. Since his run was likely longer than one now and that length usually comes at the cost of combats, let's round that down to 6000 extra meat. With seltzer that's an extra 750 MP, for a total of 6800 gained.
Damage: Damage is tricky. I'll do my best.
Geyser: 1.5*(40+.35*M+10+L) 10 is from flavour, L is level from intrinsic. I'm more concerned with how this plays late-game, early game casting stream a bunch worked fine in Type II, so I'm going to pick level 10 and M = 50. So this is 115.5 damage per cast of geyser, for 6 MP.
Stream: 1.5*(3.5+M/10+10) = 27.75 damage/cast from spell damage alone. Also you'll take 9 damage from splashback for every cast. Damage from the bander will be B/2-B, for an average of 0.75*B. If we assume 36-lb bander for easy math, that's an additional 27 damage, for 54.75 damage/cast for 2 MP and 9 damage to yourself (if you're running flavour).
MP for combat: As you can see, your combat strategy is going to dictate how much MP you'll get per turn. Let's see how much damage you'll need to do in, say, the airship. The monsters have ~100 HP normally, and if you're going all-out you'll be running 40+30+15+15+20 = 120 ML (gown at 40 because that's average over moxie days).
So that's 220 HP. So you can kill with 4 streams, 2 geysers, 5 cannons, or some combination of those. 4 streams is optimal, although I personally found casting geyser was necessary because of jala's MP regen. You could also fiddle with not using flavour, it would probably require an extra cast of stream (1 geyser, 3 streams), but save you 18 HP/round. If you're a DB you'll get smust, which means 4-5 casts shouldn't be problematic. Of course, 3 casts of stream is equal to 1 of geyser, so you may as well just cast geyser twice and end combat faster. If you do that you'll be losing 1 MP/turn. If you're casting 5 streams you're looking at losing 3 MP/turn (2*5+3+1) since you'll need noodles and smust to not get hit constantly.
Out of all of these casting geyser twice with jala on (so noodles is unnecessary) was the method I preferred, so that's what we'll assume, for 12 MP/combat.
Bottom line: Out of 550 turns at 70% combat you'll be spending about 4620 MP in combat. But that's ignoring extra buffs. Flavour is 8 MP/40 turns, jaba is 8 MP/10 turns, and jala is 3 MP/10 turns. These buffs add another 715 MP, bringing it to 5335 MP.
But this overestimates things, since any non-hot elemental monster will only take 1 geyser to kill. These include the cyrpt, much of spookyraven, the entire battlefield, and various monsters in locations around the game. Estimate 90 turns for all of those (you're spending a lot of time in the ballroom), and that's 540 MP saved, bringing your total for combat down to 4795.
So it looks like late-game the best you can hope for is being MP-neutral, as long as you're not fighting elemental monsters. The upside is you can end combat in as few as two clicks (if Jala is healing all the damage monsters do to you just cast geyser twice), and easily in 3. That's a lot fewer than the HK, making this significantly less fiddly. The downside is this is extremely dependent on the bander, as soon as you switch to another familiar you're a lot less effective. Since basically all of those turns are with a fairy, using a slimeling will help you significantly (both the MP and the damage), and the feast being equipped will help a bit too. Additionally since you're presumably running a bit less ML during these times combats will be less difficult.
Finals, for those keeping score:
Stats: 3.75/turn + 350/run
MP: ~2000 net.
HP: 62 HP/turn (for 2 casts of geyser)
Misc: 10% items, 15% meat, fast and easy combat with bander.
Stats: 3/turn from the PS.
Damage: Since HK's regen is based on your bonus weapon damage, we need to figure that out first. So we'll assume the same stuff as before. 1.5*(MAX(50-Monster Defense),0) + 16 + 15 + 7 + 10). Since the monster's defense is always going to be greater than your muscle, this reduces to the weapon power and bonus weapon damage only (where I'm assuming major/minor claws and rage in addition to PS's bonus), or 72 damage/cast. Shieldbutt adds an additional 36 damage/cast. The bander, after giving smust, gives the same 27 as it did with stream. So this is 108 damage first cast, and 135 afterward. That's enough to kill in two hits at 100 ML pretty much every time. Although it will likely require a third hit during most of the MacGuffin/war (in my experience this is the case).
MP: Unfortunately the spring raindrop's exact mechanism is not known (there is a lot of data on it here). Fortunately we don't need to know the exact formula. I took 136 data points with the exact setup here (HK in mainhand, PS in offhand, rage and both claws, as well as high ML). I found that the HP gain/turn averaged 19.67 and the MP averaged 19.26, varying between 12 and 30. So we'll go with a 20 MP/combat averave, which gives us 7700 total.
But we need to include all those casts of rage, which is 8 MP/10 turns. Over 550 turns that's 440 MP, for a net of 7260.
MP from combat: We have two combat situations in the late game, shieldbutt->raindrop->shieldbutt, or noodles->raindrop->shieldbutt->shieldbutt->stream/shieldbutt. In the former case that's 10 MP/combat, in the latter it's 16. Assuming 550 turns at 70% combats again, that's 3850 total for the former, 6160 for the latter. To take the easy way out we'll assume a 50-50 distribution on these, that's 5005 total on combat.
So that's 2255 net MP from the katana.
MP: ~2300 net.
Misc: Extra HP and weapons drops from PS.
Code: Select all
| BoT+Sickle | HK+PS | Difference |
Stats | 1800 | 1160 | +640 |
Total MP | 6800 | 7260 | -460 |
Net MP | ~2000 | ~2300 | -300 |
HP | 62/turn | 31/turn | +31/turn |
Misc |10% items, 15% meat|50% weaps/15% Musc| NA |
While this comparison is a bit tricky without comparing actual in-run numbers from similar ascensions against each other, it looks like you're trading about 300 stats from the sickle and 350 from the bag for about 300 MP. That MP is worth 2,400 meat in seltzers, not a whole lot. Additionally, because you're getting so many extra stats you could drop antiphon after some period, which costs 3.8 MP/stat. You would need to cast it only 9 fewer times to make up the MP difference, which is only 90*0.7 = 63 fewer stats, more than made up for by the gains. So by casting antiphon fewer times (especially late in the game) you can very easily erase the MP gap while getting far more stats.
The end result is that you're trading 300 MP for a bunch of stats and 10% items (plus more +items if you can time the buff right). The downside is the strategy is unproven, and any small error in this analysis could drastically affect the conclusion (several times I ran the numbers with slightly varying assumptions and found the BoT did much worse).
How about other scenarios?
The BoT would be a hands-down favorite in other situations, such as a myst run (although a myst run might be able to get away with sickle/PS). Additionally, pulling a scarf could make the BoT even more appealing as it's more stats and more damage (both from the bonus spell and the 15% myst). HCB lead-ins could become more popular if the BoT is viable and turncounts drop just a bit more. The extra stats might reduce reliance on puttying ascii. And in any run longer than 2 days I'd say the BoT is very serviceable, the downside is the much larger skillset needed.
So which one wins?
They're quite even. In fact, the BoT/sickle seems better, but as it has yet to be tried in Type I I view it as a riskier strategy. Which one you choose depends mostly on personal preference, it seems. But the idea that the BoT is completely inferior should be dead.
Would you ever use a BoT?
In Type II definitely, it worked pretty well there. But you were running less ML and had no other choice. In type I it appears better, and if my analysis here is accurate, it might be optimal (resulting in slightly faster runs due to about 6 fewer turns powerleveling). But it's a sight risk since it's unproven, and most of my alternate analyses swung in the direction of "very not worth it", meaning any flaw in an assumption here could be fatal.
The most bottomest line:
"Quite possibly better but unproven". Until someone does a Type I run using the BoT/Sickle setup, it will remain tantalizing but unproven tech.