Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
n720 wrote:
Now do a zero-time artiblast depolymorph, and plunge into some water. Crawl out onto a land mine on some ice, which explodes. This breaks a lit potion of oil, causing a fiery explosion. The ice you are on is melted, and you drop into the resulting water and crawl out again -- onto another land mine on ice. This also explodes. It breaks a potion of unholy water, triggering lycanthropy, and then puts you into a pit. The explosion of the potion of oil is not done yet, and now it melts this ice, dropping you into water yet again. This time, you have to ditch items before crawling out, because of the difference in carry capacity between your lycanthropic and normal forms and a fine-tuned inventory weight. You drop a boulder; because you are in a pit, you get squished under it, losing rnd(15) HP, which reverts you to normal form. This calls drown again, in something similar to bug C343-193, with a similar result of crawling out twice in a row. Land on the magic portal and a polymorph trap in either order, and turn into a flying form again and leave the Quest
I really want this to work but don't think it adds up.
Trying to follow this exactly, the double-drown here relies on losing HP from emergency_disrobe → flooreffects. In order for flooreffects to be able to call drown(), you'd need to be on a water tile. However, dropping a boulder on water turns off the handling for dropping a boulder on a pit (because the boulder/water interaction is checked first), and even if the boulder "sinks without a trace" the game still doesn't check the pit.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
I think there are two fundamental problems with a 2003-turn run:
a) How to get into the Sanctum before the end of turn 2000;
b) How to get into the correct polyform by the end of 2002,
n720 wrote:
Now do a zero-time artiblast depolymorph, and plunge into some water. Crawl out onto a land mine on some ice, which explodes. This breaks a lit potion of oil, causing a fiery explosion. The ice you are on is melted, and you drop into the resulting water and crawl out again -- onto another land mine on ice. This also explodes. It breaks a potion of unholy water, triggering lycanthropy, and then puts you into a pit. The explosion of the potion of oil is not done yet, and now it melts this ice, dropping you into water yet again.
I've mostly been looking for a way to do a). Unfortunately, although your plan is very impressive and deals with a number of things I didn't realise were problems, there are two big issues with it. One is that detonating landmines will cause wounded legs status, interfering with many actions we need to perform, and I don't think it can be cured without spending a player action. The other is that the timing rules for explosions are different from what you imagine here (explosions do damage to terrain left to right (for some reason) but damage to the player last). There seems to be a much simpler solution if we're using landmine chains, though; after a 0-time unpolymorph that drops us into water and crawls out right onto a landmine on ice, have a fiery explosion (from a lit potion of oil on the square) melt the ice on that square and the square to the right, and have a magic portal and polymorph trap there (in that order) on the square to the right and the square to its right. We don't fall in once and crawl out twice; rather, we fall in once, crawl out once, then fall in again when the explosion catches up to us, pushing us onto the polymorph trap. In a bit of coding that was really annoying for most of my plans but helpful for this one, magic portals (unlike level teleport traps, which trigger immediately if it's the player's turn) never activate until all effects of the current action have been resolved. Now we're a flier in the upper Dungeons over water, and still have a "oLS equipped; there are multiple ways to get to the VS level in zero time from there.
EDIT: This doesn't work on Home:1 for a wizard as on hard-floor levels you can't dig on squares where a trap already exists. That's a ridiculous restriction, but it prevents us getting ice underneath a polymorph trap. This may be fixable. (It could be fixed by playing valk, but this would require redoing the whole run and give us fewer, maybe zero, options on Fire. There's hopefully an easier way to fix it.)
The amount of luck in the plan is astronomical as worded, though; a 200 teleport fail on an unvisited level which we can't migrate monsters onto have odds that are unlikely to be satisfied on any RNG seed that exists, unless we're in a swimming polyform that can't exist on land (this makes all the squares on the level mostly-invalid targets, meaning that most teleport attempts will naturally fail). Swimming polyforms, however, can't pick up items; so we couldn't pick up the Bell on action 2 and although it's easy enough to get migrating monsters to carry it around, I can't see a way to get the Bell dropped some time during the zero-time portion of action 3 so that we can autopick it up (other than by catching the Dark One in the explosion, but if he drops it on a square "behind" us in the explosion it'll be a water square so we can't pick it up, and if he's on a square "ahead" of us in the explosion he'll still be alive so won't have dropped it). At least there's a lot of scope here for creativity, so there's likely some way to do it.
n720 wrote:
Coming into turn 2002, use a delayed instadeath to end helplessness, and polymorph into a queen bee. Because you were in natural form, you get only two actions. Put on an amulet of life saving and jump into the portal to the Plane of Water.
However, b) is also a problem, and this is one of the best attempts at it I've seen. One huge problem here is the amount of luck needed. The portal on Fire would have to be on some very specific locations; (67, 12), (64, 15), (64, 16), and (64, 17) appear to be the only four map tiles that work (out of the entire very large map), and even then, we need a push from a monster to reach any of those (they're all out of the range of a 4-square jump from the starting location of (69, 16)).
Coming onto Water, nh_timeout() runs before movebubbles(), so my basic idea was to flood the area with monsters so that there was no room to move the player inside a bubble. Unfortunately, after inspecting the code, this seems not to work; the player turns out to take precedence over monsters when the bubble moves, so there's no way to escape the bubble as a result of a turn boundary; the monsters will be mnexto()ed around in order to make room. Even if that could be fixed, Water doesn't seem like a fun level to do a 200 teleport fail on.
Assuming that we fulfil a), which now seems likely to be possible somehow, we'll be in the Sanctum on the last player action of turn 2000. We have one free action then to trigger a monster turn, and can spend it fixing wounded legs or putting on a "oLS (but not both). Even assuming we could do both, though, we can't leave the Sanctum before the first action of turn 2001 (although we can reach dlvl 1 on the subsequent monster action); can't reach Earth before the second action of turn 2001 (slight exception: if we're in natural form and an artifact generates on the entry square on Earth, we can "oLS-cancel it, but this doesn't seem useful (due to the need to be in natural form) and is also incredibly unlikely, and I'm not sure it's even possible – it would be another matter if we could migrate a trice corpse there but we can't); almost certainly can't reach Air before the third action of turn 2001 (there's no timing reason why a monster action couldn't get us from Earth to Air on the second monster action of turn 2001, but I don't know of an action that would actually accomplish that without spending an additional player action to #monster for enexto); so because we used the third player or third monster action of turn 2001 to reach Air, we can't reach Air with more than one action left.
I'm considering a hypercharge engulf to be too unlikely to succeed here (in addition to the fact that it would be very slow and tedious to set up – possibly taking more time than we had, because setting it up requires turn boundaries – and the fact that I have no idea how it interacts with portals, it moves the player in a random walk and a random walk from the entry square of Air hitting the portal is very unlikely if the walk's of a reasonable length). So we can't reach Fire on this action, except possibly via overflow hurtling (in which case we end up on Fire in natural form).
I think the best approach might be to go back to the old, old plan for the Planes; if we're on Air on the last action of turn 2001, the best approach might be simply to manually polymorph into an air-E, use the turn boundary to flood the level with eggs, and enexto to Fire on the first action of turn 2002. We're now on Fire with three actions left, which beats the two we'd get from an overflow hurtle solution; even though we get there later we're moving faster, so actually have more time left. Polymorph into air-E, equip Eyes, step, jump is four actions, but (if we had time to equip a "oLS and didn't need to heal our legs) we could "oLS-cancel one of them. A push from a monster might theoretically mean that we didn't have to step, which would give time to heal our legs too (making the jump legal). That would take us to Water with four actions left, which is enough.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
The way I see things (which IIRC agrees with the way SDA sees things) is that a major skip is a sequence break that "reconnects the level graph", i.e. it makes it possible to get around a mechanic that gates progress from one part of the game to the next. Most games have rules for when major sections of the game become available (e.g. in Super Mario Bros. 1 completing a level or using a warp pipe connects to a specific other level, in Super Metroid you need to defeat 4 bosses to unlock the final area), and these rules are normally very clear and obvious to the player. A sequence break that nonetheless respects these rules (e.g. teleporting from the start of a level to the end) is a minor skip. A sequence break that violates the rules (e.g. wrong-warping in Yoshi's Island) is a major skip.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
Oh, I see; you're trying to get a monster on the Quest portal level to push the player without taking an action. That might work; I'll have to look into how schedule_goto works if it triggers twice in the same action, because I don't think that's an intended case in the code.
I suspect that if this works at all, there's a simpler way than the way you described, but that key idea is one that I haven't really looked into yet, so it might yet lead to an action saving on the way down (which is when we really want it).
EDIT: As far as I can tell, deferred_goto is called once at the end of the monster turn, then not called again before the player acts (unless the player is helpless, in which case it's called at the end of helplessness). That in itself is not a problem: we could trigger the level change using a zero time action. However, the specific case of getting a monster to push us off the portal seems to be missing a spoteffects() call, meaning that the game wouldn't notice if we landed on a trap, water, etc. At least spoteffects() is sufficiently widely used that there might be some way to do it indirectly. (It's actually done indirectly in n720's plan, via using the depolymorph code.)
Actually getting a monster onto the Quest portal might also be difficult. We can't do it from outside – if there's a monster on the portal we couldn't use it ourself, so how could we have entered the Quest? – but we might be able to do it via letting a monster leave the Quest via the portal. When changing level, the player is placed before any monsters which migrated there since the player was last there; however, the migrating monsters can legally be placed on the same square as the player, having a chance to push them. So that seems like it might work too.
As a sanity check, it's worth verifying that all this actually saves time:
2000 player action 1 or monster action 1: unlock Quest
2000 player action 2: level teleport to Quest
2000 monster action 2: kill the Dark One, teleport player onto Bell, teleport player onto levport trap (very difficult – remember the level's only just been generated! – but theoretically possible)
2000 player action 3a: zero-time drop into water, crawl onto portal, get pushed onto water (requires ridiculous polyform management as mentioned by n720 earlier)
2000 player action 3b: zero-time drop into water, crawl onto levport trap (or trapdoor, I guess), levport to Valley
2000 player action 3c: manual level teleport to VS level, "oLS-cancelled
2000 player action 3d: start reading Book
2000 player action 4: finish reading Book
2000 player action 5: enter Sanctum
2000 monster discharge: retrieve Amulet, etc.
Yep, it seems like if all of this can be made to work, it's an action faster.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
Unfortunately, when you change level during a monster action, monsters don't move after the level change until either you've had a turn, or a turn boundary happens. I thought of leaving the Quest the intended way, but I can't see how it's any faster simply because we need to spend one of our own actions, or a turn boundary, on every second level.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
I didn't know how much I wanted to know what the site registration process was like nowadays until I saw this.
Also, that "original CAPTCHA" is amazing. I'm assuming that the current one mostly works simply because it's hard to adjust the API of existing CAPTCHA-breaking libraries to it, rather than because it's hard for computers to break in an abstract sense.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
I think the description would be improved by explaining why each item is needed. For example, is the deku stick required to complete the game? Or is it just a choice you made for an arbitrary item to duplicate over when anything would do, and it was just the most convenient?
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
IIRC, RBA can be used to refill bombchus but regular item drops are the only fast way to get regular bombs (which are needed in some situations).
Also, this category isn't as arbitrary as it gets. It's pretty quick and simple to describe, after all; really arbitrary runs tend to have a lot more descriptions. (I do have to admit that it is fairly arbitrary, though, but I see it as more of a "conduct run" – a run which has a single arbitrary restriction to make it more difficult/interesting – than a completely arbitrary one.)
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
c-square wrote:
Check this out: A human SMW ACE submission! And he's looking for an opponent.
I bet a TASBot 11 Exit vs. Human ACE would be very well received. People there would be especially excited if the Human beat TASBot for once (gotta give them something to celebrate, eh?).
It strikes me that this could be billed as "a human using a TASbot-style route, versus TASbot using a human route". When you put it that way, it can be very uncertain as to who is faster and might well lead to a close ending.
EDIT: Even better: "To make this fair, TASbot's going to use the human route [pause] and the human's going to use TASbot's route".
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
My own view on this is that the fastest any% and 100% completions of any specific version of a game should be accepted at least to the Vault. (We may want to think about vaulting all but one of the runs, even if they're entertaining, if they're very similar.) That way, we'd at least have a baseline for knowing what the time would be on any specific version of a game, and a record for trying to beat. The purpose of the Vault is to serve as a record repository, after all; why should we stop tracking records just because similar runs exist?
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
Attempted fix for the actions-on-turn-2000 issue: we can farmove on the last action of turn 1999 (actually moving, and stopping at turn boundary due to an adjacent hostile monster; we might need the monster to hit us, but that's no big deal), kill a monster at turn boundary with a scroll of stinking cloud, then use the first action of turn 2000 (which would otherwise be a dead action) to set up equipment removal storage.
Unfortunately, though, I've done more analysis of equipment removal storage and I'm not certain we can make use of it this early (which is yet another way the current plan is broken).
Although it's easy to set up equipment removal storage with a ring (I misunderstood how it's done so it's wrong in the current plan, but it can consistently be done by attempting to simultaneously unequip a ring and shield and then getting hit by a monster on the following monster action), this can't be used to cancel helplessness; the buggy codepath is specifically looking for armour, shirt, cloak, boots, helmet, gloves, or shield (presumably because it needs to list each case individually and the people writing the code weren't aware that it's possible for a ring to be the currently-being-removed item).
Now, we can get equipment removal storage with body armour via taking off body armour and shield simultaneously, then getting hit by a monster. There's actually a secondary bug that works in our favour here: normally if a worn item is stolen it causes helplessness for a number of turns equal to the item's delay, but the code neglects to run that bit if the item is flagged as currently being removed anyway (possibly on the basis that it'd look strange to have to remove the item again from scratch if you were already halfway through removing it).
The big problem with all this, though, is that unlike rings, most wearable armour has a tendency to fall off when we polymorph. For example, an air elemental can't wear items in any of the seven slots that are capable of cancelling helplessness. While we're maintaining equipment removal storage, we can't use fast polyforms (because the equipment would fall off), and we can't use steed speed either (because the "actually moved at least one space" action required to use the steed's speed rather than our own cancels equipment removal storage). As such, if we're using an overflow hurtle at all, we may either have to set up equipment removal storage only just before the hurtle, or else cancel it using lifesaving (which leaves us helpless until the next turn boundary). The fastest we can set up equipment removal storage for armour if we've previously been in air elemental form is probably two actions (one to equip, say, a cloak; the other to unequip cloak and lenses at the same time), and two actions is a lot when you only have seventeen total.
Anyway, that's about where I am at the moment. Fitting all this into 2003 turns is beginning to look very difficult indeed, although it may well still be possible.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
I did some more research on farmove storage, trying to understand exactly how it works.
There are two relevant variables, flags.mv and multi. flags.mv is what I'm calling "farmove state"; it's intended only to be set during a farmove command (given that we're using numpad controls, that's commands like "g4", "56", "M8", etc.). multi, meanwhile, has three possible values: 0 is a neutral state, positive is command repeat state (i.e. a state in which the player is doing the same thing over and over again), and negative is for helplessness (i.e. a state in which the player is doing nothing over and over again, waiting for the helplessness to end). I call this the "command repeat state".
Now, farmove commands set the game into farmove state (pretty obviously), but they also set the game into command repeat state (which makes sense, as a farmove consists of moving repeatedly). The command repeat state eventually ends, and there are two intended codepaths for this: a) if it ends because the "remaining repeats" count goes from 1 to 0, it clears farmove state at that time; b) the function "nomul" alters command repeat state (to zero or negative), and also unconditionally clears farmove state. "nomul" is heavily used in the NetHack codebase, and serves two purposes: it's the main function used to inflict helplessness (a "negative" command repeat state); and it also has an alternative mode (inflictin "0 turns of helplessness") to cancel multiturn commands in cases where the player is surprised or can't continue the action (e.g. is attacked by a monster, runs into a wall, etc.). Farmove storage itself can be set up using a glitched codepath, in which we lifesave from a delayed instadeath in the middle of a farmove; that clears command repeat state but not farmove state.
Anyway, the important thing to note here is that although the game is pretty good about clearing farmove state when the player is in command repeat state, it mostly ignores it when the player is not in command repeat state. As long as we can avoid doing something which would cause a nomul() call (and unfortunately, although most things that cause a nomul() call are fairly specific or obscure, there's a really long list of them), we can sustain the farmove state for a pretty long time. For example, I was wrong a couple of posts ago: reading the Book is an occupation, so it does not cancel farmove storage.
Unfortunately, the potential gains from this aren't as large as they might seem. Staying in farmove state permanently means that the steed is permanently at speed 53. However, our movement point allocation sometimes uses our own speed and sometimes uses our steed's; the requirement for using the steed's speed is that the most recently entered player command is a movement command and actually moved the player at least one space. The "actually moved at least one space" requirement doesn't apply if the player is drowning in lava, but that doesn't seem like a particularly helpful alternative case.
Another huge problem with this is that the same codepath that sets the "actually moved at least one space" flag clears the "in equipment removal" flag. This means that if we want to get an extra action from farmove storage, we necessarily have to lose equipment removal storage in the process. The current 2003 turn plan is invalid because it tried to get a fifth action on turn 2000 via a farmove that doesn't move; that's a problem, because that won't actually give a movement point boost. We can't change it into a farmove that does move, either, because that would clear equipment removal storage.
I don't have any fixes planned for all this yet; perhaps we'll think of something. But I wanted to get the research out there for the record and in case someone else has an idea.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
I'm beginning to wonder in how many games you could find the RNG routine simply by searching the ROM for 0x41C64E6D. It's not a very commonly used number in most contexts, but it *is* very commonly used in RNG algorithms. Obviously, this wouldn't work for every game, but it wouldn't surprise me if it worked in a fairly large proportion of them.
64-bit RNGs are becoming more popular, but unlike 32-bit RNGs, there isn't a consensus value for a yet (which is normally the most distinctive value in an LCRNG algorithm). So unfortunately, those may be harder to find by this method.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
An overflow hurtle is a glitch which happens when you hurtle via kicking a monster while levitating, but you have a negative total weight in your inventory. The formula for the distance travelled breaks down quite badly upon seeing the negative numbers, and the end result can lead to very large distances if the weight of the player and monster are chosen appropriately (although it can also lead to zero distances surprisingly often; as the name suggests, integer overflow tends to be relevant here).
Moloch's Sanctum being no-teleport is a notable problem for the current plan, although probably not an insurmountable one (after all, we have a spare action!). I considered plans a while back that make use of a c!oGL in monster inventory to take the Amulet out of the Sanctum that way; the current plan seemed simpler (and seemed to work), but I might have to revisit that. One problem with trying to use farmove storage is that I'm not aware of any way to set it up without spending an action, and reading the Book necessarily cancels it (meaning we can't just set it up beforehand). Additionally, I'm not immediately aware of a way to set it up within the Sanctum within 1 action and without the help of a turn boundary (although it's a fairly new glitch and I haven't comprehensively looked for ways to set it up yet). (Incidentally, this would make the 200 teleport fail sequence simpler to set up on the way into the Sanctum, but more complex on the way out, for polyform management reasons in both cases.)
If there are artifacts that work for zero time unpolymorph that are more usable than the PYEC, that would be an obvious improvement (not in gametime, but to our sanity trying to set things up).
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
2000:25 Read the Book (action 1).
2000:M1 Hostile quantum mechanic steps next to us. (Does this cancel
Book-reading? If so, there could be problems.)
2000:13 Read the Book (action 2, taken automatically).
2000:M2 Hostile quantum mechanic hits. 200 teleport fail onto
polytrap, become air elemental, 200 teleport fail onto
levport trap, enter Sanctum. Conflicted trice, quantum
mechanic follow (they don't need a large charge, the required
charge is measured in subpixels here).
Just noticed a potential fix to these problems if they do occur: luckily the quantum mechanic actions happen as the last thing in the turn, so we can if needed start the Q one space further away and give the monsters one level of charge extra.
This does badly depend on the hypercharge glitch actually existing, though. It's a theoretical glitch that's never been tested. (The idea is to cause time to advance on a level without monsters moving, causing them to gain more and more subpixels without any actual consequent movement, so that on a later turn, the movement can happen all at once.)
EDIT: Just noticed that I have a different fix a couple of posts ago. Guess that shows that this is the most concerning part of the run :-)
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
boct1584 wrote:
I like the idea of TASVideos collecting single-level records for games like this where you CAN pick one level to play at a time. Maybe make these sorts of runs auto-Vault? Or a new tier?
SDA has a rule that single-level runs are allowed, but only if one exists for every level (making a full "individual level table" that serves as a substitute for a full-game run). Once the full table exists, the various levels can be obsoleted individually.
Something like that might make sense for TASvideos, too. The rule that every level needs a run before individual levels can be submitted individually seems to have a good reason for existing, though.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
OK, I think I've made up my mind as to where I stand on this.
The first thing to think about is the discussion that lead to the creation of the Vault; we want to be able to publish any%/100% records even if they have no other redeeming qualities. (Incidentally, I argued that Vault and Moons should be seen as complementary categories, allowing a game to be in both, but I was outvoted and that lead to the current tiers system.)
As such, from my point of view, there's no doubt that both the PAL and NTSC versions of the run are legitimately vaultable (regardless of whether or not they're also viable for Moons); and if they currently aren't for some reason, the rules should be changed so that they are. In general, I'd be in favour of allowing both the PAL and NTSC versions of any game separately into the vault if a) they obey the other valut rules, and b) there's a noticeable difference in gameplay between them (i.e. the PAL record isn't just a straight copy of the NTSC record played back slightly slower). Both of these are definitely true statements for this Super Mario Bros 1 situation.
The remaining question is which (if both) of the runs should be in the Moons tier, rather than the Vault (an issue that's necessitated by the fact that the site sees Moons as "above" vault rather than the two as separate tags). I'll leave that up to the judges because I don't hugely mind how it's handled; the distinction is fairly arbitrary anyway.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
Just to explain my no vote: this run is really hard to follow and I couldn't tell what was going on and whether it was impressive or not. So I didn't find it particularly entertaining to watch.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
I was actually really entertained by this, likely because it's so bad (rather than in spite of it). It creates a sort of constant horror as to what will happen next.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
I'd argue that because Super Mario 64 runs can be synched between emulators, and between emulator and console, via automatic lag compensation on polling, that it really doesn't matter what emulator is used; after all, all you're submitting is the input file. Come up with an emulator-agnostic format for it if you need to, to make it really clear that the run isn't attached to any given emulator. There are games where the emulators act differently from each other in terms of sync-stability, and therefore someone has to make a decision about which emulator each run is on, but IIRC that isn't the case here.
The actual time should be determined via the most accurate available method of replaying the run (which in the case of Super Mario 64 is TASbot, because none of the existing emulators handle lag correctly); this goes for both the displayed time, and for determining whether a change is actually an improvement. Producing the encode is up to the encoders. (For what it's worth, I would not object at all to a rule that states that runs submitted on Mupen can be encoded on Bizhawk.)
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
I don't think this is the same situation as with Spikestuff's links. Minimum presses of any button looks really arbitrary to a viewer, and can't really be evaluated with an input display. However, minimum presses of a button that does something specific and obvious within the game (such as jumping) can be much more interesting, as it becomes more of a low%-like category.
What do the A and B buttons do in this game? That's highly relevant here. For example, if they're jump buttons, the category should be "minimum jumps", defined objectively in the submission text via the use of banned control inputs.
Additionally, I think a run like this should aim for speed as a secondary goal. A LOTAD can be entertaining, but seeing the same goal optimized is typically much more entertaining, as you get all the usual TASing fun in addition to seeing how the puzzles were solved. Additionally, low%ish / "puzzle" runs of a game greatly benefit from detailed commentary explaining what's going on, the intended ways of doing areas, how banned button presses were bypassed, and the like; I've never seen a LOTAD of nontrivial length that doesn't have one of these, and an optimized run would benefit from it too.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
I still believe this particular choice for what to include in the movie file should be flagged as a speed/entertainment tradeoff; the game's either completed as soon as it enters a postgame state, or it isn't legitimately completed at all, and adding extra input to show the credits is thus ending input late.
I'm against the idea that entering a cheat code can ever be considered part of a legitimate completion of a game. (It can be considered a legitimate way of unlocking a game mode or character, at the most, but in such a case that's happening outside the actual gameplay.)
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
It's not random, it's always a Mythril Sword. You can get more from random encounters but that seems like too much a waste of time, so I guess you'd just have to put a human in Soldier or bangaa in Warrior for a bit, which shouldn't be too hard. The thing about combos, though, is that as long as you have at least 1 helper and a bunch of JP, they do massive damage even if nobody else helps (and in fact, you'd want to stay restricted to 1 helper for animation length reasons). Also, only JP on the initiator matters, not on anyone else. If you have two units adjacent to an enemy with 10 JP each, they can each combo in turn and end up doing 10× the damage of a regular attack.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
hejops wrote:
Facing doesn't affect damage.
I never bought consumables; I was selling them.
Yes, my laws were not planned out optimally. Time was wasted due to bad laws at Desert Patrol, Hidden Vein and To Ambervale - the last set of laws is particularly horrid. Antilaws could be a great idea, but I'll need to check if story missions give out Antilaw cards.
I think every mission after Antilaws gives out at least one law card. Not 100% sure on that, though.
Famfrit is Magic based, and Montblanc stops gaining significant Magic after switching out of Black Mage, so I doubt Famfrit would be useful past Diamond Rain.
OK, that makes sense. It's a pity that Gadgeteer stat growth is so awful. (Switching classes would help, but you're being pulled in too many directions there to boost all the stats you'd want.)
The second Human (Rooster) became useless towards the end, so I'm definitely scrapping him, or replacing him with someone more useful. However, having a party of 5 (initial 4 + Bangaa from Dueling Sub) makes me ineligible for Race Wanted missions, so I'll have to find another way to get a recruit.
Kotetsu can be obtained very early (Type 2 = earliest after Desert Peril). And I agree, the Jagd Hunt turtle was terrible. Again, the only sane way to get an Assassin is through Race Wanted, so that would require me to rethink the party, especially since Huntmoon starts only on the 61st day. But Assassins are that good, so I just might consider...
I was thinking of maintaining a party of 3-4 through most of the game, hiring people for special occasions and then firing them soon afterwards. In other words, you'd have a small core team, plus a "newbie slot" that you could just top up using Race Wanted whenever you needed a particular class. In RPG runs, it's normally fastest to use less than a full party, in order to avoid splitting up experience; this game might be different because you don't want to waste time with enemy turns, although if you're throwing Chroma Gem around, maybe not.
What enemies do you have in mind for Poison? None of the early game enemies struck me as exceedingly difficult. Sick Knuckles arrive at the first shop upgrade (in my run, after Antilaws).
I was initially thinking of Famfrit, but that's probably too early (although Swarmstrike is available off zero shop upgrades and zero mastered skills, so it's not entirely impossible; Fencer is a common class, so you could recruit one specifically for the mission by using the mission before). Later in the game, against units with a lot of damage, combos are likely to be the better option (if you initiate a combo with 10 JP, and at least one other unit participates, you're doing at least 5× regular damage).
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
Hmm, another possibility: on your first shopping trip you bought some consumable items, but then you didn't use them before your second shopping trip. Buying consumables at the last possible moment means that you have more cash on hand, potentially allowing you to buy more missions at the same time and thus save on menuing.
Other ideas:
- Walking around to manipulate laws seems like it's more time-conssuming than it could be. How fast would it be to manipulate and use an appropriate law card instead? Additionally, it feels like in a few battles, a Target Area ban would help prevent the enemy using time-consuming animations, whilst also helping to grind up JP on your Gadgeteer (perhaps giving more uses of Famfrit).
- I have a feeling that you have more party members than you really need. New hires come in at the average party level, so you could dismiss someone, funnel all the experience to a subset of your party, then hire someone at the new higher level, also saving all the turns that the dismissed members would have, plus the need to train them up. You could also hire people into expert jobs directly (e.g. get yourself your own Gadgeteer so that you don't have to use Montblanc).
- How early is it possible to manipulate a Petalchaser or Kotetsu? In some missions (e.g. Jagd Hunt), you spend a lot of time trying to kill enemies who aren't vulnerable to regular physical attacks. It strikes me that it might be faster to hire an Assassin and just let them use an OHKO ability from the weapon directly (no need to grind up AP or experience on them). Obviously this won't work on bosses, but there are many missions where you have to kill tough non-bosses.
- Is poison a cost-effective way to do extra damage to enemies earlier in the game? Against enemies that take many turns to beat, poisoning an enemy is likely more effective than damaging them directly. (Finding a source of it might be hard, though; Green Gear may be the best option, but I'm not sure how early you can get the Sick Knuckles.)