Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
You can show that the hint is that room 8 is not empty as follows:
Before the king gives the hint, the lady could be in room 1, since the sign on room 1 is true in this case, and the truth or falsehood of the other signs is irrelevant (the other rooms could be empty). Similarly, the lady could be in room 7.
(You can also work out that the lady could be in rooms 3, 4, or 5, but only two cases are needed here.)
So without the hint, the prisoner cannot determine where the lady is. If the hint was that room 8 is empty, then this gives no new information at all (no signs refer to the contents of room 8) and the prisoner could not have deduced the location of the lady. So the hint has to be that room 8 is not empty. Then (following BrunoVisnadi's post), the prisoner deduces that the lady is in room 7.
(The prisoner can also deduce that rooms 1 and 6 are empty and that rooms 2, 8, 9 have tigers.)
Edit: Changed the wording above. Actually you can tell that the sign on room 8 is false, but that can be deduced even without the hint.
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
p4wn3r wrote:
Out of curiosity, has there ever been an instance where this caveat raised problems in an actual tournament?
It's common practice that when you're ahead, but have to play a long endgame to take the victory, you start repeating positions (although obviously not three times) to quickly reach move 40, so that you get more time on the clock. I can picture that someone might confuse three-fold repetition rules and accidentally concede a draw, or a player might claim a draw where one does not exist. Has this ever happened?
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
Oh, now I get it. The major discrepancy between U and J versions is not the unskippable cutscenes themselves (though that adds some time), but all the voice acting in the J version.
Now that I remember, I based my initial assumption that it would take hours based off the U version voice acting in the intro (that is seriously one long intro). Had I bothered to go any further, I would have realized that the U version actually has relatively little voice acting (unlike the J version).
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
Didn't know it was possible to complete this game in under 30 minutes. I thought it would have taken 3 hours! (The U version cutscene skips and the glitches make a difference, apparently.)
I liked this run until about the 16-minute mark which is when the reality of fighting end-game bosses at extremely low levels kicks in. I wouldn't be against a glitchless run.
I'm hardly familiar with the Saturn version so I can't comment about whether this can be improved. I'm more familiar with the SNES version, which is a conventional turn-based RPG rather than an action RPG.
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
p4wn3r wrote:
Let A, B and C be the angles of a triangle, which satisfy the relation
cot(A) + cot(B) + cot(C) = sqrt(3)
Prove that this triangle is equilateral. Bonus points if you can do it without calculus!
Let u=cot(A) and v=cot(B). Then:
cot(C) = cot(180°-(A+B)) = -cot(A+B) = -(cot(A)cot(B)-1)/(cot(A)+cot(B)) = -(uv-1)/(u+v).
Plugging into the above equation gives:
u + v - (uv-1)/(u+v) = sqrt(3),
or, after multiplying by 4(u+v) and rearranging:
4u2 + 4v2 + 4uv - 4sqrt(3)u - 4sqrt(3)v + 4 = 0.
To solve this equation, we will complete the square three times. We can write:
(2u+v)2 + 3v2 - 2sqrt(3)(2u+v) - 2sqrt(3)v + 4 = 0.
Substituting w=2u+v:
w2 + 3v2 - 2sqrt(3)w - 2sqrt(3)v + 4 = 0.
Then complete the square in w and v:
(w-sqrt(3))2 + 3(v-(sqrt(3)/3))2 = 0.
Since squares of real numbers are always non-negative, this gives us w=sqrt(3) and v=sqrt(3)/3. Then 2u+v=sqrt(3) and so u=sqrt(3)/3. This gives cot(A)=cot(B)=sqrt(3)/3, and so A=B=60°, and C=60°. Therefore the triangle is equilateral. QED
By the way, it's a lot easier with calculus; you can just plug in C=180°-(A+B) and differentiate, or use Lagrange multipliers. You can even simply say that if any two angles are not equal, say, A<B, then since A<180°-B, the derivative of cot(x) (which is -csc2(x)) at A is less than the derivative at B and so you can decrease the value of cot(A) + cot(B) + cot(C) by increasing A and decreasing B. Therefore cot(A) + cot(B) + cot(C) is bounded below by the value at A=B=C=60°; that is, cot(60°) + cot(60°) + cot(60°) = sqrt(3).
Edit: Fixed some signs.
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
Oh, looks like Ledah's initial wait does matter.
I did this strat in my most recent file:
Link to video
Here, the initial waits are Ein 10, Ledah 18, Red Dragon 21. (RNG delays are Ein +1, Ledah +7, Red Dragon +7).
(Also, the initial wait difference between Ein and Red Dragon can't be more than 12.)
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
I had a look at the code that generates random numbers to see why it was very difficult to manipulate initial waits other than shifting up or down.
The game generates an additional random initial wait delay from 0 to 9 for all party characters you could possibly have in the game (regardless of whether you have them or not) and 6 enemies you face in battle (not that there are ever that many). These 12 values are generated as:
((RNG+VCOUNT+SIMPLECOUNTER) mod 128 ) mod 10
where
(first is system bus, second is IWRAM)
RNG address = 30034E4 (34E4)
VCOUNT address = 3004FD4 (4FD4)
SIMPLECOUNTER address = 3001FD0 (1FD0)
However, there is a catch. RNG updates itself each time it is called (and only when called), whereas VCOUNT and SIMPLECOUNTER do not change during the calculation of these 12 values.
Since RNG only updates itself when called and VCOUNT and SIMPLECOUNTER don't change in the middle of calculation, the result is that it is difficult to manipulate initial waits for turn order unless you change RNG itself (which not a lot of actions do). Delaying for different VCOUNT and SIMPLECOUNTER only move initial waits up or down, although some values can be made to be off by two (because of the double mod in the formula).
As an aside, RNG updates as:
RNG = swap(RNG*41 mod 65536)
where
swap(n) = (n mod 256)*256 + floor(n/256).
Judging from Flameberger's posts and Google doc, I'm sure he knew most of this stuff already, but I wanted to put this down just in case it is useful to anyone in the future.
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
Sorry for the delay; I was very busy in real life.
I uploaded some of my test TASes here:
http://www.mediafire.com/file/1pxh71ja10cna94/riviera.zip
The U TASes use VBA v22 and the J TASes use VBA-rr-svn480.
The U TASes (the ones I could find) are pretty awful but maybe there are some strats there that could help. I don't know. The J TASes are a little better done but still a lot of improvement there (note that the J TASes use J glitches so they start to diverge after the Aghart fight where I first use the stat glitch).
By the way, two other things:
- Against Boulder Dragon (Red Dragon in Chapter 1), it is possible to have Red Dragon use 1x Normal skill (Bite) and 2x Rage skill (Saberfang) and still have everything else play out the same. You just need Ein to start with a much lower initial wait than Red Dragon (maybe 10 or 11 lower or more). I think it saves 1-2 seconds.
- I remember that getting a Game Over in battle allows you to retry with the enemy at less health (first time 75%, second time 50%, iirc). Have you ever considered using it as part of your strats?
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
BrunoVisnadi wrote:
This channel posted some weeks ago this 'joke proof' of the 4 color theorem. It seems so simple, but I've tried it for a while and I couldn't find the mistake yet:
https://www.youtube.com/watch?v=Ryoipkksktg
The place where the proof fails exactly is at 11:54: "... we can recolor the left blue square to green, the right one to yellow, and make the center blue."
To be less pedantic, there are four-color colorings of the second 5-neighbor case (at 11:16) which satisfy all the conditions supposed by this case, except that after switching blue/green, the right blue square cannot be switched to yellow without affecting the yellow square on the bottom-left. In particular:
* The red/yellow and red/green connections can cross, because they both share red.
* In this case, it is possible for a green region in the red/green connection to be contained within the red/yellow connection, connected through green/blue to the left blue square.
* Then recoloring the left blue square changes this green region to blue. This breaks the red/green connection above and possibly enables a blue/yellow connection between the right blue square and the yellow square.
* In that case, there is no way to recolor the right blue square to yellow without changing the yellow square to blue.
An example that shows this is in the image below.
Regions AEFCH surround the ? region with the same colors as in the video. AB is a blue/green region surrounded by red/yellow connection HDGE. Similarly, CD is a blue/yellow region surrounded by red/green connection HBGF.
The "proof" says we can switch the blue/green region AB, then the blue/yellow region CD, in order to color the ? region blue. However, if we switch the blue/green region AB, then B becomes blue. This creates the blue/yellow connection CDBE. Then switching the blue/yellow region containing C switches E to blue, making it impossible to color ? blue.
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
Ranking in detail (no spoiler tags because these are not spoilers):
30 Super Mario Bros 2 Peach in 8:20.82 (http://tasvideos.org/3640M.html) [4882pt]
29 Densha de GO!One Handle Controller No. 715 TAS [5278pt]
28 Fate/UnlimitedCodes Rider [5638pt]
27 SD Gundam Gachapon Senshi 5 [5773pt]
26 Mario Sunshine stage comparison [6110pt]
25 Super Mario Bros Warpless in 18:37.46 (http://tasvideos.org/3665M.html) [6404pt]
24 Oishii Korokke o Tsukurou in 29.96 [6718pt]
23 Image Fight 2nd round [7877pt]
22 Yumenikki Uboa in 1:48:27 [8149pt]
21 Gradius IV Boss Rush [8404pt]
20 Momotarou Matsuri in 31:13.71 1/2 [8655pt]
19 Runabout 2 in 24:30.33 [8967pt]
18 Strikers1945 2nd round [9021pt]
17 Ko~ko wa Do~ko no Hakoniwa Ja? 01:54:41 [10493pt]
16 R-TYPE II 2nd round [11080pt]
15 Kamen Rider Kurabu 1/2 aim for 37m14s [11196pt]
14 Densha de GO! [11619pt]
13 Nights: Journey of Dreams [13254pt]
12 Metroid in 7:57.38 (http://tasvideos.org/3666M.html) [13329pt]
11 Super Momotarou Dentetsu III 5th year greatest earnings TAS [13574pt]
10 Puyopuyo if it were played by two shogi masters at 6-dan (part1) [13967pt]
9 Tales of Destiny Leon WIP part2 [14475pt]
8 Sonic & Sega All Star Racing part 1 [15281pt]
7 R-TYPE II 2nd round [15410pt]
6 Super Mario All Stars SMB2 in 7:16.47 (http://tasvideos.org/3644M.html) [16605pt]
5 Mega Man Zero 2 in 49:47.57 (http://tasvideos.org/3617M.html) [17557pt]
4 Super Metroid any% in 35:58.31 (http://tasvideos.org/3653M.html) [18331pt]
3 Fate/UnlimitedCodes Lancer Arcade Mode [23480pt]
2 Super Mario Sunshine any% in 1:07:52.96 part1 (http://tasvideos.org/5873S.html) [210599pt]
1 Final Fantasy 6 1-victory clear [322141pt]
Translations are off because I did this very quickly and I just want to get the main stuff down.
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
Flameberger, if you want, you can cancel this submission. (Though I think this submission would probably be accepted even if you had no intention of improving it.)
If you like, I can upload some of my older files. It will take a while to find them, because I did them a very long time ago during the VBA days.
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
I only noticed this after looking over your TAS again, but in 6-4, it is faster to fail the minigame for the jump(the one right after getting Leyte Claw); it takes you directly to 6-5.
I like the strat you have for 5-4; didn't think about using LXA Ankhs by going to the red Crystal Sanctuary twice. How much faster is it than using LTA Ankhs (which is the intended route)?
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
I never expected this TAS. I thought this game was such a pain that nobody would have bothered.
This game is pretty interesting if you're playing casually, but this game is extremely difficult even to make a route for it; I made no less than 7 test TASes for this game on the U version (the one without the useful glitches) and never got to the end in any one of them. RPGs that aren't named "Pokemon" are complex beasts after all. (Pokemon is still complex; don't get me wrong. It's just that Pokemon isn't as complex.)
I still think doing the glitched J version would be an interesting exercise. It has the buff glitch (if character dies when buffed and you win the battle or in practice mode end it, that character permanently gains that buff) and the HP doubling glitch (if you lose in practice mode, the characters you used will double their HP). I made a not-very-good test TAS on the J version shortly after learning about the glitch (about two years ago), completing it with a time of about 3:02. With polish and optimized strats it could be reduced to 2:45 or so. (Might be too optimistic a guess.)
I like many of the strats used in this submission's TAS (some which I never thought of). I noticed some other things though. Some notes:
* It is a pain to manipulate initial waits (turn order and stuff), but it is possible. I think it is easier if you delay frames from some time back. Still not sure how it works.
* Against Aghart, I think it is faster to let Ein die after the first Flare Tempest, so that Ledah uses a Divide in place of Ein using Double Slash twice (and Aghart doesn't use Flare Tempest again). Not sure of the timing.
* I think Slumber Ivy works just as well instead of Spider Web on the Black Wyrm in Jet Black Wind (although you'd have to play around with wait times). If you use Slumber Ivy, then you don't need to do the event to get Spider Web and you can skip one Rock event. The events cost about 34 seconds.
* I typically use Fia->Longbow->(Lv2)Cleave Arrow rather than Lina->Longbow->(Lv3)Arrow Rain as the finisher for Black Wyrm because it doesn't take as long to level for a Lv2 skill compared to Lv3. Don't know how this changes the rest of the TAS though if Lina needs to be levelled.
* Spiked Hammer may be useful somewhere (low-accuracy attack that does 75% current HP damage when used by Lina and 50% otherwise). Obviously not in a RTA because of its extremely low accuracy.
* Perhaps the Tarot is useful somewhere (it's an HP% attack and with the right luck it does massive damage to everyone at once when used by Serene, IIRC), even though I never ended up using it. However, it costs a lot of time to get, before you gain access to Wyvern Wings which are so much better.
* Hourglass is useful against Isher's first (invincible) phase; have everyone throw a Scythe or something on the first round of attacks then have the next character to act use Hourglass over and over. This cuts down on the number of moves your party needs to make.
* I think it is faster to get poison gas (or arrow trap?) on a trapped chest and fail it. IIRC the HP penalty can be assessed to anyone in your party and isn't even permanent.
* In Chapters 4 and 5, you have unlimited access to Wyvern Wings (because you can get them from Wyvern-type enemies in the two practice mode battles Air Riders and White Plateau Wyvern, respectively). Maybe it would be better to go out of the way to farm Wyvern Wings? Don't know how many is necessary using this general route.
* Maybe the Killer Beehive has some use? It does 88x8 fixed damage if you get it to hit the enemy.
* Yeah, the boss battles in the final two chapters are a big reason why I never completed my older TASes on the U version. I still remember some really weird strats I tried such as putting Cierra in the back row vs Ledah (at low levels she's the only one who can survive Ledah's Burning Galley and everyone else dies horribly) and using Slime Ball to poison Malice while surviving with Serene using Ice Emblem (Malice is the only late-game boss which is not immune to poison; a poisoned combatant takes 15% max hp damage every time its turn comes up). Needless to say all my runs pretty much end somewhere during the boss rush in Chapter 7.
Flameberger wrote:
Purify is actually a glitched status which has no effect. However, even a glitched debuff would overwrite buffs and cancel my Power Wrist.
I thought that was the whole point of Purify: to remove your buffs and prevent you from using any buffs. Then again, the other statuses also do that and have some other kind of effect, so I don't really know.
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
Don't know if the title screen music is glitched in BizHawk, but this is how it is supposed to sound:
Link to video
(yes, I know it's not any better...)
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
Memory wrote:
What do you define as 100% in Super Mario Land even?
I actually thought of it some time last week. Maybe get everything there is to get in this game? (coins, extra lives, mushrooms, flowers, stars)
I assume deaths would have to be ruled out; otherwise you'd obviously get infinite loops.
Edit: Actually, IIRC it's not possible to get all coins, but most of the coins in the game can be gotten. There is also the possibility of going down and up pipes to "restore" some coins and item boxes. Only works once per pipe though (thankfully).
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
Pretty well optimized; in fact, it's only 1.5 minutes slower than my Pokemon Yellow submission which uses L100 Nidoking glitch and is run on an emulator that has far less lag!
Unfortunately I'm guessing the lack of general interest in Pokemon Yellow is because it is seen as an inferior version of Pokemon Red/Blue, which is faster. I can't blame anyone for seeing it that way.
A part of me wishes that red bar skip never existed. However, the other part of me finds it funny enough that red bar skip breaks a lot of the audio in this game (though if I remember correctly, red bar skip works even better on Red/Blue because it also skips opposing Pokemon cries too).
TiKevin83 wrote:
The easiest way of explaining the category difference is that "glitchless" allows S+Q (not relevant for TAS) and allows exploiting oversights that cannot lead to ACE (like Poke Doll on Marowak) whereas "glitchless classic" disallows an arbitrary set of non-ACE-causing glitches to experience more of the storyline.
In my opinion, the simplest explanation is that the non-JP RTA community has always allowed Pokedoll skip and the JP community has always banned it. That is a valid explanation, and both rulesets are equally valid.
I'm fine with Pokedoll skip since I don't like the Rocket Hideout, and the game is not obviously broken because of that. It's not like the Sonic the Hedgehog "no zips" TAS, which didn't go over well with a lot of people just because of two "ejections" (quoting Aglar here) that are so obviously glitches that clashed with the glitchless feel of the rest of the TAS.
TiKevin83 wrote:
I'd be careful with that Youtuber. The channel reposted a 1:35 Pokemon Sapphire run last year, in complete ignorance of the 1:27 that existed for a long time. In fact, the Ruby TAS is faster despite being run on the English version!
At least the run posted appears to be the same as the run on nicovideo which I think is the fastest one.
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
Oh, nice. I had actually considered making this run (not to such a perfect extent though) when I first saw it on speedrun.com. Good thing I didn't actually work on it (beyond a single test run I made in a couple hours).
A bit surprised you were willing to do all this work in a category where every single run is in danger of being completely obsoleted by a single discovery.
What's great about this TAS: No save corruption, no item menu corruption (those are way too easy and have been done millions of times before). Instead, just an extremely well-planned glitch that suddenly warps you to the end. And it looks like there was a secret Bug Catcher Hall of Fame in Viridian Forest all along. :D
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
I uploaded an updated version of Fortranm and mkdasher's script that includes some additional things like autofire AABB, battle analysis for crits/misses, and a "minimal mode".
Autofire AABB should only be used at low emulation speed (0.5x) and not when going through level-up stat messages.
Battle analysis makes it easier to check for optimal (minimum total) delays for crits/misses.
I also have a WIP up to Floaroma Town. Some little optimizations like pressing direction keys to cancel the last dialog are used here (faster than bumping).
Fortranm wrote:
Good to know, but switching to L=A shouldn't really waste any time since you need to choose "confirm" to actually save the options anyway.
You can just wrap over the top to get to "confirm".
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
Here is my WIP of Pokemon Diamond (J) up to Jubilife City.
http://tasvideos.org/userfiles/info/45437177458522052
I have to redo some things because I forgot that you can press directional buttons on the last dialog of some dialogs, saving time instead of turning.
I'm working off of seed 0x160e0f55; my tests already show that this seed is very advantageous for avoiding encounters in Oreburgh Gate and Oreburgh Mine.
Chimchar has IVs h0 a31 b7 c30 d0 s26 (Atk+ SpDef-) and the wild Starly has Def- nature as a bonus.
Fortranm wrote:
For instance, every time you use an item from the menu, you would need to wait for one non-lag frame before you can press A another time. This wouldn't happen if "L=A" is chosen.
You don't need L=A in that case; in menus, you can click on the center of the Pokeball on the touch screen. It's the same as pressing A.
You can also use the Pokeball touch screen to scroll down a list very quickly, if you have 4 or more items to scroll through.
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
Chatterbox wrote:
1UP Mushrooms in X-1 (except 1-1) only appear when we get all the coins in (X-1)-3 ( For example 1UP Mushroom in 4-1 will appear when we get all the coins in 3-3).
Hm, I never knew about this before. Explains why you did what you did in the X-3 levels.
Guess even the earliest games have some neat things programmed in.
Editor, Experienced Forum User, Published Author, Skilled player
(1941)
Joined: 6/15/2005
Posts: 3247
Fortranm wrote:
It would be very helpful if you upload your script with crit. hits and misses detection. Thank you very much.
http://tasvideos.org/userfiles/info/45171049311956410
Red is for both critical hits and max range; blue is for 95% miss; magenta is for both red and blue.
Note that crits and range and hit/miss are determined before the move message saying that Pokemon used move. Range is determined one frame after crit (2 RNG values) and hit/miss three frames after crit (5 RNG values). If you want both crit and max range, you need two red squares two apart from each other.
Fortranm wrote:
I made Monferno get poisoned during Gardenia fight in my glitchless run because it speeds up many upcoming battles. I'm not sure if it's worthy for Jupiter fight alone.
Regions AEFCH surround the ? region with the same colors as in the video. AB is a blue/green region surrounded by red/yellow connection HDGE. Similarly, CD is a blue/yellow region surrounded by red/green connection HBGF.
The "proof" says we can switch the blue/green region AB, then the blue/yellow region CD, in order to color the ? region blue. However, if we switch the blue/green region AB, then B becomes blue. This creates the blue/yellow connection CDBE. Then switching the blue/yellow region containing C switches E to blue, making it impossible to color ? blue.