I'll wait a little longer for comments on the start of run thread before making rules changes and retiming. If you want to retime your own runs, go for it.
Before retiming, I have to wonder, what about the end of the run? As in, what does "final hit" mean?
- @Dragondarch ends at start of the death sound playing.
- @Krayzar ends very slightly earlier than the sound starting.
- @bizarreluck ends at start of death sound, however the rest of the Hermit animation continues to play. This saves 4 seconds.
- @Spockrocket on Ianuki ends after the final hit lands, before the feedback damage or death sound. This saves 1-2 seconds.
Everyone timed with the final rule correctly but in interpretations that were fastest. I realize it's hilarious to be all legalese here over a few seconds. Issue is 1-4 seconds is significant in a 1 stage map. I don't like the idea of an attack being better if it's the final hit either.
My proposal and open to other proposals: "final hit" means start of death sound and completion of attack animation. As in, if Chariot/Devil/Hermit is used, card has to finish animating.
If you time with damage number disappearing instead of death sound starting or however Krayzar did it, that's fine. Maybe a runner here one day is deaf or hard of hearing.
When I time, I usually time it on the first frame that I see the boss health value transition to becoming 0 like you say - as a few of the videos seem to use that as the metric, but now that I look through, you're right it does seem to be less than consistent.
I usually try to use consistent visuals where possible as that tends to keep things more clear for timing, but it doesn't much matter to me. There's visuals for the final hit (either unit shake or hit flash are standard on all versions) and the start of the damage numbers (visible and pretty consistent on all versions) too that could work just as well.
Health value becoming 0 seems like a good visual cue. I'm also ok with NewSchoolBoxer's proposal. I have no strong feelings on this one, I'm happy to go along with whatever gets agreed upon. I don't want a timing method to give an unfair edge, especially in Fireseal where 1-2 seconds can make the difference for WR. I didn't realize my timing was that inconsistent with other runners, apologies for that!
I just went with the death sound because it's easier for me to sync up with a sound effect than a visual cue. But it looks like both the sound effect and the HP bar changing happen at basically the same time.
As for waiting for a card animation to finish, I'm not entirely sold on it. I'd rather have it be something consistent rather than having a couple fringe cases with certain cards. That would likely lead to some confusion with new runners.
Hi Dragondarch, welcome back!
I recorded on PS2 on Dragon's Heaven that my capture card or OBS deinterlaces to 30 fp. VSDC seems to redo the deinterlacing and put back into 60 fps. Music off so I could discern when death sound plays. 22.6 MB file:
- Damage numbers fully appear from Hermit at 1:40.266.
- 30 frames later / 500 milliseconds, damage number(s) and attack name, if attack has a blue bar name, both disappear.
- 19 frames later / 317 milliseconds, the HP bar updates to 0.
- 2 frames later / 33 milliseconds, the pixel decay on Albeleo and death sound start.
- 288 frames later / 4.8 seconds, Hermit is fully over and WINNERS is displayed.
If two runs are off by 1-2 seconds, they're going to be retimed anyway using the same start and end points. At least I would. I like runner's choice of 0 HP update or death sound, STR card issue aside. Defeats Ianuki time save.
Seems a battle loads 18 frames / 300 milliseconds faster with music turned off. I think a mute run could be rejected under suspicion of splicing versus needing an explicit music on rule.
That's a good point for accessibility making rules easy to understand. I can't think of a compromise solution that is worded concisely. WINNERS doesn't show on SNES or Saturn Diablo.
Thanks Spockrocket for being supportive of 0 HP / death cue! No apologies needed. Everyone followed the rules. I'm okay with no retimings for existing runs that result in a longer time, barring a tiebreak or what is decided for bizarreluck's run.
My Saturn Sarjem Island runs unveiled a bug that's even buggier than the PlayStation's version: bonus card duplication.
PSX can pull World from the 6 questions and again from the bonus card. Odds are rare but seemingly more common than the 1 in 62 from World not being removed from the bonus card pool. Happened about once per Fireseal rage quit session. World isn't meta in speedrunning before the last two stages but does give about the highest Opinion Leader stat boosts to STR & INT and ALI & LUK reduction.
Buggy Saturn is where it's at! I have video proof of double Chariot, Hermit, Justice, Death, Devil, Tower, Star, Judgment and World and didn't record Empress. Looks like no cards are removed from the pool when you draw the bonus. Possible exception for Magician sitting first in memory. Odds would be 6/22 (27%) for a dupe event and it does occur around that often in reality.
I know just what you're thinking: Devil run starting with 2x Devil and Joker-ing a third on the way to give Mizal the thematic death she deserves for being the greatest threat in the Zeteginian Empire.
Saturn's RNG appears as weak as SFC/SNES' in that confirming Male/Female locks in all 7 starting cards. Krayzar demonstrated this on emulator and now we have console proof.
Yeah after running on both SNES and Saturn, this is definitely true. I NEVER got a duplicate on SNES. On the Saturn, I would get multiple duplicates a night. I also found that I was getting the exact same cards on resets of the Saturn way more often than the SNES with no intentional RNG manipulation.
Was just discussing this on stream last night - I've observed this behavior as well, particularly when I was on the Fireseal grind, but I haven't been able to reliably reproduce it on demand. This could be huge for improving the consistency of runs on Saturn. In your observations so far NewSchoolBoxer, does it seem to be timing-dependent, or is there some other factor(s) you've noticed?
It is definitely timing dependent. But given that Saturn, like Playstation is a disc based system, you'd need to time after loads for it to be reliable. Disc drives are impressively non-deterministic. I used sound cues for my Fireseal runs on Saturn to try to do this since they are reliably paced.
There's some evidence in the SNES code (supposedly) that the card redraw effects that happen in Jokers and Warren's final card are not actually supposed to run. If you've ever run SNES and got a pattern you were looking for but all of the sudden the last card isn't what it should be, you've seen a card redraw. That happens when your leader's luck is modified enough by the starting cards that it triggers an incredibly small chance to redraw that card based on an algorithm that selects certain "bad" cards to be redrawn or certain "good" cards to be redrawn based on the Leader's luck value + or - 50.
I can tell you it does happen, but due to the inconsistencies on hardware compared to emulator on SNES, it's difficult for me to pin down exactly how the redraw works or if it can be manipulated based on any specific thing the player does. On SNES, that algorithm is (supposedly) functionally the same as the main one due to an oversight and thus keeps track of the last drawn card, but it may not have been intended to. So the behavior we see on Saturn and sometimes Playstation may have been intended as it was meant to be truly random.
I keep saying supposedly because this information comes 3rd hand from translated ancient Japanese webrings that pulled their info from a decompilation project in 2007 that is basically vaporware now. I can't even find the mods that that were supposedly produced with it after days of fishing on Internet Archive.
I think I meant to post in the RNG Research but I had them both open. My bad. I'm with Sheex that I noticed the repeat of the starting 7 cards on Saturn the first day I tested. That's way beyond how it goes on SNES and suggests a weaker RNG. I don't have an explanation of why ~78 frame apart produced the same cards 7 for 7 on Saturn. We know it's a bugged system that duplicate cards are common but impossible on SNES and only possible for World on PSX.
I haven't wrapped my head around Saturn but here's my SNES RNG understanding:
I looked at Dragondarch's spreadsheet in the Getting Started thread. Card List (Alpha) has what must be the RNG variable that loops 0-255 one tick per frame so a ~4.26 second cycle. All 256 seeds are laid out for a name of "!". Or I assume it is considering the speedruns. Also suggests the seeds were from emulator and more or less verified on real console. The exact timing might not the same but close enough to manip.
I have cart and flash cart but used emulator for better comparison. I waited extra long and named "!". First time, matched seed #150 except the 6th card was Judgment instead of Moon. Second time, matched seed #88 except bonus card was Devil instead of Temperance.
Third time I waited about the same time as the manip intended and got seed #23 exactly 7 cards for 7. I think the 0-255 landing is the most important factor and the total frame count or 256 cycles is a small factor. What about the name then?
I don't have TAS patience so I hooked up my Hori Super Famicom Fighting Commander. I clocked the turbo to 30 fps, which is the max possible if you think about: on-off-on-off. Middle settling spams the botton, top setting only when I hold the button. I used middle setting on A with no save file so defaults to New Game. Opinion Leader gets named eight 0's in a row.
Emulator isn't so smooth to use since the max USB polling is 1 kHz. Once per 1ms doesn't sync with once per frame so can register a button 1 frame late every now and then. Given ~60 frames per second, this happens. Anyway, about 8 Tarot cards sets are possible, and in every case, 1 card is different: either the 6th or 7th bonus card. I explain by being one or two 0-256 cycles ahead of the spreadsheet. What's interesting are the seed numbers: 124, 125, 247, 248, 249, 255 for the first six I recorded. Some repeats.
I think the name is strictly an offset to the starting seed. We know the hex values of each character. Not hard to prove that right or wrong.
Oh and none of the Saturn starting cards I looked up matched the SNES spreadsheet even 2 cards in a row.
Just something to add, by my manual count, SNES emulation via bizhawk is 3 frames faster than a real SNES for OB. That difference came up in 20 or so manual timings, so I think it's pretty accurate.
So, there is one thing I want to add to this research.
When I was mapping out all those cards way back when, there was an odd variable I found that loved to change the final card: manually dismissing the Quest logo.
I have no clue why this affects anything, but my research was fairly inconsistent until I just stopped dismissing it altogether, and let it dismiss itself after however long it takes for it to do so.
@Dragondarch - it's definitely an intentional quirk as there's versions of the game where it noticeably doesn't happen. Like on Saturn, where the intro is a movie and the timers can't run during playback, or PSX which has a reset key press that doesn't trigger the opening logo and would make it impossible to trigger and therefore useless to have. Saturn and PSX seem to handle that by making sex the trigger, though you wouldn't notice anything on PSX unless you played with it under emulation due to it's psychotic methods of randomization.
@Sheex - is this on SNES? There seems to be a variance with both Sun and Moon - I've gotten patterns where my mapper scripts predicted Moon in emulation, but upon running it again, or comparing to hardware, it was suddenly sun. That tends to be around the 5th card in a lot of patterns. I think there were another two cards that did that but I don't remember.
Yeah it's on SNES. I haven't run any other system quite enough to say that is replicated anywhere else. I know the Saturn likes to duplicate the 7th card.
