So I have tried speedrunning 15 puzzle category recently, and almost immediately noticed, that I'm getting same boards over and over again Like this one: https://www.speedrun.com/final_fantasy_nes/runs/zq35erry I've got that same board in my run as well:

It seems like there are a limited amount of possible boards in the game, and there is a very little amount of them, have not really count them, but I'm getting same ones like every other game, I already improved my time on this by a few seconds (my first attempt was 38.5, now it's like 34, i'll upload soon)

So Is it known how this game rng works for the puzzle game? Can you just submit runs done at same boards? If no, what should I do then? Since I got the same scramble as the other person, it really feels like there are no random boards involved, so I basically have to use the same ones

I might try to write down all of the possible boards to prove/disprove my theory, should not really take a lot of time, as the amount seems very limited

Ok I did some testing.

https://docs.google.com/spreadsheets/d/1FY5qzaNUQeLjOWLsSQnc8dSE42sa-mxTfPW4KCh6PmU/edit?usp=sharing

Here are my results

I have scrambled puzzle 100 times, (2 times i closed it by accident before writing down)

So out of 98 boards only 45 were "unique", but many of those "unique" boards are not truly unique, I have tried to get some extra data by normalizing scrambles (solving them by just moving blank to the corner), getting around 35-37 unique scrambles after normalizing blank positions, since many scrambles are just different by 1-2 moves

But even after normalization, there are very few truly different states, many of them can be grouped together even by simple sorting, as they often start with same first row

7/46 start with 1 15 4 12

5/46 start with 1 15 8 9

4/46 start with 13 15 12 8

4/46 start with 5 15 4 12

4/46 start with 6 15 8 9

etc. with a bunch of smaller groups like that as well, some of which, as I already mentioned, are the same exact boards, different by 1-2 tile moves. Some of the differences in the scrambles are harder to notice, because the change is, for example, just one cycle of 3 elements, so it's still possible to combine scrambles into even larger groups

Now just for the fun, let's try to find each boards from the 46 unique I've found to the runs on the leaderboard, including obsolete, and mine not yet verified 38.5s one

(I'm going to use numbers from the 2nd list of the sheet, not from the raw data)

• mine is #26
• 00:47.470 by Shanezell - it is 1 short move away from #35
• 00:48.633 by deathbychipmunk - "unique", but 2 moves and a cycle away from #42 (3 2 10 cycle + right column)
• 00:48.933 by Bison_H - #26, as I already mentioned
• 00:56.923 by Leonis07 - #2, exactly that one
• 00:57.093 by Soloyourtrashkid - VOD unavailable
• 00:57.300 by cambark - 1 cycle away from #6 (4 6 15)
• 01:07.133 by deathbychipmunk - 1 short move away from #27
• 01:08.120 by Avin_Chaos - could not really find the match
• 01:09.600 by cambark - 1 short move away from #12 (or from #13, same thing)
• 01:22.494 by buffalax - 1 short move and a cycle away from #29 (maybe 2 cycles?)
• 01:25.010 by Avin_Chaos - ~2 moves and 2 cycles away from #41
• 01:43.297 by mammaru339 - exactly #45

So I could not find exact matches for most of the runs, but almost for all of them there is a very similar scrambles, I think I could have found more matches with a larger data collected than 100 attempts, but it is already very clear, that the amount of possible scrambles is very limited, and those scrambles, which are unique, tend to be split into several very similar groups, in which scrambles are different by 1-2 blank moves or some of the tiles are cycled, so it's hard to say, which is actually a unique one, and since it works this way for everyone, and not just me, it is just how the game works.

Some of the notes I want to point out based on that little research:

• It should be OK to do runs on same boards, because otherwise you would just run out of scrambles to speedrun on (although it might Not to be ok to do that using rng manipulation or save states)
• As a result of that, It is worth learning best possible solutions to scrambles beforehand, and reset until getting specific board / group of boards
• Some of the boards are objectively better than the others, you can filter out many of the possible scrambles I think, but finding truly best scramble might be a fun challenge, as you would also have to take obscure game controls style into account when choosing scramble, but I think it should not be very far from just a scramble with a shortest possible solution in wide-moves notation, but not exactly that, also, someone would have to actually find all possible boards, there might be around 100 of them I think, if we include all the 1-move variations etc., maybe even more, my best guess is something nice like "64" or "128"

P.S. I personally would have enjoyed it much much more if boards were truly random, learning solutions for specific scrambles, trying to find them etc., and waiting for them seems very not fun and not how I usually do sliding puzzles outside of this game

Also, I think this issue might be in other version of FF1, because after about 1000 solves done in 3DS version, I could not get very lucky scrambles, and my personal best seem to have a lot of moves, more than I would expect for a personal records, but they seem to be a lot more random than this NES version, at least I could not notice that while playing, and not getting same boards over and over again like in here.

After some very quick testing with an optimal solver, potentially good boards are:

• 1 15 4 12... (like the ones that close to 57.300 by cambark)
• 6 15 9...... (like 48.633 by deathbychipmunk)
• 15 4 8 0... (like mine or 48.933 by Bison_H)
• 5 15 4 12...

But that's really just a first test on the little amount of data, that I have, and I also used short-moves notation solver, not the long-moves ones, I should probably do that later and collect more data as well.

Bad news is that the shortest scrambles in short-moves notation are 44-movers, maybe there are better ones obviously, but there are probably no extremely lucky boards possible

When I was doing some puzzles not so long ago I also noticed quite a few of the puzzles being very similar. I would say that as long as no save states are used and you are just loading up a puzzle and finishing it asap...that's what matters. We all know that FF1 NES is limited in what it does with RNG.

@Shanezell I have a meme speedrun idea btw - Fastest time to unlock the 15 puzzle minigame / get to the ship, from the start of the game to the minigame activation

I have unintentionally did that like 10 times already, trying different ff1 versions and emulators at multiple platforms, just to speedrun 15 puzzle minigame, now I wonder how fast can it be done and if it has been done before.

I have like 0 knowledge of the game so I was basically doing this:

• go 4 warriors, magic feels completely useless in early game (or at least complicated to use for someone like me)
• buy everyone strongest weapons and armor, you have just enough money for that
• grind nearby monsters up to level 3-4, go sleep to heal from time to time
• go to the castle, and immediately fight the boss
• go heal once again, and proceed to the next city
• go sleep to heal,
• fight pirates, they are not strong at all,
• go to the ship, at last

update: well it is probably would be pretty much the same as any% first split

In the normal Speedrun we already rush Garland and the pirates, it can be done under 5 mins 30 seconds in a standard run. Might be and to cut off a bit with only trying for that in mind but I don't see it as a worthwhile category to create.