It also made me wonder if a 100% Richter/Maria run on the Alucard scenario (meaning Richter/Maria beats every single boss, including Alucard's exclusive ones) is possible.
Yes such a run is possible, I might attempt one later.
Because they was very lazy, they made it so that the only difference is when you start new game with richter or maria, they have 5 time attacks unlocked directly.
-defeat minotaur, defeat death, defeat sucubus, save richter, meet maria (all set at 00:00:01)
if you set those to 0, you will do the bosses
when you start with alucard and bug switch to maria or richter, those events are still at zero (locked)
note that doing sucubus result in a soft lock, the cutscene doesnt proceed.
note also like in this run, if maria or richter try to "meet maria", the access to olrox will be locked as the cutscene cannot proceed like it should.
so per say, doing this will get you 2 extra boss, death and minotaur/werewolf
(and maria boss too, but she can be access just by doing the statue glitch, shes not part of the 5 unlock, they forgot it I think).
maria and richter will inherit alucard stats, so you better not turn on luck mode or you will have 25hp/1mana
if you have 75hp/30 mana, then you will be able to perform all of maria spells, except the invulnerability one
Zakatos wrote:
a lot of suboptimal choices of his own
Ok just to be clear, theres indeed a lot "choices" that are suboptimal, but the only one that matters is going by the reverse clock tower instead of chapel, the rest are worth only a couple of frames each(prologue, items, wolf)
Also, the run doesnt aim at been the fastest possible, but rather at completing the game without killing or hurting anything, by definition this is a slower behavior than just kill everything, and require extra steps, yet, ive tryed to be as fast as I could, while trying to be a little entertaining, it is more fun I think for exemple that I use richter + maria instead of just maria, or use some wolf instead of just backdashing in the early part.
This is more a playaround than purely a speedrun. But the action is at optimal speed within its parameters, or at least to my knowledge extent.
Like ive explained on the sub, Im considering those elements (vase/candles/switchs) as scenery and not related to monsters/ennemies, the run aim at lowest interaction with monsters, it is already a very big restriction that the psx run doesnt do, the difference been speed, there is multiple occassion where you could just bounce on monsters to go faster.
Yet your question is very valid and just to be clear, it took me a while to figure how to handle this.
It is totally possible to do a run where you interact with nothing, technically even if you dont destroy a candle you still "touch" it, so you could just fly with mist for 20+ minutes and touch nothing, theres nothing superplay or tas-worthy in this idea, anybody can do it.
So to answer your question, it wouldnt be hard to avoid breaking candles, but for speed strategy its going to mean to walk instead of using wingsmash, also for exemple in the second castle you will have to fly with mist for many entire rooms, because you dont want to touch or hit a monster, and without the wingsmash, it mean slow mist flying around.
It is most the time a case where you have to aim precisly the wingsmash to get where you want, so avoid candles is out of question, im only bouncing on a few candles but there again, it will only be slower to avoid it or heavly cripple your capacity to speed.
I believe excesive mist usage or walking wont be publishable material.
If people absolutly want to see the difference, I mean I can always try to make something but, im not sure its worth the efforts.
Edit: just to make sure, but around the 6min mark im hitting nothing (?), its mostly later in the run, some candles look like chandeliers, the only "vase" (not a candle) im hitting is the one with the garnet gem in outerwall. (Originally it was "3 hits" but due to how im excluding scenery from the run goals, well its only 2 hit on dracula)
You might as well just start your TAS in the save file where you last saved before Labs, because the only difference I see is how the save room gets glitched out.
This would defeat the purpose of the bug, the difference is you start a new game, and not just load an existing one (in that case you could just save at dracula)
It can be seen as a password, or by other definitions (cf the donkey kong run), but it is not just loading end game and redo the end fight, it use a "new game"
A richter or maria bare run will be faster regardless, for two main reason, the glitch itself takes time as it involve to save and soft reset (thats many thousands frames worth), it also take more time to complete alucard scenario, you get dracula (who has 10'000 hp).
Factoring the dialogues in and you get an inherent extra 2min30 to the run (thats what replay mode does skip)
If you try to start with richter instead of maria; you will be 32hp and 10 mana, the problem with this is that with only 10 mana you wouldnt be able to do the wingsmash or super jump like we doing in the rejected run.
A full level maria... so in maria or richter gameplay, you do get experience and levels, but they arent changing your stats (hp, mana, strenght ect...), the only thing that can happen is increase your HP with hpmaxup
But it is very likely that if maria or richter happen to have -1 luck like alucard does in this run, they will also deal 9999dmg, that is possible by simply loading alucard and do the glitch then start new game with maria or richter.
But because the soft reset take long, and increase loadings by a lot, it will still not beat a run that just go directly as maria, the bosses still die quick so its unlikely to be faster to have max dmg. (this is an assumption, I could be wrong, but we are within few seconds range at best)
Yesterday I was talking to paulo, and there is one thing ive overlooked while making the movie(s), its not important for them, but its important for the glitch capacity to alterate routing and make stuff that would be impossible otherwise.
You _have_ to use a save room, but you can glitch your position in the entrance further than the save room, for instance, you could start alucard, switch to maria, go into the marble gallery cdroom, reload alucard and boom, access to marble gallery, this is a scenario that wont make it in a tas here as it wont beat any speed record, it require more soft reset, but it allow for an insanly destructured run where alucard beat no monsters and no bosses, among the possible feats.
The botom line is that yes you can alterate your start position with alucard in the entrance, but you cant start near dracula, and we are already in the most optimized position for speedrunning, it does look like arbritary but it happen to be the right place (and the fastest method).
Among the few things that are important:
-the 3:29 run doesnt even use maria at all.
-the route alucard is taking was never used in any of the tas we have. (btw the entrance represent less than 2% of the map, the run finish with a bit over 10%, thats not 50% of the run but less than 20%)
-comparing the bug to selecting peach in smb ... I mean its a bug not a feature... you sayd "switch" to the princess... not bug the game to do it... your not suppose to be able to do that in the first place!!!
If it was just about that, it would be fine, but like you mentioned it unlocks everything, wich makes it the most interresting bug in sotn since ACE... I guess?
thats letting aside the fact that alucard gameplay is not comparable to maria, while mario vs peach... its not like peach transform into cakes and tea and mario doesnt.
Now its true the 3:29 run was posted a few days after the 2 others, but it was updated twice, and it was very noticable in the thread, yet no words on it... so ive resubmited it.
No words either on skiping dialogs or not, while it makes a 2:30 difference, but whatever, I suppose taking a totally different route and beating the time of both psx published run (with and without dialogs) doesnt matter after all.
It is also important to mention that rejecting something only on the premise that we have too many run is toxic.
I believe this is a prime exemple, maybe this isnt as popular as some other game, but there should be no question of if its worth more branches, what matters is if the run bring something new, something unexcepted, something interresting.
Does this run check those boxes? I think it does, and so it does also deserve to be published.
The labbeling doesnt matter, and wether or not it obsolete something else that is close or far to compare with shouldnt matter either.
I can understand the need to limit and organize things, but to be fair, there isnt a million game where you can do shenigans like this one (and some others), but on the vast majority you dont even have to ask yourself that question because there isnt material (bug...) enough to make those kind of run anyway.
Ive been told also that by giving as much info as I could it triggered the rejection "reason", the 5 unlocked events; this totally discourage people to actually show their finding, if you present less knowledge then what... publish because you dont know what your talking about... go team.
Im also over the part where "I should have update the psx run", because thats only a confusion, the premise of that come from the assumption that I could have done this on psx or that theres a major time saver discovered for it, wich is not the case, only small optimization can be made for psx.
You wanted to know under what branch this should be classified; but does it really matters? is it what it come down to? labeling... I think tas are more than just labels
But if you want my answer, I dont see whats insulting or problematic with "saturn alucard", or "saturn fastest alucard" or "saturn alucard bug" or "saturn save bug", or anything resembling that.
The day you will have psp tasing the same thing will happen all over again, sram nosram and all that stuff will also have to be taken into consideration, but essentially it will be the same equation than the one here.
3 runs were presented here, each having its own merit and well made (even if its mostly copy/paste it doesnt make them less good), not even picking one sound a bit exagerate, you had the possibility to pick one over the others for whatever consideration you could have to allow a new branch, and you decided to not even make a branch at all... this is quite extreme
Here a partial answer to whats happening
1. defeat minotaur and werewolf
2. defeat sucubus
3. defeat death
4. save richter
5. meet maria for the first time
You can see that the 5 top are unlocked at 00:01, so this is right when you start the new game as maria, then the 3 last are revelant to the speedrun itself, defeat hippogryf, defeat medusa, reach last save point
This is probably why theres nobody in the throne room, it is also probably why maria/richter does not get to fight sucubus/death/minotaur
But when you do the opposite, load maria into a (clean) alucard game, she will have to collect the relics, cant open the blue door, richter will be in the throne room but will do nothing, and you can fight death and sucubus....
Edit:
Ive reworked the sram run to include the 9999 dmg bug, with this you can defeat shaft in one hit and dracula in 2 hits, to make this work you need to have -1 (or less) luck stat, this is done here by having 2 duplicators and the blood cape (2x-10 and -1)
It is a good alternative than to have a lvl99 alucard, as it does the maximum dmg possible.
The verif movie (~15/20min) essientially beat dracula (get replay mode) just like the nosram, then get 2 duplicators + the knuckle (happen to crit easier than the rest), and unlock the revelant teleporters on the way as well
Shield is use to manipulate the crits, and the axe is use to make a second hit after the 9999 dmg
Link to video
I think this should really look like the closest to a perfect run you could have, assuming you can only start from entrance savepoint because of how the bug works, if you can load at last save point somehow it would be extremly faster.
To be fair/frank, I dont know the exact answer, im tempted to say you need a savefile to gather your stats/relics/map unlock from, but maybe you dont even need a savefile at all, just soft reset and start new game seem to be fine.
Theres only one thing im sure off, this doesnt corrupt the ram, everything is left untouched, by now we would have witness many side effects if it wasnt the case.
So this fall into the bug category and not memory corruption, maybe it can be considered as memory manipulation.
Maybe I should have mention that richter and maria does have all relics by default, they simply are turned off for most of them, and you dont have the ability to change them (theres no menu for them); when you start your new game with alucard you simply activate them (in menu), but they come directly from the maria/richter gameplay.
This simply mean all 3 characters share everything, but because in maria/richter you dont have access to the menu, you cant really figure that without ram watch.
To be more clear, it is only when you start a clean new game with alucard that the game will write "you have no relic unlocked" in ram, otherwise they are all just there.
Or if you load alucard game that doesnt have all relics unlocked yet.
Or if you die (assumption).
Whenever you new game or load richter/maria, they are all unlocked in ram.
For the map it doesnt matter who is gathering the rooms, it just carry over regardless what character you were using in the first place.
By extension, this mean you could have a maria run that have:
-maximum STR/INT stat to maximize her damage
-increased mana so that she can hadouken all the bosses
-wear alucard gear that boost stats beyond 99, like the rings, alternativly she could go for max luck to increase her critical hits
-use swords/shields (?)
-access to [USE] items (?)
(by default she wear nothing except the dragon helm, richter wears nothing at all)
This actually suggest that you dont even "need" a save room, and that it can work anywhere.
Link to video
So I mean really at this point, I can only say that it is not arbitrary in the sence we using what we know to work and is usefull, what wasnt tested may work yes, we dont know everything.
The capabilities of this glitch are streching far beyond our little speedrun route.
I would definitly invinte anybody who is interrested/motivated/curious to give a test at this bug.
Our beloved dev and coders can also look at debug and maybe share some light on this strange bug.
Im sorry I forgot to mention that info, it does only work in entrance, if you do it somewhere else, you will be locked in that glitched room, somewhere in the first castle (even if you saved in the inverted one), the only way out then is a library card.
Ok just because I could, I made a "test" run of what would be to take full advantage of sram, this is as close as newgame+ shenigans you can get.
Sorry for double posting but here we go
Link to video
So what happen is in the sram there is a clear game with alucard, all relics, unlocked teleporters, and mablung/alucard shield.
I just load this save, do the glitch (soft reset), start a new game and boom, everything is up.
From there you just casually take all the teleporters up to shaft/dracula and kill them almost instantly.
It can be optimized further with maybe a different shield? this was done very quick just to show case.
We the authors, have limited knowledge on this new bug, but its not a memory corrutpion, like ive mentioned in sub, the trick reside in the fact that the game doesnt clear the ram when you start a new game, it obviously does if you load a game tho (or you would be able to do some very bad things), this is just a dev oversight, they could have fix it by just disallowing soft reset when you load a game, that alone would prevent it even if the code still doesnt clear the ram.
Id dare to say this is 101 dev, clear the ram when you start new game or load the game, they simply forgot to lock the former.
And remember the richter "special mode", or the murasama stats, those are psx ram value that also carry over soft reset, so we can deduce they just plain sucks at reset the ram.
paulo did a video for this
Link to video
There is one thing important ive found out just yesterday while making tests with it, you actually do keep the map along stats/equipment, what this mean is you can with maria unlock the keep teleporter and with alucard just go down instead of up in the glitched room and directly access to keep like this, you can even go further with maria and unlock the teleporters in reverse keep and death wing lair, essentially making alucard run just about to use teleporters and do marble gallery.
You could also like moose mentioned get lvl99 stats, and better gear.
All this combined would make a newgame+ "concept" run that beat dracula under 2min. (and thats a very generous estimate)
And none of this would matter for the clean sram, or at least with how its presented right now; its obviously slower on the paper to unlock teleporters, it will increase your loading times by a lot.
The question is not wether we want to perform or like this idea, but only if this is cheating or not.
I do believe its cheating, but is it really? judge input on that would be nice
To what extent we can use this bug, where is the limit with cheating.
There is little difference between doing that and using a password to load last stage of a game, imo.
After saying all that, you may ask what are the bennefit of dirty sram in this movie context, if you declare all the previously mentioned techniques cheating, well the dirty sram presented here isnt, and the difference is just that it allow to skip the dialogues, it cost no frame and is totally equal to the clean sram, but its "faster" by definition.
On the other hand, if its allowed to make a "newgame+" nonsence with lvl99, all relics, all teleporters unlocked, AND the replay mode, then this sram movie is all wrong.
Here a run that does unlock replay mode (and use a similar strategy to what the real run would have), its gonna be necessary in order to skip shaft and dracula dialogues
Link to video
Just linking the youtube ive got this info from for reference:
Link to video
and the dude in comments redirect to this:
https://gamefaqs.gamespot.com/boards/196885-castlevania-symphony-of-the-night/77346267
(I havent read everything but I dont think theres more to dig, I could be wrong tho)
Edit: It is suggested this would work in reverse entrance as well, so maybe you have to get maria first in the second castle, shes much faster movement than alucard so it would make sence, the question is just does the extra ground to cover make it worth it.
It is also suggested that everything carry over the soft reset, including your level/stats; what that mean is you can have a premade alucard lvl99 and one shoot dracula/shaft
Ive got a proposal for others teams:
You voted for something unknow, how about we vote for something we know.
Im proposing to pick the saturn game, it isnt very long (30min) and there is a RTA
Link to video
I realize this isnt really in the rules, but should the others teams agree I dont see a real issue with it.
My main reason for proposing this is the current game is an atari clone of lone runner with absolutly nothing appealing in it, not even musics, maybe it make up for an excellent puzzle solving, but the saturn game is so much more pretty, it has accelerations, triple jumps and what seems like cool stuff overall for a tas, id dare saying it might even be entertaining, unlike those 21 lvl coin-picking 8bit graphics.
This is getting pretty positive voting, but not a lot of convincing thread feedback. Is this worthy of Moons? It's certainly an incredibly impressive run on a technical level, but does removing all of the enemies actually make it more entertaining of a watch?
Im not gonna answer about moons because thats something im not sure about.
I want to thank everybody who voted yes or contributed to the run, it is possible that ive had slightly more yes vote because of my "return"
But for the removal of ennemies, it sure does remove some action of the run and make it a little dull, but this tas is all about fighting the lag and trying to reduce it, so on that front I think its a success and it display something that cannot be achieve real time, it is noticable that removing monster make you go faster(nearly a minute!), im not sure its more entertaining but rather shorten the torture maybe.
I could go for another run that doesnt skip monsters, but im not sure it would find a good reception from audiance, and lets face it, the game is very much trash.
Im definitly going to rework the run in the future with newer version of bizhawk, and finish the grey run too (even if its not publishable), but im not sure its worth a second category, unless ppl get vocal about it, also, console verification would be really interresting.
http://tasvideos.org/userfiles/info/65478467165677605Link to video
The run is optimized up to the end of autoscroller, it is over 1k frames faster than the sub, stage2 boss lost some 30 frames but I havent searched much on how the spawn works. It is almost certain the run can be done under 16min.
At this point im not sure if its worth to work further on grey, that just depend if grey and color can co-exist in the vault.
I'll finish this run anyway, but maybe my time would be better spent on color for now.
Notice there are slight emulation difference with 2.3.2, its sublte tho.
This would be the understatement of the year.
Within 5 minutes he basicly figured the candles are working the same than monsters.
There is still investigation to be done but this alone is just amazing.
Here a fool-around proof of concept:
Link to video
One another important bit about the lag mystery is that apparently, when you pause, the cpu load drop to 80% .... hence why the lag sequence after a pause is quite different than if you dont pause...
You can have a look at my latest stage1, I dont believe im gonna revisit it soon (but feel free to try to improve it), at least for the grey im please with this version and will keep going like this, while still using the sub as base of research of course.
It is about 2 seconds faster than the sub.
Link to videohttp://tasvideos.org/userfiles/info/65389261794911803
this input file use bizhawk 2.4.2, gambatte with gbc syncs, theres an inherent 13 frames slower at loading due to emulation diff.
Few things:
-theres no frames rules in the game after all, any improvement is an improvement, but instead theres lag rules, that mean lag comes in a sequence, that is altered by pressing inputs, stuff drawn ect... like we already know, but whats also important is pressing pause does alter the sequence a lot, and that can increase the lag by itself, it was clever to hit candles because this does effectivly mitigate the lag, with this method without pause, you stop the scrolling so automaticly you are slower, if you try to hit the candle, you will slow down even more, yet in the end, it is faster than pausing, provided that you dont hit candles. (even if it makes no sence, thats whats happening here)
-a lower lag count doesnt always beat a frame count
-theres nothing that carry over screens, but the lag sequence from the previous room can make the loading faster in the next one, it is trivial to test with ropes, for exemple just jumping a frame too late can lead to a faster loading of next room.
-the invulnerability make you move faster, supposly because the sprite is half the time not drawn on screen, this is the only item that make it worth to hit a candle, if theres more on the way in other stage its probly best to remind me, it took me many tries to figure what candle was worth it, actually none does, except that one (and ive tested litterally every of them before this one)
-rooms with no scrolling arent affected, a few monsters spawn must be removed with pause, like in room2 or last scroller before boss, they cannot be manipulated with just the scrolling.
A new test strategy for stage1
This is possible by not using pause to remove monsters, instead im simply moving backward for 1 frame at the right moment.
Note that hitting candles with this method is slower, regardless how you try to manage the lag, you also cannot do it while jumping (in case its not obvious already)
http://tasvideos.org/userfiles/info/65368766340652844Link to video
This still need more working, but the first room alone is 82 frames faster.
If taking damage happen to be viable, then probly should not upgrade the whip at all and maybe save a few more frames in first room, and use directly the eyeball in second room.
Its gonna take some iterations to verify if its gonna really work or not, and before reaching the boss its hard to predict, but otherwise I think nearly a second and a half in first room is the best I can figure for now. Not using pause will also save much time trough the run.
Havent tested on color yet but excepting it to work the same (or better).
Available for duty, at least in theory, if no shit happens.
If any teams wants a low knowledge player you can pick me!
Note: Ill be disapointed if theres no candles to destroy
Ok so to make sure about whats happening there.
Ive redid stage4 room1 of your movie.
By proper use of the pause trick and avoiding candles, ive saved 110 frames in that room. (here)
I then tryed to figure whats going on with that unholy candle.
This is the result ive got
So you might be right, there might be something about that candle, the time saved is as you can see, very minimal, its within 15-20 frames range.
For further reference, it costed 4 frames to hit the candle in the first room.
Its not like ive not warmed you on discord about everything you are complaining right now, I did multiple times.
If I was using hacks, it would be inside the movies too.
As arukAdo mentioned, it was a "test run", meaning that the purpose was just demonstrating a specific better strategy, and not attempting to make a new improved submssion.
Well unfortunatly it is my intention now, provided that he will not run the gbc version, and im not sure if hes whilling to fix the thousand(s?) of frames remaining to be saved in the grey run.
The only reason ive submited nothing is because I wanted to investigate things, you can also see now how controversial it would have been, without verifying that the color version doesnt have this candle nonsence for exemple, but im sure we are not yet at the end of the tunnel.
MOD EDIT: fixed and resized image link! <3 samsara
Ive followed instructions (stage4) and the candle is still there, it makes no difference. (except I lost frames...)
The corridor in question has almost no lag past the first ennemy, btw. (I think its 30'ish lag)
The only lag present is from the monsters trying to spawn (theres still the logic running in ram)
The whole idea that a candle content is lagging the game is just wrong, im not gonna bother to explain in details how ive figured this (HINT RAM WATCH, only when you destroy a candle the game will draw its content...), you can always proove me wrong, in that case, you will also have to proove how its saving any frames (ie show 2 different movie that does showoff the frames saved)
Other than that, this is just a sterile talk. (I fail to understand what im suppose to change beyond those 3 candles, and how much im suppose to save overall, I know ive got limited knowledge on the game so if someone else could figure something im suppose to do, plz feel free to call me an idiot)
This is also a good reason to have ghost script, and to do the grey tas too, so that there is nothing left to debate or guess. (this is outside considerations for publishing/acceptation/rejection)
Edit: on the discord it was mentioned that the trick could be considered as a different category, because it remove all monsters, and pause so often (annoying kinda), I dont believe the game is worth 2 category, but should that happen, the old published run would not fullfill either criteria (it does pause few times...) and I would have to basicly redo a movie where you dont pause at all, thats fine for me as long it is clearly stated in the rejection message here (im not doing a 3rd movie if its uterly pointless, 2 is enough), im begging our overlords for a definitive decision on this matter.
Im not even asking about grey vs color version, because this is something that was asked, voted, rejected (!) and published already, so Im assuming/hoping it will stick.
We now know that the game has the "double speed" flag set in the code, so it is very obvious the reason the game lag less than grey is this. (thx whoever found that, sry ive forgot your name)
There is a jap cart of konami collection (older release date than EU), but in there its not a gameboy color (so it doesnt have the double speed flag) but instead its a super gameboy (so its no longer grey but red... with a border belmont art... its not awesome tho), so theres no point to use that I think (and hope).
Edit2: For the record ive voted meh, in a couple of minute while using ram watch ive managed to shave 70 frames from his first room, while I admire the work to figure without ram watch how to remove that many monsters, this is sub optimal at best, later on its about 600 frames saved in stage3 begin.
Here a test run with better spawn manipulation, just stage3 first block is already 10 second faster than any previous run, it is 2803 frames faster than published movie with 2320 less lag.
Link to videohttp://tasvideos.org/userfiles/info/65283066503537296
To improve further;
-ghost script (cannot be used between grey/color, unfortunatly)
-script to tell you on screen when a monster is spawning (check ram range)
-bot to verify every candles
-cleanup small mistakes
Note that you can improve jumping by making sure you dont land on a lag frame, thats something that isnt possible everywhere and pretty tedious to produce, also if theres too many lag after the landing it might not have any benefical effect.
Because I was curious to test things, ive made a little movie and it save 70 extra frames on the first room. (took me only 5min)
http://tasvideos.org/userfiles/info/65245845160899640
The difference is just proper use of the bug and avoiding to hit candles.
Originally, destroying candles was faster, because combined with monsters it creates extra lag, but because we basicly now remove all monsters, its slower to hit them, that is because jumping is just slightly slower than walking. (thats very easy to test)
Edit: This doesnt mean you shouldnt destroy any candles, but should avoid destroying those that are not giving you any advantage or removing lags. (it could be very tedious to verify each candles interaction/lag)
There is frame rules in this game so there is no way to exactly predict how much you gonna save trought the run.
1. defeat minotaur and werewolf
2. defeat sucubus
3. defeat death
4. save richter
5. meet maria for the first time
You can see that the 5 top are unlocked at 00:01, so this is right when you start the new game as maria, then the 3 last are revelant to the speedrun itself, defeat hippogryf, defeat medusa, reach last save point
This is probably why theres nobody in the throne room, it is also probably why maria/richter does not get to fight sucubus/death/minotaur
But when you do the opposite, load maria into a (clean) alucard game, she will have to collect the relics, cant open the blue door, richter will be in the throne room but will do nothing, and you can fight death and sucubus....
Edit:
Ive reworked the sram run to include the 9999 dmg bug, with this you can defeat shaft in one hit and dracula in 2 hits, to make this work you need to have -1 (or less) luck stat, this is done here by having 2 duplicators and the blood cape (2x-10 and -1)
It is a good alternative than to have a lvl99 alucard, as it does the maximum dmg possible.
The verif movie (~15/20min) essientially beat dracula (get replay mode) just like the nosram, then get 2 duplicators + the knuckle (happen to crit easier than the rest), and unlock the revelant teleporters on the way as well
Shield is use to manipulate the crits, and the axe is use to make a second hit after the 9999 dmg
Link to video
I think this should really look like the closest to a perfect run you could have, assuming you can only start from entrance savepoint because of how the bug works, if you can load at last save point somehow it would be extremly faster.
So you might be right, there might be something about that candle, the time saved is as you can see, very minimal, its within 15-20 frames range.
For further reference, it costed 4 frames to hit the candle in the first room.
Its not like ive not warmed you on discord about everything you are complaining right now, I did multiple times.
If I was using hacks, it would be inside the movies too.