To the two people maybe reading this: you're good guys, but you waste a lot of time checking dead websites.
So my last blog post was almost three years ago. It wasn't long after that I ended up quitting Touhou altogether. I indeed ended up never PBing again, which means that even to this day, my Imperishable Night extra PB is standing at 3.15b. It is, at the same time, for some reason still the world record.
In February of this year, I restarted playing with the only category I really know - Imperishable Night extra. There have been numerous improvements in routing since January 2016, and even more improvements since I restarted two months ago. As it stands right now, the theoretical maximum score for Youmu should lie somewhere between 3.31b and 3.32b. I don't doubt that this will go higher eventually. Maybe in some years someone will actually produce a run that scores that high - when grazing is more optimized, spell strategies are improved and stage routes are better than they are now.
However, this is not my goal. Right now, I'm simply looking to improve the world record by just over 40m, meaning I want to get a score that lies in the 3.2b range. This is definitely possible as I've improved on many things and the only place where I should notably lose score is Keine's second spell Ichijou as that was simply incredible in my PB. What I want to do with this blog entry is take everyone reading this through a potential run that would not die anywhere. I may also describe the reason for the improvement if it's worth mentioning.
I'll start by saying that the eventual score could be anywhere from just barely reaching 3.2b to as high as 3.24b, depending on how well things go.
We start the stage with some pretty boring stuff, the furballs, the frame perfect cancel I never hit and the big fairies. Afterwards, however, we have three of the hardest cancels in all of Imperishable Night. The thing is that in the extra, familiar killing isn't a huge issue compared to the main game - unless you're playing at a high level, or in my case try to. The two five fairy cancels and the ten fairy cancel all combine very precise damage control, timing and positioning. The ten fairy cancel in might just be the most complicated and precise cancel in all of Touhou that isn't just RNG based. I never counted frames on these cancels so I don't want to name some random frame numbers. Just know that the window for the best cancel is not only dependent on the right frames but also your positioning. If your position is bad, then their spawning bullets will have an awkward angle, which will translate into less time for your cancel. This reason for that is, despite the familiars coming in at the same angle, is that you have only a small window in which you can actually cancel those bullets with the slightly trailing familiars (they're shot at the same time but the bullets' spawn movement is much faster). That means that your positioning needs to be far away from the fairies that you want to kill for clearing the clusters, but you can't just sit in the corner. Because for all three cancels, you want to basically cancel out the remaining clusters with the second waves of familiars shot. Important to note that if you damage control was not on point here, you will break familiars and greatly reduce the potential time gain. With teams, this can obviously be mitigated as you can just bypass the familiars but doing the cancels in human form is optimal. If every cancel up to Keine was absolutely perfect, I estimate Youmu could gain anywhere from 15,100 to 15,300 time. This will basically never be reached but it should be noted. The world record gets 13,800 time entering the mid boss, so this is a big point of improvement. you can calculate about a million score gain per 100 time for this point. As the cancels are overall pretty difficult, I usually end up with less than 14,300.
Keine's first spell has actually not changed since 2016. I sometimes looked at it again to try and see if there was an improvement but I haven't been successful yet. Maybe I'll find something eventually but I don't think it would be a good idea anyway. It would just be connected with more precise positioning and repositioning. This extra stage is already tight enough as it is, if you ask me. This spell has basically no variance either so it should always give around the same amount of score.
Keine's second spell obviously did not change either, but this time, there is a lot more variance to it. On a good day, I can match what I gained in my PB (over 2700 graze) but on bad days, I might go as low as 2450 or so. I already mentioned that this is where I will take a score loss, but not that big. My aim for this spell is around 2600 graze, more is welcome of course. This amount of score loss amounts to about 3.5m per 100 graze. That means if I mess up and get only 2500 graze, it's over 7m score lost compared to the world record. It's a very precise spell at that level. The important thing to note on this pattern is that it slowly rotates on itself, but there is some RNG involved in the starting position of the rotation. Some starting positions allow for very easy grazing (the one in the world record is basically optimal) and some are very hard to graze. This graze can be separated into three stages, if you will. Because the pattern rotates on itself, the way you have to move also changes slightly. You start out going into the gap between 2 lanes and 1 lane. I move to the right so the 2 lanes are on my left. This lasts only a very short time because the most profitable movement is what comes right after and I jump into it earlier than is recommended. In this phase you're basically moving along the ball of bullets. Basically that's where the graze is - the difference between 2500 and 2700 graze is how much time you spend hugging that ball. You want to do this as long as possible. You'll have to move out of the way of the bubbles that Keine shoots though. The final phase is going between the 2 lanes sets. This phase is worth the least because you spend the most time away from the ball. You want to try and keep that one very short. It is possible to get over 2800 graze on this spell, but I don't think that 2900 is very realistic for human play. Maybe with perfect movement or something. Incidentally, Youmu has an advantage on this pattern because you need the time gained from the graze to be able to push you to full human for the item collection. As Youmu goes from -50 to 50, instead of -100 to 100, this is much easier with her. If you're new to the extra stage and play a team, capping the spell 2-3 seconds earlier might be useful for the collection. Alternatively, you can use the time from the spell capture to give you full human, but it will obviously lose some spell bonus (and time, of course) on the following pattern.
Now for the black sheep of this mid boss. God, what have I done to this. Let me start by saying that this is one of the biggest improvements I have in this run and it's also a very consistent improvement that doesn't have a big graze fluctuation on a normal capture. What my PB does is very old at this point, but also incredibly easy. In 2016, t1100 showed an improvement worth about 10m to the spell, but it ended up being much harder. Basically they stream the jellybean bullets higher on the screen and therefore graze both streams of bullets, rather than just one as it used to be the case. This means that you have to reposition yourself every wave at a certain vertical height and you only have very weak visual cues like the moving pale rice bullets. However, as they are moving and seeing how you are not always moving the same, you will sometimes not have the correct height and die while streaming, which is why I consider the visual cue to be weak. At some point you develop an intuition of sorts as to what will work, but it's never truly consistent for that. Gut feeling can take you only so far. However, this was not all there was to the attack. It was in the end possible to improve it a bit more, which is what I did. In the section between positioning yourself for the streaming part of the spell, when Keine spawns the pale rice, I changed the route a little bit. Rather than just trying to almost touch all the small pillars, I try to sniff Keine's skirt. The area where I sit for just a little bit is quite small and disappears quickly. It is truly an improvement that fits into this extra stage thematically. I find that scoring in this extra can be summarized as very precise and this improvement is part of that now. It's become a very difficult spell and currently I'd put it on the same level of difficulty as Ichijou. It is probably harder to get a full timeout of this spell now than it is for Xu-Fu. It is not as difficult as Honest Man's Death or Imperishable Shooting though. My consistency on this may be worse right now but I'm much more practiced on the other extra spells. This spell now gains around 15m more score compared to the world record. If you have not seen the new spell route yet, I uploaded a video of it here.
What follow is the second half of the stage portion, which is easier than the first half on a technical level. The only actually dangerous section here are the purple jelly bean fairies. With Youmu it is possible to collect a lot of items without using her bomb. In fact, Youmu has to since her bomb is pretty bad for this section. You can match the item gain from the bomb just by doing it without it. That means that sometimes you will die if you mess up the movement on this section. But normally this shouldn't be a problem once you have learned how to move. For those interested, hugging the fairies for the pale jelly beans is actually very easy. I will fail it once out of 500 times. That kind of easy. The part after is another five fairy cancel but these fairies have less health than the ones in the first half of the stage portion. Basically nicking them is all you need to do to set up your cancel. Be careful though, if you miss one of the fairies with items, you may as well reset your run. Sometimes the one second from the right will run away, but it fortunately doesn't carry any items. Nicking that one's health isn't easy, and it's more important to concentrate on the whole cancel, rather than that one fairy. The death fairy is static and always gives 2200 time. A good second half for the stage portion will yield about 11,000 time and 343+ items. Do remember that every 2 extra items increase every following time orb by 10 score. This may sound completely insignificant, but then also remember that I gain about 80,000 time orbs on Mokou. Together with the score from the item (about 900,000 in the second stage half) and the end game bonus (items * 50,000 * 2) makes every extra item worth almost 1.5m extra score on average, only actually taking effect with every 2nd extra item though.
Now comes the boss fight. The opener is well known for spinning. A good opener will gain 4000 graze, but Youmu will not get all that in time because she starts at -50% gauge and first has to come to full youkai, which takes a bit longer than with a team or a solo youkai. Usually you can consider 3800 time to be successful. However, what is theoretically possible is to graze the wave you're usually grazing as well as grazing some of the bullets in front of you by dashing forward unfocused and then going back immediately to continue with the regular lane. This is not super difficult to do per se, but the problem here lies in the gauge control. Getting 4300 or 4400 graze with the strategy is possible and quite doable. But getting 4100 or 4200 time (minus 200 for the transition) is very difficult. So difficult in fact that I don't even consider it as a potential improvement. If I got a fantastic opener with this strat and gained 4400 graze with 4200 time gain, it would be an improvement of over 4m, maybe 5m. But more likely is that I'll gain that extra 400 graze at less than full youkai, so it's only like 1m extra score. A very tiny improvement considering you're risking your life with Youmu's unwieldy unfocused speed. Do note that this isn't trivial to do thanks to Mokou's hitbox. It's a vertically oriented rectangle that extends past her sprite. It's so large that it can kill you even in normal scoreplay. Always beware of fat bosses.
Iwakasa is a very uneventful spell that a player with a high powered shot should have a capture rate of over 90% on. I angle my option to hit the familiars on the right for a time gain. If I remember correctly, it was actually 200 to 300 extra time compared to just shooting her as the world record does. Considering that the spell lasts insignificantly longer this way, this is basically a free improvement of about 1.5m-2m.
The second non spell used to be an easy and free pattern but shortly after the current world record was set, it was suggested to me by Arf and Chirpy that I try spinning spinning it. At first I thought it was a stupid idea but as I fiddled around with the starting position and the pattern, I realized that this is actually a huge improvement. The spinning start above her and goes anti clockwise. The problem with this pattern compared to the opener is that if Mokou gives you bad movement RNG (vertically), then you can not continue spinning and are forced to quit. It is possible to reenter but I consider it to be a bit too risky to be worth. The good thing about this movement RNG is is that it is predictable when it is too dangerous to continue as Mokou will always move when you're below her. If she moves too high on the screen, with maximum height being certain death and close to maximum being RNG, you should stop. Another thing with this spin compared to the other one is that the lanes are messier and you have to actually move though a single lane of red amulets (exit and reenter) during the pattern. That might kill you. With Youmu, you need to be at full human going into the following spell or else you will lose significant spell bonus. So for Youmu, and most team shots as well I imagine, it is important to get back to full human on this non spell. What I do is that I spin the first 30 seconds of the non spell and take the rest of the time to go full human again. If RNG allows for this, I gain somewhere between 800-900 graze. With 800 graze, it's an improvement of just over 10m overall. And even if she doesn't allow for a full 30 seconds, it will still be worth more than doing what the world record does.
The route for Flying Phoenix hasn't changed since 2009 and I don't think it will for solo humans and teams. I tried a few strats where I grazed it, but the time gain was too low. I lost too much spellbonus and it was really hard too. It might be worth considering for solo youkai shots but even then, spellbonus loss might just be too much. For Youmu the cancel timing is important. You want to aim for a big bullet cancel. Fortunately with Youmu's power, this is quite easy to do quickly.
Third non spell is completely uneventful. Milking it for graze unfocused gives around 1m worth of score but I don't do it. It's too little for 40 seconds extra dodging, trivial as it is. I don't need to increase my chances of failing a pattern even more. Incidentally, you could start to spin this if you start on her left. Going anti-clockwise you gain around 90 graze. This is stupid and improves the run by literally less than half a million score. Pretty funny though.
Flawless Shrine is similar to Iwakasa in the sense that you should rarely fail it. Sometimes it will happen, but it shouldn't. I also angle my option here for ~100 extra time compared to my PB. It is not a big improvement but it's worth doing. Unlike Iwakasa though, this pattern lasts a bit longer because of this so if you're uncomfortable with this pattern for some reason, you might opt out of doing this.
The fourth non spell is what really starts the run. It is not a very difficult pattern to dodge and it has always been milked. Unlike on the second non spell, staying full youkai here is the good play, as obviously Xu-Fu comes after it. There isn't much to it other than not dying to bad movement, which occasionally happens.
Xu-Fu is a difficult pattern in some ways and a fairly easy one in others. The easy part is that once you understand how you have to move, the capture rate is going to be rather consistent for such a long pattern (a whopping 90 second for some reason, all other spells until the survival have only 70-78). The amulets are static and the red rice is not going to be in your way with correct movement. Which means that gaining 11,000 to 12,000 time is going to be very consistent even at the start. The hard part of the spell is the optimization. It is possible to gain over 15,000 time on this. That's up to 1000 graze variance compared to someone who just learned it properly. The highest time gain in any properly uploaded run I've seen is ASL's Youmu run at 13,600 time. The world record gains 13,000. I speak in time on this pattern on purpose because the cancel for the spell has huge variance to it. Sometimes you will gain 1200 or 1300 time, and sometimes close to 2000. Quite frankly, I haven't looked into the cancel enough to be able to tell when the optimal cancel timing is. Just know that it is significant. However, as the world record has a rather poor Xu-Fu, this spell should by all means be an improvement points. In practice, I tend to be somewhere between 14,000 and 14,300 time for this spell. At the lowest that means (with 1000 time) a gain of over 9m, and over 10m improvements are definitely going to be on the table here. But how does this huge difference come from? It's obviously in the movement - when you move towards the edge of the screen, hitting the corner of the swastika gives a fair bit of extra graze. On top of that, you need to make sure to hit the red rice rings, and whenever you're moving towards the left on the pattern, make sure you graze the dropping blue amulets as they open up the way for you. These are all small factors that individually don't lose you that much score but it accumulates during this long spell. The first thing to look for if you want to optimize this spell is the left movement with the blue amulets, because those are the biggest of the three. The edge of the pattern is next. Overall, this is the second hardest spell in the extra stage from a pure optimization standpoint, because it's not always clear where you lost graze.
The fifth non spell is much like the fourth non spell, except harder. I've died to this a lot in recent times. So much in fact that I considered not doing it, but at the same time, it is unfortunate but a full milk on this is 5m score. I can't really pass up on it. My score goal doesn't allow for that much leeway if the run is overall below average.
Honest Man's Death (HMD for short from here) is infamous for its difficulty and is probably the #1 thing keeping people away from this extra stage. It is stupid looking and is stupid to play too. It is, however the thing that gives the most score in all of IN. Doing this pattern is a whopping 250m score (spell bonus included). Obviously the gain is influenced by your height, up to a certain point. This can be clearly illustrated with three numbers: ASL's 5300 graze, the world record's 6000 graze and my practice PB gaining 6700 graze. There is an incredibly large difference between the amount of graze you can gain from this. What most people might not realize however, is that most of the score does not come from the blue tentacles, but the small black bullets aimed at you. Optimally, you should be grazing all of them (the TAS does do this and gains over 8500 graze). This also proves the tentacles aren't as important because when you are that high, you gain no graze from them at all. This is obviously not human viable but the principle still applies for the most part. In current human play, you will try to graze 5 lines at of them at all times, with the tentacles brushing against you too, effectively. At the height I'm doing (with Youmu's head stuck in the phoenix' tail), inaccurate taps lose graze from the tentacles (geez, so they do matter after all!) but also from the black bullets as you will only be grazing four lanes for a short while. This, again, adds up even if individually not big. At this point in time, I will stick to my current height without trying to increase the lanes I'm grazing - but this is a potential point of improvement if 3.3b should be reached at some point. I think by going up more, you may end up being able to graze 6 or 7 lanes at once. Would have to test if those are worth more than the tentacles overall though, as you'd basically get nothing from them anymore then. Furthermore, you would have to make many more taps. Currently I move after every 5 bullets, but any higher, you'd have to move after every 4 bullets, which would be another reason why the spell would become more difficult. My typical graze here falls between 6200 and 6500, which is anywhere from 6m to 15m. A potentially big increase. Incidentally, the cancel timing on this spell has a variance of ~300 time. But considering how fast the bullets are moving, and how difficult this spell is already, I don't think it's something to worry about for now.
The non spells following HMD are very random. It would be more optimal to milk them, though a cancel in full human is required for all of them. The 6th non spell is actually still very doable but can be quite scary. I'm not good at dodging so I only try to finish it off unfocused. Youmu's time gain is still very good, after all.
Woo used to be a spell that gave me a headache. I died to it a lot 6 years ago. But fortunately not anymore. You want to kill it as quickly as possible, which can be done with Youmu when she is shooting the many waves for a second time. Optimally, you want to kill as there are 9 familiars of the screen but 6 is also acceptable for me. Wouldn't be if the goal was higher but that is irrelevant for now.
7th non spell is the 6th but harder. The red amulets actually really reach you and can cause some chaotic dodging when you have to get away from purple amulets too. It's potentially a very dangerous pattern, but it doesn't last very long so that's fortunate.
Phoenix Tail is a big improvement over the world record for two reasons: The first one being that in my PB, I actually go above full human and stop gaining time from shooting, which obviously loses a lot of spell bonus and PiV. The other reason is that I again angle my shot. Optimally I basically don't hit Mokou directly at all, killing the spell by only shooting her familiars. My past calculations resulted in me gaining 6-7m score on this compared to my PB if I don't mess up. This is a huge increase. It does make last the spell twice as long though, so dying on this attack is a very realistic possibility, especially for me.
The final non spell is so hard to me that I'm probably going to bomb it. A bomb here is 11-12m score lost, but my skill level doesn't really allow for the risk of capturing it. It's akin to some guy who can't 1cc EoSD trying to force a book capture for his first 1cc. Complete nonsense. I'll make the decision whether to bomb it on the fly in every case but I assume most of the time I'll end up using one. It's just too hard.
Fujiyama Vulcano is the easiest attack in the second half of this boss fight, and in fact the only one I would call easy since Flawless Shrine. Important to note is, however, that there's a big score increase in this pattern too. I've talked about my cancel like a hundred times already, but let's do it once more: What is commonly seen is people killing the spell when Mokou shoots a single explosion. What is much better, however, is killing the spell when she's shooting her multiple explosions just before. It is important to know that you mustn't kill it with more than 2 explosions on screen (10 familiars each) because otherwise you will lose you point items. This stupid cancel has huge variance to it, obviously. Because there are very many fast moving bullets onscreen when you cancel, the bullet cancel can have variance on over 4m. However, even at its lowest, it still gains score. Not to mention that you also gain more time. A more or less perfect cancel with the common strategy ends the spell with 100-200 time less than a good cancel with the dumb strategy. I call it dumb because the optimal window is like 2-3 frames thanks to the disappearing explosion bullets and the arrowheads, which leave the screen very quickly. Not to mention that if you're early with the cancel, you lose 45 items, each worth 1.5m at this point. However, this cancel gains 6-7m score. You gain a much bigger bullet cancel, more time and since you finish the spell earlier, your spellbonus will be higher. You need to deal enough damage before you get to that point, though. If your damage output wasn't high enough, resorting to the common cancel (like the world record) is necessary.
I don't think the route for Possessed by Phoenix has ever changed. It still hasn't. It's still the same. A bit of care during the tapping and you can capture the spell. Be careful not to ram into anything while you're shooting the familiars for time.
Hourai Doll is a personal run killer. Or used to be. I haven't actually gotten there on a live run since 2016. The spell is somewhat precise if you want good numbers, but the variance is not nearly as high as on other spells. A normal capture will gain about 3000 graze, and my practice PB isn't even 3300, although 3300 is definitely possible. This isn't a spell where I'll gain any score, I don't think but I also shouldn't lose much if any. For those wondering: When dodging, your main focus will not be the red/blue balls or the arrowheads but the yellow bullets. Those are random so you have to pay attention to them, while the rest is easily sidestepped since it's all aimed. It's certainly nerve wrecking this late, though.
Imperishable Shooting (IS) is probably my favorite spell in all of Touhou. It's also the most difficult spell in Imperishable Night (main game included) if you want it to be. The route used in the world record is obsolete and very bad. The current route looks very different and uses the screen much better. It also stretches the gauge more, making it easy to drop out of full youkai. In 2015, 2016 it was well accepted that any practice PB over 40m was really good, and that 41m spell bonus was incredible, so getting 38m in a run was great. Then the route was completely revamped by st. st used team shots, so they had an easier time with gauge management (as they gain youkai gauge by grazing focused). His results ended in the area of 45m, his scarlet capture was 46m spell bonus. When I saw those replays I was really excited. I spend thousands of attempts polishing his route into what it is now. I adapted some movement to adjust for Youmu's faster unfocused speed, changed some minor things to allow for more graze and managed to raise the spell bonus to over 49m in practice with 50m definitely being possible. Explaining the route in detail is pointless and very hard to visualize because it will mostly be "I go in circles", which is why you can watch the replay of my current practice PB here. It is an immensely fun spell, even if it is difficult. I don't think st rerouted the spell for the routes to be used in full runs because it's so difficult but such thing will eventually happen to any pattern improvements. For full runs, my goal spell bonus is anything over 46m. That translates to around 1500 graze, I believe. The improvement is obvious, as it would be 8m in spell bonus alone. It's 1000 extra time, and each orb is worth 5300~ score (plus 2000 endgame bonus), so that's another 7m. The graze also gives around 1.5m, which means a 46m Imperishable Shooting would gain me 15-16m score compared to the WR. And more is always possible (as is less, of course).
Putting all that together, it is easy to see why the potential variance for my next PB (if it ever does happen) is so large. Some patterns could give me more than expected (like Xu-Fu being 14,700 time, rather than just 14,200), others could give me less (Imperishable Shooting being 44m spell bonus). I could decide to not bomb the final non spell and suddenly have an unplanned buffer in my score or I could lose score by not being able to spin the second non spell for the full 30 seconds. All sorts of things could happen, which is why I have a window of about 40m score. In a good run, the improvement compared to my current PB would be somewhere over 75m (resulting in over 3.23b), but realistic is at the very least the ~12m loss from using the bomb. To give a more accurate prediction, I should probably be somewhere between 3.21b and 3.22b with my final score.
However, the current problem is that I am not a good enough player to execute the route with any consistency. I will have to do a lot more practice and improve my skill level. Naturally, this includes a plethora of credits. My earliest estimation would put me somewhere in fall 2021, so in about 18 months. I don't expect to get a PB before that. This extra stage was already very difficult to begin with and I've changed some things to be multiple times as hard as they were in the world record (Keine third, HMD, IS, 2nd non spell, Xu-Fu) and I think it's possible that I will change even more things to make it even more difficult (like, say, I figure out how not to mess up the gauge on Mokou's opener). Arrogant as it may sound, a score of 3.2b in IN extra is so difficult that it would take even the best players serious dedication to get it (although they'd obviously not take 2 years of actually playing). I changed the route enough that I think I've gone past the point of stupid improvement but this is what I need to do not only to make 3.2b a reality but to also facilitate entry to 3.3b (which I think will be achievable with human play eventually). Because if I don't do it, no one will.
I've tried getting people to play this extra stage with memes, by being honest about it and by trying to point out the good things about it. I'd always be willing to offer help, too. At this point, I've given up on getting people to seriously attempt to push it. I've come to accept that it is unpopular and I can see why potential players would want to avoid this extra, as unfortunate as it is for me. I think better scores could be achieved, since other players think differently and would be able to find thing I didn't even think about. It is lonely all alone.
Thanks for reading. Sorry if it was boring.
Now comes the boss fight. The opener is well known for spinning. A good opener will gain 4000 graze, but Youmu will not get all that in time because she starts at -50% gauge and first has to come to full youkai, which takes a bit longer than with a team or a solo youkai. Usually you can consider 3800 time to be successful. However, what is theoretically possible is to graze the wave you're usually grazing as well as grazing some of the bullets in front of you by dashing forward unfocused and then going back immediately to continue with the regular lane. This is not super difficult to do per se, but the problem here lies in the gauge control. Getting 4300 or 4400 graze with the strategy is possible and quite doable. But getting 4100 or 4200 time (minus 200 for the transition) is very difficult. So difficult in fact that I don't even consider it as a potential improvement. If I got a fantastic opener with this strat and gained 4400 graze with 4200 time gain, it would be an improvement of over 4m, maybe 5m. But more likely is that I'll gain that extra 400 graze at less than full youkai, so it's only like 1m extra score. A very tiny improvement considering you're risking your life with Youmu's unwieldy unfocused speed. Do note that this isn't trivial to do thanks to Mokou's hitbox. It's a vertically oriented rectangle that extends past her sprite. It's so large that it can kill you even in normal scoreplay. Always beware of fat bosses.
Iwakasa is a very uneventful spell that a player with a high powered shot should have a capture rate of over 90% on. I angle my option to hit the familiars on the right for a time gain. If I remember correctly, it was actually 200 to 300 extra time compared to just shooting her as the world record does. Considering that the spell lasts insignificantly longer this way, this is basically a free improvement of about 1.5m-2m.
The second non spell used to be an easy and free pattern but shortly after the current world record was set, it was suggested to me by Arf and Chirpy that I try spinning spinning it. At first I thought it was a stupid idea but as I fiddled around with the starting position and the pattern, I realized that this is actually a huge improvement. The spinning start above her and goes anti clockwise. The problem with this pattern compared to the opener is that if Mokou gives you bad movement RNG (vertically), then you can not continue spinning and are forced to quit. It is possible to reenter but I consider it to be a bit too risky to be worth. The good thing about this movement RNG is is that it is predictable when it is too dangerous to continue as Mokou will always move when you're below her. If she moves too high on the screen, with maximum height being certain death and close to maximum being RNG, you should stop. Another thing with this spin compared to the other one is that the lanes are messier and you have to actually move though a single lane of red amulets (exit and reenter) during the pattern. That might kill you. With Youmu, you need to be at full human going into the following spell or else you will lose significant spell bonus. So for Youmu, and most team shots as well I imagine, it is important to get back to full human on this non spell. What I do is that I spin the first 30 seconds of the non spell and take the rest of the time to go full human again. If RNG allows for this, I gain somewhere between 800-900 graze. With 800 graze, it's an improvement of just over 10m overall. And even if she doesn't allow for a full 30 seconds, it will still be worth more than doing what the world record does.
The route for Flying Phoenix hasn't changed since 2009 and I don't think it will for solo humans and teams. I tried a few strats where I grazed it, but the time gain was too low. I lost too much spellbonus and it was really hard too. It might be worth considering for solo youkai shots but even then, spellbonus loss might just be too much. For Youmu the cancel timing is important. You want to aim for a big bullet cancel. Fortunately with Youmu's power, this is quite easy to do quickly.
Third non spell is completely uneventful. Milking it for graze unfocused gives around 1m worth of score but I don't do it. It's too little for 40 seconds extra dodging, trivial as it is. I don't need to increase my chances of failing a pattern even more. Incidentally, you could start to spin this if you start on her left. Going anti-clockwise you gain around 90 graze. This is stupid and improves the run by literally less than half a million score. Pretty funny though.
Flawless Shrine is similar to Iwakasa in the sense that you should rarely fail it. Sometimes it will happen, but it shouldn't. I also angle my option here for ~100 extra time compared to my PB. It is not a big improvement but it's worth doing. Unlike Iwakasa though, this pattern lasts a bit longer because of this so if you're uncomfortable with this pattern for some reason, you might opt out of doing this.
The fourth non spell is what really starts the run. It is not a very difficult pattern to dodge and it has always been milked. Unlike on the second non spell, staying full youkai here is the good play, as obviously Xu-Fu comes after it. There isn't much to it other than not dying to bad movement, which occasionally happens.
Xu-Fu is a difficult pattern in some ways and a fairly easy one in others. The easy part is that once you understand how you have to move, the capture rate is going to be rather consistent for such a long pattern (a whopping 90 second for some reason, all other spells until the survival have only 70-78). The amulets are static and the red rice is not going to be in your way with correct movement. Which means that gaining 11,000 to 12,000 time is going to be very consistent even at the start. The hard part of the spell is the optimization. It is possible to gain over 15,000 time on this. That's up to 1000 graze variance compared to someone who just learned it properly. The highest time gain in any properly uploaded run I've seen is ASL's Youmu run at 13,600 time. The world record gains 13,000. I speak in time on this pattern on purpose because the cancel for the spell has huge variance to it. Sometimes you will gain 1200 or 1300 time, and sometimes close to 2000. Quite frankly, I haven't looked into the cancel enough to be able to tell when the optimal cancel timing is. Just know that it is significant. However, as the world record has a rather poor Xu-Fu, this spell should by all means be an improvement points. In practice, I tend to be somewhere between 14,000 and 14,300 time for this spell. At the lowest that means (with 1000 time) a gain of over 9m, and over 10m improvements are definitely going to be on the table here. But how does this huge difference come from? It's obviously in the movement - when you move towards the edge of the screen, hitting the corner of the swastika gives a fair bit of extra graze. On top of that, you need to make sure to hit the red rice rings, and whenever you're moving towards the left on the pattern, make sure you graze the dropping blue amulets as they open up the way for you. These are all small factors that individually don't lose you that much score but it accumulates during this long spell. The first thing to look for if you want to optimize this spell is the left movement with the blue amulets, because those are the biggest of the three. The edge of the pattern is next. Overall, this is the second hardest spell in the extra stage from a pure optimization standpoint, because it's not always clear where you lost graze.
The fifth non spell is much like the fourth non spell, except harder. I've died to this a lot in recent times. So much in fact that I considered not doing it, but at the same time, it is unfortunate but a full milk on this is 5m score. I can't really pass up on it. My score goal doesn't allow for that much leeway if the run is overall below average.
Honest Man's Death (HMD for short from here) is infamous for its difficulty and is probably the #1 thing keeping people away from this extra stage. It is stupid looking and is stupid to play too. It is, however the thing that gives the most score in all of IN. Doing this pattern is a whopping 250m score (spell bonus included). Obviously the gain is influenced by your height, up to a certain point. This can be clearly illustrated with three numbers: ASL's 5300 graze, the world record's 6000 graze and my practice PB gaining 6700 graze. There is an incredibly large difference between the amount of graze you can gain from this. What most people might not realize however, is that most of the score does not come from the blue tentacles, but the small black bullets aimed at you. Optimally, you should be grazing all of them (the TAS does do this and gains over 8500 graze). This also proves the tentacles aren't as important because when you are that high, you gain no graze from them at all. This is obviously not human viable but the principle still applies for the most part. In current human play, you will try to graze 5 lines at of them at all times, with the tentacles brushing against you too, effectively. At the height I'm doing (with Youmu's head stuck in the phoenix' tail), inaccurate taps lose graze from the tentacles (geez, so they do matter after all!) but also from the black bullets as you will only be grazing four lanes for a short while. This, again, adds up even if individually not big. At this point in time, I will stick to my current height without trying to increase the lanes I'm grazing - but this is a potential point of improvement if 3.3b should be reached at some point. I think by going up more, you may end up being able to graze 6 or 7 lanes at once. Would have to test if those are worth more than the tentacles overall though, as you'd basically get nothing from them anymore then. Furthermore, you would have to make many more taps. Currently I move after every 5 bullets, but any higher, you'd have to move after every 4 bullets, which would be another reason why the spell would become more difficult. My typical graze here falls between 6200 and 6500, which is anywhere from 6m to 15m. A potentially big increase. Incidentally, the cancel timing on this spell has a variance of ~300 time. But considering how fast the bullets are moving, and how difficult this spell is already, I don't think it's something to worry about for now.
The non spells following HMD are very random. It would be more optimal to milk them, though a cancel in full human is required for all of them. The 6th non spell is actually still very doable but can be quite scary. I'm not good at dodging so I only try to finish it off unfocused. Youmu's time gain is still very good, after all.
Woo used to be a spell that gave me a headache. I died to it a lot 6 years ago. But fortunately not anymore. You want to kill it as quickly as possible, which can be done with Youmu when she is shooting the many waves for a second time. Optimally, you want to kill as there are 9 familiars of the screen but 6 is also acceptable for me. Wouldn't be if the goal was higher but that is irrelevant for now.
7th non spell is the 6th but harder. The red amulets actually really reach you and can cause some chaotic dodging when you have to get away from purple amulets too. It's potentially a very dangerous pattern, but it doesn't last very long so that's fortunate.
Phoenix Tail is a big improvement over the world record for two reasons: The first one being that in my PB, I actually go above full human and stop gaining time from shooting, which obviously loses a lot of spell bonus and PiV. The other reason is that I again angle my shot. Optimally I basically don't hit Mokou directly at all, killing the spell by only shooting her familiars. My past calculations resulted in me gaining 6-7m score on this compared to my PB if I don't mess up. This is a huge increase. It does make last the spell twice as long though, so dying on this attack is a very realistic possibility, especially for me.
The final non spell is so hard to me that I'm probably going to bomb it. A bomb here is 11-12m score lost, but my skill level doesn't really allow for the risk of capturing it. It's akin to some guy who can't 1cc EoSD trying to force a book capture for his first 1cc. Complete nonsense. I'll make the decision whether to bomb it on the fly in every case but I assume most of the time I'll end up using one. It's just too hard.
Fujiyama Vulcano is the easiest attack in the second half of this boss fight, and in fact the only one I would call easy since Flawless Shrine. Important to note is, however, that there's a big score increase in this pattern too. I've talked about my cancel like a hundred times already, but let's do it once more: What is commonly seen is people killing the spell when Mokou shoots a single explosion. What is much better, however, is killing the spell when she's shooting her multiple explosions just before. It is important to know that you mustn't kill it with more than 2 explosions on screen (10 familiars each) because otherwise you will lose you point items. This stupid cancel has huge variance to it, obviously. Because there are very many fast moving bullets onscreen when you cancel, the bullet cancel can have variance on over 4m. However, even at its lowest, it still gains score. Not to mention that you also gain more time. A more or less perfect cancel with the common strategy ends the spell with 100-200 time less than a good cancel with the dumb strategy. I call it dumb because the optimal window is like 2-3 frames thanks to the disappearing explosion bullets and the arrowheads, which leave the screen very quickly. Not to mention that if you're early with the cancel, you lose 45 items, each worth 1.5m at this point. However, this cancel gains 6-7m score. You gain a much bigger bullet cancel, more time and since you finish the spell earlier, your spellbonus will be higher. You need to deal enough damage before you get to that point, though. If your damage output wasn't high enough, resorting to the common cancel (like the world record) is necessary.
I don't think the route for Possessed by Phoenix has ever changed. It still hasn't. It's still the same. A bit of care during the tapping and you can capture the spell. Be careful not to ram into anything while you're shooting the familiars for time.
Hourai Doll is a personal run killer. Or used to be. I haven't actually gotten there on a live run since 2016. The spell is somewhat precise if you want good numbers, but the variance is not nearly as high as on other spells. A normal capture will gain about 3000 graze, and my practice PB isn't even 3300, although 3300 is definitely possible. This isn't a spell where I'll gain any score, I don't think but I also shouldn't lose much if any. For those wondering: When dodging, your main focus will not be the red/blue balls or the arrowheads but the yellow bullets. Those are random so you have to pay attention to them, while the rest is easily sidestepped since it's all aimed. It's certainly nerve wrecking this late, though.
Imperishable Shooting (IS) is probably my favorite spell in all of Touhou. It's also the most difficult spell in Imperishable Night (main game included) if you want it to be. The route used in the world record is obsolete and very bad. The current route looks very different and uses the screen much better. It also stretches the gauge more, making it easy to drop out of full youkai. In 2015, 2016 it was well accepted that any practice PB over 40m was really good, and that 41m spell bonus was incredible, so getting 38m in a run was great. Then the route was completely revamped by st. st used team shots, so they had an easier time with gauge management (as they gain youkai gauge by grazing focused). His results ended in the area of 45m, his scarlet capture was 46m spell bonus. When I saw those replays I was really excited. I spend thousands of attempts polishing his route into what it is now. I adapted some movement to adjust for Youmu's faster unfocused speed, changed some minor things to allow for more graze and managed to raise the spell bonus to over 49m in practice with 50m definitely being possible. Explaining the route in detail is pointless and very hard to visualize because it will mostly be "I go in circles", which is why you can watch the replay of my current practice PB here. It is an immensely fun spell, even if it is difficult. I don't think st rerouted the spell for the routes to be used in full runs because it's so difficult but such thing will eventually happen to any pattern improvements. For full runs, my goal spell bonus is anything over 46m. That translates to around 1500 graze, I believe. The improvement is obvious, as it would be 8m in spell bonus alone. It's 1000 extra time, and each orb is worth 5300~ score (plus 2000 endgame bonus), so that's another 7m. The graze also gives around 1.5m, which means a 46m Imperishable Shooting would gain me 15-16m score compared to the WR. And more is always possible (as is less, of course).
Putting all that together, it is easy to see why the potential variance for my next PB (if it ever does happen) is so large. Some patterns could give me more than expected (like Xu-Fu being 14,700 time, rather than just 14,200), others could give me less (Imperishable Shooting being 44m spell bonus). I could decide to not bomb the final non spell and suddenly have an unplanned buffer in my score or I could lose score by not being able to spin the second non spell for the full 30 seconds. All sorts of things could happen, which is why I have a window of about 40m score. In a good run, the improvement compared to my current PB would be somewhere over 75m (resulting in over 3.23b), but realistic is at the very least the ~12m loss from using the bomb. To give a more accurate prediction, I should probably be somewhere between 3.21b and 3.22b with my final score.
However, the current problem is that I am not a good enough player to execute the route with any consistency. I will have to do a lot more practice and improve my skill level. Naturally, this includes a plethora of credits. My earliest estimation would put me somewhere in fall 2021, so in about 18 months. I don't expect to get a PB before that. This extra stage was already very difficult to begin with and I've changed some things to be multiple times as hard as they were in the world record (Keine third, HMD, IS, 2nd non spell, Xu-Fu) and I think it's possible that I will change even more things to make it even more difficult (like, say, I figure out how not to mess up the gauge on Mokou's opener). Arrogant as it may sound, a score of 3.2b in IN extra is so difficult that it would take even the best players serious dedication to get it (although they'd obviously not take 2 years of actually playing). I changed the route enough that I think I've gone past the point of stupid improvement but this is what I need to do not only to make 3.2b a reality but to also facilitate entry to 3.3b (which I think will be achievable with human play eventually). Because if I don't do it, no one will.
I've tried getting people to play this extra stage with memes, by being honest about it and by trying to point out the good things about it. I'd always be willing to offer help, too. At this point, I've given up on getting people to seriously attempt to push it. I've come to accept that it is unpopular and I can see why potential players would want to avoid this extra, as unfortunate as it is for me. I think better scores could be achieved, since other players think differently and would be able to find thing I didn't even think about. It is lonely all alone.
Thanks for reading. Sorry if it was boring.
