Sonntag, 3. Juli 2022

Personal best update and scoring guide (part 2)

The boss fight is the main meat of the extra stage. The first pattern is already a pretty hefty milk with several viable strategies. The commonly used one is what I'm doing here as well. Spinning the boss while grazing the bullets following you. With teams, you use your youkai character to get to full youkai pretty quickly, but that is obviously not viable for solo Youmu as she gains no youkai meter when grazing. Instead the best way to get to full Youkai immediately is to shoot her for a second to generate as many time orbs are possible and move up to get into position while in focus. This will lose you 100-150 time orbs from grazing but there's nothing else you can do. Generally, this gives you up to 4000 graze and 3900 time if Mokou's movement RNG is good. For me it was passable here at a roughly 3900/3800 split. It's worth about 50m and pretty consistent so a good start to the stage if you want to score this extra. One important detail for solo Youmu is that before you finish the spell, you want to graze the last swab of bullets unfocused so you actually end up in full human for the first spell. If you do this well, you go from 50% to -30% immediately. If done badly, you will lose up to 10m on the first spell, so practicing the transition is recommended.
The alternative strat for this non spell is much more difficult, but obviously gains score , but only if done well. I haven't tested this on slower shots but at least with Youmu (as well as Ghost team and Magic team by extension), it's possible to not only graze the bullets following you but also the ones leading you - at least to an extent. By dashing forward unfocused and immediately returning to the lane following, you are able to get extra graze for the bullets leading you. This is very risky for two reasons, however. The first is, obviously, the fact that you can just ram either the bullets or Mokou and die. The second reason is that, with solo Youmu, it's easy to lose full youkai too much. If your gauge management is bad, you might end up with less than 3800 or 3900 time, even if your graze is higher than normal. You'd risk your life to lose score in that scenario. I haven't tested this to the extremes, so I don't know how much can be gained but by my estimation it's about 5m. Dashing forward isn't something you can do at any given time, as you have to make sure you can return; and you have to make sure Mokou isn't going to move as you're dashing back and forth, or else she might just walk into you.

The first spell is a pretty quick affair but not without its problems. I angle my option to her right wing and I'm sitting not before Mokou but below her left wing she spawns on each spell. The reason I do this is because she spawns 3 familiars on each wing that spin around. Doing this gains you a lot more time than it would if you just shot her normally (about 500 time). This, however, means that you gain time at the highest rate possible in the game while shooting unfocused. The result of that is that it's incredibly easy to make gauge management mistakes that cause you to rise above -30%, losing you score. You can only make very short taps, and they can't be in too rapid succession. 
I enter the spell at -27%, so my graze/gauge management on the first non spell wasn't perfect, but fairly good  Unfortunately I was forced to make a lot of short taps in quick succession so I ended up going all the way to -11% on the gauge, which is basically unrecoverable. Best choice here is to just kill the spell as quickly as possible. Fortunately this happened reasonably late so I only lost 2m-2.5m spell bonus, instead the worst case scenario of 10m. I gained about 2100 time on this spell, out of the possible 2350. Overall, a quick gauge management mistake cost me about 4m score. Pretty hefty.

Non spell 2 is another attack with several viable strategies. One is the quick kill. It's the standard way of dealing with it. With Youmu's unfocused time gain, it gives about 600-650 time. The second way would be to milk it for graze unfocused. It's honestly worth almost nothing so I wouldn't recommend it for any goal. You gain maybe 1m through that extra graze. The third and final way is spinning it much like the first non spell while in focus. It is by far the most profitable way of handling it, but also the  most dangerous. I chose the fastest way in this run as I recently changed my route to be a bit easier on myself. 
The spins give you over 1000 graze, and time with the way I did them before. This is a score gain of roughly 15m. I exited the spin at about 30 seconds left on the time to dodge the pattern unfocused in the corner while dealing as little damage to Mokou as possible. That way I would be able to finish the spell while gaining time from shooting, making it so you only have to outscore the initial invincibility time she has at the start of each pattern. It is, however, possible to extend the spin by another 15 seconds, and it worth a relatively large amount of score. You'd have to kill Mokou while dodging completely unfocused so you enter the second spell as you go full human again, though. So it's riskier in 2 ways, since dodging in the corner is easier than doing it directly underneath the boss.
The main problem for the spins is that they're not as clean as on the first non spell. You will have to exit and re-enter the spinning spot through the red lane that is in front of you right of the bat, and will be forced to do it at least once more, maybe twice. Mokou's movement RNG is important here and there are spots where, if she is near the top of the screen, can kill you unless she moves down. Sometimes she moves to the very top (max height). That's pretty much always bad. She can wall you in almost any scenario from here, so long as you have to go over her. It's actually complete ass when that happens.
For the scenario where she is near the top, on the second spin above her, you can choose to opt out of the spin so she doesn't wall you. This obviously loses graze, but it's better than dying. Once the pattern make another revolution, you can re-enter the spin but the loss can not be made up. I recommend doing this only if your score goal is over 3.26b.

The second spell, flying phoenix, is one Youmu wants to kill after the 2nd cycle. She starts out by shooting 2 phoenixes on their own, and then a wave of 5 in a fan shape. You need to damage Mokou quickly so you can cancel the pattern on the 2nd set of 5, as that will gain you the largest bullet cancel possible. Anything over 7m yellow cancel is good, but almost 8m is possible. The cancel has some RNG to it in that that Mokou's height matters. She will be in the middle of the screen since you misdirect her horizontally, but if she's low on the screen, less bullets spawn. She was high here, so I got lucky. One more thing that I don't actively aim for is that if you kill the spell as she starts her third cycle (that is, she shoots out a single phoenix), you gain a better cancel; but only if you kill her immediately as she shoots the phoenix. None of the previously spawned bullets will have left the screen yet and you gain the bullets of a 6th phoenix (only the glowshit bullets and the jellybeans she shoots). This is what can push the cancel to 7.9m yellow bonus. I actually got this cancel in this run. Basically perfect execution and RNG. This was pretty lucky because as I said, I don't actively try to kill it at that specific time since being slow means you lose cancel instead. Could lose a huge amount of yellow bonus if you end up missing bullets. Killing it early is only a couple hundred thousand, so it's the better concession.

The third non has no interesting milk. You may choose to milk it in human for graze but it again gains basically nothing and is not recommended. Killing it unfocused is honestly the only reasonable choice. Spinning it is not possible either outside of a quarter turn before she moves that gains about 90 graze. But shooting unfocused with Youmu simply gains more score as time is worth more than graze.

The third spell, flawless shrine, is a bit like the first spell in that shooting the option at an angle gives better time, but it's not as pronounced here because not only do you not shoot familiars at all given times, there's also less of them. I did mess up my option placement here a bit, but the loss was barely noticable. You should gain 2150 time with the angled shot; I gained 2100 so it's whatever. Spell bonus was okay too. Maybe a few hundred thousand lower than optimal but negligible. 

Since time immemorial, the fourth non spell is a pattern that is started in full human and you quickly transition into full youkai for graze. The reason for that is not because it's a hugely profitable pattern to graze (it's about 500 graze, and roughly 4-5m score), but because it's followed by a spell you graze. 
One thing I haven't mentioned anywhere about this pattern is that it's possible to gain more graze. This is not done by being more aggressive with height, forcing you to make tighter dodges, but instead is done by dashing. The way this pattern spawns is that Mokou shoots slow red amulets in a circle away from her, and slightly faster purple ones in a similar fashion. The pattern overlaps in such a way that it is possible to dash is a half circle in the dodging area of the spell. Basically you do what the TAS does but not near bullet spawn but near the bottom of the screen. I don't know how much graze can be gained through this as I've only done this 2-3 times for testing to see if it was possible. All I know is that this is incredibly risky and unbelievably stupid. I doubt it will find any application if full runs, but optimizing this for the sake of a sum of best for this stage is a possibility.

The fourth spell, xufu, is the first really difficult pattern of the boss fight (unless you're doing the second non spins). There are 3 ways to graze it, although 2 of them are stupid - for different reasons. The first stupid way is doing it above the line you want to graze. This is stupid because it's just a bad strat. It loses a huge amount of graze and was fixed by ASL in 2009. If you use AM's replay for reference for some reason, don't copy that. 
The normal way is the only non stupid way. It's what ASL introduced and what I still do to this day. A regular timeout that was should give at least 11,000 time. I use time here instead of graze consciously. It is incredibly easy to miss time because you momentarily drop below 30%. It's not very common that you stay below that threshold, however. At the very least, I've only seen it happen during freak occurrences where a lot of things went wrong. I gained 14,200 time in this run, which is decent for me. I can't realistically ask for much higher, even though more is definitely within the realms of possible. The way this works is that you start by following the forming pattern because the grazing line. What is important to know is that the talisman spawns are static and that the rice in the ring spawns are, too. The other rice is aimed at you. What you want to do is to misdirect the aimed rice in a way that it doesn't get in the way, and well as have the rings with a more closed circle be on cycles when you aren't on their outside, because you're have to make up that ground to not die. Sometimes you will  be able to graze their insides, and sometimes you won't be in contact with the rings at all. You want to touch the corners of the pattern with your swings, if at all possible (though that's risky). 
the way to start is by making a large swing left, position yourself slightly left of the middle and move back to the starting position, let the blue amulets pass, and then go back left again. This start serves for extra graze, but also sets the cycle for the red rice, both aimed and circle rice. From here on, the graze pattern stay the same. You move left (after the 2nd long left swing) as far as you dare, return to where you started this cycle and then move left again and return, but this time to the middle.. By now, the aimed rice will almost have reached you, and will pass you to your left. Now you move right until you hit the corner of the pattern (optimally), and position yourself at the starting position of the entire pattern (after you followed it when it formed) and make a swing to the left until the middle and return, but one amulet further to the left (as a ring will pass through the starting gap). You now move right to the corner of the pattern again and return. By now, the aimed rice on the left will have left the screen and you move to the very left again with a large swing as the blue amulets sort of open up a path for you. Optimally you graze the blue amulets below you as you move. The pattern repeats from here (left, left, right from the middle, left, right, large left). My personal best with this strategy is just over 15,000 time, but more is possible with better gauge control.
The second stupid strat for this spell gains you several hundred more time, but is also much much harder, inconsistent and also poses a lot more gauge problems. It's stupid because it's just so unbelievably hard. No doubt turns xufu in the most difficult attack of the stage, but it's a very large improvement, with 16,000 time being confirmed possible, and 17,000 time being likely possible. How much the strat can be stretched is unknown, and also how much time can be expected from doing it, as a practice PB helps calculate score potentials, but not score averages. This is something I have to look into once serious attempts for 3.3b are on the table. But for now, it's something that exists only as documentation. I believe learning it eventually is necessary, though. 

The fifth non spell is about the same as the fourth.You Graze is not because it's super profitable but because the following spell is grazed and dodging this is still humane. It gains about the same amount of score as the previous non spell, but does not have the option of risking your life with the unfocused dashes. The purple amulets are too fast, so the pathway you'd dash through closes too quickly. Or at the very least, I'd like to think so. I haven't actually tried, because as I said, it was already stupid on the previous non spell. It'd only become dumber here.

The fifth spell, Honest Man's Death (HMD for short), is what people often think is the hardest attack in the stage. And it's true to an extent. In the learning phase, HMD is very straightforward, but also very hard. But because the movement is so simplistic, consistency can be gained to some extent. This graze consists on 78 microtaps, each of which you have to do with relatively strict timing. How much graze you get is mostly dependent on your height - the variance is huge. At higher levels, the tap length also becomes an important factor. There are 3 possible tap lengths, if you will. I will call them short, middle and long. The middle length is optimal, always. The short tap may kill you if the laser is approaching as you do not move far enough to dodge the entirety of its hitbox in that case. Make sure to note that if you practice this. The long tap is always safe, but moves you away from the middle, which is where you gain the most graze. You start out slightly off center (to the left, as the first laser starts spawning on the right). You move back and forth until the laser spawns, and then make 2 taps in the direction of the laser. The only exception to this is the 4th laser that spawns. You only need to make 1 tap in its direction (you will be centered for this if you did it right). As the pattern speeds up, you will have 4 taps per laser cycle. 
Do note that being too high on the screen actually loses graze. In Imperishable Night, you do not gain graze from bullets the instant they spawn. The bullets have a refractory period, if you will. It takes a few frames before they become worth anything other than a potential death (as they can still kill you). The tentacles are so fast that being too high means they pass you before you gain any graze from them. In the middle of the pattern at the best height, you gain graze from all sides of the tentacles (as well as from the black shit you're dodging anyway). Once you leave the middle, you're losing at least 1 side of the rough diamond shape. If you really fuck up you'll lose 2, and the tentacles open up the further you're away from them so if you somehow find yourself to the far side of the pattern, you not only miss graze from the bullets spawning directly above you (because they have no grazebox yet), but only because not as many bullet enter your graze box from the opposing side. Make note of that. In this run, I got just over 6200, which is pretty bad, but fortunately for 3.2b, whether I get 6600 graze or 6000 is not very relevant. 
With Youmu, about 6800 seems to be the hard limit. If you cancel it perfectly, maybe 6850 graze is possible, but because there is a hard limit to the spell in such a way, unlike with other patterns in the stage, and because the movement is simplistic, HMD can not be considered the hardest pattern in the stage, despite what many people believe to be the case. With current routes, it's the 3rd hardest pattern after Imperishable Shooting and Ichijo. Subjectively speaking, the final non is also harder for me but that's just because I can't dodge.

The sixth non spell is not a pattern where you have to do much, besides surviving (unfocused). You could gain more time by shooting not Mokou but exclusively her wings; it should a roughly 200 time increase. This is very risky, however as it forces you to dodge more aggressively (since she follows you and you don't want her above you). That could either mean death or failing gauge control. My current stance is that it's not worth it 

Woo is a fairly easy pattern, but a little inconsistent. For want to go for a very specific strat that allows you to speedkill it while having 9 familiars on screen. The way to do it is to direct her to either side of the screen after the first wave (which is always in the middle) and then go through the lanes closest to the corresponding side of the screen (which means you go through the rightmost gap between the lanes on the right side of the screen). Problem is that if the first wave gives you terrible RNG it may block you from going through the lanes. Reading ahead to see whether the gap will be closed or not is the only way to get this consistently. After going through the lane you misdirect the waves she shoots to the left (if you started on the right) and once the cycle is over, she starts in the middle again. This time you misdirect the 2nd wave of the cycle into the right side of the screen as you do not want to go above Mokou this time. You misdirect to remove the bullets so you can comfortably kill her now. If you did this right, you should have about 41.5m SCB and a 9 familiar cancel for this spell. 

The seventh non is basically the same as the sixth non. The same possibilities and strats apply. 

Phoenix Tail is generally speaking not a very difficult spell to survive. However, with solo human characters, you will be shooting familiars more than Mokou herself. And with solo Youmu, you will optimally not shoot Mokou at all. This spell spawns a lot of familiars that you can shoot. In Youmu case, you will sit below on wing and angle your shot to shoot the other one. Normally this doesn't give you any extra time but just like on Iwakasa, the familiars that spawn around the wings are their own targets, so Youmu's options deals double damage and gains more time. The time gain is immense compared to just shooting below while staying below her. The reason this is not trivial is because first of all, this means you take longer to kill Mokou. You start dodging at 62~ seconds left on the timer and have to dodge for 25-30 seconds if you do it correctly. This is doubly annoying because of the second reason: The gauge. With the absolute assload of time you're gathering through shooting, failing gauge control is incredibly easy, so you can only make very short dodges. that means your dodging options are limited. Furthermore, Mokou shoots aimed mentos bullets at you. These are no threatening but they force you to move around. Dodging them by going below them (since they come at an angle) is dangerous because the glowshit she shoots hides behind the mentos so you can just get hit by a bullet you couldn't see. So I advise against going below unless necessary. Which means you have to go around the bullet above it. Doing that without messing up the gauge or option positioning is difficult. I hit Mokou a whole lot this run because my angling wasn't flat enough. 

The final non is impossible. I bomb it. I bombed it here too. Sometimes I bomb and then still die. I hate this pattern. It's not something that should be in an extra stage. The whole schtick with shooting her wings for extra time is possible here too  but actually doing it is probably something no one will ever bother to do. Every player would be happy to just live through this shit.

Fujiyama Volcano.is likely the easiest attack after HMD. There are several ways to approach the spell, but they all want to kill it as quickly as possible. Important to note is that because she spawns a huge amount of familiars, an arbitrary cancel can cost you your items. On solo Youmu, you can not have more than 20 familiars on the screen (that's 2 of the explosions) without losing your items. The safe way to do this spell is to cancel the pattern when she starts a new cycle (she cycles through shooting 1 explosion, then 3 and then 5 is rapid succession). 1 Explosion is always safe and it's impossible to mess up the cancel. The problem with that cancel is, however, that it's slower first of all. You lose more spell bonus by doing that. Secondly, if you're allowed 2 explosions, then doing only 1 is obviously also a time loss. The, to my knowledge, optimal way of cancelling this pattern is finishing it at the end of the 2nd cycle, during the 5 explosion phase. After the third explosion disappears, you're left not only with all the bullets from the explosions but also still have 2 explosions left over. Cancelling at that specific timing gives you the biggest bullet cancel, a better time cancel, and you have higher spell bonus since you didn't wait for 3-4 seconds doing nothing. I don't think I would recommend this for people starting out but definitely for higher level play. It's been a long time since I made a direct comparison but if I remember correctly, the score gain was about 7-8m overall. The yellow cancel alone is over 3m more score; not to mention the spell bonus and extra time. How much extra score you gain is obviously dependent on timing though. You will, however, never score lower with this strat. The bullet cancel will always be bigger.

Possessed by Phoenix is pretty straightforward. Shoot unfocused for phase 1, graze the arrowheads for phase 2 and then shoot unfocused again for the rest of the spell. Trying to optimize the time gain from shooting is very difficult. It's the kind of difficulty that you will not see unless you have played this yourself. There are three things that can go wrong for time gain (besides getting hit). The first one is that is exclusive to Youmu: You don't control your option well. Not controlling it well means you're shooting the same target as your main shot. Avoid this at all costs. If you're slowly moving down, then allow the option to trails in front of you (so it shoots the familiars behind the one your main shot is hitting). The second one is the transition between phase 1 and 2. You want to be at 30% exactly when the arrowheads start coming out. This is very difficult to time. You can not just start focusing at xx seconds, because the amount of time flowing into your character will never be exactly the same, considering that your option control plays part in it. The third problem is the reverse of the second. Going back to full human from full youkai. While it's hard to miss it, it's easy to do it too early and lose some graze because of that. These are mostly small things but on a survival that doesn't have a spell cap, and doesn't count down, missing 10 graze can already be felt in the resulting spell bonus. 

Hourai Doll is a really good final spell. Too bad it's not the actual last attack though. It starts out easy and piles up the pressure very nicely. Not in an overwhelming way, though. You always have full control. If you die, it's always your fault. Everything you want to graze is aimed at your hitbox upon spawn. That means the red and blue balls are aimed at you long before they start to move. Keeping that in mind is important if you want to optimize this. The yellow rice is random so that's what's going to kill you if you do get hit. There's two ways to start grazing this spell. It's a decision you make at the start: by going up or down you grazing the way the balls fly at you. I prefer going down but I don't think it makes a real difference. Visualizing the pattern in my head, it feels like going down should be better but it might just be my imagination; I haven't tested this, after all.
Incidentally, I specified that it's a really good final spell because it's only good in runs. It's really annoying to practice, because you have to go through the start of the pattern every time before you get to the part you actually need practice for. I don't really do any practice on it for that reason and it really showed in my grazing here. Normally you'd expect about 3000 graze from this attack, but I didn't even manage 2650. In my defense, this was the first run since 2016 that got here, but it's still a huge score loss that should be avoided in the future. 

The last spell of this extra is also the hardest pattern in the category (with my route). Runs in the past used a very rudimentary and quite safe route to graze the spell to some degree but not to point where it would be particularly risky. It used to be your average graze you do in all sorts of categories. IS is maybe slightly unique in the sense that it doesn't repeat itself in any way, but that's really it.
In 2016, sometimes in spring I believe, the pattern was revamped by st and coco, two incredibly skilled players for the lower difficulties of IN. They turned what used to be a relatively basic pattern into something actually really fun. But this fun came at the cost of consistency. They also caused IS to become an incredibly difficult pattern, second only to the dumb strat on xufu that no one does. But even then, it's workable. Both coco and st had spell captures ranging from 44m to 46m. They worked out the route for the spell, but didn't optimize it greatly. Event he magic team capture is only 45m (46 was scarlet team). After properly learning the route I realized this and took it upon myself to try and squeeze out more than they did. As is well known, I succeeded in raising the spell bonus for this spell to 50m with Youmu, and more is possible. I have ideas for this spell. But these route changes would make it harder, if they work out so I don't want to mess with my consistency on this just yet. Obviously, using my current route for this spell is unnecessary at most levels. If you're just learning the stage, getting a cap itself is good enough. You will always get some time from the graze. 
Trying to explain the way to move in this would be completely silly, so check out a replay if you wish to. Both ASL's and my replay from 2016 use the easier route. Watch my shitty TAS if you want to see the full route for this attack. Or my uploaded 50m capture. Both work. This run doesn't work, naturally, since I failed the attack. The failure cost more 66m-67m score. It was a truly tragic event that is actually quite likely to repeat itself, but maybe in a different fashion. My caprate on this attack is just over 25%, so on average I need to get 4 runs this far to actually finish with a capture. I had hoped to luck out on it (like I did in 2016 with my old PB, even if the route was easier then), but it wasn't meant to be a lucky first try. 

With my current route, a projected NMFS (with 1 bomb on the final non spell) should be somewhere between 3.24b to 3.26b. If the run is below average, 3.23b is also possible, but higher is unlikely. I am missing a fair number of strats that I should be using to gain a higher score. Even 3.26b is incredibly unlikely and requires a great run overall. But by definition, most runs will average out in the end. 

By no means is what I posted in these last two posts everything there is to the extra stage, and playing is sometimes the only way to actually find out how things work, but I hope that this is comprehensible enough to help get people started, should they find interest in the extra stage. If you are reading this, it means you most likely know either my twitter account or my discord, so feel free to dm if about any questions you may have about this category. I'll answer as best as I can.

Thanks for reading. That's all this time. Next update will be the 3.2b run. When it will be, I don't know, however.