Valfaris: "Gazer Guard" boss analysis (phase 1)

Valfaris' "Gazer Guard" is a mid/late-game boss with a compact, elegant design. The first phase of the battle (see video) is composed of three sequential tests: reading, vision, and reaction.

NOTE: The video depicts a successful run of "Gazer Guard's" first phase, so some of the interactions described in the following paragraph aren't seen here. Valfaris expects the player to risk death during experimentation, so defeating this boss for the first time may take multiple restarts, refining and hastening the eventual successful run.

The majority of enemies in Valfaris take direct damage from weapon fire, but "Gazer Guard" does not, setting up a unique boss fight. When "Therion" fires upon the central eyeball which appears to be the boss's body, its blue shield aura* and static health bar indicate that damage is not being inflicted. This feedback is identical regardless of "Therion's" equipped weapon, ruling out direct attack and suggesting a more patient approach. While the eye remains invulnerable, six floating beacons dash around the screen, pause, and repeat in an ostensibly randomized pattern. "Therion" cannot interact with the beacons while they are moving; though they can be attacked during the final and longest pause. Attacking a beacon won't kill or damage it, rather it triggers one of two events. When a beacon is struck, the boss eye either fires a single aimed laser directly at "Therion", or it recoils and its health bar decreases by a fixed decrement. Afterwards, the beacons resume their pattern of shuffling and the eyeball remains invulnerable to direct attack.

The first test for the player is reading the scenario as laid out. Once they've experimented enough to determine that the beacons are vulnerable and not the eye**, they need to figure out why sometimes damaging a beacon hurts the boss and other times it triggers a counterattack. They may deduce that given six beacons, only one is a correct target, or they may simply continue observing and experimenting until they come to the same conclusion. This should draw their attention to the shuffle, at which point they need to observe that one and only one beacon glows blue at the beginning of each cycle, just before the shuffle. At this point, the player has passed the reading/observation test; their reward is the knowledge of which beacon to target.

Now the focus shifts to the shuffling pattern. The target beacon is indicated before the shuffle, but the vulnerability window comes after. The player needs to track the (now identical) target beacon as the pattern speedily and repeatedly reconfigures. It's a game of three-card monty, a fairly rudimentary and likely familiar test of vision. This is the second test; the reward for successfully tracking the beacon through the shuffle is again knowledge, this time knowledge of the target location when it becomes vulnerable.

The third test barely registers as such, it should be so ingrained into the player at this point. As the beacons finish shuffling and are fixed in place, they lose their purple glow, indicating a change in state that experimentation will have revealed to be their vulnerability window. If the player has passed the first two tests, they know now which beacon to target and where it is. If any other beacon is attacked at this point, the window ends, the eyeball attacks, and the cycle restarts. Striking the target beacon without damaging any of the others is the third test. This can be accomplished with a melee weapon by running and jumping to its location, or with a ranged weapon by aligning "Therion" with his target along one of his eight axes of aiming. This action needs to be accomplished in a short period of time (1-2 seconds), otherwise the cycle restarts. To pass the test the player needs to be able to quickly attack an arbitrarily placed element (the target beacon) while aiming around its cover (the other beacons). This time the reward is damage to "Gazer Guard" - irrevocable progress in the boss fight.

This ends the suspension of challenge and the cycle of tests returns to the beginning. Invoking the rule of threes, the player needs to complete this three-part test three times to end phase one of the battle. The beacon movement during the shuffle becomes faster through progressive cycles, increasing the difficulty of the middle test while leaving the other two consistent. This produces a linear difficulty  curve. Phase two will be discussed in the next video.

What makes this part of the "Gazer Guard" battle particularly elegant is the inverse relationship of complexity and suspension as the tests progress. The first test is highly specific to this boss and requires processing and filtering novel information. The subsequent shuffle challenge is a new way of testing visual tracking, a skill underpinning the entire game. The player is applying a developed skill to a novel scenario, which they ought to be more comfortable with than they were with the initial analysis. At the peak of suspension, the third test, the player needs to aim and shoot at a static target, something they've been doing against every enemy in the game. As more successes are suspended on the line, the abilities tested are increasingly fundamental - this gives the game cohesion and identity: victory in Valfaris is ultimately determined by running and gunning skills, not isolated puzzle solutions.

* the blue shield aura is an indicator the player should have learned by this point in the game, as it is used on every enemy that receives an attack but no damage (guarding).

** you could say experimentation is the first test, but I think that's implicit in any scenario where the player does not yet know how to progress.

A brief consideration of tells in shoot-em-ups

I was considering the application of "show &" tells to shoot-em-ups the past few days while playing Raiden IV, Mushi-F, and Lightening* Force; the idea that enemies should visually indicate when an attack is about to happen (e.g. Bowser sucking in air before breathing fire). It should be taken into consideration that I have not cleared any of the three games I'm about to discuss, although I'm on the final stages of Raiden IV and Mushi-F. Lightening Force I've played for like 3 hours so take that with a grain of salt.

As I said before, Raiden is so positionally-oriented that precise shot timing doesn't really matter. If there's a group of enemies above me that I know fire single tracking bullets, I also know that I will be safe sweeping left until I clear the bullets. I can start the sweep any time before the first bullet reaches me and end it any time after the last one fires. If there's also an enemy left of me, then I should probably sweep up-right to get away or down-left to get under it and fire. These scenarios are pretty static across runs. Timing doesn't really matter that much - what matters is doing these movements in such a way as to sync up with targets while also setting up the next dodge. Which is quite enough to make it more complicated than "sweep left". I did think it would be helpful on a few enemies who shoot starburst or static patterns - and I also noticed these enemies do have animated tells (e.g. a mine blinking faster before it explodes).

Mushihimesama Futari yeah nah. It's all sight-dodging. I'm looking at the bullets for paths through them, or for enemies under them to shoot and consume them. Also, everything pretty much shoots immediately as it comes on screen and doesn't stop til it's destroyed. Herding is a pretty big deal in situations where aggression fails - essentially, you can treat enemies as though they are firing at all times and this would not significantly change the strategy.

Lightening Force I can definitely see the applicability. Not for bullets so much, but the game is plagued with enemies that blast on screen and smash into you so fast there's barely a second to react. Some of them do have foreshadowing, swooping across the background or burrowing through the desert sand, but not nearly enough. Blazing Star solves this in the simplest way possible, with "incoming" arrows where enemies are about to emerge. Crude, but effective. The bosses desperately need tells for everything they do, because they move around jerkily without pattern and again fire extremely fast bullets without warning. Maybe need is too strong - they'd be altered, for sure. As is, playing it safe and staying completely out of their targeting zone is pretty much the only road to success. The anxious fidgeting makes sure that even that isn't particularly easy.

So yeah, I'm inclined to think something is wrong with Gradius if it would really benefit from visual tells, and I'm starting to question if it actually would. The most troubling factors seem to be malingering enemies like the footbots who hang around for 10 or 20 seconds and can mostly be ignored except for those three random times they shoot, and the crowded spacing that means sustained motions like I described re:Raiden lead to constant collisions (I've said in the past I don't care for walls in shmups). It's like the series carves out a weird space between Thunderforce, where you always have to play it safe, and Mushihimesama, where you always have to wait and see, but without having the elegant level design of Raiden IV which realizes that it's in that space.

*light·en·ing

/ˈlītniNG/

noun

noun: lightening; plural noun: lightenings

1. a drop in the level of the uterus during the last weeks of pregnancy as the head of the fetus engages in the pelvis.

Baby clears: Prehistoric Isle 2

Well that didn't take long. Still, it's nice to get a quick clear here and there, especially after weeks toiling away on arduously boring Raiden V or vengefully punishing Psikyo "games". Nicer still to land at #4 on the Switch leaderboards, though that's a meager accomplishment if ever there was one (only about 15 players appear to have hit the 1CC so far. Judging from their score, the 100th ranked player hasn't even cleared Stage 2!). Prehistoric Isle 2 was never going to amount to much anyway, being a plainly stripped down, late-period aesthetic showcase. The reward on offer is clear: fluidly animated prerendered prehistoric creatures inhabiting dreamy ruins, jungles, and volcanoes. After struggling under the bitter tyranny of hellrank stoking the fires of Strikers 1945, the opportunity to memorize a couple patterns, learn a handful of scoring techniques, take a couple risks and finish the ride is just nice, like a blowjob for fifty bucks. Notes:

• I've said it a thousand times before, I'll say it a thousand more times before I die, I don't like rank (PI2 has none). If I want to increase the difficulty of the game, I'll go into the settings and increase the difficulty. I did exactly that with Preshistoric Isle 2 after my first attempt was a startling 3CC, then realized that I'd rather play at the default difficulty and strive to increase score, interacting with more of the game's systems without having to worry about an ever-fluctuating survival threat. Build dynamism into your design (chains, score items, whatever) and you can allow the player interesting choices, rather than railroading them onto the hardest path.
• The difficulty curve is just right - Stage 1 is pretty easy, leaving most of the challenge first to learning mechanics and later to maximizing score; Stage 2 gets the enemies more active, making sure there are always bullets onscreen; Stage 3 is almost like a bonus level, a really simple straight shot that also has the prettiest backdrop - the calm before the storm; Stage 4 is where the difficulty peaks, introducing lots of new enemies and constant pressure; Stage 5 is the gauntlet, revisiting all the previous challenges and providing very few opportunities for recovery; and Stage 6 is the denouement, a protracted boss battle of the R-Type battleship form.
• The Japanese version has a 5-HP health bar versus Worldwide's 3-life stock; this doesn't change the effect of getting hit (main weapon and subweapon single-level power-down, one rescue lost), but it does change the appearance of restoratives. In the Japanese version, delivering a certain number of rescues (more than three?) to a pre-boss evac chopper generates a +1HP item. At the beginning of Stage 4, defending a certain number of evacuees (all 30?) grants another. (That's at least +5HP total over the course of the game, and I may have missed some). There is a 1UP that appears somewhere in Worldwide, but I haven't been able to find it. This was a clear of the Japanese version, which is obviously somewhat easier, although I also chose that route because....
• The health bar makes the game more interesting. Delivering rescues to the evacuation chopper can be used to get back health, but it means sacrificing the pretty hefty stage-end bonus you get for carrying them through the boss fight (500k per rescue + another 500k for all five). This is a neat twist on the traditional score-extend - you have to pull off the rescue to get either benefit, but at stage-end you get to decide whether it'll be a score bonus or recovery - and whether you want to risk losing the bonus entirely if you make a mistake on the boss. The North American version of the game trades this for... nothing. Delivering rescues still grants some power-ups, but that isn't nearly as meaningful as health/1UPs. The game is simply boringer without that element of risk/reward.
• The design of bosses is a clear highlight. They last for minutes, moving around the screen, attacking from different directions, almost never repeating attack patterns. Having a fully powered up weapon or using bombs to shave off a few seconds can make a huge difference if it means skipping the last phase or two. Particularly vicious is the Hornetpede who snakes unpredictably across the screen while firing two-phase bullets.
• The bomb doesn't seem to offer any invulnerability, or if it does, it's so short as not to matter. No panic-bombing. This is a pretty fair trade-off in a game that offers at least one health restorative per stage. 
• I played with the green chopper (spread shot) using the red (chain gun) weapon. Subweapons I swapped a lot, it's hard to hang onto them for long, but super missile seemed to be the best. Napalm was invaluable in the final boss confrontation. 

Anyway, that's a solid recommend for $8. Now I guess it's Blazing Star, unless I want to scream some more at Strikers.

People hate Raiden V, right?

Picked up Raiden V: DX for a pretty good sale price at $13, gave it a spin just to see how it would run on my new PC. Lightning fast, you might say, although I've never played nor imagined a game so dedicated to the furiously denounced flaws of the widely derided Sine Mora. Can you say BLOOM LIGHTING EFFECTS? And while Sine Mora was able to light the shmup world on fire with just optional screen shake and two bullet colors (orange and blue), Sine Mora 2: Raiden V has individual enemy volleys that can mix as many as five colors (red, blue, green, yellow, pink!), not to mention any other enemies on screen and the player's two hues of their own (blue and your choice of red, blue, or purple). It even has a camera that will zoom in and out mid-combat! Boy there's nothing I love more than watching the playing area shrink to half scale (and double size) while an approaching bullet pattern become so small and orange that it's completely obscured by the sun glare blooming off the orange buildings in the orange desert and the whole game becomes a gleaming white blob.

Seriously it's a nightmare. It's also baffling that anyone called this game "burdensomely latched to hoary conventions of shmups yore", when it's in fact so modern that it barely feels like a Raiden game at all (unless this is what the Fighters subseries is like). It has a long-form story mode with incessant voiceover during gameplay (you may say - hey, isn't that a Star Fox thing? but only Sine Mora and Raiden V are smart enough to pair it with subtitles covering part of the gameplay area), save points, branching paths, three ships with different stats and fully customizable weapons loadouts (three choices each for red, blue, and purple, for a total of 81 different profiles!), achievements, and a goofy-ass networked parallel-play power-up system. Like, Raiden III and IV are 90% the same game, IV is a loose remake of II, and the original Raiden has like eight variations and ports. This is a series about on par with Mega Man for variation, so continuing from IV to V feels like going from Mega Man 4 straight to ZX Advent without the 15 years of connective tissue explaining how you got there.

Did I mention the lifebar!!!

Okay I just hit up YT and WP and yeah looks like most of the new weapons and the branching paths are just carryover from Fighters. Why not now, nearly 20 years after the final entry?

New 1-credit clears! Axelay, M.U.S.H.A., Cho Aniki, and Other Nonsense Words

Spring is such a wonderful season for shoot-em-ups. God knows you can't play them all year round, but in the spring, you can. I'm a terrible player, but I've finally developed the method and Schedule I prescriptions necessary to clear extremely easy shooters in roughly five years apiece. Mostly that means sticking to old 16-bit stuff on Wii VC; I'm fonder of later arcade and XBLA games like Strania or Under Defeat, but I haven't improved to the point that I can clear them. Plus, fuck rank.

This week I finished a somewhat controversial 1CC of Axelay, one probably not destined for the record books at Win Galaxies. So whenever I pick up the game I play out the remaining credits of my last suspend state for practice; I was busy all weekend, but when I picked up last week's suspend I ended up clearing the game. Some key notes:

• The game lets you map Shot and Missile to different buttons or to the same (like later Gradii). Naturally, most people will put them on the same button - but they'll be displeased when they reach the final boss's mirror phase, where he fires shots when the player does and missiles when the player does. And if the player fires shots and missiles at the same time, so will he, and that'll create an undodgeable attack pattern. So ten hours in I had to relearn to play the game with two buttons (harder than it sounds, because there's a lot of trigger fluttering).
• I like the weapon select, but it's oddly balanced. Once you get a new weapon in each category, you'll never use the previous ones - they're not objectively worse, but they don't complement each other correctly. And one weapon, Spinneroo, I couldn't comprehend in the slightest.
• The rotating vulcan weapon that starts firing backward and angles forward as you hold the button is enough to base a game around. It puts timers and technique into an independent aiming mechanism, like a cross between dual stick (because it's independent of ship movement) and reverse options (because it has its own traversal time).
• The vert stages would be better if they weren't multiple screens wide. The breadth is meant to offer different paths through the stages, but it ends up being like Life Force's split routes, where you just choose one early on and always take it. Ultimately it just adds confusion and ambiguity to an already difficult-to-read style of presentation. 

Overall I liked it a lot. Naturally it's reminiscent of Gradius and Life Force, but the weapon-swapping and management (where getting hit costs the currently equipped weapon, like M.U.S.H.A.) also adds an element of Contra III. I definitely prefer a shooter without any powering up or recovery phases; saves all the frustration and jerky pacing without dismembering the difficulty. The rules are simple, the design is lean (not a hint of popcorn), the weapons are self-consciously weird, the visuals are terrific (the artwork, not the Mode 7), and the difficulty lives in that perfect spot where screwing up hurts but doesn't undercut the momentum.

Cho Annunaki is interesting to look at too, with an aesthetic somewhere in the neighborhood of 15th century European painting crashing a train into 20th century manga. Even the buff dudes at this point feel more Classical Antiquity than Homoerotic Gaydiquity. Notes:

• It's tempting to call the game pure popcorn, with enemies and even bosses rarely hanging onto life for more than a few seconds, giving it a really soft, smooth flow. 
• Each stage has a unique enemy set that tends to conform to the same balanced triumvirate: one enemy creates barriers (the lenses in Stage 2, the alarms in Stage 3, the nautiluses in Stage 4, the helms in Stage 5), one pumps bullets onto the screen (the chess knights in Stage 2, the jerkoffers in Stage 3, etc.), and one that dive-bombs toward the player (the birds, the bees, the planes). It's a conservative style, but it does highlight the subtleties in each incarnation, like the way the lenses can be pushed back and the nautiluses fire when passed.
• This is the most tautly suspended power-up system I've ever seen. Basically the player has ten power-up levels to scale and five for each of two options, only each level takes more and more pick-ups to achieve (1 for the first, 3 for the second, 7 for the third, 17 for the fourth, 28 for the fifth, so on - the game doesn't display any of this, and only the first, third, and ninth (I think?) level-ups even result in a visual distinction). On a good run, I reach full power by the final boss of stage four (of five), and any single death shaves a good chunk off that (maybe half?). So if I've died twice, no way I have full power on the final boss. And since it takes so long to power up, I ended up essentially competing with my options, as splitting things three ways drastically slows down the leveling up, and isn't worth it in distributed firepower. 
• No extends! No hidden 1UPs either. This is true of all Masaya games I know (Gley Lancer, Cybernator, Gynoug). Extends shouldn't be underrated as a mechanism to draw the player into the scoring system and reward them for perfecting early parts of the game.
• Was there a lot of subtly sexual language in what I just wrote?

Then we've got M.U.S.H.A., my first victory over Compile (unless - is Recca considered crypto-Compile? It at least came from the same Carnival that generated Spriggan). Anyway:

• On the subject of extends, I didn't really notice them here, but the option supply is kinda accomplishing the same thing. It's really quite odd that the player can build up a huge stock of options to deploy later, never affected unless they use a continue, and it can definitely be put to use. Personally I stuck with 3-Way, Back, or Reverse for the early levels, then once I had 30+ in stock and things got bitchy, I switched to Lock. These little fellas eat bullets, too, and they're graciously fixed in place for reliable dead zones.
• Splitting the weapon power level such that picking up a secondary weapon provides an extra hit point (and getting hit costs the secondary weapon) alleviates recovery as I discussed favorably wrt Axelay, but M.U.S.H.A. goes a bit too far and strays into easy mode. With a pretty sizable invincibility frame and frequent weapon drops, the max challenge is really just to go ten or fifteen seconds at a time without getting hit. Of course, if the main power drops enough and your option supply runs out, you're pretty much fucked. So it's a very low difficulty hurdle with a steep dropoff. 
• Black hole bomb* is a crazy and awesome weapon concept (pulling enemies! in a shmup!), but it generates so much chaos on screen I ended up not using it much. At the same time, white-green laser loved covering up bullets approaching from the front, so it needed to be paired with Reverse option to last. 
• The graphics are too big.

So yeah. All three decent, fun games that I'd recommend, but I liked them in the order listed here. 

* The idea that angular momentum and energy may be transferred from a rotating black hole to a particle being scattered by it was proposed by Roger Penrose in 1971. The first discussion of a runaway effect, the black hole bomb, was explored by W. H. Press and S. A. Teukolsky in 1972. If such an effect were to spontaneously occur, it may point to new physics beyond the Standard Model, and showing that black holes have "hair", as pointed out by a paper from 2017, by William E. East and Frans Pretorius.

Deathsmiles part 2: The tweaks and twixes of Mega Black Label

K got the slightly-non-scrub clear of Deathsmiles (all stages Rank 2+, beat the penultimate boss for real and beat more of the final boss than I expected (only used 3 bombs I think)). Ultimately it's definitely not a game I'd call bullet hell (**danmakuu** \'>_<'/ kawaii densetsu gaiden, senpai!! *\^o^/*) - the multiple aiming directions and handful of bullet-clearing techniques make it much more about planning how to keep bullets from piling up than about herding or tap dodging or reading geometry. In fact, sans the final two bosses, if there are ever enough bullets on screen that I'd have to go into weave mode, I just bomb. That's how rarely it happens (if you're playing decent at least). Nearly all of my character movement is with the goal of orienting my option or lining up a shot - in the final playthrough I made good use of Sakura with her reverse option, alternating locking it into place as Double or Tail. This time I was playing Mega Black Label, a mode that tweaks the scoring and adds some content but not much else (I'll be referring to the base game as "vanilla").

Before we start discussing the updates, you might want to check out my full run-down of Deathsmiles' score items and Power-Up mode. There's one detail of the base game that I left out, which is that the Overall Counter maxes out at a value of 10,000, after which all item drops will be crowns (which can still break apart into tiaras and skulls when hitting the ground). I've seen this state called "Frenzy mode", but nothing in the game or manual uses that terminology. I didn't mention that because I forgot to, but it also has no bearing on gameplay incentives - Power-Up scoring is two orders of magnitude greater than Normal even without it, and raising the Overall Counter as quickly as possible is inherently beneficial.

Mega Black Label's two biggest changes are meant to send scores into the tropoplosphere: first, Power-Up can be deployed at any Item Counter value above 500, and second, the Overall Counter is never reset. The scoring is so inflated that the second extend now sits at 300 million, far above my top All-Clear score for vanilla. Better still, the Overall Counter is now displayed directly below the Item Counter (just like I suggested in my previous post!). There's also a new playable character, a new optional Chapter 4 stage, and Rank 999, a fourth selectable difficulty alternative for each stage.  I'm not going to go into the new content, but I ought to note that playing stages on the highest available rank is necessary to maximize score, because higher rank means more enemies, and more enemies means more items. Since I'm interested only in scoring within the confines of a 1CC, I pick stage rank based on what I think I can (learn to) survive, and work on score only after.

Into the weeds on the first point first: the lower threshold for Power-Up. Transparently, this opens up a lot more flexibility in planning where to deploy the transformation, and it likewise makes doing so less costly. It's quicker to charge and it's quicker to recharge. This has obvious survival utility (Power-Up still boosts attack and clears bullets when transitioning), but also allows burning off some Item Counter value before a big item flow is known to be coming up to provide a refill (for instance, I know the ogres in Chapter 5 are going to spill gallons of items, so I make sure I'm always coming up to them on empty). At the same time, there are new costs to consider. Fever mode, which you'll recall lasts as long as the Item Counter is kept at 1000, remains unchanged - from Item Counter values of 500 to 999, the player character remains in Normal mode, despite Power-Up now being available. And if the transformation is triggered in this range, the Item Counter will immediately drop to 500 and count down from there, meaning any progress toward Fever is lost and the transformation duration is artificially shortened. Because of that counterbalance, it's still worthwhile to think ahead to get the most out of the Item Counter.

This wouldn't be as meaningful if the Overall Counter was resetting to 0 at each of these refills, as it was previously established that OC is the driver of exponential scoring. A shortened Power-Up would mean lesser Overall Counter values and probably wouldn't be very consequential. Likely with that in mind, Mega Black Label drops the Overall Counter reset (that in vanilla Deathsmiles occurs at the end of Power-Up) and also drops the cap value of 10,000. Now the Overall Counter is free to grow and grow throughout the entire run of a game, ramping up the suspension of a scoring run while also sacrificing most of its atomic nature. Although it's value is preserved throughout, it's still only active (i.e. applied to item pick-ups and able to grow) during Fever or Power-Up modes. If the player loses a life, the Overall Counter takes a big hit (maybe 30%? hard to track precisely during gameplay). More significantly, for the entire duration of boss battles the value quickly drains. Bosses need to be killed as quickly as possible or milked for items (which is only possible in Power-Up) to keep the counter up. This also seems to discourage the all-Fever style of play I briefly mentioned in the previous write-up.

These changes are interesting, but as my descriptions probably convey, I haven't fully wrapped my head around the best ways to use them. More flexible planning is ultimately more complex planning, and the full playthrough suspension of the Overall Counter throws massive flux into potential scores. I remarked about the vanilla game that you can completely blow a stage and still get consistent results in the next - that's no longer the case in Mega Black Label. There's more happening in parallel, more choices available at every point and more potential for tradeoffs, all making it harder to identify where one strategy is outperforming another. At some level of play I could see myself enjoying this granularity, but at the moment I find it more satisfying experimenting with routes and timing my Power-Ups in vanilla, getting that more straightforward feedback.

A review of Deathsmiles' scoring

This isn't a guide to scoring in Deathsmiles, nor is it a review of Deathsmiles. It's just an attempt to reconstruct the scoring system as a big picture that is neither as cryptic as the manual nor as prescriptive as a guide.

When it comes to score in modern arcade shooters, the classic question is always "does it really need that many zeroes?". When I started playing Deathsmiles, I was reaching the final boss with scores around 3 million, wondering why the extends were sitting up at 20 and 45 million. After taking a few minutes to read the manual, I was instantly able to adjust my play such that I was grabbing both extends and reaching that same finale with scores over 50 million. What's odd about this is that it didn't at all involve dying less or killing more enemies, and barely even affected my routes on screen - the initial change was completely orthogonal to the core survival challenge of the game.

Scoring in Deathsmiles is driven entirely by the collection of point items dropped by enemies; drastic variability in scoring derives from the player's ability to determine how many points each item is worth. This stems from a transformation dynamic; in Normal state, the player will be collecting items worth hundreds of points, and thus only total a six- or seven-digit score (hundreds of thousands or millions). In the transformed state, aka Power-Up Mode, the same items are worth thousands and tens of thousands, and will thus accumulate to eight- or nine-digits (tens or hundreds of millions). Therefore, scoring in a Normal state will be completely drowned out by Power-Up scoring and ultimately does not matter. A single 30-second burst of Power-Up at the beginning of stage 2C (the volcano) can score twenty times as much as an entire single-credit playthrough without it, and serves as a good illustration of how useful this technique can be (incidentally, the manual points out this location explicitly).

Anyway, the way it works is that you have an Item Counter, 0-1000. Picking up point items dropped by killing enemies increases the counter; when it reaches 1000*, press B to go into Power-Up Mode. While powered up, the counter gradually decreases, and at 0 the player returns to Normal. There is no way to prematurely exit the state, but it (and the counter value) do carry over between levels (i.e., if the player kills a boss in Power-Up with their counter at 350, they will start the next stage in Power-Up with their counter decreasing from 350). The Power-Up state lasts about 30 seconds, during which: 1.) player shot coverage and damage is substantially increased 2.) collecting items will not increase the Item Counter, and 3.) the score value of items is compounded with an additive modifier (Overall Counter) which continues to increase as more items are collected, thus creating an exponential effect (each item increases the score AND increases the value of the next item). The modifier, Overall Counter, doesn't exist outside Power-Up and is reset to 0 each time Power-Up ends*; it's therefrom implied that this modifier, and thus the quantity of items collected during Power-Up, is the key to scoring.

As long as the Item Counter sits at full (waiting to be deployed for Power-Up), the player character will be in Fever Mode. In this state, all enemies drop crowns (the most valuable point item). The Overall Counter also becomes active, but builds very slowly. Fever Mode does not affect any gameplay aspect outside of score, and will continue indefinitely as long as the player doesn't Power-Up or take damage. It mostly acts as a way to prime the Overall Counter while waiting for a planned transformation opportunity; there isn't any strong incentive to stay in Fever (it scores worse and leaves the character weaker than Power-Up), and the incentive to reach it is redundant with the incentive to reach Power-Up*. That said, supposedly the proest of the pros play the entire game in Fever and never use Power-Up, benefiting off the long-term Overall Counter growth.

The type and number of point items dropped by an enemy is determined by the enemy type and which weapon it's killed with - rapid, focus, or lock-on shot. Each enemy has a defined 'weakness' that has to be discovered via experimentation; that is, a specific shot type that generates extra items (rather than doing extra damage). You'll usually see dozens of instances of each enemy in its given habitat, so deducing weaknesses isn't too hard until things get frantic at the end (actually remembering them can be a bit tougher - and I would swear a few enemies have different drops for different player characters). Of course, there are also practical/survival incentives for using different shot types, so this sets up a balancing act for the player. (Mushihimesama Futari has a similar, simpler system: when the item counter is Blue, use focus fire to get the most items, when it's Green, use rapid fire; the color alternates at fixed counter values).

There are three types of point items of increasing value: skulls, tiaras, and crowns, the last of which only appear in Fever or Power-Up modes. Items fall to the ground from the point where they're spawned; upon landing, the larger items break into a multitude of skulls. Each item collected adds to two independent sums simultaneously: the player's score and the Item Counter (in Normal state) or the Overall Counter (in Fever or Power-Up)*. Probably the most confusing aspect of the game is that a collected item adds a different value to each of the aforementioned sums, e.g. a skull adds (+1) to the Item Counter, (+100+OverallCounter) to score, and something like (+log(OverallCounter)) to the Overall Counter, while a tiara adds (+5) to the Item Counter, (+800+OverallCounter) to score, and the same (+log(OverallCounter)) to the Overall Counter. So at an Overall Counter value of 250, three skulls = 3*(100+250) = 1050 are worth the same amount of points as a single tiara = 1*(800+250) = 1050, and are worth three times (3*(log(250) compared to 1*(log(250))as much to the Counter itself, which will increase the value of the next pickup. Since a tiara breaks into three skulls upon hitting the ground, this means that in Fever or Power-Up, it's better to let it break apart before collection - i.e., the highest quantity of items drives the highest score.

It's definitely complicated, but it's not AS complicated as it at first seems, due to the confusing presentation. The Item Counter should just be a super meter, because presenting a number so prominently suggests that the numerical value matters, which it does not (all the player needs to know is whether it's full and how soon it will be); the Overall Counter should be named something like Item Bonus, because it does not count anything "overall", in fact doesn't really count (the rate at which it increases is exponential), and sounds too much like Item Counter; the Overall Counter should be displayed where the Item Counter* is such that the player can actually read it (presently it's displayed as a +XXXX next to the item acquired, meaning it's in the middle of the gameplay area and somewhat hard to track), and enemies should somehow indicate that you killed them with the correct weapon (e.g. in MushiF, using the right weapon generates large gems and the wrong one small gems. In Deathsmiles it's more like the weakness gives you 3 crowns and 2 skulls, and the wrongness gives you 5 tiaras, so you need to mentally track your expected reward while determining weapon choice).

Overall it's well done, if maybe a little too meta-level for my tastes. It all comes down to deploying Power-Up at the right time, something that carries over between levels and therefore creates suspension (especially with a level select in play), yet also is self-contained enough that the barrier for entry to experimentation is low (you can totally butcher or totally ace the early stages and still get consistent results later). Contrary to my first impression, the counters do tie survival strategies to scoring, both because they can be exchanged for weapon power and because the type-weakness aspect of enemies is a counterpoint to which weapon might seem practically useful. The stage design is a bit monotonous and the bosses are too easy until they're too hard, but turning up the rank helps a little. Still, it's the abstract systems that drive a game like this, not the design.

*See MBL