The scrub rules of STG game design

[Rastan78]

So this is a list I came up with along with some help from Prickly Angler. The scrub rules are consist of attitudes and ideas I've observed over the years from players regarding what good or bad game design looks like within the genre. Some of these principals of course may be good or bad depending on player taste and context.

My question is should players and designers reevaluate these ideas? If we reach a consensus about what polished and balanced shmups should be does that limit us? Doesn't a game's identity also come from its idiosyncracies and rough edges?

On to the scrub rules!

1. Shmups should never have surprise kills that require you to have seen the level layout before to pass. Good (scrub) design dictates that a player with lighting relexes could in theory no miss a game on reaction during the first playthrough. No memo plz!

2. Games shouldn't expect you to learn esoteric or hidden mechanics. A game should explicitly, within the game itself, communicate how it is meant to be played. Or else there is a failing on the part of the developers.

3. Dying should never be beneficial to the player. Using lives as a resource to reduce rank, build bomb stock, checkpoint milk, extend chain combo etc. breaks an unwritten rule of good game design. The player should never have to die.

4. Visibility should always be number one even if that means the aesthetic design of the game is in service of that ideal. Desaturated backgrounds and brightly colored bullets are the ultimate solution.

5. You should never be forced to miss powerups to play optimally. Items are there to be grabbed and power progression should be smooth and always in the player's favor. No downgrades or dodging troll powerups.

6. Checkpoints are bad and are simply a relic of a prehistoric age. Any game would be improved by adding instant respawn. Checkpoint milking was an oversight and bad game design.

7. The meta game should not be a result of a happy accident where designers have created systems with unforeseen consequences and complexity. They should be in control of how their game is played in the intended way and patch out safespots, exploits, and glitches.

8. Good balance requires a smooth difficulty curve from easy to challenging wihout noticable peaks and valleys. Spikes such as a long, tough boss fight earlier in the game are result of poor balancing.

9. If you have a shield you should still never be forced to take damage. If it's nearly impossible to avoid shield damage then the game is badly designed. A shield was tacked on a a band-aid to paper over the imbalances.

10. View should be direct top down or side view. Angled perspectives like those in Raystorm make it too difficult for the player to judge dodges.

11. The ideal length of a game is 20 to 40 minutes. We can't have shmups that consist of multiple 10 min routes or longer campaigns with saves.

12. No item shops. They interrupt the action.

13. Hitboxes should be significantly smaller than sprites. Only outdated and archaic games have hitboxes roughly as big as the actual player sprite.

14. Gradius syndrome: the player shouldn't be punished severely for a single death no matter how OP the powerup system is.

15. No high HP enemies during levels to antagonize the player. They disrupt the flow and only belong in euroshmups.

16. No randomness ever. Unpredictability is a scourge that should be stamped out.

Anyway there are more which I may have time to add later. Thanks again to Prickly Angler and Shatterhand for adding to the list.

[jehu]

Almost hard to complain about this being bait when you can so plainly see the hook, the line, the pole, and the two ornery Darius players peering over the edge of the boat to see if they’ve gotten any bites

[BareKnuckleRoo]

Response is kinda long so under a spoiler tag:

Spoiler

1. Shmups should never have surprise kills

Agree, but some nasty surprises with a "warning" indicator such as enemies from behind can still be a thing without being too awful.

2. Games shouldn't expect you to learn esoteric or hidden mechanics.

This is hard to avoid as many games even unintentionally develop unintentional esoteric meta that becomes beneficial to know. I think a good in-game manual is key to minimizing this.

3. Dying should never be beneficial to the player.

Yagawa players won't like this. I generally agree that most games which encourage suicides to rank control would be better without this present.

4. Visibility should always be number one

I don't like these kind of absolute statements. Visibility is important, but you can sometimes have a great game that doesn't have impeccable visibility. Look at monochrome shmups like we had on the GB something like Monospace.

Generally though, I agree that enemy bullets and player shots work best if they're high contrast colours or do something like make player shots desaturated or transparent.

5. You should never be forced to miss powerups to play optimally.

6. Checkpoints are bad and are simply a relic of a prehistoric age.

Agreed.

7. The meta game should not be a result of a happy accident where designers have created systems with unforeseen consequences and complexity. They should be in control of how their game is played in the intended way and patch out safespots, exploits, and glitches.

At some point you gotta stop patching and accept there's unintentional scoring exploits that will sneak in though. Ah well.

8. Good balance requires a smooth difficulty curve from easy to challenging wihout noticable peaks and valleys. Spikes such as a long, tough boss fight earlier in the game are result of poor balancing.

Futari 1.5 Ultra has the first stage as being one of the toughest compared to the fourth stage. I think this generally holds as a good thing to aim for, but having a breather, more scoring oriented stage isn't necessarily bad.

9. If you have a shield you should still never be forced to take damage. If it's nearly impossible to avoid shield damage then the game is badly designed. A shield was tacked on a a band-aid to paper over the imbalances.

I agree in the case of limited use shields like Darius or Gradius where shields make your hitbox bigger. Giga Wing and Eschatos don't necessarily force you to use the shield there but it's more of an active shield rather than a passive one, so I think this is different from what your wording implies, and I suspect you're intending to refer to a passive kind of shield shield.

10. View should be direct top down or side view. Angled perspectives like those in Raystorm make it too difficult for the player to judge dodges.

Hard disagree. Games are allowed to experiment with weird view angles. It doesn't always work, and that's fine, but there are plenty of good games out there that aren't direct top down or side view such as Eschatos, Viewpoint, XSynergyGate, etc.

11. The ideal length of a game is 20 to 40 minutes. We can't have shmups that consist of multiple 10 min routes or longer campaigns with saves.

Hard disagree. There's nothing wrong with a shmup having extra routes or being designed around more bite size chunks of gameplay, or even both as in the case of Touhou games having Extra modes that are tough but not as long as a run of the main game.

I do think arcade shmups should probably be an hour max if it's a 2 loop game. Infinite looping is not a good thing because score becomes less about the score mechanics and more about pure survival time (and because someone obsessive like me will play it for multiple loops and it's not healthy to play that long in one sitting...).

12. No item shops. They interrupt the action.

They're hard to implement in a non-intrusive way, but Drainus' arcade mode proved it could be done well, so I disagree with writing them off entirely.

13. Hitboxes should be significantly smaller than sprites.

Also: player ship sprites should also always be drawn below the enemy bullets. Judgement Silversword and early versions of Monolith were otherwise good games that did this and it wasn't a positive thing.

14. Gradius syndrome: the player shouldn't be punished severely for a single death no matter how OP the powerup system is.

Agreed; there's a lot of games that are still fun even with this in play but they're fun in spite of this, not because of this.

15. No high HP enemies during levels to antagonize the player. They disrupt the flow and only belong in euroshmups.

Disagree; the wording doesn't give any examples and is too vague for me to agree with. What examples are you looking at that overuse this? Obviously any shmup has to be designed so you can't get flooded with nigh-invulnerable zako enemies, but there's nothing wrong with games having tanky little miniboss style enemies periodically as long as they're designed well. Needs clarification with examples of what's bad, good, etc.

[Rastan78]

Just to be clear. These aren't my own ideas or beliefs. These are things I've heard stated by many other players over the years. My idea is that some of these should be questioned. Although some are more in bad take territory and some are more widely accepted.

Yeah the title is a bit of a troll.

Coming from the FGC a scrub is not based on being good or bad as a player. It's more of not having a learning/adapting mindset vs blaming the game and its design, other outside circumstances.

Is it the role of designers to make a game that fits every player like a glass slipper or to push their boundaries and subvert their assumptions? Of course the truth is somewhere in the middle.

[Bassa-Bassa]

It's like if the list was made by an Ikaruga fan.

[Lemnear]

I should add this to that post I made, sort of a formula for the perfect SHMUPS

Spoiler

1. Shmups should never have surprise kills

Do sniper tanks that shoot as soon as they enter the screen (even if not completely) count? Or enemies that come out from the sides at full speed? Those from behind?
The "WARNING" is always welcome, at least for me.
Anyway yes, these are really cheap tactics to increase the difficulty.

2. Games shouldn't expect you to learn esoteric or hidden mechanics.

I agree.

3. Dying should never be beneficial to the player.

How can you lower your RANK then?

4. Visibility should always be number one

Absolutely yes, but I think as graphics get better this will become more and more difficult.

5. You should never be forced to miss powerups to play optimally.

Agreed.

6. Checkpoints are bad and are simply a relic of a prehistoric age.

I hate checkpoints, unless it's an R-Type or R-Type-Like, in which case I feel it's necessary, otherwise you can use a life to get through a difficult section without ever having to learn it. If an R-Type game didn't have checkpoints it would be breaking Point 3.

7. The meta game should not be a result of a happy accident where designers have created systems with unforeseen consequences and complexity. They should be in control of how their game is played in the intended way and patch out safespots, exploits, and glitches.

Agreed.

8. Good balance requires a smooth difficulty curve from easy to challenging wihout noticable peaks and valleys. Spikes such as a long, tough boss fight earlier in the game are result of poor balancing.

It depends on how difficult these peaks are. Certainly a game with a softer difficulty curve is more appealing to a wider audience.

9. If you have a shield you should still never be forced to take damage. If it's nearly impossible to avoid shield damage then the game is badly designed. A shield was tacked on a a band-aid to paper over the imbalances.

It seems more like a EUROSHMUPS thing rather.

10. View should be direct top down or side view. Angled perspectives like those in Raystorm make it too difficult for the player to judge dodges.

Correction, the view can be angled as much as you want, it's the BULLETS that must NOT be tilted in strange ways, otherwise you don't understand when they actually enter the playable field. If it's ONLY the view that's diagonal/bent, then that's fine.

11. The ideal length of a game is 20 to 40 minutes. We can't have shmups that consist of multiple 10 min routes or longer campaigns with saves.

Agreed. (even less for me, 15-30min).

12. No item shops. They interrupt the action.

I think this is a design choice, not all SHMUPS are a succession of uninterrupted levels, you can very well go back to base and shop for weapons and gadgets (Air Carrier Wing/Ordyne etc.)

13. Hitboxes should be significantly smaller than sprites.

I think this is a design choice too. (from what I've experienced with "test players", it's more intuitive to have the hitbox the same as the sprite)

14. Gradius syndrome: the player shouldn't be punished severely for a single death no matter how OP the powerup system is.

Agreed. It's masochism.

15. No high HP enemies during levels to antagonize the player. They disrupt the flow and only belong in euroshmups.

Absolutely. You lose the satisfaction of destroying enemies if they have too much health, they make the game sometimes forcedly difficult, and they also break the rhythm a bit.

Spoiler

[Personal Extra Point:] the game must be a synaesthetic sensorial experience, it must saturate the synapses, between audio, visual and gameplay. Batsugun and CAVE in general do it.

However, almost every point on this list clearly says “Battle Garegga, fuck you!”

[Acid]

As a fan of Battle Garegga, Trouble Witches, and Ginga Force, I disagree with most of these. To each their own.

[Lethe]

Evidence continues to mount that Mushihimesama is the ultimate scrub game.
(unless you're playing it for score but what kind of idiot would do that olololol)

[XoPachi]

I have a moderate case of color blindness to the point it affects my work on occasion. I really think that games with high density obstacles and hazards need to provide contrast options. It doesn't affect me in these games often, but there are occasions like Seibu games where it becomes more of a pain than it needs to be to keep track of hazards.

[Shatterhand]

Battle Garegga has been voted #1 shmup of all time a few times in our annual poll here, and it breaks at least rule 2, 3, 4 and 5.

Gradius V is also very often on that list and it breaks rule 11

Rule 10 is also denied by games like Eschatos, Einhander, Silpheed... damn, even Philosoma and the mentioned Raystorm (and Raycrisis)

Rule 8 is also broken by a lot of excellent games, but I'll just speak for myself... I *do* like difficulty spikes at specific points of a game, and then the game gives you some time to breah just after that with an easier section. I really enjoy when some of the spikes are midlevel, which are surprising and also make the end of level boss fight a little less tense and more enjoyable.

[Lander]

Ooh, fun thread idea. Brain food

I propose Scrubrule #16: No randomness ever. Unpredictability is a scourge that should be stamped out.

Good (scrub) design

Hol' up, isn't that an oxymoron?
Though on reread, I suspect the implication is 'Good design for Scrubs', rather than 'Scrub' == 'Good'.

My question is should players and designers reevaluate these ideas? If we reach a consensus about what polished and balanced shmups should be does that limit us? Doesn't a game's identity also come from its idiosyncracies and rough edges?

I'd say the best rules account for bending and breaking; guidelines for general goodness, but without dogma that states violators are subject to immediate dismissal.
Breaking convention can result in fascinating ideas that work out well, even if it's no guarantee, whereas sticking to it religiously results in the current (ugly, borderline-deformed) face of the games industry

1. Shmups should never have surprise kills that require you to have seen the level layout before to pass. Good (scrub) design dictates that a player with lighting relexes could in theory no miss a game on reaction during the first playthrough. No memo plz!

A tricky proposition for any game that's not a randomized roguey abomination, since memo and familiarization is a fundament of the learning process! Even the most fair challenge, designed expressly without any bushwhackin' waves from behind, sudden background -> foreground hitbox transitions, or other such surprises, will be gradually chipped away by familiarity.

Really, I think this one is a case of failure to articulate on the part of the scrub community. The key is in fairness, but grasping what's actually unfair versus simply challenging is its own problem.

2. Games shouldn't expect you to learn esoteric or hidden mechanics. A game should explicitly, within the game itself, communicate how it is meant to be played. Or else there is a failing on the part of the developers.

I'm okay with this sentiment, though I think there's a degree of flex with what defines esoteric or hidden. By nature, those ideas are more applicable to secrets and bonuses that are not expected of the player, but included as a reward for the inquisitive and thorough.

Case in point, consider the recent-ish Animal Well; not a shmup by any means, but a textbook example of the designer trusting the player to be a halfway competent individual, capable of fiddling around, pressing buttons, parsing implication, and figuring out everything from basic commands to nuanced systemic interactions. It'd be a markedly less impressive work if it felt the need to beat them over the head with each element as it came up.

3. Dying should never be beneficial to the player. Using lives as a resource to reduce rank, build bomb stock, checkpoint milk, extend chain combo etc. breaks an unwritten rule of good game design. The player should never have to die.

*snaps fingers* the lives system is now a life bar, and 1UPs are health pickups. QED.

Personally, I think any resource given to the player is fair game for risk/reward abuse. Lives correlating more directly to credits than other kinds makes them psychologically more valuable, but they're just a number at the end of the day.

[Rastan78]

Almost hard to complain about this being bait when you can so plainly see the hook, the line, the pole, and the two ornery Darius players peering over the edge of the boat to see if they’ve gotten any bites

[Rastan78]

Rule 8 is also broken by a lot of excellent games, but I'll just speak for myself... I *do* like difficulty spikes at specific points of a game, and then the game gives you some time to breah just after that with an easier section. I really enjoy when some of the spikes are midlevel, which are surprising and also make the end of level boss fight a little less tense and more enjoyable.

This could also apply to scoring for elements that are really tricky or random. Maybe even better if they come by mid game if they will result in restarts. The boss milking on Guwange cat boss comes to mind and supposedly led players to quit scoring the game. At least it's stage 3 iirc and not the final boss.

I think I heard the anti difficulty spike mind set coming mostly from game reviewers. A lot of players will appreciate peaks and valleys since it creates tension and release, and also notorious and memorable moments.

Futari stage 3 is another example of a spike found in a highly regarded game.

[BareKnuckleRoo]

Futari stage 3 is another example of a spike found in a highly regarded game.

Stage 3's difficulty is pretty overrated. It's harder, but not absurdly so, and the hardest part of the level, the shots fired by the pillbugs, is a lot easier when you realize it's aimed and manageable with tap dodges if you can't wide dodge and speedkill yet. I can think of a number of other CAVE shmups where by stage 3 it's a much more significant jump in difficulty (DOJ, Dangun, then there's Espgaluda 2 and Ketsui where it's harder than Futari, but not much harder than their respective stage 2s because the difficulty ramps up faster).

[Some-Mist]

I love da small hit boxes...

Spoiler

but I love big ones too

[Shatterhand]

I propose Scrubrule #16: No randomness ever. Unpredictability is a scourge that should be stamped out.

That was a rule that I used to live by as a game designer. Sophstar which I designed has a total of 0 calls to Random functions because of that.

But I've seen big games doing it a lot, and it looks some people do like that. Our next game is toying with a bit of Randomness, which, by now, is just in the order of the patterns of the bosses. But I'm thinking if I should add a little bit of randomness on some of the bullet patterns too. Like, a spinning pattern maybe shouldn't always start at the same angle, or stuff like that.

[Rastan78]

I should add no randomness to the OP for sure

[PC Engine Fan X!]

I propose Scrubrule #16: No randomness ever. Unpredictability is a scourge that should be stamped out.

That was a rule that I used to live by as a game designer. Sophstar which I designed has a total of 0 calls to Random functions because of that.

But I've seen big games doing it a lot, and it looks some people do like that. Our next game is toying with a bit of Randomness, which, by now, is just in the order of the patterns of the bosses. But I'm thinking if I should add a little bit of randomness on some of the bullet patterns too. Like, a spinning pattern maybe shouldn't always start at the same angle, or stuff like that.

Randomness, especially on an end-stage boss, would keep a skilled ace player on his or her toes quite easily.

PC Engine Fan X! ^_~

[Lethe]

The interest of randomness to scrub talk is centered on things being "unfair" over "unpredictable". We've had threads before which have touched on ideas like binary versus continuous randomness which ties into that supposed distinction. The bullets are in different places = often fine. You get a random score from a medal or the boss randomly picks between easy pattern and impossible pattern = usually complained about. What's funny is the same people who'll complain about RNG will also refuse to play something 100% routing-based like Psikyo because it's also "unfair" in a different way.

But I've seen big games doing it a lot, and it looks some people do like that. Our next game is toying with a bit of Randomness, which, by now, is just in the order of the patterns of the bosses. But I'm thinking if I should add a little bit of randomness on some of the bullet patterns too. Like, a spinning pattern maybe shouldn't always start at the same angle, or stuff like that.

Since this is a pet subject of mine I'm going to talk your ear off about random and "random" implementations. There is a lot of depth to this from game to game.

One of the more interesting cases I found was in looking really closely at Hellsinker. The input variables have been carefully chosen to be consistently manipulatable but also opaque if you're not paying attention. In other words, substituting pseudo-randomness for making the whole thing so fucking convoluted that you undergo a similar learning process anyway, without the risk of getting plain screwed over. (Well, in most cases...)

• Most of the time the out-variables (e.g. timings and positions of enemy spawns) are based on framerule timings instead of frame-perfect determinism. So the game's behavior is under your control as long as you're consistent within a reasonable window.
• Then it uses greater timing precision for elements more precisely under the player's control, such as triggering a new boss phase. Or for a few things that are not intended to be routable.
• Similar rules are extended to the player's screen position. There's high-precision subpixel coordinates, down to low-precision screen chunks, and they're both used to affect triggers in different situations.
• Weirder elements like the physics system seem to be affected by the player's resources, such as shot power level, so they also become predictable to some extent if you know what you're doing.
• In result, "random" outcomes are procedural and can be anticipated. When things get combined with each other it can create funky situations, and you need to develop experience to know how being slightly off-target somewhere is going to affect the game elsewhere. It's hard to play precisely enough that everything goes exactly right, but you can play well enough that things land within predictable margins of error.

[To Far Away Times]

I think I am quite possibly the biggest scrub in the world, lol.

Spoiler

On to the scrub rules!

1. Shmups should never have surprise kills that require you to have seen the level layout before to pass. Good (scrub) design dictates that a player with lighting relexes could in theory no miss a game on reaction during the first playthrough. No memo plz!

Absolutely agree with this. This is what separates R-Type from a lot of other memorizes. Prior knowledge helps a ton, but everything feels fair the first time.

2. Games shouldn't expect you to learn esoteric or hidden mechanics. A game should explicitly, within the game itself, communicate how it is meant to be played. Or else there is a failing on the part of the developers.

Fuck Garegga.

3. Dying should never be beneficial to the player. Using lives as a resource to reduce rank, build bomb stock, checkpoint milk, extend chain combo etc. breaks an unwritten rule of good game design. The player should never have to die.

Fully agree with this. I would think except in very niche cases this would be bad game design.

4. Visibility should always be number one even if that means the aesthetic design of the game is in service of that ideal. Desaturated backgrounds and brightly colored bullets are the ultimate solution.

When I was making my own shmup I prioritized visibility to the point where I made sure the color pink was never used on anything except bullets. Visibility was given a ton of priority, and I picked colors that maximized visual clarity, even at the expense of cooler visual details.

5. You should never be forced to miss powerups to play optimally. Items are there to be grabbed and power progression should be smooth and always in the player's favor. No downgrades or dodging troll powerups.

A power up that is a power down is mega dumb. Here's looking at you Darius 1 and Gaiden. I only put up with that because of those game's incredible aesthetics.

6. Checkpoints are bad and are simply a relic of a prehistoric age. Any game would be improved by adding instant respawn. Checkpoint milking was an oversight and bad game design.

Agree. There's a reason this mechanic is almost never seen in modern shmups. Though I would say a checkpoint system that has no power downs would be fair.

7. The meta game should not be a result of a happy accident where designers have created systems with unforeseen consequences and complexity. They should be in control of how their game is played in the intended way and patch out safespots, exploits, and glitches.

I come from the Fighting Game Community, so using glitches and exploits is what you're supposed to do. No problem with this.

8. Good balance requires a smooth difficulty curve from easy to challenging wihout noticable peaks and valleys. Spikes such as a long, tough boss fight earlier in the game are result of poor balancing.

I disagree with this one. Some of my favorite games have peaks and valleys. Mushihimesama Futari, DOJ, R-Type. That push and pull pace is paramount to keeping players from tiring out.

9. If you have a shield you should still never be forced to take damage. If it's nearly impossible to avoid shield damage then the game is badly designed. A shield was tacked on a a band-aid to paper over the imbalances.

Forced to take damage = bad game design 100% of the time.

10. View should be direct top down or side view. Angled perspectives like those in Raystorm make it too difficult for the player to judge dodges.

I like Raystorm and Viewpoint, so I would disagree with this one.

11. The ideal length of a game is 20 to 40 minutes. We can't have shmups that consist of multiple 10 min routes or longer campaigns with saves.

Agree with this. Though I'm cool with a 15 minute run time too. I think some of the Darius Burst AC routes clock in around that time, and that feels good to me.

12. No item shops. They interrupt the action.

I like item shops in games like Trouble Witches, but my Euroshmup alarm does tend to go off at the sight of them.

13. Hitboxes should be significantly smaller than sprites. Only outdated and archaic games have hitboxes roughly as big as the actual player sprite.

A smaller hitbox does feel better, but it is not neccisarily a requirement.

14. Gradius syndrome: the player shouldn't be punished severely for a single death no matter how OP the powerup system is.

Gradius syndrome sucks so bad. May as well just make the games have one life.

15. No high HP enemies during levels to antagonize the player. They disrupt the flow and only belong in euroshmups.

This one definitely trips the Euroshmup alarm as well, but I can also think of games that do tanky mid size enemies very well like Cho Ren Sha 68K.

16. No randomness ever. Unpredictability is a scourge that should be stamped out.

Players should be rewarded for perseverance and repeat plays by being able to lean on memorization. I'm okay with some randomness, but generally prefer it is kept to a minimum.

[Lander]

That was a rule that I used to live by as a game designer. Sophstar which I designed has a total of 0 calls to Random functions because of that.

But I've seen big games doing it a lot, and it looks some people do like that. Our next game is toying with a bit of Randomness, which, by now, is just in the order of the patterns of the bosses. But I'm thinking if I should add a little bit of randomness on some of the bullet patterns too. Like, a spinning pattern maybe shouldn't always start at the same angle, or stuff like that.

That's a development conviction I can get behind though I've backed off my once-hardline position with time. Nowadays I draw the line somewhere around "does the designer understand what's happening under the hood?"; if they can be handed appropriate context of the game state that led up to a point, and reasonably explain why a given outcome happened without falling back on "it's just random", then it's fair game.

Shuffling order is an interesting edge case, like Tetris' 9-bag randomizer. Since the frequency of a given outcome remains fixed, there's a clear problem space that - while likely not consistently predictable by its implementor - retains enough definition that you can make definitive statements like "outcome X won't happen again until outcomes Y, Z, and W have occurred", so you avoid losing all reasoning structure and ending up at the mercy of the tides.

The interest of randomness to scrub talk is centered on things being "unfair" over "unpredictable". We've had threads before which have touched on ideas like binary versus continuous randomness which ties into that supposed distinction. The bullets are in different places = often fine. You get a random score from a medal or the boss randomly picks between easy pattern and impossible pattern = usually complained about.

Ah, but I contend that a good scrub has a poor grasp of the distinction between fairness and predictability, choosing to latch onto the latter as a convincing platform for criticism!
But yes, the bottom line is ultimately the feeling of "I got cheated!"

There's a certain threshold where I'd consider pseudo-randomness to still fall under the umbrella of determinism; if the player can become conscious of it and consistently manipulate the system to achieve a known result, then all you've really done is add an (admittedly abstruse) layer of indirection to the reasoning. That's probably the happy medium for me; afford the experienced player coarse control, but with enough margin for error that the game state can butterfly-effect off into interesting and unknown new permutations following an error in execution.

The point where it crosses the line into perceived true randomness - let's call it The Calrissian Test for lols - is where things get wholly out of hand. Monster Hunter is a key example; the logic behind drops is convoluted to the point of seeming truly unpredictable, and the duration between spins is lengthy. Thus, disappointment can become so frequent that the playerbase has collectively hallucinated a 'desire sensor' built into the game, designed specifically to extend progression by denying needed materials!

Per-frame determinism is an interesting idea; in the past I've dabbled with modeling an STG as essentially an animation timeline, which can be freely scrubbed ( ) back and forth like a video without losing any coherence. Though as you'd expect, it's not so simple when a player is added to proceedings, and ends up quite rigid without special machinery to account for choice.

Still, I think there's some fruitful investigation to be done on that front. There are research languages out there that model functions in such a way that they become reversible, which essentially hoists the 'timeline' idea up high enough that you can implement all the randomness and external input you like and still be 100% deterministic. Bit of a technicality, given that the end result would be indistinguishable from aggressive save-state rewinds to the untrained eye. Though fanciful notions of STG-VHS aside, the bones of it - immutability, functional purity - have interesting implications for guaranteeing sane and reproducible progression of state.

[BareKnuckleRoo]

That was a rule that I used to live by as a game designer. Sophstar which I designed has a total of 0 calls to Random functions because of that.

But I've seen big games doing it a lot, and it looks some people do like that. Our next game is toying with a bit of Randomness, which, by now, is just in the order of the patterns of the bosses.

I think a lot of games with replay functions don't use true RNG but rather use some elements that on the frame the call is made will always be the same like score or something, use that as a seed to modify the pattern. Makes it easier to implement replay recording, guess this would be considered per frame determinism?

A mix of fixed patterns, aimed patterns, and randomly spraying patterns never hurts. I like how Progear has bosses where it picks from a "pool" of 4 attacks or so (S2 boss phase 1, S3 boss phase 2, S4 boss phase 1) but I've seen people complain about it since RNG can cause you to get the hardest attacks rather than the easiest. In my experience this isn't a huge issue in Progear as bosses generally have enough health to see 3 out of 4 attacks in these phases, and the difficulty is such that the attacks are all about even in difficulty, but other people seem to disagree and would prefer Progear not use any RNG in patterns at all.

RNG in a shmup isn't necessary awful but it needs to be done in a controlled way, essentially.

[Lander]

Generally, pseudo-RNG is implemented by setting a seed value from a volatile source at the start of play (say, the system clock) then munging it around with a bunch of arbitrary arithmetic to produce a sufficiently-obfuscated result, and changing the seed so the next invocation is different. Using a gameplay variable like score is a nice way to add extra obfuscation, since it's a moving target rather than a fixed progression, but risks accidentally introducing nondeterminism if whatever feeds the score is itself nondeterministic. It's kind of infectious, in that sense - you have to take great care not to let it creep in anywhere.

Formally, per-frame determinism (as I understand it, at least) is the total guarantee that - given the same inputs - the simulation will produce the same outputs, from start to finish.

Simple-sounding on paper, but easy to screw up. For instance, using floating point (i.e. decimal) numbers immediately violates the constraint unless your target platform provides IEEE-compliant deterministic operations, as does offering a variable framerate without extra concessions to independent simulation and render ticks, or having done something nondeterministic on any previous frame. As soon as anything diverges, all bets are off.

My animation timeline idea was an attempt to address that analytically; instead of saying "move thing in this direction by this distance" every frame to update old state → new state, you toss the old state at the start of each frame and go directly from time (+ timestamped events for spawns, despawns, etc.) → new state for every entity in the game. Handy for avoiding the aggregate inaccuracies inherent to discrete integration, but very different from the conventional approach.

[BareKnuckleRoo]

Formally, per-frame determinism (as I understand it, at least) is the total guarantee that - given the same inputs - the simulation will produce the same outputs, from start to finish.

As far as I'm aware this is the conventional definition. Assuming the same known starting point, you always get the same results if the inputs per frame line up. But yeah, it's really hard to implement. I know of at least one doujin with a replay option disabled in the menu because I believe they tried and were struggling to implement it properly.

The original Final Fantasy for NES used a random number table where a bunch of numbers were pre-generated in a table and each frame it cycled to the next number in the table. Any rolls of 0 are a "critical success" which allow you to do things like hit the final boss with status effects it's otherwise resistant to (same goes for your own team). In theory it's only a 1/201 chance of happening, but in practice because the table was seeded with three 0's, the chances of a this kind of critical hit are 3/256.

[Steven]

I imagine Elitist Underground probably agrees with everything or almost everything on this list. In fact, if someone said this entire list came directly from him, I'd believe it.

That said, I kind of sort of partially somewhat agree with a few things here, like game length and visibility. For example, Cygni comes to mind, as you can't see anything in that game because it's a giant mess of everything being the same color. I think it's also like an hour and a half long or something like that for just one loop, and that's way too long.

[Sumez]

Is this whole thread just a big jab at Greg? XD

[Steven]

The original Final Fantasy for NES used a random number table where a bunch of numbers were pre-generated in a table and each frame it cycled to the next number in the table. Any rolls of 0 are a "critical success" which allow you to do things like hit the final boss with status effects it's otherwise resistant to (same goes for your own team). In theory it's only a 1/201 chance of happening, but in practice because the table was seeded with three 0's, the chances of a this kind of critical hit are 3/256.

It's still better than Shin Megami Tensei's Fiend's weapon drops. The Fiends have a 1/256 chance of spawning, and once they do spawn, they have a 1/256 chance of dropping the best weapons in the game when defeated. You can probably save scum with save states on an emulator, but good luck getting those weapons on real hardware.

[cody]

Would you play a game where all the enemy bullets were literally invisible?

[Shatterhand]

I think a lot of games with replay functions don't use true RNG but rather use some elements that on the frame the call is made will always be the same like score or something, use that as a seed to modify the pattern. Makes it easier to implement replay recording, guess this would be considered per frame determinism?

Yeah, that's whay I'm actually doing. Get score + player position to generate the "random" order of attacks. I still think it's "random enough", I'd be surprised if someone managed to reach the boss with the same exact score, on purpose, two times.

[m.sniffles.esq]

16. No randomness ever. Unpredictability is a scourge that should be stamped out.

Love this

Games that continue to be exciting and interesting ARE KILLING THE INDUSTRY. You must all PLEDGE upon the alter of whatever pig god you worship to make games that are as fucking boring as humanly possible. If people can replay old games they will NEVER buy new ones and WE SHALL PERISH

Edit: while I realize the entire list is a manifesto extolling bland, stale, cookie-cutter design--for whatever reason--that last one really threw me for a loop.