It should be theoretically possible to ascend any character


Although the central place for design discussion is ##crawl-dev on freenode, some may find it helpful to discuss requests and suggestions here first.

Crypt Cleanser

Posts: 714

Joined: Saturday, 5th December 2015, 06:56

Post Friday, 19th February 2016, 22:50

It should be theoretically possible to ascend any character

My feeling is thus: for design strength, it should be theoretically, but maybe not humanly, possible to win at least a 3 rune in every game of DCSS without being struck down by RNG.

I'm not talking every race/class combo, btw, just... the bulk of them. Maybe 50%+?

Imo, it's very slight changes needed to open this up to theoretical infinite-winstreak.

1) Don't make adder spawn on d1. Or make them very slightly less dangerous (either reduce speed, or lower hp/evasion, mb)
2) Change shaft to remove insta-death possibility vs lowbie character (many ways to achieve this).
3) Move centaur minimum level to spawn numerically up by 1 or 2 levels

There may be a few other changes needed, not sure, but I think this alone would do it.

For this message the author lethediver has received thanks:
zxc23

Sar

User avatar

Ziggurat Zagger

Posts: 6418

Joined: Friday, 6th July 2012, 12:48

Post Friday, 19th February 2016, 22:54

Re: It should be theoretically possible to ascend any charac

Can you provide any pro arguments, apart from your feeling? I mean, that might sound like a stupid answer, but Crawl manual actually makes an argument that supports the existence of unavoidable deaths as a feature of Crawl. Here's that section:
  Code:
The possibility of unavoidable deaths is a larger topic in computer games.
Ideally, a game like this would be really challenging and have both random
layout and random course of action, yet still be winnable with perfect
play. This goal seems out of reach. Thus, computer games can be soft in the
sense that optimal play ensures a win. Apart from puzzles, though, this
means that the game is solved from the outset; this is where the lack of a
human game-master is obvious. Alternatively, they can be hard in the sense
that unavoidable deaths can occur. We feel that the latter choice provides
much more fun in the long run.

Dungeon Master

Posts: 625

Joined: Thursday, 23rd October 2014, 03:08

Post Friday, 19th February 2016, 22:58

Re: It should be theoretically possible to ascend any charac

Where's this list of over 50% of the allowable combos that can't win, anyway.
Last edited by CanOfWorms on Friday, 19th February 2016, 22:58, edited 1 time in total.

Ziggurat Zagger

Posts: 8786

Joined: Sunday, 5th May 2013, 08:25

Post Friday, 19th February 2016, 22:58

Re: It should be theoretically possible to ascend any charac

Now that the die-in-abyss-on-turn-0 bug is fixed, do you have evidence that this is not already the case?

Crypt Cleanser

Posts: 714

Joined: Saturday, 5th December 2015, 06:56

Post Friday, 19th February 2016, 23:06

Re: It should be theoretically possible to ascend any charac

Sar wrote:Can you provide any pro arguments, apart from your feeling? I mean, that might sound like a stupid answer, but Crawl manual actually makes an argument that supports the existence of unavoidable deaths as a feature of Crawl. Here's that section:
  Code:
The possibility of unavoidable deaths is a larger topic in computer games.
Ideally, a game like this would be really challenging and have both random
layout and random course of action, yet still be winnable with perfect
play. This goal seems out of reach. Thus, computer games can be soft in the
sense that optimal play ensures a win. Apart from puzzles, though, this
means that the game is solved from the outset; this is where the lack of a
human game-master is obvious. Alternatively, they can be hard in the sense
that unavoidable deaths can occur. We feel that the latter choice provides
much more fun in the long run.


"This goal seems out of reach."

This is the part I disagree with. I think the goal is already 99% of the way there and only need 1% to complete. Crawl is very well balanced in terms of difficulty.

Why should every game be winnable? For many reasons. Because the point of playing a game (other than fun) is to win. Because a well designed game always have a path for the player to do this. Because games where you can always win (even if very rare or very hard) are a true test of skill, and not merely luck. Because it may someday result in really fucking impressive streak.

Halls Hopper

Posts: 77

Joined: Tuesday, 15th May 2012, 10:12

Post Friday, 19th February 2016, 23:09

Re: It should be theoretically possible to ascend any charac

And we are back into Optimal Player startegy topic once again. I'm getting sick and tired of this.


Having completely winnable games means having always a path in the decision tree that leads to win. Unfortunetly that means reducing RNG involvement ( and therefore allowing player to take that path on his own) into game hence makes game less exciting.
Last edited by carbonbasedlifeform on Friday, 19th February 2016, 23:13, edited 1 time in total.
What about the forests?
Nope

For this message the author carbonbasedlifeform has received thanks: 3
nago, Sar, Turukano

Crypt Cleanser

Posts: 714

Joined: Saturday, 5th December 2015, 06:56

Post Friday, 19th February 2016, 23:11

Re: It should be theoretically possible to ascend any charac

duvessa wrote:Now that the die-in-abyss-on-turn-0 bug is fixed, do you have evidence that this is not already the case?


It notoriously difficult to prove a negative. I welcome an argument on the subject though, but, unlucky shaft death seem logically pretty hard to avoid. Not enough perma food to wait for RNG to generate monster to step on all tile.

Crypt Cleanser

Posts: 714

Joined: Saturday, 5th December 2015, 06:56

Post Friday, 19th February 2016, 23:13

Re: It should be theoretically possible to ascend any charac

carbonbasedlifeform wrote:And we are back into Optimal Player startegy topic once again. I'm getting sick and tired of this.


Why? Optimal strategy discussion is bad when it leads to not-fun game addition. If you have some argument for why the very small number of unavoidable death provide more fun than the already plentiful just-barely-avoidable death opportunities, please state.

I personally, am sick of forum member stating what they are sick of. Everyone sick of something here. You cant make any thread without atleast 1 guy saying, "Not this again."

Solution to being sick and tired of bieng sick and tired: stop being sick to begin with. Just my 2c.

Crypt Cleanser

Posts: 714

Joined: Saturday, 5th December 2015, 06:56

Post Friday, 19th February 2016, 23:16

Re: It should be theoretically possible to ascend any charac

CanOfWorms wrote:Where's this list of over 50% of the allowable combos that can't win, anyway.


Any combo CAN win, the question is, can they win 100% of the time if played perfectly?

Can ANY of the classes do this? Even with shaft, early adder, and early centaur?
User avatar

Abyss Ambulator

Posts: 1249

Joined: Sunday, 18th September 2011, 02:11

Post Friday, 19th February 2016, 23:18

Re: It should be theoretically possible to ascend any charac

My feeling is that Crawl has already moved close enough in this direction to have fulfilled its goals, as evidenced by 30+ game winstreaks by several players. Any further changes have more potential to actively reduce Crawl's difficulty in undesirable ways than to substantially improve the odds of winning every game.

Dungeon Master

Posts: 625

Joined: Thursday, 23rd October 2014, 03:08

Post Friday, 19th February 2016, 23:20

Re: It should be theoretically possible to ascend any charac

lethediver wrote:Why should every game be winnable? For many reasons. Because the point of playing a game (other than fun) is to win.

Well known popular game Animal Crossing, known for its deep victory conditions.

lethediver wrote:Because games where you can always win (even if very rare or very hard) are a true test of skill, and not merely luck.

In what sense is it a true test of skill, and not a true test of being able to read spoilers.

lethediver wrote:Because it may someday result in really fucking impressive streak.

Apparently a 43 game streak is not "really fucking impressive".

lethediver wrote:Can ANY of the classes do this? Even with shaft, early adder, and early centaur?

Berserker.

Besides, the only real way to guarantee that it's always possible to win is to either remove the RNG, because as long as the RNG exists the hypothetical miss every attack against the first enemy you choose to fight while it hits for max damage every attack until you're dead is possible, or make every species spriggan speed so they can reset encounters.

For this message the author CanOfWorms has received thanks:
MIC132

Ziggurat Zagger

Posts: 4055

Joined: Tuesday, 10th January 2012, 19:49

Post Friday, 19th February 2016, 23:24

lethediver wrote:It notoriously difficult to prove a negative.

I think it's actually your side that is the easy one, here. All you have to do is find one single example of an unwinnable game. The negative here is "no unwinnable games exist".

Anyway, I think it is probably theoretically possible to win every character now that turn-0 death is gone. I think it should be possible to rng-manipulate d:1 monsters into not killing your character (since nothing can generate in los of where the player enters the dungeon), and then after that it should be very easy to rng-manipulate away deaths since you have access to lots of staircases you can use.

Unlucky shaft death is obviously always theoretically avoidable, since there is no tile in the game that can contain a shaft that a player must step on (shafts cannot generate on stairs or under the runes or orb).

But you don't seem to actually be asking about "theoretically winnable" in this sense, so perhaps you should define what you are actually asking, first.

Note that by the same argument it is basically impossible to make it so that if the player cannot do any rng manipulation (i.e. you are a human and are not savescumming) that every situation is winnable. You can quite easily "anti-rng-manipulate" many monsters into "unavoidably" killing the player early on, especially since you can also manipulate away any potentially helpful consumables.

For this message the author crate has received thanks:
Arrhythmia

Crypt Cleanser

Posts: 714

Joined: Saturday, 5th December 2015, 06:56

Post Friday, 19th February 2016, 23:27

Re: It should be theoretically possible to ascend any charac

CanOfWorms wrote:
lethediver wrote:Why should every game be winnable? For many reasons. Because the point of playing a game (other than fun) is to win.

Well known popular game Animal Crossing, known for its deep victory conditions.

lethediver wrote:Because games where you can always win (even if very rare or very hard) are a true test of skill, and not merely luck.

In what sense is it a true test of skill, and not a true test of being able to read spoilers.

lethediver wrote:Because it may someday result in really fucking impressive streak.

Apparently a 43 game streak is not "really fucking impressive".

Besides, the only real way to guarantee that it's always possible to win is to either remove the RNG, because as long as the RNG exists the hypothetical miss every attack against the first enemy you choose to fight while it hits for max damage every attack until you're dead is possible, or make every species spriggan speed so they can reset encounters.


Animal Crossing; So? Getting into high level theoretical discussion about the point of ALL video games is besides the point. It is clear DCSS fall into the "point of game is to win" category of games, and not endless grinding game. You have a document from the dev team stating that, making the game theoretically always possible to win is admirable goal, so Im not sure why you are overcomplicating things. I have no patience for argument just to argue, if you want to do that, find a dev and get them to change their stance. Until then I am immovable.

True test of skill and not reading spoilers -- not sure your point. Skill include many things. Reading. Learning from experience. Striving to not die again to same thing twice. You dont need to read, you can learn many ways. But, when you learn, and yet death still not avoidable, its no longer a matter of skill. When it is possible, then skill is involved. Fairly simple.

43 is very impressive, but why not more? Stop twisting words. I am not insulting anyones achievements. Why focus on the one area where you can make that implication, and ignore all else?

For this message the author lethediver has received thanks: 3
duvessa, Shard1697, zxc23

Ziggurat Zagger

Posts: 8786

Joined: Sunday, 5th May 2013, 08:25

Post Friday, 19th February 2016, 23:28

Re: It should be theoretically possible to ascend any charac

lethediver wrote:
duvessa wrote:Now that the die-in-abyss-on-turn-0 bug is fixed, do you have evidence that this is not already the case?


It notoriously difficult to prove a negative.
No, you are being asked to prove a positive: that unavoidable deaths exist in DCSS. All you need to do is find one example of an unavoidable death.
lethediver wrote:Even with shaft, early adder, and early centaur?
By the time any of these things can actually occur, the possible states already exceed the number of atoms in the universe, making the chance of the game actually being unwinnable essentially 0 (and uncomputable anyway). You need to look much earlier (game that consistently ends in a single-digit number of turns) to find a demonstrably unwinnable game. For example, this game was unwinnable and it is trivial to prove this since there was no entropy at all: it was over before the player could take any action at all. However, the bug that caused this game was fixed.

edit: my god how did you guys make 4 more posts while i was typing this

Halls Hopper

Posts: 77

Joined: Tuesday, 15th May 2012, 10:12

Post Friday, 19th February 2016, 23:30

Re: It should be theoretically possible to ascend any charac

CanOfWorms wrote:Besides, the only real way to guarantee that it's always possible to win is to either remove the RNG, because as long as the RNG exists the hypothetical miss every attack against the first enemy you choose to fight while it hits for max damage every attack until you're dead is possible, or make every species spriggan speed so they can reset encounters.


Thats what I'm talking about.

Rng is involved into almost everything ( loot/dungeon/monster generation, shafts placement, types of monsters which spawn attached to stairs,damage, accuracy and so on). Proposing DCSS to be a completely winnable game naturally pairs with RNG reduction and we will have roguelike chess game ( note that even chess or less complex games are not solved to always have an optimal strategy).
What about the forests?
Nope

Crypt Cleanser

Posts: 714

Joined: Saturday, 5th December 2015, 06:56

Post Friday, 19th February 2016, 23:37

Re:

crate wrote:
lethediver wrote:Unlucky shaft death is obviously always theoretically avoidable, since there is no tile in the game that can contain a shaft that a player must step on (shafts cannot generate on stairs or under the runes or orb).


I address this because I think it gets to the heart of our disagreement.

Yes. It is theoretically possible character will not die to shafts because he got lucky. That's not what I meant to convey though.

I want it to be theoretically possible that a character can play DCSS, and not die to anything that does not involve some degree of human error. If playing perfectly -- without scumming, or having magic psychic power to know which tile is trap tile - then player should be able to win.

Shafts make this impossible, because while the player would avoid them if they were lucky, it is not possible for a human being to be lucky eternally. Eventually the RNG that is life, will cause him to step on a trap. The fact that he does not have to step on specific trap tile is irrelevant, what matters is that he must step on tile that MIGHT be trap tile with no way to out-skill the RNG. His option is to step on it and pray, at least on first few floors where player not have high enough hp/stats to be invincible with perfect play yet.

Sorry if that is complicated, but you asked complicated question. Player should always be able to win if playing perfectly WITHIN the confines of having the human limitation of not knowing what is under each tile in the game. Does that make sense? perfect but not superhuman or supernaturally lucky.

Ziggurat Zagger

Posts: 4055

Joined: Tuesday, 10th January 2012, 19:49

Post Friday, 19th February 2016, 23:39

It's impossible for what you want to be the case, unless you want it to be the case that the player can basically never die. Naga, no potions generate, hits tele trap on d:2 and then lands in orc priests who smite for max damage every turn, etc. If you want that RNG to be winnable, then crawl would be completely trivial.

(It's rng-manipulation in reverse; just as it is possible to rng-manipulate every "bad" situation into a non-lethal one, it would be equally possible to "rng-manipulate" seemingly fine situations into unavoidably lethal ones. The player cannot escape the latter, just as the player cannot ensure the former.)

edit: actually, it is possible for the player to never actually deal any damage to a monster. Melee and ranged attacks can always deal 0 damage. Additionally, every spellcasting background starts with a non-0 chance of miscasting. So, I think it's pretty clear that unless you radically change how crawl works, ensuring that any character can win crawl despite the worst possible rng is quite clearly futile, since you can never kill a single monster, and by lategame there are lots of monsters that will kill the player in a single action.
Last edited by crate on Friday, 19th February 2016, 23:47, edited 2 times in total.

Dungeon Master

Posts: 625

Joined: Thursday, 23rd October 2014, 03:08

Post Friday, 19th February 2016, 23:40

Re: It should be theoretically possible to ascend any charac

lethediver wrote:It is clear DCSS fall into the "point of game is to win" category of games, and not endless grinding game. You have a document from the dev team stating that, making the game theoretically always possible to win is admirable goal

And they say they choose to take the alternative,that unavoidable deaths are more interesting than theoretical wins, so perhaps it is not so clear that DCSS falls into the "point of the game is to win" category.

lethediver wrote:True test of skill and not reading spoilers -- not sure your point. Skill include many things. Reading. Learning from experience. Striving to not die again to same thing twice. You dont need to read, you can learn many ways. But, when you learn, and yet death still not avoidable, its no longer a matter of skill. When it is possible, then skill is involved. Fairly simple.

That's kind of the definition of a spoiler though, using knowledge from previous attempts to improve on a new attempt. There are in fact games that are designed to be beatable by skill instead of knowing all the intricacies because everything is presented up front, e.g. Sudoku.

lethediver wrote:43 is very impressive, but why not more? Stop twisting words.

How am I twisting words when you literally say "it may someday result in really fucking impressive streak". It's already possible to get impressive streaks. The only thing in making wins "theoretically possible" is undervalue these kind of streaks.

Crypt Cleanser

Posts: 714

Joined: Saturday, 5th December 2015, 06:56

Post Friday, 19th February 2016, 23:43

Re: It should be theoretically possible to ascend any charac

CanOfWorms wrote:
lethediver wrote:It is clear DCSS fall into the "point of game is to win" category of games, and not endless grinding game. You have a document from the dev team stating that, making the game theoretically always possible to win is admirable goal

And they say they choose to take the alternative,that unavoidable deaths are more interesting than theoretical wins, so perhaps it is not so clear that DCSS falls into the "point of the game is to win" category.

lethediver wrote:True test of skill and not reading spoilers -- not sure your point. Skill include many things. Reading. Learning from experience. Striving to not die again to same thing twice. You dont need to read, you can learn many ways. But, when you learn, and yet death still not avoidable, its no longer a matter of skill. When it is possible, then skill is involved. Fairly simple.

That's kind of the definition of a spoiler though, using knowledge from previous attempts to improve on a new attempt. There are in fact games that are designed to be beatable by skill instead of knowing all the intricacies because everything is presented up front, e.g. Sudoku.

lethediver wrote:43 is very impressive, but why not more? Stop twisting words.

How am I twisting words when you literally say "it may someday result in really fucking impressive streak"


Saying that it may result in really impressive streak someday does not mean that all streaks up to this point in time have not been impressive. Saying that event A may lead to event B does not negatively reflect on all previous events similar to B. Not unless you are making slight leap in logic in order to paint me as a jerk, because that is your goal.

You are arguing in bad faith, so I no longer feel obligated to respond.

For this message the author lethediver has received thanks: 3
duvessa, Shard1697, zxc23

Dungeon Master

Posts: 625

Joined: Thursday, 23rd October 2014, 03:08

Post Friday, 19th February 2016, 23:46

Re: It should be theoretically possible to ascend any charac

Well, you're arguing if you do A then you might get B, but B is already possible without actually doing A so why bother doing A in the first place, unless you also add the premise that B is not possible without A.

Crypt Cleanser

Posts: 714

Joined: Saturday, 5th December 2015, 06:56

Post Friday, 19th February 2016, 23:46

Re:

crate wrote:It's impossible for what you want to be the case, unless you want it to be the case that the player can basically never die. Naga, no potions generate, hits tele trap on d:2 and then lands in orc priests who smite for max damage every turn, etc. If you want that RNG to be winnable, then crawl would be completely trivial.

(It's rng-manipulation in reverse; just as it is possible to rng-manipulate every "bad" situation into a non-lethal one, it would be equally possible to "rng-manipulate" seemingly fine situations into unavoidably lethal ones. The player cannot escape the latter, just as the player cannot ensure the former.)


This is why I dont believe it should be possible with every race/class.

Can you go into more depth on your 2nd paragraph? What are you defining as rng-manipulation?

Crypt Cleanser

Posts: 714

Joined: Saturday, 5th December 2015, 06:56

Post Friday, 19th February 2016, 23:47

Re: It should be theoretically possible to ascend any charac

CanOfWorms wrote:Well, you're arguing if you do A then you might get B, but B is already possible without actually doing A so why bother doing A in the first place, unless you also add the premise that B is not possible without A.


Because more of B is good. You have constructed very binary scenario. Either i do not post about making win streak more theoretically possible, or i am shaming every win streak to exist thusfar. Bad premise, therefore, bad discussion for me to continue in.

For this message the author lethediver has received thanks:
duvessa

Vestibule Violator

Posts: 1601

Joined: Sunday, 14th July 2013, 16:36

Post Friday, 19th February 2016, 23:48

Re: It should be theoretically possible to ascend any charac

lethediver wrote:Why should every game be winnable?

No need to ask this; I (and I presume most of the forum) would agree that all other things being equal, it would be best if every game was winnable.

But that's not the whole story; making changes to achieve that goal usually doesn't keep all other things equal. So one also needs to ponder is the question "Why should every game be winnable, given the sacrifices needed to achieve that?"

Usually, trying to achieve one aim (removing unwinnable games) to the exclusion of all other concerns is a bad idea, especially when taken to the extreme.

---

This reply is not meant to imply any particular judgement of any specific change being good or bad.

Dungeon Master

Posts: 625

Joined: Thursday, 23rd October 2014, 03:08

Post Friday, 19th February 2016, 23:49

Re: It should be theoretically possible to ascend any charac

Well it's clearly possible to get more streaks; There's a current active streak of 30 wins. The only thing making streaks more common would accomplish is undervaluing a streak.
Last edited by CanOfWorms on Friday, 19th February 2016, 23:49, edited 1 time in total.

Ziggurat Zagger

Posts: 4055

Joined: Tuesday, 10th January 2012, 19:49

Post Friday, 19th February 2016, 23:49

Re: It should be theoretically possible to ascend any charac

See my edit above. It's entirely possible that the player gets rng such that he is never able to deal any damage to a monster at all. In fact, this lets you more-or-less prove that it is impossible to perfectly ensure (without being able to do rng-manipulation yourself) a win with any race/class combination.

Tomb Titivator

Posts: 885

Joined: Sunday, 28th June 2015, 14:44

Post Saturday, 20th February 2016, 00:45

Re: It should be theoretically possible to ascend any charac

I'll point out that "having 100% probability of winning a character with optimal play" and "having >0% probability of winning a character with optimal play" both fall into the category of "possible to win this character."

Blades Runner

Posts: 546

Joined: Saturday, 7th May 2011, 02:43

Post Saturday, 20th February 2016, 01:36

Re: It should be theoretically possible to ascend any charac

crate wrote:See my edit above. It's entirely possible that the player gets rng such that he is never able to deal any damage to a monster at all.

This is true, but perhaps shouldn't be (assuming enough attempts at dealing damage, and an actual theoretical ability to deal nonzero damage). Dice rolls are a decent way of constraining / loosening statistical variance, but if need be, there are methods to constrain statistical variance more strictly.

Halls Hopper

Posts: 77

Joined: Tuesday, 15th May 2012, 10:12

Post Saturday, 20th February 2016, 02:32

Re: It should be theoretically possible to ascend any charac

savageorange wrote:
crate wrote:See my edit above. It's entirely possible that the player gets rng such that he is never able to deal any damage to a monster at all.

This is true, but perhaps shouldn't be (assuming enough attempts at dealing damage, and an actual theoretical ability to deal nonzero damage). Dice rolls are a decent way of constraining / loosening statistical variance, but if need be, there are methods to constrain statistical variance more strictly.


Just for reference

By the way, there are many 'other' things which fell into same category thus making the proposal delusional

-random teleportation,shaft traps ( which put player into potentially no escape situation)
-random monsters attached to stairs which can potentially 1-hit any character upon entering new level ( paralyze involved for example)
-random monsters appearing within LOS ( such as centaurs) in the situation where you have no teleportation/blink and there are the staircases are too far away to successfully escape

etc...
etc...
What about the forests?
Nope

Vestibule Violator

Posts: 1601

Joined: Sunday, 14th July 2013, 16:36

Post Saturday, 20th February 2016, 02:47

Re: It should be theoretically possible to ascend any charac

savageorange wrote:
crate wrote:See my edit above. It's entirely possible that the player gets rng such that he is never able to deal any damage to a monster at all.

This is true, but perhaps shouldn't be (assuming enough attempts at dealing damage, and an actual theoretical ability to deal nonzero damage). Dice rolls are a decent way of constraining / loosening statistical variance, but if need be, there are methods to constrain statistical variance more strictly.

... and they have their own drawbacks; it really isn't enough to just say "do something along these lines".

You can't guarantee a player can ever hit a minster unless you either remove misses, or you make the to-hit rolls not independent. If you do the latter, some pitfalls are:
  • If the dependence is strong enough to have an observable and occasionally predictable effect, players will game the system (and doing so is usually tedious and un-fun)
  • If the dependence is too weak, then it won't affect anyone's gaming experience, so it wouldn't be worth implementing

You can reduce the variance, but Crawl is intentionally a high-variance game. The developers explicitly do not want to narrow things down enough so players can do things like reason out "It needs 4.0-6.0 auts to kill me, but I can kill it in 2.1-3.5 auts, so it's safe to fight".

Blades Runner

Posts: 546

Joined: Saturday, 7th May 2011, 02:43

Post Saturday, 20th February 2016, 03:51

Re: It should be theoretically possible to ascend any charac

carbonbasedlifeform wrote:Just for reference

Any statistical property desired can be implemented. This includes forcibly eliminating outliers. The idea that the Infinite Monkey Theorem is relevant is based on the false idea that 'random' rolls have to be completely RNG dependent.

Hurkyl wrote:You can reduce the variance, but Crawl is intentionally a high-variance game. The developers explicitly do not want to narrow things down enough so players can do things like reason out "It needs 4.0-6.0 auts to kill me, but I can kill it in 2.1-3.5 auts, so it's safe to fight".


An example is a dice roll where the distribution changes over time according to whether the statistical 'history' is going in the intended direction.
Say you have 50% ((100 - 1d100) <= 50) chance to hit an enemy. and have missed 4 times consecutively. The dice roll can automatically rise (1d100 -> 2d50 -> 4d25 -> 5d20 -> 10d10 ..) over time as the game notes these consecutive misses. You can even completely flatten it (100d1) if need be.

Though I admit I was thinking more of things that are renowned for annoying unreliability, like acquirement and troves, rather than outliers in general gameplay, like missing a monster 10 times in a row.

Crypt Cleanser

Posts: 746

Joined: Thursday, 5th December 2013, 04:01

Post Saturday, 20th February 2016, 04:30

Re: It should be theoretically possible to ascend any charac

I feel like the current discussion's gotten very abstract. I feel like discussing the idea of theoretically making inevitable death impossible in this game won't get us anywhere, and maybe we're better off discussing specific mechanics that some believe cause inevitable death.

Specifically: 99% of the time I see a discussion of inevitable deaths occur here, someone mentions D1 adders, or sometimes jackals or gnolls. It seems that, if unwinnable games exist, those are clearly one of the main causes. So I feel like the obvious question here is: Why are adders, jackals, and gnolls allowed to spawn on D1 if they're one of the biggest causes of inevitable death in the game? What do they add that is worth the "unwinnable" games they cause?

I feel like a discussion of this question will help a lot. Even if they stay in the game, I think it would be good to hear someone defend them, since I only ever see them get mentioned in this context.

The main other topics I see come up when it comes to "unwinnable games" are ones that have, for the most part, already been discussed to death (particularly traps and banishment).

Ziggurat Zagger

Posts: 8786

Joined: Sunday, 5th May 2013, 08:25

Post Saturday, 20th February 2016, 04:34

Re: It should be theoretically possible to ascend any charac

Quazifuji wrote:Specifically: 99% of the time I see a discussion of inevitable deaths occur here, someone mentions D1 adders, or sometimes jackals or gnolls. It seems that, if unwinnable games exist, those are clearly one of the main causes.
it only seems that way to people who are bad at the game
i have asked for, and seen other people ask for, actual examples of D:1 adders, jackals, gnolls, hobgoblins, goblins, etc. causing an unavoidable death
none of us have ever gotten a single one

For this message the author duvessa has received thanks:
Lasty

Vestibule Violator

Posts: 1601

Joined: Sunday, 14th July 2013, 16:36

Post Saturday, 20th February 2016, 05:12

Re: It should be theoretically possible to ascend any charac

duvessa wrote:
Quazifuji wrote:Specifically: 99% of the time I see a discussion of inevitable deaths occur here, someone mentions D1 adders, or sometimes jackals or gnolls. It seems that, if unwinnable games exist, those are clearly one of the main causes.
it only seems that way to people who are bad at the game
i have asked for, and seen other people ask for, actual examples of D:1 adders, jackals, gnolls, hobgoblins, goblins, etc. causing an unavoidable death
none of us have ever gotten a single one

To go down that aside....

I have had a Spriggan whose first action in the game was to autoexplore into the range of gnoll with a halberd. Probably should have been an OHKO. Luckily, it didn't wake up (and also luckily, the starting vault had another exit... but I was unfamiliar with it, and had a long think over whether I had better odds trying to stab the Gnoll).

I've had a SpEn get OHKO'd by a hobgoblin with a club on D:1. Avoidable, but it brings up the question of whether we are expect XL:1 characters with melee killdudes to avoid with hobgoblins with weapons (nowadays, I do avoid with XL:1 Sp's when possible). This was back when said clubs could have positive enchantment, so this can't happen anymore with Sp. It could happen with Felids, though.

I've had a few characters have their first encounter be a jackal, and have no doors to retreat to; many combos can't win this reliably. I find this especially harsh with Wz, due to the miscasts and low damage. (but last time poked at things, I convinced myself that retreating to gain mana to try to cast more Magic Darts is still preferable to engaging in untrained unarmed combat for some races)

I've had a few other similar situations but with a door you can retreat to. The jackal usually goes away after hitting 5 a couple of times, but that's not guaranteed either.

Swamp Slogger

Posts: 137

Joined: Wednesday, 14th December 2011, 16:11

Location: Australia

Post Saturday, 20th February 2016, 05:53

Re: It should be theoretically possible to ascend any charac

I've got a 31 win streak going, and I've thought a lot about this topic. At this stage, my only concern when starting a new character is I will face an unavoidable and unfair situation in the very early game. It's possible that I will make a fatal mistake later on but that will be entirely my fault, and also a lot less likely I believe.

duvessa wrote:
Quazifuji wrote:Specifically: 99% of the time I see a discussion of inevitable deaths occur here, someone mentions D1 adders, or sometimes jackals or gnolls. It seems that, if unwinnable games exist, those are clearly one of the main causes.
it only seems that way to people who are bad at the game
i have asked for, and seen other people ask for, actual examples of D:1 adders, jackals, gnolls, hobgoblins, goblins, etc. causing an unavoidable death
none of us have ever gotten a single one

I haven't gotten one but that doesn't mean they don't exist. My last GrGl fought an adder on D2 or D3 and went to half HP, which would've been more than fatal if that occurred on D1. I also came across an almost adjacent Ijyb on D:3 after turning a corner, which could've been fatal if he had a decent wand. Here is a video clip where the octopode player opens the door of the D:1 entrance vault and is greeted with two enemies including one with a venom short sword; if magic dart were miscast or failed to kill it, a single hit from the short sword would likely have been death. I also remember during the 0.16 tournament Zooty's 14 win streak was ended by an adjacent Sigmund upon entering D:2. http://dobrazupa.org/tournament/0.16/players/zooty.html http://crawl.xtahua.com/crawl/morgue/Zo ... 162219.txt . You can watch the the tv, !lg zooty dsas killer=sigmund 1 -tv . He didn't play optimally, yes, because he didn't have the blowgun wielded when entering D:2, but it would've made little difference. Perhaps going back upstairs would've been a better shot but it's still entirely up to chance.

Here are changes that I would make to heavily reduce RNG deaths in crawl. This would not make RNG deaths impossible, but the game would go from about 95% winnable with optimal play with a fairly weak combo to about 99%:

D:1

- A close look at D:1 entry vaults to remove particularly poor ones
- Weapons with brands don't spawn on D:1
- Gnolls, adders removed from D:1

D:1-3

- Tele traps and shafts removed from D:1-3
- Strong wands (hex, bolt) removed from D:1-3

Lair

- Remove zot traps from Lair (or remove them entirely)

OOD

- Minimum OOD timer extended to 2000 or 2500 turns (was 1000 turns)

New floors

- Shafts can't place the player within two spaces of an enemy, or place the player into the final floor of a branch (includes dungeon, depths)
- First-use downstairs can't place enemies adjacent to the player OR the player gets a free action to go back up without enemy action

Hexes

- MR-based paralysis takes effect after one turn and can't be removed as a status OR replace all paralysis with petrification and remove/lower AC bonus from petrification
- Paralysis length capped at four turns
- Chain confusion removed
- Erolcha spawns no earlier than D:12 / prevent first turn banishment

Xom

- Xom banishment (at all times except abandonment) made impossible before XL 12

Convenience

- Enemies wielding distortion or chaos weapons always give a warning in message log when sighted


These changes make the game easier but that is necessary to make the game less random. They are also focused on the very early game, where the overwhelming majority of RNG deaths occur. Regarding the OP, the depth for centaurs and killer bees could be pushed down one perhaps. Not sure about jackals.

crate wrote:Unlucky shaft death is obviously always theoretically avoidable, since there is no tile in the game that can contain a shaft that a player must step on (shafts cannot generate on stairs or under the runes or orb).

Unfortunately there is no facepalm smiley.
Last edited by zxc23 on Saturday, 20th February 2016, 06:10, edited 1 time in total.

For this message the author zxc23 has received thanks: 2
Hurkyl, Shard1697

Vestibule Violator

Posts: 1601

Joined: Sunday, 14th July 2013, 16:36

Post Saturday, 20th February 2016, 06:06

Re: It should be theoretically possible to ascend any charac

zxc23 wrote:D:1

- A close look at D:1 entry vaults to remove particularly poor ones

That would be nice. Regarding just the survival aspect, I hate the ones with wide visibility that usually let a half dozen or more enemies see you as you work your way out of them. I'm undecided if I hate the open ones (you can get surrounded) or the mazes (you can't try to slip away from the crowd) more.

- Gnolls, adders removed from D:1

Aren't adders, at least, restricted to random spawns and never from initial monster placement?

- Minimum OOD timer extended to 2000 or 2500 turns (was 1000 turns)

Maybe a better change is to make the grace period depend on the area of the map; my only recollection of 1000 turns being too short is for very large maps that can easily take thousands of turns to explore, especially if you need to rest a lot.

Ziggurat Zagger

Posts: 8786

Joined: Sunday, 5th May 2013, 08:25

Post Saturday, 20th February 2016, 06:37

Re: It should be theoretically possible to ascend any charac

zxc23 wrote:I've got a 31 win streak going, and I've thought a lot about this topic. At this stage, my only concern when starting a new character is I will face an unavoidable and unfair situation in the very early game. It's possible that I will make a fatal mistake later on but that will be entirely my fault, and also a lot less likely I believe.

duvessa wrote:
Quazifuji wrote:Specifically: 99% of the time I see a discussion of inevitable deaths occur here, someone mentions D1 adders, or sometimes jackals or gnolls. It seems that, if unwinnable games exist, those are clearly one of the main causes.
it only seems that way to people who are bad at the game
i have asked for, and seen other people ask for, actual examples of D:1 adders, jackals, gnolls, hobgoblins, goblins, etc. causing an unavoidable death
none of us have ever gotten a single one

I haven't gotten one but that doesn't mean they don't exist. My last GrGl fought an adder on D2 or D3 and went to half HP, which would've been more than fatal if that occurred on D1. I also came across an almost adjacent Ijyb on D:3 after turning a corner, which could've been fatal if he had a decent wand. Here is a video clip where the octopode player opens the door of the D:1 entrance vault and is greeted with two enemies including one with a venom short sword; if magic dart were miscast or failed to kill it, a single hit from the short sword would likely have been death. I also remember during the 0.16 tournament Zooty's 14 win streak was ended by an adjacent Sigmund upon entering D:2. http://dobrazupa.org/tournament/0.16/players/zooty.html http://crawl.xtahua.com/crawl/morgue/Zo ... 162219.txt . You can watch the the tv, !lg zooty dsas killer=sigmund 1 -tv . He didn't play optimally, yes, because he didn't have the blowgun wielded when entering D:2, but it would've made little difference. Perhaps going back upstairs would've been a better shot but it's still entirely up to chance.
I see we are using wildly different definitions of "unavoidable death". I am taking it to mean "there is no sequence of player inputs for this game that would result in a win." You seem to take it to mean "any death that I think is unfair".

For this message the author duvessa has received thanks:
ydeve
User avatar

Slime Squisher

Posts: 338

Joined: Wednesday, 20th November 2013, 11:37

Post Saturday, 20th February 2016, 07:37

Re: It should be theoretically possible to ascend any charac

lethediver wrote:it should be theoretically, but maybe not humanly, possible to win at least a 3 rune in every game of DCSS without being struck down by RNG.

But it is already possible!
Proof

Swamp Slogger

Posts: 137

Joined: Wednesday, 14th December 2011, 16:11

Location: Australia

Post Saturday, 20th February 2016, 07:47

Re: It should be theoretically possible to ascend any charac

duvessa wrote:I see we are using wildly different definitions of "unavoidable death". I am taking it to mean "there is no sequence of player inputs for this game that would result in a win." You seem to take it to mean "any death that I think is unfair".


I speak out primarily against 'RNG deaths' which are deaths for which there was no sequence of player inputs given a particular situation that arose from optimal play that could have led to survival 100% of the time. If you come across D:2 Sigmund who immediately notices you, it's up to chance whether he confuses you or not. You can get lucky and escape, sure, but it's still up to chance whether or not you survive that situation. You can stack the odds in your favour (manual exploring heavily utilising the stairs, hardcore luring, paying attention to shouts and noise) but it won't be 100% or anywhere near that. Getting rid of all RNG deaths is perhaps unrealistic for a game like this, but we can certainly do a lot better than now with just a few changes that don't even affect 95% of the game.

Ziggurat Zagger

Posts: 8786

Joined: Sunday, 5th May 2013, 08:25

Post Saturday, 20th February 2016, 08:13

Re: It should be theoretically possible to ascend any charac

But you could have taken different actions on D:1 to generate a different D:2, and different actions on D:2 to prevent Sigmund from successfully confusing you.

If you disallow RNG manipulation, well, crate already explained why that necessarily leads to unavoidable deaths.

Autoexplore deaths are stupid, obviously, but they are not the same as unavoidable deaths and they do not have similar solutions either.

Swamp Slogger

Posts: 137

Joined: Wednesday, 14th December 2011, 16:11

Location: Australia

Post Saturday, 20th February 2016, 08:29

Re: It should be theoretically possible to ascend any charac

duvessa wrote:But you could have taken different actions on D:1 to generate a different D:2, and different actions on D:2 to prevent Sigmund from successfully confusing you.

This is the most ridiculous argument I've ever seen on this forum. You're suggesting optimal play involves savescumming or a degree of psychic ability on the part of the player.

For this message the author zxc23 has received thanks: 3
Blade, Hurkyl, Shard1697

Dungeon Master

Posts: 625

Joined: Thursday, 23rd October 2014, 03:08

Post Saturday, 20th February 2016, 08:43

Re: It should be theoretically possible to ascend any charac

Well one issue with fixes like "Shafts don't place you within two spaces from a monster" is that due the RNG's involvement in generation you can construct extremely implausible lethal situations (this shaft places you in view of 10 orc priests by some freak RNG) that can still occur and I doubt there's a clean way of excising these scenarios from the game without adding special-cases in the game generation, which would make it harder to maintain.

Ziggurat Zagger

Posts: 8786

Joined: Sunday, 5th May 2013, 08:25

Post Saturday, 20th February 2016, 08:49

Re: It should be theoretically possible to ascend any charac

zxc23 wrote:
duvessa wrote:But you could have taken different actions on D:1 to generate a different D:2, and different actions on D:2 to prevent Sigmund from successfully confusing you.

This is the most ridiculous argument I've ever seen on this forum. You're suggesting optimal play involves savescumming or a degree of psychic ability on the part of the player.
The best play possible would surely involve some level of RNG manipulation, would it not?

Swamp Slogger

Posts: 137

Joined: Wednesday, 14th December 2011, 16:11

Location: Australia

Post Saturday, 20th February 2016, 09:28

Re: It should be theoretically possible to ascend any charac

CanOfWorms wrote:Well one issue with fixes like "Shafts don't place you within two spaces from a monster" is that due the RNG's involvement in generation you can construct extremely implausible lethal situations (this shaft places you in view of 10 orc priests by some freak RNG) that can still occur and I doubt there's a clean way of excising these scenarios from the game without adding special-cases in the game generation, which would make it harder to maintain.


This is why I say it's unrealistic to eliminate all RNG deaths. It may be impossible to prevent all RNG deaths but that's not an argument against reducing their number, which we can do quite easily and I suggested a few such changes above.

Crypt Cleanser

Posts: 714

Joined: Saturday, 5th December 2015, 06:56

Post Saturday, 20th February 2016, 10:47

Re: It should be theoretically possible to ascend any charac

duvessa wrote:I see we are using wildly different definitions of "unavoidable death". I am taking it to mean "there is no sequence of player inputs for this game that would result in a win." You seem to take it to mean "any death that I think is unfair".


For what it is worth, I agree with 31 win streak guy's assessment of unavoidable death more so than yours. The point of my thread here is to eliminate death that a non psychic, non magician player could not have avoided. If like you, I assume the player has developed psychic power and magically avoids shafts even tho he didnt know where they are, then yes, I have no reason to post.

But, I am not operating under that assumption because it doesn't reflect real life.

You, and anyone else who argue that this thread has no point because IF player was psychic or a magician, he could have avoided death, seems to be besides the point.

Crypt Cleanser

Posts: 714

Joined: Saturday, 5th December 2015, 06:56

Post Saturday, 20th February 2016, 10:51

Re: It should be theoretically possible to ascend any charac

duvessa wrote:
zxc23 wrote:
duvessa wrote:But you could have taken different actions on D:1 to generate a different D:2, and different actions on D:2 to prevent Sigmund from successfully confusing you.

This is the most ridiculous argument I've ever seen on this forum. You're suggesting optimal play involves savescumming or a degree of psychic ability on the part of the player.
The best play possible would surely involve some level of RNG manipulation, would it not?

I think i should point back to the definition in my OP that states "humanly possible" and not "possible for alien or comic book hero that has power to see inside computer circuit and know which tile is trap, or which moment to press key so RNG fires correctly."

Crypt Cleanser

Posts: 714

Joined: Saturday, 5th December 2015, 06:56

Post Saturday, 20th February 2016, 10:53

Re: It should be theoretically possible to ascend any charac

CanOfWorms wrote:Well one issue with fixes like "Shafts don't place you within two spaces from a monster" is that due the RNG's involvement in generation you can construct extremely implausible lethal situations (this shaft places you in view of 10 orc priests by some freak RNG) that can still occur and I doubt there's a clean way of excising these scenarios from the game without adding special-cases in the game generation, which would make it harder to maintain.


You know, looking over the whole thread, I think i am ok with this.

I think i should alter my initial statement. Maybe getting rid of all unavoidable death is not possible without an unreasonable amount of tweaks to the game. However, i would be satisifed with simply making a confined set of small changes to early game only that would make them several orders of magnitude less likely.

For this message the author lethediver has received thanks: 2
chequers, zxc23
User avatar

Zot Zealot

Posts: 991

Joined: Monday, 15th April 2013, 15:10

Location: Augsburg, Germany

Post Saturday, 20th February 2016, 11:24

lethediver wrote:"This goal seems out of reach."

This is the part I disagree with. I think the goal is already 99% of the way there and only need 1% to complete.

lethediver, you are quite new to crawl and to this forum. Please consider that most people who answered to your topic have much more experience than you. (This is not meant to offend you at all. Should you feel offended nevertheless: please read this post again in two or three years.)

You can't remove unavoidable deaths entirely without reducing RNG too much, this has been said several times. carbonbasedlifeform made an excellent comment about this.

carbonbasedlifeform wrote:Having completely winnable games means having always a path in the decision tree that leads to win. Unfortunetly that means reducing RNG involvement ( and therefore allowing player to take that path on his own) into game hence makes game less exciting.

But imo it's interesting how this discussion went on. You started with adders / centaurs / shafts being the main problems in early game, zxc23 elaborated your ideas with more precision. This topic could result in reducing unavailable deaths in early games.

*******

And another thing: if you have problems with duvessa's posts either ignore them or put duvessa on your foe list.

Dungeon Master

Posts: 3618

Joined: Thursday, 23rd December 2010, 12:43

Post Saturday, 20th February 2016, 11:32

Re: It should be theoretically possible to ascend any charac

lethediver wrote:
Sar wrote:Can you provide any pro arguments, apart from your feeling? I mean, that might sound like a stupid answer, but Crawl manual actually makes an argument that supports the existence of unavoidable deaths as a feature of Crawl. Here's that section:
  Code:
The possibility of unavoidable deaths is a larger topic in computer games.
Ideally, a game like this would be really challenging and have both random
layout and random course of action, yet still be winnable with perfect
play. This goal seems out of reach. Thus, computer games can be soft in the
sense that optimal play ensures a win. Apart from puzzles, though, this
means that the game is solved from the outset; this is where the lack of a
human game-master is obvious. Alternatively, they can be hard in the sense
that unavoidable deaths can occur. We feel that the latter choice provides
much more fun in the long run.


"This goal seems out of reach."

This is the part I disagree with. I think the goal is already 99% of the way there and only need 1% to complete. Crawl is very well balanced in terms of difficulty.
I wrote that paragraph quoted from the manual and you do it no service by quoting only half of it: it is trivial to make every game winnable. The problem is to keep up the challenge.

Why should every game be winnable? For many reasons. Because the point of playing a game (other than fun) is to win.
Crawl has a score, so achieving the best score for a very tough seed works for me.

Because a well designed game always have a path for the player to do this.
How do you expect to get away with this claim? Chess and Go are arguably well designed games and have no-win states (draw in Chess; for Go, jigo with even komi, or triple ko). Crawl is a single player game, but is has randomly generated content. Ensuring wins means we have to cut off a huge part of the evil end of the random tail... and in my opinion, that's the most interesting part of the game you would lose.

For concrete examples: games where the RNG throws OOD after OOD at you, or which are resource starved, e.g. the first potion of heal wounds is on D:12, and so on. Yes, many players will perish, but this is precisely the opportunity for really good players to shine.
You argue to remove the most dangerous early game killers, which is something different. However, I object to your thrust: instead of artifically removing dangers from the game, we should make sure that later parts of the game become more dangerous, so provide opportunities for good play to shine.

Because games where you can always win (even if very rare or very hard) are a true test of skill, and not merely luck. Because it may someday result in really fucking impressive streak.
This is the worst point in your list, in my opinion. As an ad-hoc definition of the depth of Crawl, I take the maximum winrate over all species/background combinations, assuming optimal play, running over all players. It's a number between 0 and 1, and quite close to 1. You are argue it should be 1.
(This is inspired by the notion of depth for abstract board games which applies to Chess, Go and the like.)

The crucial bit is: making every game winnable reduces depth. I am very much interested in a deep game, so I don't want that. The puzzles mentioned in the pas-de-faq have very little depth (I know people who argue that these are not actually games, but that is another issue.)

In concrete terms: I want to be able to distinguish mediocre, good, very good and excellent players. This is not possible if you lower the risk threshold too much. (I'd argue it is too low in current Crawl.)

In my opinion, we should actually strive to make the game harder, streaks be damned. I don't know what a good winrate one should aim for (say among viable species/background combinations), but I am certain is should be more like 80% rather than 100%. If every game is winnable at the outset, you forfeit a big chunk of potential depth.

For this message the author dpeg has received thanks: 4
Blade, Sar, Shard1697, ydeve

Crypt Cleanser

Posts: 724

Joined: Tuesday, 29th November 2011, 11:04

Post Saturday, 20th February 2016, 12:03

Re: It should be theoretically possible to ascend any charac

Crawl is a game you play alone, not against someone so I don't see a reason Crawl has to be 'fair' or 'winnable'. Every race/class combo is winnable if you're lucky, but losing is part of the fun unless you care about useless things like score or winrate.
"Damned, damned be the legions of the damned..."

Swamp Slogger

Posts: 137

Joined: Wednesday, 14th December 2011, 16:11

Location: Australia

Post Saturday, 20th February 2016, 12:10

Re: It should be theoretically possible to ascend any charac

dpeg wrote:For concrete examples: games where the RNG throws OOD after OOD at you, or which are resource starved, e.g. the first potion of heal wounds is on D:12, and so on. Yes, many players will perish, but this is precisely the opportunity for really good players to shine.
You argue to remove the most dangerous early game killers, which is something different. However, I object to your thrust: instead of artifically removing dangers from the game, we should make sure that later parts of the game become more dangerous, so provide opportunities for good play to shine.

I agree that often RNG results in interesting situations and opportunities for good players to shine. Your hw example is a good one. I also agree with making later parts of the game tactically harder, and I like many of the enemies that have been added of late.

However:

dpeg wrote:The crucial bit is: making every game winnable reduces depth. I am very much interested in a deep game, so I don't want that. The puzzles mentioned in the pas-de-faq have very little depth (I know people who argue that these are not actually games, but that is another issue.)


Do adders and gnolls on D1 really provide depth? What about a D1 kobold with a distortion dagger waiting outside the entrance vault? The changes I would like to see are focused on D1-3, which is both the most shallow portion of the game and the most RNG. The deepest and most sophisticated tactics on these early floors are in fact luring, pillar-dancing and stair-dancing, which are the very things you've said you want to fix in the luring thread.

For this message the author zxc23 has received thanks:
Rast
Next

Return to Game Design Discussion

Who is online

Users browsing this forum: No registered users and 32 guests

cron
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group.
Designed by ST Software for PTF.