Sil on Roguelike Radio

[Mikko Lehtinen]

V is deterministic, so that it's not difficult to estimate the danger of any situations once you know what you're doing. I'm not sure Sil is actually harder in this regard, but it does force you to descend at times when you may not have found optimal gear yet. So even though you might be able to calculate the odds of risky moves and find them not to your liking, you don't have much of a choice in the matter.

I prefer games like Sil where you can't avoid the danger, but where you can choose between different strategies and tactics, some riskier than others.

I can't bring myself to see Angband's choice of slow descend versus diving as a strategic choice. For me it's as much a strategic choice as choosing to pay $5 real life money for getting better equipment for my character. In Angband I'm not paying money, but I'm trading my time and enjoyment for better chances of survival (I enjoy it more when it's genuinely dangerous). That doesn't feel like a strategic choice for me. It's a meta choice, happening outside of the game. (OK, I'm exaggerating a little, some of it is strategic. )

[Mikko Lehtinen]

I always play with just one life in ToME. Same in other games where death is kinda optional (like PrincessRL). It's just not the same for me without proper permadeath! But I understand the desire of others to not have to repeat so much or play so carefully.

As a designer, I prefer multiple lives because then you can raise the difficulty level! For me the most fun moments in roguelikes are near-death experiences, where you survive only because of good tactics and some luck. With only one life it's much harder to design a roguelike that repeatably produces these kind of situations.

But really, lives are just a different kind of hit points. Would this mechanic feel better to you: There are two kinds of hit points, called Life and Stamina. After your Stamina is lowered to zero, attacks start to drain your Life, which does not regenerate at all. Once Life is lowered to zero, you die.

[bisonbison]

Isn't this what the interface outside the combat rolls window already shows you (though the damage done only flashes up temporarily)? I only find myself looking at the rolls when something is going wrong, mostly to find out whether I've just been unlucky or if I have little or no chance to hit and do damage.

Well, if you're gonna have a combat rolls/results window, I think it makes sense to have the information cleanly presented so that anyone can understand it. Having a simple and verbose mode would let people choose their level of detail, and if it could be logged to disk or at least cached so that the last 10 or 100 turns were scrollable/browseable, it would be a lot of valuable information.

[fizzix]

As a designer, I prefer multiple lives because then you can raise the difficulty level! For me the most fun moments in roguelikes are near-death experiences, where you survive only because of good tactics and some luck. With only one life it's much harder to design a roguelike that repeatably produces these kind of situations.

I definitely agree here. It lets the designer get away with "unfair" situations. The stupidly high damage from AMHD/drolem poison breath, or since this is a Sil thread, an Ururauko melee, is less problematic if the player can learn about the danger without losing a character. Obviously, once the player learns the danger, they won't repeat it. But the goal is to keep new players from getting frustrated from a, "how was I supposed to know about that?" death. I can speak more to the V situation, and in a one-life permadeath game, I still think that 600+ poison breath is a horrible gameplay mechanic.

But also as Darren says, every player is going to prefer something slightly different. So it's probably best to provide a pure-roguelike option, a multiple-life option and maybe even an infinite-life option. Obviously the shorter the game, the less need there is for extra lives. If the whole game takes an hour, then one life is certainly reasonable even for a very unfair game.

[Darren Grey]

As a designer, I prefer multiple lives because then you can raise the difficulty level!

Hah, that's just an excuse for bad design! :P Stretched to extremes it ends up like Dark Souls, which lots of people like but is rather different from the procedural experience of roguelikes.

There's other ways to toy with near-death experiences. The main thing is keep HP always low (Sil does this well) - most certainly avoid the crazy HP and damage scaling of most *bands. You can go all the way and make a 1 hit point roguelike for the ultimate in tension

But really, lives are just a different kind of hit points. Would this mechanic feel better to you: There are two kinds of hit points, called Life and Stamina. After your Stamina is lowered to zero, attacks start to drain your Life, which does not regenerate at all. Once Life is lowered to zero, you die.

That's not the same as multiple lives, where being brought back to life normally means a restoration to full HP and potentially appearing in a different zone.

The wound system could work well. One thing that could really work well is if you had extra powers whilst your Stamina was zero. So the closer you are to death the greater your abilities are, making you feel both strong and fragile. I've got a bit of this in Rogue Rage - at low HP your Rage bar goes up far faster, giving you access to major super-powers. The player ends up teetering on the edge of death a lot, which can be exciting.

[Mikko Lehtinen]

Hah, that's just an excuse for bad design! :P

I disagree. You said previously that you don't like chance or risk management in roguelikes, because if you lose even one fight, it's game over man.

With multiple lives or some other mechanic that makes the stakes in any given fight smaller than "everything", the game designer can introduce challenges where the win/lose ratio for a good player is, say, 75/25. With only one life in a mid-length roguelike, all fights need to be winnable (for a good but not excellent player) at least 90% of the time.

Do you also think that classic platformers and shoot-em-ups with three lives are "bad design"? If not, what makes them different?

For the record, when you get wounded in Halls of Mist, hit points are restored to half of the maximum, adrenaline burst cures most negative effects, and one or two of your stats are damaged temporarily. Deeper in the dungeons some of the stat damage may be permanent.

I personally think that the main thing that separates multiple lives and a wound system is flavour. In a roguelike there needs to be an illusion of a single life, so you need to wrap multiple lives inside a mechanic that pretends to be something else. There may be other differences, too. But the main benefit of any alternative "multiple lives" system is that it allows you to lose less than everything when a fight goes badly.

[Mikko Lehtinen]

The wound system could work well. One thing that could really work well is if you had extra powers whilst your Stamina was zero. So the closer you are to death the greater your abilities are, making you feel both strong and fragile. I've got a bit of this in Rogue Rage - at low HP your Rage bar goes up far faster, giving you access to major super-powers. The player ends up teetering on the edge of death a lot, which can be exciting.

This sounds fun!

[Mikko Lehtinen]

Almost always when I get wounded in Halls of Mist I think "oh shit, that was my own fault, I could have avoided that".

Multiple lives or a wound system are definitely not needed just because they function as a "bad luck buffer"! Mostly they're needed because the tactical challenge level of the game is set so high that a human with a normal-sized brain can't cope.

If the challenge level was set lower than that, there would be much less room for skill improvement. Truly dangerous situations are fun because they force you to think. They also force you to burn your resources. Without dangerous situations the resource management subgame wouldn't really work.

[Darren Grey]

I disagree. You said previously that you don't like chance or risk management in roguelikes, because if you lose even one fight, it's game over man.

With multiple lives or some other mechanic that makes the stakes in any given fight smaller than "everything", the game designer can introduce challenges where the win/lose ratio for a good player is, say, 75/25. With only one life in a mid-length roguelike, all fights need to be winnable (for a good but not excellent player) at least 90% of the time.

Which is fine. I personally believe in the ethos of "perfect play will always win, but there's no such thing as a perfect player". Make a game hard but guaranteed (or as guaranteed as possible) to be beatable with the right choice of moves. I agree that some people find this draining across a longer game, but this can be countered with the right pacing.

Do you also think that classic platformers and shoot-em-ups with three lives are "bad design"? If not, what makes them different?

Real-time games need a failsafe for dexterity-based mistakes. In roguelikes we should be more punishing to those who make mental errors Also sometimes those lives are needed because of cheap boss tricks you can only learn of by dying to them - I dislike the use of such mechanics.

One other thing with lives systems is they can become irrelevant in a game. Often when you lose one life to a part of the game you quickly lose the next two, as you simply aren't good enough to get past that challenge. This would be especially the case in roguelikes where death is often because of mismanagement of resources or reliance on luck earlier in the game. Once you hit a bigger obstacle without growing correctly as a player you burn through those lives easily, making 3 lives the same as 1. I've seen this happen a lot with ToME players, who then blame the game instead of themselves.

A single life teaches the lesson better, I think.

In a roguelike there needs to be an illusion of a single life, so you need to wrap multiple lives inside a mechanic that pretends to be something else.

I can get behind this, and your wound mechanic works well for this. For me it spoils the immersion to have something like the ToME4 lives system. If the life/wound system is presented with its own game mechanics then it becomes a part of the game rather than something which pulls you out of it. Plus there can still be the challenging of beating the game without ever getting wounded

There may be other differences, too. But the main benefit of any alternative "multiple lives" system is that it allows you to lose less than everything when a fight goes badly.

Yeah, this is something most roguelikes do badly. FTL is quite nice for having many systems go down when partially damaged. More roguelikes need to have something more interesting than an adventurer as the main character.

[Derakon]

Regarding the lives mechanic, any thoughts on a party-based system where the game isn't over unless all party members are incapacitated? Then if your frontliner eats it to an unexpectedly powerful attack (or a status ailment you didn't know could happen, etc.), you're still in the game, albeit at reduced fighting capacity. Depending on the game, said frontliner might be rivivifiable or not. I could see things being particularly interesting if you had to have your reduced party retreat, regroup, and then try to recover the body while not having access to that character's abilities.

I know the vast majority of roguelikes only let the player control one character, for simplicity's sake, but that's not the only way things have to work.

[HallucinationMushroom]

Make a party based roguelike, and I'll nab a few extra arms and another keyboard...

Other than reminiscing on my gold-box d&d days (I let them die and left them forever), my next thought in-line is Killer 7, as you mention controlling only one player at a time. It's an odd duck, but you essentially have about 6 characters with different special abilities. The special ability of one of them (Garcian) is that he can revive the other characters should they die.

Anyway, yeah, nothing useful to add... I just like to ramble.

[Scatha]

I'll return a little to the issues with combat. I'd separate these as:

(1) Combat is high variance;
(2) It's hard to work out what the odds are.

It seems that (1) is more of a matter of taste. Having (2) pointed out, we could well look for useful ways to communicate more information to the player. I realise that (1) feeds in here: while ADOM has a less tractable combat system, variance is lower so this isn't so important.

With regard to the variance in Sil, I think that the variance in combat is moderately high but not overwhelmingly so. Almost every time I lose a character I can ascribe it to player error (or an underpowered character). Actually it's possible that the worst time for this is the start of the game, where there are fewer options and you are more dependent on finding good equipment (particularly for difficult builds).

Actually the variance can function a little like multiple lives systems: you can think of it not as stochastically letting you die when you deserve life, but stochastically letting you live when you deserve death. And by being memoryless, it doesn't induce the temptation to restart if you lose a life early. (It's also a little less clear about providing feedback to the player when they made a mistake, but at least ending up on low life is a clue.)

I think that the variance in Sil is about right (and this is dictated by flavour reasons as well as gameplay ones: it helps to give combat a feel of the dangerous). But maybe it would be worth testing a version with significantly different variance to compare! If it turned out much better we'd need to put some more thought into combat again ...

[Scatha]

I personally believe in the ethos of "perfect play will always win, but there's no such thing as a perfect player". Make a game hard but guaranteed (or as guaranteed as possible) to be beatable with the right choice of moves.

It's a reasonable ethos. Sil provably doesn't fit this, but I don't think it's very far off. I'd guess that our best players, if attempting to win, would win something like 80-90% of their games? (I could be way off here.) Of course perfect play will be better. This is closer to your ideal than it is to a lot of forms of Solitaire.

FTL is quite nice for having many systems go down when partially damaged. More roguelikes need to have something more interesting than an adventurer as the main character.

I agree with this! And I think there's plenty of design space for a good party-based roguelike (but also lots of pitfalls to be avoided).

[andrewdoull]

I'm not either, and I should have challenged him on that as it didn't make any sense to me with how the stat increase system works, as well as with how rolls can max out and strength effectivity is limited by weapon weight. I quite like how Sil's systems are balanced out - everything is kinda useful to everyone, so you always feel like you want more. This is much better than games where you have no choice, or the choice is obvious, or you always get everything you want. The difficult decisions in character building are perhaps what I enjoyed most in Sil

I made this comment primarily having played Smithing characters where the difference between not having 'improve item stats' and having it is significant.

It feels wrong to somehow use the guaranteed forge on level 2 to make basic items - but at the same time I wonder if moving this forge to the room the player starts in would address some of the criticisms that Darren makes about starting equipment.

[half]

By having variance in the turn by turn damage you end up obscuring these longer term variations. Or at least making them less relevant, and harder to plan around. Too much randomness obscures tactics, in my opinion.

Thanks. That makes sense.

I wanted to show that random damage system weigh towards longer time to kill an enemy with a set amount of HP, due to last blow overkill. In particular contrasted with a deterministic system where the player can optimise against overkill.

That also makes sense: random damage gives asymmetric time-to-death distributions. This asymmetry is much more pronounced in cases like that where you could kill it quickly (c.f. a case where you do 1d10 damage each turn and it has 1000 HP). When I was pointing out the other side of the things (that it could take less time) I therefore didn't pick a great example. Better examples would be cases where the monster is just a bit too easy and randomness makes them the right challenge level, or when they are a bit to tough and randomness makes them the right difficulty level. Of course it can do the opposite too, but I'm just pointing out that your example was a bit one-sided (you picked the case where it was the worst, but there are compensations).

However, this isn't the reason to have randomness. As Scatha has mentioned, it is more to do with wanting the game to have a certain feel. A deterministic version would seem too puzzle-like and sterile to me. I would enjoy small doses of it (like with Go or Chess) but couldn't play it for long. I'd also find it much more of an abstraction of the theme (The Silmarillion) than I planned. It wouldn't feel like walking the halls of Angband. Also, as debo said, it wouldn't feel appropriately 'epic' to me. This is clearly a matter of taste and you are really an outlier here in the roguelike community. Indeed, I think I'm in the abstract/deterministic half of the roguelike community but you are so much further in that direction. I appreciate that taste, but it really is a matter of taste.

I should also add from a skill perspective that probabilistic games are not any less skilful than deterministic ones (I'm not sure if you were implying they are). They are not any more skilful either. It is just a different approach, to do with increasing your probability of winning instead of making yourself win. Consider Backgammon and Bridge. Plenty of skill there (and very abstract). In some ways it is a bit of a shame that the probabilities all have to collapse to a win or a loss (or a death at level X...). It is nice to be able to say that you made the unique best play in Bridge that gave you a 2/3 chance of making the trick and a bit sad if you then lose that trick.

Sil is particularly vulnerable to this as it has a very low floor to damage, and the random armour makes even that floor uncertain.

The armour makes the floor = 0 in most cases, but that is already the floor from missing, so it is not that different. If I'm making characters/monsters miss too often or do no damage too often that is bad, but it would be my fault for balancing it like that, not the system's fault (i.e. it could be solved with some +melee to everyone and reducing the protection values of monsters/equipment).

In general, Sil's combat has quite a few dice rolled. However, I'm not sure from your posts if you are aware that the extra dice often *lower* the relative variance. For example, two dice are rolled to see if you hit, this makes the tails of the distribution wider in absolute terms (39 possibilities instead of 20) but makes possibilities near the middle much more likely. Similarly, instead of just rolling a set of dice for damage, I subtract off the roll of another set of dice. This lowers the relative variance. Basically having 1d20 vs 1d20 is mathematically the same as rolling 2d20-21, and having 6d4 damage vs 3d4 protection is like doing 9d4-15 (but with a hard floor of zero, further lowering relative variance).

This comes down heavily to personal opinion. I have blogged before about my great distaste towards such random mechanics.

Ah, I'd forgotten you said this when writing my comments above. Sounds like we agree about the taste issue.

Sil's combat is perhaps the worst I've seen for this pet peeve of mine in how the modifiers effects the dice rather than the end result, and the armour is random each turn rather than flat. Most roguelikes have modifiers that give +1 to overall damage instead of 50% chance of +1, which has the effect of providing more guarantees in combat. In Sil it seems the whole combat damage system is set up to maximise the thing I hate!

Yes, modifying the sides instead of a flat bonus does produce more variance. I can see why you wouldn't like that. It is quite nice in a number of other ways though, right? It makes points of strength help more with heavier weapons very naturally and means that everything is of the form XdY instead of XdY+Z.

Heh, I know ADOM isn't great. It is a lot more reliable in combat than Sil, with many of those modifiers getting simplified immensely during play - in particular the damage dice on weapons get heavily outweighed by flat modifiers. Combat is mostly very predictable. But it still has a lot of bad design in combat, and in particular a lot of overly detailed systems that end up having very minor effects on play.

Good point. I was arguing that the ADOM system is less elegant and you are pointing out it is more reliable.

If I praise ADOM it's because I've been playing it for many years and it's hard to forget your first love. You might feel the same about Angband.

I can assure you I don't! I probably like ADOM more than Angband actually. What makes ADOM shine is the world, the quests, and the special abilities that make the classes feel unique and exciting. Neither Angband nor ADOM have particularly elegant mechanics (and nor does NetHack for that matter!).


I think we've covered (and resolved!) most of our disagreements now. Thanks for the chat! I hope you have some enjoyable games of Sil, or at least look over it for features you like and copy them for your own games, or tell others about them. A lot of the fun of making it has been doing many things differently to other RPGs/Roguelikes to show how it can work. I therefore encourage other people to take the mechanics they think are improvements and bring them to other games.