Should monsters have a fixed max hp or variable max hp?

[Hurkyl]

I also think that variable HP is good. The less there are possibilities to exactly calculate optimal fighing strategies, the better.

Variable HP doesn't do anything more or less to the possibilities to exactly calculate optimal fighting strategies: it simply makes calculation more complicated.

I also like that some max-hp yaks enter clouds and some don't. It also brings up a point, are there any other behaviors that are influenced by variable max-hp? I

It's already randomized, I believe: I've had plenty of Yaks just stand next to a cloud for a couple turns before deciding to enter it.

The most significant behavior I think has been overlooked in this thread is the ability for a player to tell the difference between, say, a nearly dead Orc Priest and a moderate health Orc Priest, because a third of the HP bar could be anywhere from 3 to 7 HP.

[TeshiAlair]

Interesting Thing 1: Many of the complaints against variance are referring to early game issues, such as Super Quokka and orc priests.
Interesting Thing 2: It has seemed to be a pretty standard comment that the very early game is both harder and less interesting due to the swinginess.

Why not just reduce variability early on? Having a 9-21 hp orc priest is just an early character splatter, having an ancient lich taking longer to kill than usual may require a retreat or a change of plan, but occurs at a stage in the game when you DO have backup plans.

[Sar]

please don't break early game

[johlstei]

the early game is hard and fun

[damiac]

Removing variable monster HP breaks early game and makes it easy and unfun?

Try to have some content in your posts, not just thankbait.

How does an early 9 hp orc pries make the game more hard and fun?
Are 21 HP orc priests necessary for an 'unbroken' early game? I'd call a 21 HP orc priest on D2 'broken' myself, especially if I'm 'unspoiled' and so assuming since the last orc priest died in two decent hits, this next one probably won't take more than double the damage to go down. Especially at the part of the game where engaging an orc priest is usually a fight to the death, since it's unlikely i have any (known) escape consumables.

Would it make the entire game more fun if every monster had a small chance to do quadruple damage with no warning? Think of how cool it would be when a stone giant one hits your 150 HP character! That's hard and fun!

[neil]

Would it make the entire game more fun if every monster had a small chance to do quadruple damage with no warning? Think of how cool it would be when a stone giant one hits your 150 HP character! That's hard and fun!

Since now we're using hyperbole to argue for changes to the game: Would it make the entire game more fun if every stone giant did exactly 150 damage per hit? And if we gave each of them exactly 300 HP? No, of course not: it would make it too hard and unfun. I think that proves conclusively that fixed HP for monsters would be a bad idea.

[damiac]

My point was that to an early weak character, getting into a fight with an orc priest can sometimes be a duel to the death. If I'm making that decision to engage because the last orc priest went down in a couple of hits, and suddenly this one has double the health, it can be the same sort of surprise death as if a monster suddenly hit for much more damage than previous experience suggested was possible.

I think I made the comparison pretty clear, but you decided to just take the second paragraph out of context, so I'm explaining it in just one paragraph for you this time.

[Sandman25]

Is it possible to change morgue files to display current and max HP of the monster who killed player character?
Something like "Killed from afar by a frost giant (36 damage, 10%/130% average HP)". It would mean that the monster could be already dead if not the really high max HP.
I suspect most players die to almost dead monsters (early game).

[Sar]

it can be the same sort of surprise death as if a monster suddenly hit for much more damage than previous experience suggested was possible

this happens in current Crawl as well so

[nago]

I'm sorry getting lured by Sar's thankbait, I promise I won't when he stop to say sensible things!

[Lasty]

It has seemed to be a pretty standard comment that the very early game is both harder and less interesting

Actually, I mostly hear harder and more interesting.

I think I made the comparison pretty clear, but you decided to just take the second paragraph out of context, so I'm explaining it in just one paragraph for you this time.

How many condescension points is this worth? Are you getting close to redeeming them for a new Schwinn bike?

Just about every aspect of Crawl combat is heavily randomized. It's odd to me that this one aspect is the one that bothers you so much when just the existence of randomized damage means that an orc priest with 9 hp might go down in anywhere from 1 to 30+ damaging hits with a flail -- the high tail being the case where you score 1 damage hits separated by enough time for monster regeneration/healing to keep pace with the damage you deal. Alternately, even a 9 hp orc priest can smite you for 17 damage four times in a row thanks to spell/action randomization. If you really think each orc priest encounter should be approximately as difficult as the last one, you need to change way more than random health to achieve it.

[damiac]

I think I made the comparison pretty clear, but you decided to just take the second paragraph out of context, so I'm explaining it in just one paragraph for you this time.

How many condescension points is this worth? Are you getting close to redeeming them for a new Schwinn bike?

Worth about 1/2 of the condescension points of this one i would guess:

Would it make the entire game more fun if every stone giant did exactly 150 damage per hit? And if we gave each of them exactly 300 HP? No, of course not: it would make it too hard and unfun. I think that proves conclusively that fixed HP for monsters would be a bad idea.

I'm not talking about variable HP because it 'bothers me so much', I'm talking about it because it's 'What the topic is about'.

In fact, your example makes the point we both seem to be trying to make, crawl is already really random, so why have another layer of randomness on top of it? An observant player would quickly figure out that orc priests are capable of smiting for heavy damage whenever they want, however that same observant player has no method of knowing that one orc priest could have double the HPs of another orc priest. After hundreds of games of crawl, I didn't know about variable HP until this thread, which explains why sometimes orc priests seemed appropriately difficult for how early they appear, and once in a while there's that orc priest that just wouldn't go down.

I keep focusing on orc priests because they're already a wildly variable threat, and they're also one of the best character killers in the early game. Adding more variability only serves to further confuse and annoy players. And it also means for 'optimal play' I should assume all orc priests are going have 22 hp, and smite every turn. Which means, more boring play focused on avoiding the wildly overpowered threat any given orc priest might represent.

[Sandman25]

And it also means for 'optimal play' I should assume all orc priests are going have 22 hp, and smite every turn. Which means, more boring play focused on avoiding the wildly overpowered threat any given orc priest might represent.

You should also assume that you will deal 0 damage per turn while the priest will deal max damage (17?) every time. Yes, it means there is always room for unfair/unavoidable deaths and it looks like almost everyone likes it because it makes the game "harder and more interesting".

[johlstei]

That has nothing to do with fixed/variable hp. If you are arguing for fixed hp, assume it is fixed at the value which makes your argument worst, in this case 22 hp for orc priests. Your argument should still hold up in that case, but that does not, thus you are arguing about damage/hp amount and not damage/hp variance.

vvv just post it im not scared

[Sandman25]

deleted

Edit. Initially I didn't realize that johlstei is not replying to my post so I deleted my reply as irrelevant.

[Sar]

After hundreds of games of crawl, I didn't know about variable HP until this thread, which explains why sometimes orc priests seemed appropriately difficult for how early they appear, and once in a while there's that orc priest that just wouldn't go down.

So you didn't know about variable HP but you did know that some orc priests are harder to kill. Now you know about variable HP, which means that some orc priests might be harder to kill. Which you already sort of knew. Well, at least now you know why.

[Sandman25]

I think I should post here what I initially wrote in reply to johlstei. Some games like Homm3 try to be "fair" and provide a guaranteed hit after 3 misses in a row while attacking Wraith despite chance to hit is 50%. Crawl is very different in this regard: it is possible to get 3 miscasts in a row while having 1% miscast rate (it happened to me, the chance was 0.000001) get 3 smites in a row from an orc priest (it happened during SDC game to a very good player, the chance was 0.0005 if my understanding of AI mechanic is correct), miss multiple times with 27 Fighting and Weapon etc.
It means that no matter if monsters have fixed HP or not, there will always be room for unfair/unavoidable deaths and nobody seems to care. I am not blaming anyone here, there are different design approaches and crawl is very fun as is.

[johlstei]

I think I should post here what I initially wrote in reply to johlstei. Some games like Homm3 try to be "fair" and provide a guaranteed hit after 3 misses in a row while attacking Wraith despite chance to hit is 50%. Crawl is very different in this regard: it is possible to get 3 miscasts in a row while having 1% miscast rate (it happened to me, the chance was 0.000001) get 3 smites in a row from an orc priest (it happened during SDC game to a very good player, the chance was 0.0005 if my understanding of AI mechanic is correct), miss multiple times with 27 Fighting and Weapon etc.
And no matter if monsters have fixed HP or not, there will always be room for unfair/unavoidable deaths and nobody seems to care. I am not blaming anyone here, there are different design approaches and crawl is very fun as is.

I never knew HOMM3 did that, man I played that game for so many hours. There definitely are heuristics you can use to make things feel better, such as a fixed lower bound on realized accuracy as you said. Psychologically, missing feels really bad, I would much rather hit and do 1 damage more often to monsters with more hp than see "you missed", but changing that would require rebalancing every monsters in crawl since EV no longer makes sense. That's different from fixed vs variable hp though, where I think variable hp is fine and there's nothing wrong with a monster just being a tough one.

I've definitely gotten the "miscast 3 times in a row" thing what seems like too often, last time I was tempted to check the code and make sure it was actually doing miscasts right.

[Sar]

It means that no matter if monsters have fixed HP or not, there will always be room for unfair/unavoidable deaths and nobody seems to care.

Balance
========================================

The notions of balance, or being imbalanced, are extremely vague. Here is
our definition: Crawl is designed to be a challenging game, and is also
renowned for its randomness. However, this does not mean that wins are an
arbitrary matter of luck: the skill of players will have the largest
impact. So, yes, there may be situations where you are doomed - no action
could have saved your life. But then, from the midgame on, most deaths are
not of this type: By this stage, almost all casualties can be traced back
to actual mistakes; if not tactical ones, then of a strategical type, like
wrong skilling (too broad or too narrow), unwise use of resources (too
conservative or too liberal), or wrong decisions about branch/god/gear.

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.

[Patashu]

After hundreds of games of crawl, I didn't know about variable HP until this thread, which explains why sometimes orc priests seemed appropriately difficult for how early they appear, and once in a while there's that orc priest that just wouldn't go down.

So you didn't know about variable HP but you did know that some orc priests are harder to kill. Now you know about variable HP, which means that some orc priests might be harder to kill. Which you already sort of knew. Well, at least now you know why.

Unless you let a monster regenerate to full HP then re-engage it, even in a non randomized max HP world, some orc priests are 'harder to kill' 'anyway' because you get repeated bad rolls against them (enlongating whatever amount of hp they happen to have) and they get repeated good rolls against you (reducing the amount of time/safety you have to kill it)

[damiac]

Very true. Orc priests (and really, all enemies) are already quite sufficiently random, even in the absence of randomized max HP. So why have another layer of randomness on top of what everyone seems to agree is already quite random enough?

I think reaver pretty well explained why random max HP is totally unnecessary in the OP. It already takes a random number of hits to kill enemies. The random HP code is messy.

As a counter point: Would you be in favor of having player attack damage passed through another function where the weapon's base damage is selected from 1-actual base damage, therefore giving another level of randomness? (It would also be a massive damage nerf, but ignore that...) This is running with the idea that 'more randomness is better'. I assume nobody would be in favor of actually adding another mechanic intended to make things more random, so why keep a mechanic like that if it's causing a hassle to the devs?

[johlstei]

The amount of randomness that is appropriate is a strange, esoteric thing to debate and I don't think there exist any compelling arguments that take the form of "this layer of randomness is too much randomness so we should remove it" without any real reasons given beyond that. Reductio ad absurdum doesn't work here if you don't show why a conclusion is absurd. I don't think you've made a compelling argument for removing this layer, nor dispelled one for adding another, in that post.

What level of randomness feels the best psychologically and does the most good for the balance and longevity of the game? I don't know, but that's the sort of question that needs to be asked and answered if you think it should be changed, and it needs to be answered in an absolute sense because "it's too much right now" doesn't mean much on it's own.

I think the proposal that should generally be debated is "should we keep variable monster hp or fix monster hp at the current maximum value?". I think a lot of people complaining about variance are really complaining about the strong versions of monsters, and framing it that way neatly sidesteps that.

[crate]

So why have another layer of randomness on top of what everyone seems to agree is already quite random enough?

Unless you have compiled your own version of crawl with monster hp fixed instead of randomized you have not actually experienced fixed monster hp. I suspect that no one in this topic has done so. Thus I suggest that, in fact, contrary to what you have just said, no one actually knows whether fixed-hp crawl is "random enough".

(If you believe that you can evaluate crawl's "goodness" as a game without actually playing it, then you are free to try to argue that, of course, but I do not believe this is the case.)

What we do know, rather, is what variable-hp crawl is like. As I said in my post on the previous page, if we assume that one of crawl's goals is to more-or-less retain the same sort of playerbase, and we assume that crawl is a pretty good game right now, then it is likely that the amount of variance in current crawl is pretty close to what crawl should aim for (since, as I illustrated via looking at extremes on the previous page, either reducing or increasing variance changes combat quite significantly away from the status quo).

If you disagree with those assumptions, of course, then you will reach a different conclusion, and I will make no effort to support the assumptions.

[crate]

A related note that is not addressing randomized hp at all:

There are ways in which randomizing additional things actually decreases variance. This is easy to see if you compare rolling one die and doubling the result to rolling two dice. Naturally you get less variance in the latter case, despite "adding" randomness!

There are also ways in which randomizing additional things increases variance: compare, for instance, rolling one die compared to rolling two dice and subtracting them. In this case the additional randomness does in fact increase variance.

It's not immediately obvious to me whether monster hp being random actually makes combat more variable. It is true that if you look only at [monster_hp]-[player_damage] then this is the second case (subtracting two dice rolls), which as I said is increasing variance. But combat is not quite as simple as just doing that and nothing else.

(I do believe that it is the case that randomized monster hp increases combat variance, by the way, so you don't need to try to convince me. I'm mainly bringing this up because you cannot always assume that more "randomness" (i.e. more things being randomized) leads necessarily to more variance (i.e. the results tend to vary from average more often or by larger amounts). In fact this should be easy to see if you believe that winning crawl is significantly skill-based: despite an absolutely enormous number of things being random, player skill is still a large--possibly the largest!--factor in deciding whether you win the game. Compare to just flipping a coin when you start a game (and you win if you get heads and lose if you get tails): only one random element, but because the variance is higher player skill is, of course, less important.)

edit: to be clear, I'm keeping the mean the same for the things I'm comparing in these examples. In the case of subtraction this means really what I'm comparing is something like 3.5-1d6 vs 1d6-1d6. See my post here for a bit more explanation.

[Sandman25]

I think the proposal that should generally be debated is "should we keep variable monster hp or fix monster hp at the current maximum value?". I think a lot of people complaining about variance are really complaining about the strong versions of monsters, and framing it that way neatly sidesteps that.

Actually that would be a very good change because it would result in more predictable and thus more decision-involving battles.

[Sandman25]

So why have another layer of randomness on top of what everyone seems to agree is already quite random enough?

Unless you have compiled your own version of crawl with monster hp fixed instead of randomized you have not actually experienced fixed monster hp. I suspect that no one in this topic has done so.

I am the one who experienced that. I compiled crawl to display max HP of all monsters as soon as they come into LoS. Also I was displaying current HP of monsters after every hit and also displayed expected damage roll (for melee/ranged/spells), actual damage roll, damage reduction from AC, to-hit roll, EV roll etc. It resulted into very different game with basically no unexpected deaths.
The game was somewhat easier because of that but it was also very fun to see (un)lucky string of rolls. Major benefit is that this mode really helps with learning and prevents unfair deaths (player is always guilty).

[Lasty]

I think the proposal that should generally be debated is "should we keep variable monster hp or fix monster hp at the current maximum value?". I think a lot of people complaining about variance are really complaining about the strong versions of monsters, and framing it that way neatly sidesteps that.

Actually that would be a very good change because it would result in more predictable and thus more decision-involving battles.

To be fair, if you like that change, just assume that every monster does have that much HP. You'll never be disappointed, but you'll sometimes be pleasantly surprised. It's even better for you than actually having that change get into the game.

[Sandman25]

I think the proposal that should generally be debated is "should we keep variable monster hp or fix monster hp at the current maximum value?". I think a lot of people complaining about variance are really complaining about the strong versions of monsters, and framing it that way neatly sidesteps that.

Actually that would be a very good change because it would result in more predictable and thus more decision-involving battles.

To be fair, if you like that change, just assume that every monster does have that much HP. You'll never be disappointed, but you'll sometimes be pleasantly surprised. It's even better for you than actually having that change get into the game.

I am not sure how to achieve it. One of my characters killed an Orc Wizard in 2 shots (I was very lucky with crossbow rolls at 0 skill and I suspect the monster had minimal HP) and then it died to next Orc Wizard (no luck with crossbow rolls and I suspect the monster had close to maximal HP). The problem is that I didn't run fsim at that time so I didn't know how much damage I can deal so knowing max HP does not help without knowing expected damage. I mean I would prefer the first Orc Wizard to have maximal HP too (then it would not die in just 2 shots), without desire to calculate anything.

[crate]

I think the proposal that should generally be debated is "should we keep variable monster hp or fix monster hp at the current maximum value?". I think a lot of people complaining about variance are really complaining about the strong versions of monsters, and framing it that way neatly sidesteps that.

Actually that would be a very good change because it would result in more predictable and thus more decision-involving battles.

As I pointed out in my posts before, it is absolutely not necessarily the case that more-predictable leads to more-decision-making. In the extreme where all fights are completely predictable, there is actually very little decision-making (there is no time constraint, so you can just assure you will win the fight before you start). Similarly in the extreme where all fights are maximally unpredictable there is very little decision-making.

[Sandman25]

As I pointed out in my posts before, it is absolutely not necessarily the case that more-predictable leads to more-decision-making. In the extreme where all fights are completely predictable, there is actually very little decision-making (there is no time constraint, so you can just assure you will win the fight before you start). Similarly in the extreme where all fights are maximally unpredictable there is very little decision-making.

Current damage rolls are classic/ideal from predictability point of view (linear 1-N). Varied HP adds undesired unpredictability IMHO.

[and into]

I simply don't agree that if something involves some unknown (or not fully known) variables, it is ipso facto less fair and involves less decision-making. On the contrary making decisions in a context where you lack certainty is much more intriguing, as it results in situations in which there may not be a clear "best response" at any given moment, since you don't have perfect information. Crawl is usually fair, in that there is almost always at least one reasonable response to a bad situation, and very often there are more than one such response and one has to choose amongst them without necessarily knowing in advance that it is "the best." If one knew in advance, with certainty, what the best option was, then there would not actually be a decision. It is a difficult balancing act, but overall Crawl does a very good job of giving you enough information to make reasonable choices, but in most cases not enough information to let you know 100% the "correct" answer, and this balance is central to Crawl's entertainment and replay value.

The question is whether fixed or variable max HP better serves this balance. I think crate and others have argued persuasively that, at the very least, it is not wise to assume that fixed max HP will alter this balance in a way that improves game play, whereas we do know variable max HP serves it quite well at present.

[all before]

In the extreme where all fights are completely predictable, there is actually very little decision-making (there is no time constraint, so you can just assure you will win the fight before you start). Similarly in the extreme where all fights are maximally unpredictable there is very little decision-making.

If one knew in advance, with certainty, what the best option was, then there would not actually be a decision.

Clearly the fixed vs. variable hp issue cannot be solved until we have properly clarified what a decision is and, therefore, the nature of free will.

[Hurkyl]

I simply don't agree that if something involves some unknown (or not fully known) variables, it is ipso facto less fair and involves less decision-making. On the contrary making decisions in a context where you lack certainty is much more intriguing, as it results in situations in which there may not be a clear "best response" at any given moment, since you don't have perfect information. Crawl is usually fair, in that there is almost always at least one reasonable response to a bad situation, and very often there are more than one such response and one has to choose amongst them without necessarily knowing in advance that it is "the best." If one knew in advance, with certainty, what the best option was, then there would not actually be a decision. It is a difficult balancing act, but overall Crawl does a very good job of giving you enough information to make reasonable choices, but in most cases not enough information to let you know 100% the "correct" answer, and this balance is central to Crawl's entertainment and replay value.

The question is whether fixed or variable max HP better serves this balance. I think crate and others have argued persuasively that, at the very least, it is not wise to assume that fixed max HP will alter this balance in a way that improves game play, whereas we do know variable max HP serves it quite well at present.

This sort of thinking regarding optimal decision making comes up a lot, but it's really quite wrong: decisions are made from the information you have, not the information you don't have. No matter what that level of information actually is, there is a best response you can make to that information.

If you have 1 HP, a scroll of blinking, and max power flame tongue and an orc priest comes in view, zapping the priest is the wrong choice. Even if the priest has 9 HP and you get a max damage roll, zapping was still the wrong choice -- you just got lucky and failed to suffer the bad outcome from your bad choice.

The only objective effect the lack of information has in this context is decrease the correlation between the quality of your choices and the quality of your outcomes.

[and into]

I've read your response a few times Hurkyl and I confess I'm not sure exactly what your point is. Yes, you make decisions based on the information you have, but because the information you have is not perfect, you cannot always be 100% sure you are making the right choice. Coming up with a situation in which there is obviously a correct answer (orc priest in view when you have 1 HP and a scroll of blinking—and presumably a way to use the ?blinking such that you break LOS) really is neither here nor there. In any practical instance where such a thing happened there were probably several turns before that point when the situation was not so dire, and also not so clear cut in terms of what you next action should (really, needs to) be.

[crate]

Yes, you would always want to make the best decision anyway, but it is far, far easier to figure out the best decision in combat that is (from the player's view) deterministic (crawl combat is not deterministic since the player cannot know in advance the results of the various random things). Additionally, you don't really even need to figure out the best course of action--just find one solution that leaves you with >0 hp and the monster dead.

In theory it's of course possible to prove that something is the best action in current crawl combat. But realistically crawl is complicated enough that I'm confident in saying that this doesn't happen. (This is different from saying a certain action is "good enough", which is something humans can absolutely figure out! People do win crawl with some consistency, after all.)

It does not objectively make crawl a worse game to make combat simpler to figure out, and as I said before there are plenty of games that do that. But currently crawl is not a game that has simple-to-analyse combat, and I think this is important in making crawl what it is. If I could know for sure the result of a fight before I actually perform the actions, then crawl would not be fun for me to play. It is not, I think, an unreasonable assumption to say that crawl should try to retain this too-complicated-to-perfectly-analyse character. So decreasing randomness significantly seems to me to be a bad idea.

Like I've said, I do not know if changing monster HP to be fixed values is a significant change, and I take no stance on that matter. But I very much disagree that decreasing randomness is definitely a good thing, as some people have been saying.

(Really this whole argument is similar, I think, to my personal biggest gripe with crawl: that it has persistent levels. I'm not going to discuss here why I think persistent levels are a bad thing, but I think that crawl would be a significantly more fun game if levels were not persistent. Despite this, I acknowledge that there are some benefits to having persistent levels. It is also immediately clear to me that it is not really possible to make crawl have non-persistent levels, since such a change would make crawl into something that is not really crawl any more. So I never seriously suggest that crawl change to non-persistent levels. How much randomness is best for crawl is of course not such a binary decision, but similarly I think it's not a stretch to say that crawl should not stray too far from the status quo, as I said earlier.)

[Hurkyl]

I've read your response a few times Hurkyl and I confess I'm not sure exactly what your point is. Yes, you make decisions based on the information you have, but because the information you have is not perfect, you cannot always be 100% sure you are making the right choice. Coming up with a situation in which there is obviously a correct answer (orc priest in view when you have 1 HP and a scroll of blinking—and presumably a way to use the ?blinking such that you break LOS) really is neither here nor there. In any practical instance where such a thing happened there were probably several turns before that point when the situation was not so dire, and also not so clear cut in terms of what you next action should (really, needs to) be.

The point of the example is to clearly demonstrate the general principle that there can be an obviously best response -- and we can be 100% certain* that it is the best response -- in the face of imperfect information.

Even when that response would be obviously wrong if we had perfect information (the priest's HP, your next damage roll).

Maybe a completely unrelated example would be useful: you're playing blackjack with a freshly shuffled deck, and you're dealt a total of 13 and the dealer has a 6 up. You have imperfect information -- you have no idea what the next card in the deck is or what the dealer's hole card is. However, you still have a clear best response: you stand. That response might cost you the hand, but it's much more likely to win you the hand than any other action.

You can do that with combat in crawl too. If I wanted to, it would be relatively straightforward to write a short program where I feed in my weapon, skills, and stats and how many squares are between us when the engagement properly starts, and it tells me exactly how much HP I should have if I want a 99% chance of killing the orc priest before it kills me. Or any other number of more sophisticated things; e.g. if I feed in how much I value killing the priest relative to how much I value a scroll of blinking, it could tell me how much HP I should have before engaging the priest is the best response given my values.

The interesting decision is how much I should value killing the priest in regards to my consumables or my likelihood of surviving the conflict. But for any particular choice of those those values, there is a clear best response to encountering an orc priest (I mean doing so normally, not in the extreme example I mentioned earlier).

*: For the hypothetical, I am assuming all other information is neutral; e.g. the place to blink to is very likely to be safe, as opposed to something I think is probably in LOS of a centaur I saw wandering around

[and into]

Ah, thanks for the clarification. Really this is just semantics; by "reasonable response" I meant "reasonable game play decision based on what you do know," by "best response" I meant "hypothetically ideal thing to do in that situation assuming you had all pertinent knowledge."

Sure, in your usage, "best response" is the thing that is best to do under the circumstances, with whatever imperfect knowledge you have, and that's a reasonable definition and usage as well.

My point is that imperfect information can make for more interesting judgments and decisions; there are knowns and also "known unknowns," as the cliche goes. There are "games" (e.g., tic-tac-toe) in which one's best move is also the ideal move, because the game is sufficiently simple and there is nothing pertinent that is concealed from you. (By contrast, there is also the card game War, which is 100% chance with no skill involved. That isn't very fun, either.)

If you knew exact data on enough stuff, then Crawl becomes more like tic-tac-toe. That's taken to the extreme, of course; I don't think fixed HP would suddenly make Crawl that one-dimensional. But I'm not convinced it would improve game play, either, and I think Crawl works well now in this regard.


EDIT: vv I'd be fine with uniques having variable HP, too, if only for consistency. I don't see any reasons to randomize things like MR or to make damage formulae more complicated than they already are. The people who are skeptical of changing to fixed HP are not arguing "randomize absolutely everything LOL," they are simply expressing skepticism about changing monster HP from its current non-fixed system, around which the game is currently balanced, to a new fixed one.

[Sandman25]

I wonder why all uniques have fixed HP. Shouldn't they be changed to have varied HP if the latter is so great for crawl? Just imagine Antaeus with 400-1000HP, isn't it really fun and full of decision-making?

[crate]

Antaeus with 400-1000 hp randomly actually changes pretty much nothing unless you're a formicid (in which case lol sucks to be you), since you don't want to fight him anyway.

[Siegurt]

Another point I'd like to make here is that increased randomness in the example of of hps actually serves to decrease variablility rather than increase it, it might take you 4 hits to kill two orc priests, but you might have 4 strong hits against a high hp version or 4 weak hits against a low hp version.

The reason this is so, is for the same reason that rolling 3d4 is less variable than rolling 1d12.

If monsters all had fixed hps, then you would actually see *more* variablity in the time it takes to kill them (since you're reducing the number of rolls) you would end up with a more flat distribution rather than the bell curve of number of attacks that we're accustomed to.

I'm not sure if this is entirely clear (since some people seem to be arguing that variable hit points makes it more common that it will take an unusually long or short number of turns to kill something, rather than making it less common, as it actually does.)

Another way to look at it, is that monster's hp roll is one more *roll* to add to your chances of killing something in an average number of hits, it doesn't change the average number of hits, it just decreases the likelihood of getting an outlying result.

This is one of the reasons you're slightly more likely to die to a unique, because when comparing it to a generic monster who has the same average number of hps as the unique has static hps, you are slightly more likely over a number of kills to that unique to have that time happen when you just can't kill it, that's actually *less* likely with a generic monster with the same average hps.

[Sandman25]

I am not sure about it since you can get unlucky rolls vs high HP monster and lucky rolls vs low HP monster. If you have 1-9 damage rolls, you can kill 9 HP Orc Priest with 1 attack, it is not possible with 21 HP Orc Priest.

Edit. I mean average number of attacks is not as important as min/max number of attacks.

[duvessa]

I wonder why all uniques have fixed HP. Shouldn't they be changed to have varied HP if the latter is so great for crawl? Just imagine Antaeus with 400-1000HP, isn't it really fun and full of decision-making?

Yes. Uniques, orbs of fire, etc. should definitely have variable HP as long as other monsters do. I strongly support such a change.

[Sandman25]

I wonder why all uniques have fixed HP. Shouldn't they be changed to have varied HP if the latter is so great for crawl? Just imagine Antaeus with 400-1000HP, isn't it really fun and full of decision-making?

Yes. Uniques, orbs of fire, etc. should definitely have variable HP as long as other monsters do. I strongly support such a change.

Also I suggest to change other stats accordingly. 100-200 HP Orb of Fire, 2d43-4d43 fireball, 10-30AC, 10-30EV, 50-immune MR. We have random pan lords/abominations any way and only tactics really matters.

[johlstei]

Would that be (1d3+1)d4 or (2d2)d4?

[crate]

Another point I'd like to make here is that increased randomness in the example of of hps actually serves to decrease variablility rather than increase it, it might take you 4 hits to kill two orc priests, but you might have 4 strong hits against a high hp version or 4 weak hits against a low hp version.

The reason this is so, is for the same reason that rolling 3d4 is less variable than rolling 1d12.

If monsters all had fixed hps, then you would actually see *more* variablity in the time it takes to kill them (since you're reducing the number of rolls) you would end up with a more flat distribution rather than the bell curve of number of attacks that we're accustomed to.

I'm not sure if this is entirely clear (since some people seem to be arguing that variable hit points makes it more common that it will take an unusually long or short number of turns to kill something, rather than making it less common, as it actually does.)

Another way to look at it, is that monster's hp roll is one more *roll* to add to your chances of killing something in an average number of hits, it doesn't change the average number of hits, it just decreases the likelihood of getting an outlying result.

This is one of the reasons you're slightly more likely to die to a unique, because when comparing it to a generic monster who has the same average number of hps as the unique has static hps, you are slightly more likely over a number of kills to that unique to have that time happen when you just can't kill it, that's actually *less* likely with a generic monster with the same average hps.

I actually addressed this above and I do not believe you are correct here.

A related note that is not addressing randomized hp at all:

There are ways in which randomizing additional things actually decreases variance. This is easy to see if you compare rolling one die and doubling the result to rolling two dice. Naturally you get less variance in the latter case, despite "adding" randomness!

There are also ways in which randomizing additional things increases variance: compare, for instance, rolling one die compared to rolling two dice and subtracting them. In this case the additional randomness does in fact increase variance.

It's not immediately obvious to me whether monster hp being random actually makes combat more variable. It is true that if you look only at [monster_hp]-[player_damage] then this is the second case (subtracting two dice rolls), which as I said is increasing variance. But combat is not quite as simple as just doing that and nothing else.

[but I do think that randomized monster hp increases variance, though it's not quite so simple to prove]

[Sprucery]

Also I suggest to change other stats accordingly. 100-200 HP Orb of Fire, 2d43-4d43 fireball, 10-30AC, 10-30EV, 50-immune MR. We have random pan lords/abominations any way and only tactics really matters.

The fireball damage is already randomized, because dice is used. It doesn't do static 66 damage (about average for 3d43) , for example. The other stats could be randomized, yes, but not with a very wide distribution, imho.

(In Adom (at least in the version I last played many years ago) monster stats raised as you killed more and more of the monster type. This was very annoying, as there were summoners who summoned hundreds of jackals and after a while they had absurd stats and you just had to try to avoid them completely. That was a very bad mechanic.)

[Lasty]

A related note that is not addressing randomized hp at all:

There are ways in which randomizing additional things actually decreases variance. This is easy to see if you compare rolling one die and doubling the result to rolling two dice. Naturally you get less variance in the latter case, despite "adding" randomness!

There are also ways in which randomizing additional things increases variance: compare, for instance, rolling one die compared to rolling two dice and subtracting them. In this case the additional randomness does in fact increase variance.

It's not immediately obvious to me whether monster hp being random actually makes combat more variable. It is true that if you look only at [monster_hp]-[player_damage] then this is the second case (subtracting two dice rolls), which as I said is increasing variance. But combat is not quite as simple as just doing that and nothing else.

It's possible that I'm missing your point, but 1d6 - 1d6 has the same distribution (albeit over a different set of numbers) and thus same variance as 2d6, unless you're assuming we truncate negative numbers. If I have missed your point, perhaps you could clarify it?

[Sandman25]

Also I suggest to change other stats accordingly. 100-200 HP Orb of Fire, 2d43-4d43 fireball, 10-30AC, 10-30EV, 50-immune MR. We have random pan lords/abominations any way and only tactics really matters.

The fireball damage is already randomized, because dice is used. It doesn't do static 66 damage (about average for 3d43) , for example. The other stats could be randomized, yes, but not with a very wide distribution, imho.

Fixed HP monster with base damage 2d43-4d43 (which is subject to another roll afterwards) is MORE predictable than Orc Priest with its 9-21 HP. Why? Because the worst Orb of Fire would be expected to deal twice more damage while the worst Orc Priest is expected to survive 2.33 times more damage (thus dealing 2.33 times more damage too) and 2.33>2.

[Sandman25]

It's possible that I'm missing your point, but 1d6 - 1d6 has the same distribution (albeit over a different set of numbers) and thus same variance as 2d6, unless you're assuming we truncate negative numbers. If I have missed your point, perhaps you could clarify it?

I think you are comparing d6-d6 (or 2d6) to d12. Yes, 2d6 is more predictable than d12. But we are comparing d6-d6 to 1d6 (removing varied HP leads to removing one of those d6) and I believe 1d6 is more predictable than 2d6 or 1d6-1d6 (as you will never get a value not in 1-6 range for the former).

[Lasty]

I think you are comparing d6-d6 (or 2d6) to d12. Yes, 2d6 is more predictable than d12. But we are comparing d6-d6 to 1d6 (removing varied HP leads to removing one of those d6) and I believe 1d6 is more predictable than 2d6 or 1d6-1d6 (as you will never get a value not in 1-6 range for the former).

Right, so crate offers several contrasting examples:
1a: roll a single die and double it (1d6*2) [lower randomness, higher variance]
1b: roll two since and add them (2d6) [higher randomness, lower variance]
2a: roll one die (1d6?) [lower randomness, lower variance?]
2b: roll two dice and subtract them (1d6 - 1d6?) [higher randomness, higher variance?]

I read crate as then identifying [monster_hp] - [player_damage] as being a case of 2b, which I believe crate claims is a case of increased randomness also increasing variance. What I'm trying to question is whether 2b really is higher variance than 2a (it doesn't look like it), and thus whether [monster_hp] - [player_damage] is a variance-increasing form of randomness.

Given that the claims about 2a and 2b seem to be obviously false to me, I'm asking if I've misunderstood the nature of examples 2a and 2b.