Opaque skill system is severely wounded

[b0rsuk]

I didn't pay attention to DCSS for the last several years. I'm pleased to say I like the direction it's evolving towards, with a few exceptions*. Much of the cruft and dull monsters/items are gone. Menus provide much more information. I can examine a monster to get a general idea how tough it would be to fight, and while listing all their abilities is a bit suspense-breaking, I think the overall effect is good.

Crawl has a philosophy of hiding unnecessary numbers so that players perform tactical and positional calculations rather than number crunching. But where it still fails is damage, evasion, AC and SH gains. It's already better because it says how much skill I need to get rid of shield attack penalty etc. I believe it can be made even better without sacrificing its design philosophy.

------------
What I really want to know is what's the relative benefit of training various skills. That's why playing Gnolls was so pleasant, and I think they should be listed as Simple rather than Intermediate species. So much mental overhead is gone.

Weapon skills list how much skill is needed for min delay. The skill description doesn't say how much damage and accuracy I could still get. Accuracy and damage formulas are complicated, I get that. But you could add these sentences:

Your weapon skill is currently at 14. With current stats and equipment, training it to 27 would increase your attack damage 1.7 times and accuracy 2.3 times.

(or whatever it actually is).
On the Shields screen:

Your Shields skill is currently at 12. With current abilities and equipment, training it to 27 would increase your SH to 17.

Alternatively, if it actually maxes out sooner, it could be

Your Shields skill is currently at 12. With current abilities and equipment, you would max your SH at skill 16.

Note my proposed descriptions don't introduce extra numbers. SH is already displayed. The proposed weapon skill description says what the relative increase in damage would be, but doesn't specify exact numbers.

Armor skill screen doesn't say how much skill is needed to completely negate armor penalties, because, as it turns out, it never completely negates, even at max level. But it could say how much extra AC and EV I would get from training it to 27:

Your skill is currently at 16. With current abilities and equipment, training it to 27 would raise your AC by 6 and EV by 2.

I imagine it might be true because my current DsFi is wearing leathers and training Dodging, Shields and Stealth. But I have no way to know for sure except to go on another wiki binge and suffer another round of analysis paralysis. This kind of information is very valuable to the player - how much impact does training Armor have when wearing a plate armor versus a leather armor. I suspect Tengu +1 Armor aptitude is a trap because they start with 2 fewer slots, but I'd like to know for sure. I would like to know how much I could potentially raise my AC and EV if I suddenly started wearing chain mails.



* Exceptions:
- ghosts running wild had their charm. If the chief problem was early dungeon levels, why not just limit ghosts to transporter vaults in D:1-9 and let them roam freely in D:10+ ? That way new players and new characters don't have it extra rough but they can still provide an adrenaline rush.
- some monsters were tactically interesting when running away, for example fast regenerating ones and those with ranged attacks.

[tealizard]

You're right to connect the complexity and extreme interactivity of crawl skills (you can change allocation of skill points at the granularity of individual kills) with the problem of mental overhead, the vastly superior gameplay of gnolls over other species, etc. This can never be adequately addressed through automation (e.g. skill targets) or display of additional numerical data as you suggest, though.

The way to substantially improve gameplay associated with the skill system is to make it less interactive. The player does not need the level of granularity the system offers. It would be fine to offer skill increases only at level ups, for example, and there it would make complete sense to say how different choices effect things like numerical defenses, failrates and spellpower, etc. You could have a menu that combines a choice of skill options with information from the % screen and the I/M menus, with changes shown when the player moves a cursor over different skill options -- that would be clean and functional. What is not clean or functional is similar numerical data splayed all over different menus and information screens, comparing deltas for X skill vs. Y stat.

Instead of skills being allocated in tiny intervals which the player can and should change the plan of allocation for continually, give the player the points in large chunks, allocated at a reasonable number of discrete points through the game, with limited ability to divide them so that the combinatorics of the choice are manageable. If you have to change the underlying mechanics of skill points to make this work, fine. The existing mechanics suck anyway.

Of course, even better than this level of player controlled customization of skills is no control at all after character creation. Somehow it never seemed like a problem to me that I couldn't influence the development of my dudes' stats and whatever in early dragon quest games and when I found I could do this through micromanagement like in FFVI, it turns out to be a real pain in the ass. I feel like 25 years later, we should have learned some lessons about this. But y'all ain't ready for that conversation...

[Sorcerous]

Weapon skills list how much skill is needed for min delay. The skill description doesn't say how much damage and accuracy I could still get. Accuracy and damage formulas are complicated, I get that. But you could add these sentences:

Your weapon skill is currently at 14. With current stats and equipment, training it to 27 would increase your attack damage 1.7 times and accuracy 2.3 times.

(or whatever it actually is).
On the Shields screen:

Your Shields skill is currently at 12. With current abilities and equipment, training it to 27 would increase your SH to 17.


Alternatively, if it actually maxes out sooner, it could be

Your Shields skill is currently at 12. With current abilities and equipment, you would max your SH at skill 16.


Note my proposed descriptions don't introduce extra numbers. SH is already displayed. The proposed weapon skill description says what the relative increase in damage would be, but doesn't specify exact numbers.

Armor skill screen doesn't say how much skill is needed to completely negate armor penalties, because, as it turns out, it never completely negates, even at max level. But it could say how much extra AC and EV I would get from training it to 27:

Your skill is currently at 16. With current abilities and equipment, training it to 27 would raise your AC by 6 and EV by 2.


I imagine it might be true because my current DsFi is wearing leathers and training Dodging, Shields and Stealth. But I have no way to know for sure except to go on another wiki binge and suffer another round of analysis paralysis. This kind of information is very valuable to the player - how much impact does training Armor have when wearing a plate armor versus a leather armor. I suspect Tengu +1 Armor aptitude is a trap because they start with 2 fewer slots, but I'd like to know for sure. I would like to know how much I could potentially raise my AC and EV if I suddenly started wearing chain mails.

It's an interesting proposal. Like you already explained, the benefits of transparency could definitely help players of any skill level deal with meaningful numbers. I'd be on board for something like this to happen. Do you already have a unified theme in mind for the defensive/offensive benefits?

The weapon example you gave above has multiplicative benefits shown in terms of skill, but it doesn't account for the enchantment or base weapon accuracy. Or it already does, which is like the fail % for spells that just shows the end result. I can imagine the numbers would shift with a different weapon equipped. Would the default skill rank be 27, and would an already set skill target (15, 18, 24, whatever) be used for a specific comparison?

Edit: somewhat related to the topic, there is an external tool you might like from this thread. It's nice stuff.

[b0rsuk]

The way to substantially improve gameplay associated with the skill system is to make it less interactive.

A way, definitely, but not necessarily the way.

Instead of skills being allocated in tiny intervals which the player can and should change the plan of allocation for continually, give the player the points in large chunks, allocated at a reasonable number of discrete points through the game, with limited ability to divide them so that the combinatorics of the choice are manageable. If you have to change the underlying mechanics of skill points to make this work, fine. The existing mechanics suck anyway.

The existing mechanics are a relic of the times when Crawl tried to have a learn-by-doing skill system, and it worked reasonably well for a few skills but for most others it promoted victory dancing. That is, triggering skills you wanted to have right after killing a monster (when you would get unallocated xp).

I was never a fan of 12.1 style displays and 12.5 stat increases - the number 27 is granular enough, it turns out too granular in many cases because players keep asking how many levels they need to increase a skill to gain a point of AC. Over the years I've become a fan of systems where each assigned point is a meaningful increase, and not only that - each skill must be able to enable a distinct playstyle. Because at the end that's what we care about - how many ways there are to play (stealthy stabber, armored warrior, conjurer, summoner...), not if a weapon is called a spear or an axe and works almost the same.

But while granularity is unnecessary, I don't think it's confusing. Even if you awarded skill points at level ups only (which 15 rune players would immediately object to, seeing as skill levels can now keep raising once you hit XL27), you would still have the problem of uninformed choices. You still need to explain to player what effect their skill allocations are going to have. It's not necessary to know exact numbers, but IMO it is necessary to know relative strengths of stuff, such as:

- am I going to benefit more by training Fighting, or Maces&Flails?
- will my mace strike deal more damage than a than an arrow from a bow?
- will a +0 freezing trident deal more damage than a +2 glaive?

In a roguelike game, the common way of finding out is trying, but in a game with permadeath this can be very unrewarding. Particularly in a game like Crawl which cares a great deal about availability of escape tools.

I won a few straightforward fighters, and since then I'm getting my fun out of trying bizarre combinations and hybrids. I'm trying to see what unorthodox ways of play I can find. High granularity doesn't stop me from that. Insufficient information about choices I'm making does. A couple of years ago I tried Draconians of TSO to improve their aim with breath weapons, and back then Evocations skill sucked. Now Draconian breath sucks as much as before, but evokables and missile weapons can make up for that. The improved skill screen, in particular the "cost to raise" column, has enabled me to make more informed choices than about 7 years ago.

Of course, even better than this level of player controlled customization of skills is no control at all after character creation.

I think you're going overboard here. Crawl skill system lets you do quite sophisticated things, like a fighter with Shields and Ice Magic for support, but no Dodging and no Armor skill. Or a naga that tries - and fails - to use polearms from behind an ancestral spirit. Not all of them are smart, but it can be fun to try out. Shoals is a very deadly place if approached head-on by an armored fighter, but some stealth goes a long way.

[tealizard]

It can't be going overboard to bring the game more in line with better and more successful games' mechanics. Through forks, we've had a lot of experience with these kinds of changes and they turn out to be improvements pretty much every time.

The ability to make more fine grained choices under constraints (to maintain balance etc.) does not actually create more possibility, it only fools the uncritical player into thinking so -- the design of these constraints will always involve crude estimates and heuristics, usability considerations, and so on that result in only a crude approximation to what makes sense from a balance perspective. Within crawl's system of numerical qualities of a character there are far more interesting routes available to the designer than allowed to the player under the rules of skill point allocation, forget about the much smaller set of possibilities that make sense to a player serious about winning. Giving the player the power to make elementary errors through a menu full of inscrutable numbers and skill names makes the game worse, not better, and the frequency with which the player can do so (as many times as they want, between any two actions) more so.

Anyway, the level up mechanics I describe in the previous post addresses all your concerns about the impact of skill allocation on functional aspects of the player, so I don't know what you're talking about in the rest of your reply.

[b0rsuk]

Depends how fine-grained you're talking about. If you're saying 27+ skill levels is overkill, then yeah. If you think 3 levels should be enough for everyone, I disagree.

"More successful" is similarly vague. Plants vs Zombies is more successful. Flappy Bird is more successful. There's more than 1 way to design a game. A game can be very fun because it doesn't have levelling and all that stuff, but sometimes it's fun because it has it. As with many malfunctioning game systems, there are two basic approaches possible: fix it or remove it.

I'm not saying I disagree with your points, but as long as your posts are so... granular, it's like trying to nail mist to a wall.

It's an interesting proposal. Like you already explained, the benefits of transparency could definitely help players of any skill level deal with meaningful numbers. I'd be on board for something like this to happen. Do you already have a unified theme in mind for the defensive/offensive benefits?

Could you be more specific?

The weapon example you gave above has multiplicative benefits shown in terms of skill, but it doesn't account for the enchantment or base weapon accuracy.

Oh, but it does. Note the phrase "with current abilities and equipment". So if player figures out that several more Shields skill levels will only give him +4 SH, he may consider to allocate his experience elsewhere, for example train a missile weapon or magic.

A big problem with numerical skill levels is, it implies the bonus scales linearly. You would expect skill 10 to be roughly twice as effective as skill 5. This is often not the case. I think that as long as DCSS has numerical skill levels, it would be great if the numbers themselves gave you a rough idea how good it is. This is absolutely not unique to Crawl - RPG systems are notorious for being unclear on this.

Edit: somewhat related to the topic, there is an external tool you might like from this thread. It's nice stuff.

It's sad that a game needs a tool like this.

[Sorcerous]

It's an interesting proposal. Like you already explained, the benefits of transparency could definitely help players of any skill level deal with meaningful numbers. I'd be on board for something like this to happen. Do you already have a unified theme in mind for the defensive/offensive benefits?

Could you be more specific?

A big problem with numerical skill levels is, it implies the bonus scales linearly. You would expect skill 10 to be roughly twice as effective as skill 5. This is often not the case. I think that as long as DCSS has numerical skill levels, it would be great if the numbers themselves gave you a rough idea how good it is. This is absolutely not unique to Crawl - RPG systems are notorious for being unclear on this.

So, in the first post you provided examples of how the benefits could be displayed. They didn't seem unified to me upon reading (additive & multiplicative), but now I think there is a different problem: the benefits are still opaque. What does it mean to have +4SH? How many more attacks can I expect the PC to reasonably block? The benefits to attacks are similarly appealing, but also not very useful on a second read. "1.7 times more damage and 2.3 times more accuracy" means next to nothing because damage itself isn't displayed by numbers, and accuracy depends on a wide variety of factors anyway. If I have no idea how much damage can be expected now, so working towards X times more is just as vague. The usual !s to display damage are another problem because I don't feel inclined to know the numbers behind them by heart.

Maybe some kind of average, with a small delta could do the trick? Having something like "your weapon can do an average of 37-43 damage, and the brand an average of 18-21 damage" would still be dependent on enemy AC values, but we could take an educated guess to say "I can kill the giant in 4 hits, maybe 3 if lucky". Fully agree that knowing how higher skill numbers factored into the gameplay might make for a better RPG. Especially the part about benefits not being linear struck a chord with me - until recently I believed the spell #s were linear as well.

[b0rsuk]

I think an older version of DCSS had monster descriptions like "on average, it would kill you in 7 attacks. On average, you would kill it in 4 attacks." Is that correct? Did it get removed? Ultimately that's what I care about - approximately how many attacks I need to kill something and how many attacks I can expect to last. Those two sentences could encompass a lot of these.

I was about to post about monster HD. That's another hidden number, it affects monster accuracy (so damage they deal) and resistance to some spells and effects (I think mephitic cloud, prismatic spray). Crawl has had HD for many years now so we can assume it's not going away. That's one obscure mechanic (resistance, accuracy, polymorphing) and I don't know how it could be described...
Edit:
Or is that what "It looks extremely dangerous." is about??

A problem for me is not just that Crawl hides numbers. That's an interesting design approach. But crawl has so complicated calculations it's difficult to get an idea how effective something is even if you know all the formulas and numbers. I think this is one of reasons it's difficult to come up with a fairly simple description. On the other hand... if Crawl had simple calculations, why not just expose all of them.

[Siegurt]

I think an older version of DCSS had monster descriptions like "on average, it would kill you in 7 attacks. On average, you would kill it in 4 attacks." Is that correct? Did it get removed? Ultimately that's what I care about - approximately how many attacks I need to kill something and how many attacks I can expect to last.

I believe that was in a player-built modification of the game, which I suspect wasn't maintained. Afaik it's never been in any actual dev-team released version of DCSS, nor is it anywhere in the release notes as far as I can tell

[Bozobub]

Yes, DCSS DOES need more transparency, re: what the numbers you are given actually mean. I can not agree, however, with completely changing the skill system! Frankly, DCSS has already lost much of the "flavor" that keeps me coming back; much more, and I simply won't.

That's not a threat, I doubt anyone cares much about me as an individual. But getting too far from its rougelike roots will have repercussions.

[tealizard]

Gotta say, the roguelike purist (??) who can't spell "roguelike" is a good bit, though there's nothing particularly "roguelike" about the dcss skill system, outside of its obscurantism.

[duvessa]

Rogue is well-known for its complex skill system

[b0rsuk]

I don't agree DCSS has lost much flavor. In some areas it definitely gained some. Shadow is a simple and interesting enemy - backstab attack. Swamp is interesting and distinct now, not just enemies but also layout. Hydras have more ways to deal with. Lignification, cancellation, Evocations actually worth training. No two wands or evokable items are very similar, but it used to be wand of fire, wand of draining, wand of cold and they mostly just shot bolts. There are creative additions like dream sheep, Elven Halls include dancing blades and elementals for much better variety, Depths are more distinct, there's a variety of amulets and no crap like controlled flight or very niche ones like sustain abilities. Draining is still dangerous but interesting and doesn't inflict a sense of permanent loss on me. Things like Ru and Formicid are radically new and play differently. Ranged weapons are usable without very heavy specialization. Abilities matter now (although Dexterity is a bit meh).

There's a bunch of stuff Crawl used to have but wasn't used in practice.

[Bozobub]

Gotta say, the roguelike purist (??) who can't spell "roguelike" is a good bit, though there's nothing particularly "roguelike" about the dcss skill system, outside of its obscurantism.

Nitpicking simple typos won't avail you much. It's a sure sign of someone who has nothing of note to contribute to the subject at hand; they simply have nothing else. Come up with something better or reveal you have nothing to say; there is no other choice available to you.
________
"Hardcore" roguelikes are known for (at least somewhat) obscure mechanics, that require repeated play and/or research to discover. Yeah, tell me *all about* how Nethack is oh-so-simple . And yes, I can dump many more examples in your lap, if you like. Howzabout Dwarf Fortress in Adventure mode, as low-hanging fruit?

Even Tales of Maj'Eyal, which is quite simple in gameplay, has very, VERY deep mechanics, that require quite a good bit of patience to master. And guess what? It also has an extremely complex skill/class system. It may not be a requirement of roguelikes, in general, but such complexity HAS been a hallmark of DCSS since the beginning. Don't bother trying to argue otherwise; a simple perusal of old online versions will instantly contradict you.

And no, adding/changing/removing menus, map layouts, monsters, spells, and/or what-have-you is not equivalent to eliminating base mechanics that have been in the game from Day One. Yes, completely removing magic, for example, would greatly simplify gameplay; would anyone here argue that a) it's a good idea, or b) that most players would appreciate such a thing? Only a very small portion of DCSS players EVER finish the game; catering to those who find the game "too easy" is a fantastically bad move, when it comes at the expense of everyone else. While I don't necessarily love or hate the idea, this is why I find the reflexive dismissal of an XP-modifier system to modify difficuly more than a little disturbing, when it can be set at 100% (thus, no change) at any time. If your problem is the online player rankings, a simple column noting the difficulty level or, perhaps a score divisor/multiplier would be quite easy to implement. If those rankings aren't the problem, I can't see the basis for complaint; even the coding overhead would be light, compared to most changes.

I definitely agree that DCSS has seen many excellent changes; I'm not some sort of roguelike "Luddite" . But too often, I see discussion here revolving around the same core of experienced players, addressing issues only experienced players ever see, with a dismissive attitude to contributions from outside that limited sphere. That's a mistake; the player statistics don't lie.

Regarding OP: Gnolls are a lot of fun to play BUT significantly weaker than more-specialized races for much of the game in several ways; they tend to come on strong, weaken drastically in midgame, and then come roaring back much later in the endgame, from what I've seen. Their gameplay is both *simpler* AND *more complex*, in different ways, if that makes sense. Thus, I definitely agree with them being classified as an "Intermediate" race, especially when compared to more "Tab-friendly" races ^^' .

[b0rsuk]

The thing that surely can be removed with no regrets is weighted split of xp on the 'm' screen. I can press a skill letter once to enable training it, second time to make a bigger portion of xp go to it. But if I enable two skills they tend not to get 50% xp each, but some different fraction based on my aptitudes, skill levels or stats I guess. WHAT'S THE POINT. I want to focus on these and not those skills. If you say I can only put 26% of my xp into Dodging, I will turn off other skills for the time being. WORSE, the proportion changes as my skills increase, so I have to periodically revisit the skill screen to make sure it's training those I said I want trained.

Can anyone give an argument FOR keeping the part where xp is not evenly divided between skills of the same emphasis level?

[petercordia]

You can switch to a mode where xp is evenly divided between skills of the same emphasis level.

press /

As it says at the bottom of the skill screen:

  Code:

  [/] auto|manual mode 

[tealizard]

edit: lol n/m

[Sorcerous]

Can anyone give an argument FOR keeping the part where xp is not evenly divided between skills of the same emphasis level?

1) IRL Xom worship.
2) Laziness.
3) Ignorance.
4) Lacking tutorial structure.

You're welcome.

[vt]

I can press a skill letter once to enable training it, second time to make a bigger portion of xp go to it.

This is not a defense of the status quo. But you may like to know that you can turn this behaviour off with an rc option.

  Code:

skill_focus = true
        When set to true, skills cycle between disabled, enabled and focus in
        the skill screen. When set to false, they only toggle between enabled
        and disabled. Setting the option to "toggle" adds a toggle to the skill
        screen to change the behaviour in-game.


But if I enable two skills they tend not to get 50% xp each, but some different fraction based on my aptitudes, skill levels or stats I guess.

I'm not exactly sure I understand your complaint, but my understanding is that if the screen displays that you have a 50/50 split between two skills, then xp is indeed being split evenly (modulo some randomness?) between the two skills. (I'm happy to be corrected if I am wrong about the mechanic!) It's true that all sorts of things, including your aptitudes and current skill levels, may mean you are getting more value for each point in one skill than each point in another. To me that seems good---it's part of ensuring that skilling decisions have a meaningful impact on chances to win the game.

[Siegurt]

I'm not exactly sure I understand your complaint, but my understanding is that if the screen displays that you have a 50/50 split between two skills, then xp is indeed being split evenly (modulo some randomness?) between the two skills. (I'm happy to be corrected if I am wrong about the mechanic!) It's true that all sorts of things, including your aptitudes and current skill levels, may mean you are getting more value for each point in one skill than each point in another. To me that seems good---it's part of ensuring that skilling decisions have a meaningful impact on chances to win the game.

The OP's complaint is that it doesn't split XP evenly between two skills, literally the numbers on the training screen aren't 50%/50%, the reason this is so is that they're using "auto" training, which uses "amount you use that skill" to divide up XP, the solution to their problem is to select "manual" training mode as posted above.

[vt]

Oh! Fair enough. I haven't used automatic training in years. I forgot it did more than, well, automatically selecting skills for you while otherwise using the same system.

In that case it's probably worth noting for b0rsuk that the `default_manual_training = true` rc option means that you will start the game in manual training mode, i.e., you do not have to hit `/` every game to swap to it.

[PseudoLoneWolf]

I am of the impression that auto-skilling is simply a noob trap that will, in almost all cases, lead to an objectively weaker/less optimized character than manual training, in addition to adding needless confusion for people who don't know that the game is deciding for them above and beyond their own explicit commands how much XP to allocate to skills, as noted in this very thread.

Why does this even still exist, and if it is insisted (for whatever reason) that it stay in the game, why is it the default option? Even setting aside momentarily the much larger talking point of the existing skill system in general, assuming we keep the skill system the same as it is I see no reason why manual training should not be the only option for skill training.

[tealizard]

It's very weird that this vestigial feature that you turn off at the beginning of every game is still around. Seems like there's a fair amount of code going into its implementation too. Would be good to get rid of it.

[petercordia]

I still use auto-skilling, sometimes, though never at the start of the game.
Sometimes it's just easier. You don't have to switch which skills you're training as often when using it.

[damerell]

I am of the impression that auto-skilling is simply a noob trap that will, in almost all cases, lead to an objectively weaker/less optimized character than manual training,

I think it's a bit more complicated than that. Automatic skill training is worse than manual training done right... but it won't screw up as badly as manual training done wrong. Hence I think automatic skill training can be helpful to novice players, and arguably that is why it should remain the default.

[tealizard]

Thanks guys, always reassuring to get the daily reminder that bad things are actually good.

[VeryAngryFelid]

I think it's a bit more complicated than that. Automatic skill training is worse than manual training done right... but it won't screw up as badly as manual training done wrong. Hence I think automatic skill training can be helpful to novice players, and arguably that is why it should remain the default.

I have mixed feelings about it. I recently tried a lazy autotraining game as FE and my impression is that while it surely helps new players early game, it almost guarantees death late game. No HP, no weapon/evocations for emergency situations, branching into different schools is really hard, just never ending training of stealth, spellcasting and magic skills.

I guess new players are supposed to die a few times and then switch to manual training if they really want to win.

[petercordia]

Auto training also leads to some new players training 7 different weapon skills...

[damerell]

Auto training also leads to some new players training 7 different weapon skills...

You can carry out just the same swaps with manual training - and train the same 7 different weapon skills.

[Siegurt]

The main purpose of auto training is to help people who *are so new to DCSS they have not yet discovered the skill training screen and don't even know about it yet* and for a small segment of people it's also a bridge for people who "literally have no idea what to train, like not even a rough guess conceptually" (That's a vanishingly small subset of new players, it's very rare to stumble across someone who is not just new to DCSS, but also so new to the idea of an RPG that they don't know things like "train dodging to avoid being hit" and "train magic schools to make yourself better at spellcasting", maybe someone who speaks so little english that they literally don't know what the words mean?)

That's also why it's the default, because if you're literally *that* new *and* don't read the manual first (if you read the manual you could figure out what skills you want to train, and also how to turn off automatic training), there's the intent that you still be able to play the game somewhat.

[tealizard]

The reason autotraining exists in dcss is its connection with Linley's dungeon crawl skills. It does not help new players to have a bad default option. The talk about new players is at best an excuse for keeping it around.

If anyone cares about helping new players just play the game (outside of playing gnolls, a fine option imo), the thing to do is come up with a default option that is actually good. There are pretty simple default manual skill settings possible using only features that have existed for many years and the addition of skill targets considerably expands the possibilities for sensible defaults that get the new player well into the game before presenting them with any decisions. There is no reasonable excuse for autotraining to be the default and there hasn't been for a long time.

[artagas]

I agree with 99% of what B0rsuk posted in the thread.
The 1% is that i am not sure if would classify gnoll as simple. It is a very easy species, but I think from a new player's perspective no hybrid character is really "simple" (due to both gameplay and gearing/library choices). For not-new players that screen is irrelevant anyhow.

On the more serious topic at hand: Not displaying the above info only results in: 1) confusion 2) players looking up that info from a variety of sources some of which might be wrong/dated. I mean ask yourself this question: if displaying things as OP asked was an rc option would you ever turn it off? If not, then how is not displaying them an improvement to the game.

[damerell]

I mean ask yourself this question: if displaying things as OP asked was an rc option would you ever turn it off?

I'm not sure what the actual development policy is here (especially in light of the "if you wear this armour..." inventory display in vanilla), but I recently implemented a change in Stoat Soup to display what your spell failure chances _would_ be with brilliance (or without, if you currently have brilliance) - and I think to implement this sort of "hypothetically speaking, if you had Armour skill 27 ..." display is likely to be tricky, so I don't know if it isn't done because the developers think it would be bad or just because it's hard. That would be an interesting question for anyone who cares enough to submit a pull request to answer.

[artagas]

Fair enough, i did not realize this might be difficult to implement. It would indeed be nice to know if that is a big part of the reason or not.

[Bozobub]

The reason autotraining exists in dcss is its connection with Linley's dungeon crawl skills. It does not help new players to have a bad default option. The talk about new players is at best an excuse for keeping it around.

If anyone cares about helping new players just play the game (outside of playing gnolls, a fine option imo), the thing to do is come up with a default option that is actually good. There are pretty simple default manual skill settings possible using only features that have existed for many years and the addition of skill targets considerably expands the possibilities for sensible defaults that get the new player well into the game before presenting them with any decisions. There is no reasonable excuse for autotraining to be the default and there hasn't been for a long time.

I have to agree. If you *must* have some sort of automatic option — which does seem like a good idea for new players — it needs to actually be at least decent, if not excellent, or you're just tricking noobs into repeated death. The current implementation, however, simply does not result in decent characters over time. Literally not ever.

[b0rsuk]

By the way, the combat outcome estimate messages I talked about are found in Brogue. I must have tried it at some point and bounced off. When you examine a monster (mouse only, because keyboard is for cavemen) you get these:

The kobold has a 50% chance to hit you, typically hits for 8% of your current health, and at worst, could defeat you in 7 hits.
You have a 100% chance to hit the kobold, typically for 42% of its current health, and at best, could kill it in 2 hits.

[damerell]

I have to agree. If you *must* have some sort of automatic option — which does seem like a good idea for new players — it needs to actually be at least decent, if not excellent, or you're just tricking noobs into repeated death. The current implementation, however, simply does not result in decent characters over time. Literally not ever.

This simply isn't true. It'll do a perfectly adequate job on a heavy-armour monster whacker [1] - and, in particular, it can do a better job than someone who doesn't know what they're doing.

[1] Not that I'm saying it will be terrible on every other kind of character.

[tealizard]

Narrator: It simply was true.

[b0rsuk]

By the way, this is how brogue handles character advancement:

The staff of lightning has a maximum of 2 charges, and recovers a charge in approximately 250 turns.

This staff deals damage to every creature in its line of fire, nothing is immune. (If the staff is enchanted, its average damage will increase by 50%)

When examining a jackal:

Your staff of lightning will hit the jackal for between 12% and 62% of its current health.

Brogue progression, like Rogue progression, is done with items, so you get explicit and discrete character advancement. At first I thought Brogue has no progression, but cage rooms feel very much like level-ups. Typically opened with a key, they contain several valuable items, like 5 staves to choose from, or a variety of equipment. Staves in such room come pre-identified. So in essence you are advancing by choosing what items you get. There is even respec: you can put an item back on the altar thingy you took it from, and you can take another one.

"When enchanted" sentences refer to the effect of scroll of enchantment, which permanently boosts an item.

So, Brogue lets a player choose from several new abilities, shows how the ability would scale, and tells player how effective it would be against any encountered monster. You also get a form of skill point assignment - scrolls of enchantment.

Brogue gives player all the information he needs to make decisions, all the while using less numbers than DCSS. Numbers are only used to describe what portion of monster's health an attack/ability would take, and how many turns something will take to recharge.

[duvessa]

and, in particular, it can do a better job than someone who doesn't know what they're doing.

Perhaps this is less a defense of autotraining and more an indictment of the skill system's clarity.

The reason autotraining works fine is that baseline player characters are so strong compared to the dungeon that long-term character build is inconsequential. To the point that even players playing for long streaks will not bother to train skills optimally. Sometimes they'll even deliberately waste xp to get a title they like. As a dev or as a player, it's easy to be apathetic towards the quality of autotraining when it isn't making any apparent difference anyway.

[damerell]

and, in particular, it can do a better job than someone who doesn't know what they're doing.

Perhaps this is less a defense of autotraining and more an indictment of the skill system's clarity.

Perhaps it is, but that still leaves autotraining serving a purpose as things are, and suggests that to remove it, the other problem should first be addressed.

[Siegurt]

I have to agree. If you *must* have some sort of automatic option — which does seem like a good idea for new players — it needs to actually be at least decent, if not excellent, or you're just tricking noobs into repeated death. The current implementation, however, simply does not result in decent characters over time. Literally not ever.

So:
1. "over time" for a new player means "up to about D:5 or so" because by the time you can consistently get to the temple, you're no longer a "new player" to the extent that you should be using auto training, I fully expect by the time someone has played 5-10 games or so they should look at the skill menu and figure out what the heck it should do. In that range, auto-training doesn't do an optimal job, but it's sufficiently close to optimal that it doesn't effect a new player's success chances by a significant amount (in large part because, as duvessa says, early characters are strong enough regardless of long-term build optimalness, and throwing skill points at "something that you're probably using" is plenty good enough).

That there's no *in game* way to change game launch options is a larger flaw of DCSS, and if that were remedied, we could have a sensible UI thing like "the first time you go into the skills menu you're presented with a giant thing that says 'hey, this is how skills work, now that you know, I'm going to turn off auto training mode by default unless you explicitly turn it on, you're now on your own and expected to figure out what to train' which appears once and never again"

2. Autotraining is far from perfect, but it's better than *literally nothing at all* if you have a specific set of rules for automatic training in the total absence of user input that you think would work better for a new player feel free to code it up (or even post it verbally and try to convince someone else to write the code, although that's the harder route).

[vt]

"Nothing at all" is a perfectly fine alternative if the target really is people playing their first 5 to 10 games and getting no further than D:5. Nothing is needed to tide the player over until they discover the m menu. Just like we don't need an automatic weapon choosing system until the player discovers w, or an automatic spell memorizing system until the player discovers M ....

The feature makes a lot more sense if the intention is that a more casual player can choose to ignore the whole mechanic of manual skilling and still get a reasonable amount out of the game. (I don't have anything to say about whether it succeeds there.)

[Siegurt]

"Nothing at all" is a perfectly fine alternative if the target really is people playing their first 5 to 10 games and getting no further than D:5.

So you'd suggest that "leave the starting skills marked for even distribution of XP" as the new "auto training" is better than the current auto-training? (viable, absolutely, better than the status quo? I would disagree)

[vt]

No, it was earnestly a conditional.

[tealizard]

Default skill targets and settings that get the player to d:5 with better allocation than autoskill would be easy to put together for every background other than wanderer and that's just using existing features. Straightforward extensions to the skill target system (e.g. targets with priority) would make good defaults possible all the way to lair.

[damerell]

The feature makes a lot more sense if the intention is that a more casual player can choose to ignore the whole mechanic of manual skilling and still get a reasonable amount out of the game. (I don't have anything to say about whether it succeeds there.)

Conversely I have no idea if that's the intent, but I do think it succeeds there. I think people here underestimate automatic training; the gulf between it and an expert player is huge but that does not mean the gulf between it and someone who can (say) get to D:5 is huge, or even negative.