Skill training; Auto or Manual?

[Oddysee]

Ok, I haven't really thought about this until now. But what is the best way to train skills, Auto or Manual? Or does it depend on the situation? Since Auto is the default I haven't tried changing it as it seems fine on Auto. But then again, I wouldn't know. I would be grateful for any info.

[Un67]

Manual, by far. You need to micromanage skills so you don't waste EXP (for example weapons skill past min delay).

[Oddysee]

Ok, Manual makes all skills train equally as opposed to a percentage based on use which is what Auto is right? So, I set it to Manual and only have a small number of skills turned on which are then trained equally? Is that always the best thing to do?

[cerebovssquire]

There is always one thing you want most. For instance, if you plan to kill most enemies in melee combat early on, you would want to train your weapon skill to a reasonable level first.
Therefore it is nearly always optimal to train a single skill at once, then choosing another skill next, etc. The only exception: in some situations when you want a multi-school spell, you would leave on two skills. For instance, if you want Sticky Flame, you would turn on Fire and Conjurations. Now that decimal skills are implemented (ie, 1.5 of a skill does more than 1) this is optimal play, it didn't use to be.
This does not always apply, of course; if you want Iskenderun's Mystic Blast and eventually Bolt of Fire you train Conjurations first and then Fire Magic, even though your ultimate goal is a multi-school spell.
Focus is superfluous and is something for people who think that is overly tedious to open their skill screen semi-frequently... but it is practically never optimal to use it. If you think micro-management is tedious, go ahead of course, it doesn't make the game signifcantly more difficult or anything.
Also: train what you want now, not what you might want 5 levels later.

[XuaXua]

Manual Skill management could be enhanced by allowing the player to set training maximums.

I like Manual because I have found it beneficial to pound one skill to 10-12 before anything else.

[Oddysee]

Thank you very much everyone. I will definately be using Manual and only training 1 or 2 skills at a time from now on. I love how helpful and nice the DCSS community is.

[dk]

Manual Skill management could be enhanced by allowing the player to set training maximums.

....

That would be nice.

It would also be cool if you could setup some kind of career plans, like If I'm playing Berserker, then automatically...
- train Axes to 10
- then train Dodging to 5
- then train Evocations to 8
- then start training Axes again

Maybe I should start hacking a lua-script for this...

[jejorda2]

Imagine if there were two versions of crawl, as follows:
In Game A, you get to control the skill screen, but an AI picks which weapons and spells to use and when the attacks are made to kill all the monsters.
In Game B, the skill screen is controlled by an AI, but you get to kill every individual monster on your own

I'd play game A and never touch game B.

To me, the skill screen and finding artefacts are the fun of the game. I don't understand how auto mode or felids could be fun.

[Tutankham]

Focus is superfluous and is something for people who think that is overly tedious to open their skill screen semi-frequently... but it is practically never optimal to use it.

It can be helpful if you plan to kill enemies that yield high experience while you are training a skill that only needs a few points to get to the target level. For example, you are training Throwing to level 7 and don't want it at 8 because your god would give you darts as gifts then (Okawaru, Trog). Throwing is at 6 (95%). So you can "focus" on as many skills you want to train in the future so that Throwing will get just a small share of the XP.

[cerebovssquire]

"practically never" isn't "never".

Also, you probably need other skills more than 7 Throwing anyway. Train these instead.

[KoboldLord]

It can be helpful if you plan to kill enemies that yield high experience while you are training a skill that only needs a few points to get to the target level. For example, you are training Throwing to level 7 and don't want it at 8 because your god would give you darts as gifts then (Okawaru, Trog). Throwing is at 6 (95%). So you can "focus" on as many skills you want to train in the future so that Throwing will get just a small share of the XP.

You'd rather have a couple more +0 animal skins than stacks of darts of dispersal, curare needles, and highly-enchanted javelins of penetration?

I don't find that it's worth my effort to religiously train exactly one skill at a time, but I do find it's worthwhile to limit myself to a maximum of three related skills. Generally speaking, my strategic goals are not going to be things like 'fire magic to 10', it's going to be 'get Bolt of Fire online', and I'll try to make sure that everything currently training supports that goal. It is not beneficial except in extreme marginal cases to make sure to level spellcasting separately from conjurations and fire magic, for instance. Similarly, if my strategic goal is 'shore up my defenses', I'll turn fighting and dodging on at the same time. I'd want to pick up both at low levels before training one to medium levels, so I'll just turn them both on so I don't have to micromanage it. Compared to micromanaging back and forth each time one levels up, training both at once will make a difference of, at most, one hit point or EV point during the whole process, and most of the time it will make no difference at all.

[cerebovssquire]

You'd rather have a couple more +0 animal skins than stacks of darts of dispersal, curare needles, and highly-enchanted javelins of penetration?

I think he'd rather have a cloak of preservation, a plate armour of rF+ and a vampiric demon blade than darts of frost, poisoned javelins and needles of confusion. Sophism, much?

I think training two skills for Bolt of Fire is optimal as long as you don't plan to pick up Iskenderun's Mystic Blast or something on the way... but if Fighting is at 0 and Dodging at 2, which is going to be the case for many "Mage" backgrounds, it does make a difference if you leave both on till both are at 10 or train them seperately. That difference is probably not going to be lethal, but it will be noticable. Noticable enough to not put it down as something so trivial that it's too much tedium for too little gain, in my subjective opinion.
Fire Magic/Conjurations is one thing. HP and EV are two things. Saying "alrways train one skill at a time" is maybe a little too restricted because multiple-school-spells exist, but "train skills for one thing at a time" seems perfectly ight.

[Tutankham]

You'd rather have a couple more +0 animal skins than stacks of darts of dispersal, curare needles, and highly-enchanted javelins of penetration?

Depends on the character played, no? If you like a different example better: If you play a mage using a shield, and all you are ever (or at least for a long time) using is a buckler, then you might just want to train it to 5 and not spend any more points thatn that, focus can be helpful.

[KoboldLord]

I think he'd rather have a cloak of preservation, a plate armour of rF+ and a vampiric demon blade than darts of frost, poisoned javelins and needles of confusion. Sophism, much?

I was using humor to make a point. Ammo gifts aren't something a typical character should want to avoid, particularly if they're using that skill enough to bother training it. If the character is not intended to throw things in the future, the correct amount of throwing training is to stop immediately or sooner, not stop at 7. If the character is intending to throw things to support melee for the rest of the game, then needles and javelins are awesome gifts, and the more you get the better.

I think training two skills for Bolt of Fire is optimal as long as you don't plan to pick up Iskenderun's Mystic Blast or something on the way... but if Fighting is at 0 and Dodging at 2, which is going to be the case for many "Mage" backgrounds, it does make a difference if you leave both on till both are at 10 or train them seperately. That difference is probably not going to be lethal, but it will be noticable. Noticable enough to not put it down as something so trivial that it's too much tedium for too little gain, in my subjective opinion.
Fire Magic/Conjurations is one thing. HP and EV are two things. Saying "alrways train one skill at a time" is maybe a little too restricted because multiple-school-spells exist, but "train skills for one thing at a time" seems perfectly ight.

Our hypothetical kobold conjuror here wants that hp and EV for purposes of warding off ranged attacks, like yaktaur bolt spam. Both of these stats improve the character's durability for that purpose, and both are valuable but get increasingly more expensive as the associated skill levels up. It doesn't matter much what combination of both the kobold has, so long as the character can survive the maximum number possible of incoming bolts.

More fighting is better than less fighting, but raising fighting from 8 to 9 is pretty expensive. If dodging is still back at 2, then pumping that xp into dodging is definitely going to be more desirable for our hypothetical kobold than raising mid-levels of fighting a single level. There's not likely to be good reason to rush for a particular level of most defensive skills (shields being an exception because of the penalty), because the benefits are incremental and costs increase rapidly. Technically, you can constantly micromanage between fighting and dodging, switching to the other whenever the one you're using ticks up a point, but since you get these points one at a time you'll never be more than one single hit point ahead if you micromanage.

And ultimately, micromanagement isn't going to do much good if you only have a vague idea what you're doing. Tell a newbie to focus on one skill at a time, and that newbie might end up grinding that skill well beyond the point of diminishing returns, with a half dozen other skills that do a similar thing languishing. A fully micromanaged build is going to have to be recalculated for every character in every game, since it varies according to species and item drops. Better to establish a strategic goal, with an emphasis on achievability, and train a few skills that work toward that goal.

Right, of course it only makes since to train one skill at a time if you actually know what skill you want.

Ultimately, the skill you want isn't always going to be knowable. Traps & Doors is a useless skill unless a shaft or Zot trap gets in your way, so you train it to a modest level so you can minimize your chances of being affected with as efficient a cost as possible. Fighting, dodging, and shields all have overlapping effects that are more or less useful depending on the opposition, and you don't usually know in advance what sorts of attacks you'll have to deal with. Better to train them all somewhat, to ensure you can withdraw successfully, than risk getting into a situation you didn't prepare for at all.

[cerebovssquire]

More fighting is better than less fighting, but raising fighting from 8 to 9 is pretty expensive. If dodging is still back at 2, then pumping that xp into dodging is definitely going to be more desirable for our hypothetical kobold than raising mid-levels of fighting a single level.

Well, yes, that's because 10 EV is better than 10 HP in most situations. This is also why I would first train Dodging and then Fighting.

And ultimately, micromanagement isn't going to do much good if you only have a vague idea what you're doing. Tell a newbie to focus on one skill at a time, and that newbie might end up grinding that skill well beyond the point of diminishing returns, with a half dozen other skills that do a similar thing languishing.

This is why I usually include specific skill advice when I do give this advice on CiPs. In this thread we're just discussing the subject in general. It also helps to just inform people what spell success is desirable and what minimum delay is; the rest can usually be figured out with common sense.

Ultimately, the skill you want isn't always going to be knowable.

The skill that is probably going to be most useful now is knowable and that is sufficient. Possibly there won't be many dangerous traps in my path on Zot:5 and I can just pass through the terrain with "only" the monsters being dangerous, but since that isn't probable I should train the skill. Perhaps a death would have been preventable if I had raised Dodging from 18 to 20 instead of Fighting to 10, because unfortunately a monster only just managed to land the killing blow, but in general, ~20 HP are more useful than 2 EV and so it still wasn't a bad strategic decision to raise Fighting instead. Raising Dodging would have been preparing for something that won't occur with a very high chance.

Sometimes your plans can change, e.g. when you find some loot that changes your character development plans. This would affect someone who doesn't micro-manage as well, though. Also, I think that having a plan that is adaptable, possibly at the cost of losing a little experience, is much better than having a suboptimal plan or no plan at all.

[KoboldLord]

Well, yes, that's because 10 EV is better than 10 HP in most situations. This is also why I would first train Dodging and then Fighting.

Feel free to swap dodging and fighting in my example, then. Raising dodging from 8 to 9 is not preferable to applying that same xp to fighting if fighting is still at 2.

This is why I usually include specific skill advice when I do give this advice on CiPs. In this thread we're just discussing the subject in general. It also helps to just inform people what spell success is desirable and what minimum delay is; the rest can usually be figured out with common sense.

That's very helpful in the specific case in the CIP, because there's usually a skill or two that, to our relatively experienced eyes, is obviously most urgent. If we just say what skill to put xp into without explaining why, however, then we're presenting skilling as an arcane and unknowable ritual that can't be reasoned out. Better to explain the reasons, so that the poster can apply those reasons to their next CIP, too.

The difference between alternating fighting and dodging compared to training both at once is pretty trivial, assuming we're talking about a character that benefits from both. While you might very occasionally get an extra 1-point bonus from micromanagement, the degree of emphasis put on micromanagement by posters in this thread is wildly out of proportion to the benefit. While technically not incorrect, it gives less-experienced players the misleading impression that micromanagement is as significant as the amount of text applied to it would indicate. It is not. Picking a goal and proceeding directly to that goal without any distractions is the key point to learn, not dithering over 1-point differences.

Sometimes your plans can change, e.g. when you find some loot that changes your character development plans. This would affect someone who doesn't micro-manage as well, though. Also, I think that having a plan that is adaptable, possibly at the cost of losing a little experience, is much better than having a suboptimal plan or no plan at all.

Making sure your plan is flexible is not the same thing as not having a plan. I am emphatically *not* suggesting that a player should turn on everything that looks good and letting it all work itself out. Often, there really *is* only one skill that matters for a given goal. If you're hard-skilling for Haste, them you want nothing on but charms. If you're trying to eliminate the shield penalty, then you want nothing on but shields. Not every skill is like this, however. Dual-skill spells are a counter-example that does not appear to be debated, but they are not the only examples.

[TwilightPhoenix]

Personally, I usually just leave on my defensive skill (Armor or Dodging), Fighting, and work out the rest as needed, using Focus on things I prioritize (for example, focus Long Blades and leave Dodging and Fighting on regular). If I really need something online ASAP, which, outside of the very early game, is usually pretty rare, then I'll start shutting off everything. This works for me and I can't really think of any situation I've been in where I thought "If only I had trained my skills a little differently". Well, except in those cases such as where I hit min delay on my Demon Trident of Venom I found and then discover the Singing Sword or an uber randart Demon Whip of Speed a little later.

[XuaXua]

I like Manual because I have found it beneficial to pound one skill to 10-12 before anything else.

This isn't really a reason to use manual mode, though.

Doesn't auto mode passively add skill points to skills even if off?

[KoboldLord]

Doesn't auto mode passively add skill points to skills even if off?

That effect was taken out when the restrictions on training 0-level skills were removed.

[Oddysee]

It's been a very interesting discussion so far, I'm learning alot. Anyway, last night I started a Sludge Elf Venom Mage. Now, before that I was doing fine with Manual skill training to train one skill at a time (eg Unarmed Combat until 10 then Fighting until 10 on a Monk etc). However, I was a little confused about what to train first on my Venom Mage and indeed all casters really.

I didn't really know what to train first; Spellcasting or Poison Magic. So in the end I starting training both at the same time. Was that the wrong thing to do?

[slowcar]

So in the end I starting training both at the same time. Was that the wrong thing to do?

No, it is normal that you train several things at once.

[KoboldLord]

It's been a very interesting discussion so far, I'm learning alot. Anyway, last night I started a Sludge Elf Venom Mage. Now, before that I was doing fine with Manual skill training to train one skill at a time (eg Unarmed Combat until 10 then Fighting until 10 on a Monk etc). However, I was a little confused about what to train first on my Venom Mage and indeed all casters really.

I didn't really know what to train first; Spellcasting or Poison Magic. So in the end I starting training both at the same time. Was that the wrong thing to do?

Both spellcasting and poison magic work towards the same early-game goal of getting you to be able to cast lots of poison spells, so you're fine training both. Later on, you might change that goal to where you want a specific poison spell hungerless, or to where you want a specific high-level spell like Poison Arrow usable, and at that point you'll train just the most relevant one, but at the beginning of the game you need both skills.

Venom mages are an interesting case because they often cross-train into some other build fairly early on, before any high-level poison spells show up. You can get pretty far with Mephitic Cloud spam, stacking Toxic Radiance, and Sting to finish off lone monsters, but depending on your strategy you'll often prefer to minimize poison magic investment so you're prepared to take advantage of whatever loot turns up.

[galehar]

Manual, by far. You need to micromanage skills so you don't waste EXP (for example weapons skill past min delay).

The argument is irrelevant to the question. You can use auto mode and switch off skills you don't want to train. You don't have to leave all your skills on in auto mode, you can still fine tune the result by disabling unwanted skills and using focus.

if Fighting is at 0 and Dodging at 2, which is going to be the case for many "Mage" backgrounds, it does make a difference if you leave both on till both are at 10 or train them seperately. That difference is probably not going to be lethal, but it will be noticable. Noticable enough to not put it down as something so trivial that it's too much tedium for too little gain, in my subjective opinion.

I'd say the optimal thing to do to maximize defense in this case would be to train fighting to level 2 then train both at the same time. Assuming equal aptitude. I don't see what can be gained from constantly switching from one to the other. Or do you mean that dodge 10 fight 0 is better than having both at 7?

To answer the OP, I use both. In the very early game, I usually use manual and train critical skills (weapon skills for melee, spell scholl and spellcasting for casters). After that, I get some defense (dodging, armour and/or fighting). By the mid-game, If I don't have any pressing need, I use auto-mode, but not blindly. I keep an eye on it from time to time and disable some skills and focus others. It's much less micro-management than using manual mode though. I do switch to manual mode on occasions, like when finding books.

I made a change a few weeks ago so that manual and auto mode have separate configurations. You usually have few skills enabled in manual mode and many in auto mode. So if you switch to manual and disable everything but one skill, then later switch back to auto, your previous auto configuration will be restored. And when you switch to manual again, you'll get your "few skills on" configuration back, whatever you did in auto mode. If this isn't clear, just try it in trunk and you'll see. This make using both modes in the same game much more convenient in my opinion.