How Many Skills to Train

[LoopyLeo]

So, despite it being the middle of my exams, DCSS has been running consistently on my laptop I'm wondering how many skills one should train at a time, especially when playing a mage character. Fighters I generally leave on weapon, fighting and shield, with a focus on weapon skill -- is that "right"? It seems odd trying to get to min-delay whilst maintaining a shield penalty :/
Then on my mage characters is where I have the biggest trouble -- I've generally been leaving on fighting/dodging (I generally play squishier casters), whilst also training element school AND spellcasting -- I always read people saying not to bother training spellcasting that much, but I find I personally can't seem to survive without the extra MP.
I guess what I want to know is what one would consider as close to optimal for the different classes

[VeryAngryFelid]

I was going to write a detailed answer but probably it is better to write a short version "train offense first, then defense, then more offense".
The below is taken from DEIE of Sif Muna of one of the leaders of current tournament, take a look at first 10 XLs as above that you can do whatever

  Code:

Skill      XL: |  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 |
---------------+----------------------------------------------------------------+-----
Ice Magic      |  4              5  6  8     9                                  |  9.4
Conjurations   |     2  3  4                                                    |  4.4
Summonings     |           1  4  5  6  7     9 13 14       16 17 18 19 20       | 20.1
Spellcasting   |                       6     8    11 13          14 16 17 20 22 | 22.6
Fighting       |                       1  4                             7  9    |  9.2
Stealth        |                       3  5              9                      |  9.0
Dodging        |                          7 10          12 15    17 18          | 18.5
Evocations     |                                3  7  8  9                      |  9.7
Invocations    |                                      8                       9 |  9.6
Shields        |                                         6                      |  6.0
Charms         |                                              10 11             | 11.4
Air Magic      |                                               7  8             |  8.1
Translocations |                                                     4          |  4.3
Short Blades   |                                                     1  7  8    |  8.2


[VeryAngryFelid]

HuFE of Sif Muna, high score of current tournament if you need a more conventional caster.

  Code:

Skill      XL: |  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 |
---------------+----------------------------------------------------------------------------------+-----
Conjurations   |     2     3  4     5  7  8 10 11 13                14 15                         | 15.5
Spellcasting   |              3     4     6     8  9    10 11 12             13 14 15 16       17 | 17.9
Fire Magic     |              4     5     7     8  9 10    12 14    15 16 17 18 19       20    21 | 21.0
Translocations |                    1  2  4                                                     9 |  9.1
Fighting       |                             3  5  6  8 10 11 12    13    14 15 16    17 18    20 | 20.6
Shields        |                                      5                          6 10 12 14 16 17 | 17.0
Dodging        |                                      4  6  8  9    11 12 13    14             15 | 15.1
Invocations    |                                      2  4        5                               |  5.9
Armour         |                                                  4                            13 | 13.0
Summonings     |                                                                       3  6  8    |  8.9
Stealth        |                                                                                  |  2.0

I would probably train Dodging/Fighting much earlier (like the player who did DEIE) but still only after getting Conjure Flame online.

[Magipi]

Fighters I generally leave on weapon, fighting and shield, with a focus on weapon skill -- is that "right"? It seems odd trying to get to min-delay whilst maintaining a shield penalty :/

I'm just curious, when you write about fighters above, that means specifically only fighters, and not berserkers or gladiators or whatever, right? Or do you always pick up a shield with those too, and always choose shields and 1-handed weapons no matter what?

[Majang]

I'm wondering how many skills one should train at a time, especially when playing a mage character.

That's a worthy question, and whatever response you get, it won't be undisputed by others. Here is my version:

You are right about spellcasting - I leave this on until I reach level 27, often focused, because it is the one skill that helps all your spells. It has the same contribution to spellpower than any of the spell-schools involved, and for a caster spellpower matters at least as much as spell success. Ironshot at 6# spellpower doesn't hit a barn, but at 8# it is an entirely different matter.

It is right to only train your offensive skills at the beginning of the game, and these are spellcasting and your beginning spell schools. I don't even have fighting or dodging on until my beginning spells hit 4# or 5# spellpower. Then I branch out, quite dramatically so, to train fighting, dodging, one weapon, a second spell school, possibly armour, and certainly evocations on top of my spell skills. Good that I have Vehumet, so he does not need any invocations. Since I have SC in focus, this easily means that the other spells will only receive 10% of the experience. Consequently, you are looking at a stage of slow, but steady growth. Many players dislike that, but since you really need all these skills, and soon, it does not matter whether you train them all in one go, or sequentially. Usually, when I am done with Lair and Orcs, and maybe one of the side branches, I switch off evocations, the weapon skill, armour (if I don't switch it off before, usually around 5 levels). From then on the focus is again more on your magic skills, plus fighting and dodging. These I keep until my high-level spells are castable. Then again I add more of the other skills, without switching off the magic skills, as they still need to increase in spell power.

[4Hooves2Appendages]

[...] I'm wondering how many skills one should train at a time,

Training one skill at a time is strictly best (as long as that's the most useful skill to train of course). Many players don't do this because it can be a bit tedious, as it requires regular checking of the skill screen and making changes as required. I usually do that until my character feels strong.

especially when playing a mage character.

There's no real training strategy difference between a 'mage' character and a 'fighter'. Both need to train some offense in the beginning, then get some defences, then get more offence and some utility, and so on. Mage backgrounds (IE, FE, Su, etc.) often start with some really strong spells in their book. It's usually most important to get those spells memorised and to a low fail rate. Fire Elementalists benefit enormously from Conjure Flame for example.

Please note that you do not need to race to get all the spells in the book castable. Trying to go straight for Fireball is usually a mistake, for example.

In order to get a spell to the desired fail rate it is best to split XP fairly evenly between all the involved skills. For Conjure Flame that is Fire and Conjurations. For other spells it may only be a single spell school. Please note that XP is not the same as skill levels. The skill screen can show you the relative XP cost to get to the next whole level of a skill. Try to keep those numbers similar. Picking Int for stat choices will often help to get the fail rate down and the spell power up. Once you have a reliable way to kill enemies, add some defences like training fighting and dodging.

Fighters I generally leave on weapon, fighting and shield, with a focus on weapon skill -- is that "right"? It seems odd trying to get to min-delay whilst maintaining a shield penalty :/

Generally speaking, training shields in the early game is a mistake. I mainly split my XP between weapon skill and fighting in the beginning, with characters that kill in melee combat. I put an emphasis on weapon skill. You don't need to train to min-delay immediately, but it should be a medium term goal. Once you find a decent armour, like chain or plate, train a little armour skill to pick up a cheap point of AC or two. If you can pay enough attention, then you can train armour and dodging precisely to the point where you get an extra EV or AC, then decide whether to continue training, or to change to something else.

Then on my mage characters is where I have the biggest trouble -- I've generally been leaving on fighting/dodging (I generally play squishier casters), whilst also training element school AND spellcasting -- I always read people saying not to bother training spellcasting that much, but I find I personally can't seem to survive without the extra MP.[...]

See above, train to get the good spells of the background, like Conjure Flame. I would focus on that first and train fighting/dodging a little later. Finally, I usually only train spellcasting for extra spell levels, or if it is much lower than my main magic skills (like when it requires only a quarter of the XP to raise to the next level).

There can be multiple reasons for running out of MP. Training your damage spell skills (like conjurations/fire) will increase their damage and make them more MP efficient. It also increases your damage per turn. FInally it also increases your chance to hit with some spells, or your chance to overcome enemy magic resistance with other spells.

It is often best to fight enemies one at a time, partly because MP is quite limited in the beginning of the game. Once attracting an enemy, pull back to a safer area, then kill it there, wait to regenerate MP, then look for the next enemy. Many book backgrounds start with some stealth making this relatively straightforward.

[LoopyLeo]

Fighters I generally leave on weapon, fighting and shield, with a focus on weapon skill -- is that "right"? It seems odd trying to get to min-delay whilst maintaining a shield penalty :/

I'm just curious, when you write about fighters above, that means specifically only fighters, and not berserkers or gladiators or whatever, right? Or do you always pick up a shield with those too, and always choose shields and 1-handed weapons no matter what?

Fighters I actually mean the class Fighter, for which I pretty much almost exclusively go 1-handed + shield. Gladiators I go for 2 handers most of the time, and berserkers I practically never play (probs should though, seeing as I've only ever won twice).

In regards to everyone else: thanks heaps for the replies, they've all been pretty helpful. Obviously they're opinionated, there's no "perfect" answer (mathematically though, I guess there must be ), but they've all been helpful.

I'm miles from being good player, but I'll get there someday, from the ground up

[Siegurt]

You are right about spellcasting - I leave this on until I reach level 27, often focused, because it is the one skill that helps all your spells. It has the same contribution to spellpower than any of the spell-schools involved

This is incorrect. Spellcasting has 1/4 of the effect of the spell school for a single school spell.

Specifically your spellpower (and success chance) is based on the average of the relevant spell schools, plus 1/4 of your spellcasting.

For spell power purposes, spellcasting is therefore 3/4 as helpful as a school in casting a 3 school spell, 1/2 as helpful for casting a 2 school spell, and 1/4 as helpful for casting a single school spell.

Of course there are other reasons to train spellcasting than power, but training it for power alone should optimally be done only when it is cheaper than the school (xp wise) for the same effect.

In answer to the OP: personally i find that while training 1 or 2 skills at all times is usually optimum, I am lazy and typically divide my xp into quarters, spending one quarter or one half of it on the things that are important (so i will typically have 2, 3, or 4 things in progress at a time) this isn't perfectly optimal, but it seems like a good balance between optimal and not too micromanage-y for me, personally.

[Majang]

Training one skill at a time is strictly best (as long as that's the most useful skill to train of course).

personally i find that while training 1 or 2 skills at all times is usually optimum

On which basis do you make this claim? I don't understand how this can be "strictly best" or "usually optimum", as opposed to learning the three or four (or nine) skills I need next simultaneously. Numerically I don't see any advantage to first raising dodging one skill level, then armour a skill level, compared to training dodging and armour at the same time. I will arrive at exactly the same time at the point where both skills are raised a level. And I don't think it would be game breaking or making in many instances that I get one of these skills to the target level before the other. I suspect that what you call "best" or "optimal" is pure psychology, as training one skill at a time fabricates an illusion of faster progress, as your one skill indeed grows in leaps (while all your others don't move at all).

Spellcasting has 1/4 of the effect of the spell school for a single school spell.

Specifically your spellpower (and success chance) is based on the average of the relevant spell schools, plus 1/4 of your spellcasting.

For spell power purposes, spellcasting is therefore 3/4 as helpful as a school in casting a 3 school spell, 1/2 as helpful for casting a 2 school spell, and 1/4 as helpful for casting a single school spell.

You are right, I misrepresented that. This additional bonus to spell power and success rate is still nothing to sneeze at. Additionally, some popular support spells, such as apportation and blink, you can use at maximum spell power (and perfect success) with a high spellcasting skill without ever having to invest a single experience point into the relevant spell school. As spellcasting helps with the success and power of all your spells, a 1/4 to 3/4 bonus is an excellent return on your investment if you don't restrict your spellcasting to just one school.

[ThreeInvisibleDucks]

Training one skill at a time is strictly best (as long as that's the most useful skill to train of course).

personally i find that while training 1 or 2 skills at all times is usually optimum

On which basis do you make this claim? I don't understand how this can be "strictly best" or "usually optimum", as opposed to learning the three or four (or nine) skills I need next simultaneously. Numerically I don't see any advantage to first raising dodging one skill level, then armour a skill level, compared to training dodging and armour at the same time. I will arrive at exactly the same time at the point where both skills are raised a level.

I've always thought it's more optimal, for example, to first train Armour to get a point of AC, then Dodging to gain a point of EV, instead of training them at the same time which takes longer to get either.

[VeryAngryFelid]

Numerically I don't see any advantage to first raising dodging one skill level, then armour a skill level, compared to training dodging and armour at the same time.

It's not about increasing skill levels by 1, it's about increasing AC/EV by 1. It's optimal to train Armour to get 1 AC and stop immediately, then train Dodging to get 1 EV and stop immediately, then train Fighting to get 1 HP and stop immediately etc. Compare to simultaneous training of Armour/Dodging/fighting, you get no benefit until either AC or EV or HP increases and it will take longer because you train several skills instead of just 1.
The same applies to magic, you don't get any benefits for increasing Fire from 3.1 to 3.2

[Majang]

Yes, I see your reasoning now. But it is an optimality that gets indeed paid for by an insane amount of micromanagement, for a very marginal return.

[Majang]

I've always thought it's more optimal, for example, to first train Armour to get a point of AC, then Dodging to gain a point of EV, instead of training them at the same time which takes longer to get either.

...but the same time to get both.

[ThreeInvisibleDucks]

FR: Take fractional part of AC and EV into account under the hood so no micromanagement is necessary. (I wouldn't be surprised that Crawl already does that, IIRC it takes fractions into account in other places.)

[4Hooves2Appendages]

As far as I know crawl only uses whole numbers of defences for the sake of avoiding/reducing damage. I don't know if there's any rounding involved, but I suspect you don't get the convenience of rounding up.

[4Hooves2Appendages]

I've always thought it's more optimal, for example, to first train Armour to get a point of AC, then Dodging to gain a point of EV, instead of training them at the same time which takes longer to get either.

...but the same time to get both.

Sure, but for around half the time, depending on the required XP ratios, you have 1 AC or 1 EV more. Now you could argue that 1 AC doesn't make much difference, but I would disagree, especially in the early game.

[VeryAngryFelid]

If 1 AC does not count, get 1 EV first, it always counts

[Majang]

Sure, but for around half the time, depending on the required XP ratios, you have 1 AC or 1 EV more. Now you could argue that 1 AC doesn't make much difference, but I would disagree, especially in the early game.

Now, coming back to the OP's question, for a caster, who does not train fighting or dodging in the early game, there really is no reasonable need to micromanage your skills. Spell power and and spell success are not as discrete values as AC or EV.

[VeryAngryFelid]

Now, coming back to the OP's question, for a caster, who does not train fighting or dodging in the early game, there really is no reasonable need to micromanage your skills. Spell power and and spell success are not as discrete values as AC or EV.

Right, spell success is more discrete and it takes duvessa's level of knowledge to micromanage those skills.

[Majang]

it takes duvessa's level of knowledge to micromanage those skills.

It would be interesting to hear whether she actually does that...

[VeryAngryFelid]

it takes duvessa's level of knowledge to micromanage those skills.

It would be interesting to hear whether she actually does that...

I was not saying he micromanages that, I was saying you would need to be an expert in crawl mechanic to know how spell success is calculated. As far as I remember the failure chance changes at x.5 and x.0 skill levels for 1 school spell.

[4Hooves2Appendages]

Sure, but for around half the time, depending on the required XP ratios, you have 1 AC or 1 EV more. Now you could argue that 1 AC doesn't make much difference, but I would disagree, especially in the early game.

Now, coming back to the OP's question, for a caster, who does not train fighting or dodging in the early game, there really is no reasonable need to micromanage your skills. Spell power and and spell success are not as discrete values as AC or EV.

There are clear break-points for spell training, like when a spell turns from red to yellow, or from yellow to white. Or when the spellpower is maxed, as it can get quite quickly for level 1 spells. I micromanage that in every game, as I said, until the character feels strong.

I don't know if it's been changed yet, but spell success used to be granular, not continuous, making regular checking, and training to specific points, optimal.

For a 2-school spell, with both aptitudes the same, currently at the same skill level, you could make an argument that it's not worh the hassle to train them individually. There is still a benefit of course, but it's fairly small.

However, for any multi-school spell you care, especially in the beginning, train the cheapest skill to the next whole skill level, then train the next cheapest skill (or the same if it's still the cheapest) to the next whole level, then, well you get the drift. That is the most XP-efficient way to get your spell to where you want to be.

As mentioned above, you can mix in spell-casting occasionaly, if you understand that it contributes a quarter.

[VeryAngryFelid]

You still can get max severity miscast from gray spell, the color just roughly shows probability of that so it is not that important to get a spell to yellow (or green) before casting it.

[Majang]

I'm still to be convinced that this kind of micromanagement has any practical bearing on the real life of a caster. Let's envision, for the sake of the argument, a typical early mid-game EE. In terms of spells, he wants to max out his sandblast and stone-arrow spell power, while at the same time improve the castability of LRD, and get Bolt of Magma online. All these spells are situationally important for the fighting style of an EE, so that involves training three different spell schools (earth, conjurations, fire) plus spellcasting, all during a phase of the game when he prudently also works on improving his defences with fighting and dodging, and where he urgently needs a few pips of evocations in order to activate flight. None of these needs can claim any particular priority (I guess you can ignore evocations for much longer, but that does not change much). Micromanagement (based on what?) does not really address the need that there is a phase in the game where you need to develop in many directions simultaneously - it just masks the need by letting you hyperactively move dials without actually accomplishing anything meaningful.
And even if I have a spell castable, that is no reason for me to stop training the relevant skills, as castability is only a minor part of making a spell useful. I can have Iron Shot at 10% spell success, but it would be useless against all but the most inert targets, unless I get its spell power up a lot more.

[VeryAngryFelid]

I can have Iron Shot at 10% spell success, but it would be useless against all but the most inert targets, unless I get its spell power up a lot more.

I am not sure. It would be nice to see a table with
Average Conj/Earth skill - Accuracy - Spell power - Damage roll formula
for a character with specific Int and Spellcasting but I doubt it will ever be created, at least personally I am too lazy

[Majang]

My statement is more based on personal experience. Once I confidently took on St. Roka on the lowest level of the Orcs, with my freshly learned and nicely castable Iron Shot, and I learned it the very hard way that most of my shots didn't even hit the feller. Those who did caused only minor damage. Since then I tend to avoid him until my shot is at 7 or 8#. With this, he dies to three or four shots.

[watertreatmentRL]

If you really want to get the most out of your experience, you should only train one skill at a time and switch skills frequently on level up. In practice, I don't bother to do this unless I'm very serious about winning, which I basically am not and haven't been for a long time.

I play a lot of fighters lately because it's just a very straightforward game. What I normally do is train shield and weapon skill equally for a while, usually to about 7 or 8 and then add in fighting or armor, later some dodging, one at a time, and (sometimes) focus weapon until I hit the weapon and shield breakpoints I need. I generally don't want those other skills to be more than about half the weapon skill until then. Somewhere in there I'll train some invo if I've gone Oka or one of a few others. After that, depending on god choice I'll mix in throwing or evocations, and train dodging more into the late game. It's not a super optimized thing, but it's reasonably easy and consistent.

[4Hooves2Appendages]

Micromanagement (based on what?) does not really address the need that there is a phase in the game where you need to develop in many directions simultaneously - it just masks the need by letting you hyperactively move dials without actually accomplishing anything meaningful.

I think this is the crux of our disagreement. I doubt that there are characters that 'need to develop in many directions simultaneously', while at the same time not really being sure what you actually mean.

if you mean that sometimes you have different development options, then I agree with that. I disagree that they all need to be pursued simultaneously. In fact, I would strongly suggest to pick a priority and train for it, then pick the next priority, possibly review if the situation changes. That way you actually get something tangible to use, like better spell power or success for a useful spell, a couple of points of EV, lower fail rate on a useful god power, some power for your good wands, min-delay on nice weapon, some HP, or whatever you decide is most important now.

Inability to decide what training to pursue, especially in the early game (later there is a surplus of XP anyway), is usually a sign of inexperience. Sometimes it can be a refusal to bother thinking the options through, or spending any time on the decision, usually by an experienced player, or perhaps even a speed-runner. That is by no means the best way to strengthen a given character.

[VeryAngryFelid]

My statement is more based on personal experience. Once I confidently took on St. Roka on the lowest level of the Orcs, with my freshly learned and nicely castable Iron Shot, and I learned it the very hard way that most of my shots didn't even hit the feller. Those who did caused only minor damage. Since then I tend to avoid him until my shot is at 7 or 8#. With this, he dies to three or four shots.

I wouldn't treat a single fight as a proof of anything, crawl fight is too random. For reference, a long time ago I created a thread in GDD after it had taken me 17 attacks to kill a monster where fsim was showing I was expected to kill the monster in just 2 attacks.

[Majang]

I think this is the crux of our disagreement. I doubt that there are characters that 'need to develop in many directions simultaneously', while at the same time not really being sure what you actually mean.

Yeah, I guess we can leave it at that. But I do observe that people here are repeating over and over again that successive skill training is "optimal", without so far showing how this translates into tangible benefits, beyond having a pip of something a few hundred turns earlier (and something equally desirable a few hundred turns later). I can't grasp the logic of that.

[gameguard]

I think this is the crux of our disagreement. I doubt that there are characters that 'need to develop in many directions simultaneously', while at the same time not really being sure what you actually mean.

Yeah, I guess we can leave it at that. But I do observe that people here are repeating over and over again that successive skill training is "optimal", without so far showing how this translates into tangible benefits, beyond having a pip of something a few hundred turns earlier (and something equally desirable a few hundred turns later). I can't grasp the logic of that.

i dont know whats so hard to understand about this issue. Obviously its better to get 1 thing now then another thing after, rather than getting both things later. I agree that its tedius and almost no one does this though.

[Majang]

Yes, I do get this point. But people here in this forum treat successive skill training as somewhat essential to survival, and call simultaneous skill training bad, as if someone who does that is already doomed from the outset. But when you ask why, it all boils down to an extremely marginal effect, paid for with extreme tedium (which, as far as I understand the game design philosophy, is something that all want to avoid).

[duvessa]

it takes duvessa's level of knowledge to micromanage those skills.

It would be interesting to hear whether she actually does that...

She does. Also, it doesn't take much knowledge to optimize spellcasting vs. spell schools now that skill costs are shown; all you need to know is that 1 level in spellcasting is like 0.25 levels in all schools, plus the spell levels and MP. It's just a matter of deciding how much you value spell success and power compared to spell levels and MP.

[Siegurt]

Yes, I do get this point. But people here in this forum treat successive skill training as somewhat essential to survival, and call simultaneous skill training bad, as if someone who does that is already doomed from the outset. But when you ask why, it all boils down to an extremely marginal effect, paid for with extreme tedium (which, as far as I understand the game design philosophy, is something that all want to avoid).

You'll note that the phrasing used in this thread was "strictly best" and "usually optimal" i think both of those acknowledge that the effects are marginal, and don't presume it is required for survival (i have seen several threads claiming you can just turn on auto training for the whole game and that is probably good enough to win in most cases) but there are always those who will interpret anything short of perfect as bad.

Even taken to its most extreme (only train one skill at a time ever) the amount of tedium isnt incredibly high (compared to say, luring every critter back to fight it on the up stairs) and you can get about 90% of the (marginal) benefits for 10% of the tedium by limiting yourself to 4 or less skills at a time

[4Hooves2Appendages]

That's a good point about the difference between obsessive skill-changing and excessive luring. Luring takes way more time, but is quite a bit more powerful too. Hmm.

[Majang]

You'll note that the phrasing used in this thread was "strictly best" and "usually optimal" i think both of those acknowledge that the effects are marginal, and don't presume it is required for survival.

You are right about this thread. I have seen another thread, the one discussing a certain Caster Walkthrough, where you find statements like:

It recommends training something like 8 skills at once in the midgame, which I'm fairly certain is worse than automatic skilling, among other things.

[VeryAngryFelid]

I have seen another thread, the one discussing a certain Caster Walkthrough, where you find statements like:

It recommends training something like 8 skills at once in the midgame, which I'm fairly certain is worse than automatic skilling, among other things.

This is a valid complaint, training 8 skills at once is much worse than autotraining. Autotraining is rather good, even when autotraining 20 skills you still can get 60% XP go into a single skill while for manual training having 8 skills means you get exactly 12.5% XP go into every skill, which means you will get next AC/EV/HP etc. much later than you might with training just 1-2 skills manually or autotraining 20 skills.

Edit. I suggest you to play a Be in autotraining mode to see what I mean, you will be surprised how close it is to manual training of a good player. You will even see your Stealth training drop to almost zero as soon as you equip a plate armour

[Magipi]

the effects are marginal

A lot of times they are marginal, sometimes not.

Let's say you decide you want more AC and more EV. The optimal way would be to train armour until you get +1 AC, then switch to dodging until you get +1EV, then switch back. I don't think anyone does that. You simply train them both. And yes, in some periods you are -1 AC, but yeah, that is marginal.

Sometimes the effect of is much more dramatic. I've seem people find some good spell (let's say Controlled Blink) and start training for it, but they train 5 skills simultaneously. Like instead only training Tloc, they train tloc + fighting + dodging + armour + evo. This means that the spell takes 500% more time to get online. Instead of getting it online by clearing 1 floor, it takes 5 floors. This means that for 4 floors you have no cblink when you could have. This is not marginal. (Yes, you have like +1 AC and +1 EV an +4 hp, or maybe not even that, depending on the skill levels and whatnot.)

(And yes, in the latter situation training spellcasting is also most likely a mistake, especially if your spellcasting is already very high.)

[TonberryJam]

A standard method is to start with training your primary kill weapon or spell. Switch to defense training when you reach min delay or enough spellpower. Then, repeat this process when you have a better weapon or spell or need more spellpower. The transition between offense and defense is where the number of factors determining when to do this occurs over the course of the game.

Number of skills trained at a time is determined by your class/race apts./god choice.
A Minotaur can easily train 4 it's +2 apt skills and keep pace with the dungeon progression.
A Troll may train UC and/or Fighting because it has poor apts and gets the most out of these two at the start with it's high Str.
A Caster would usually train the school(s) of its primary attack at the time. Then may train spellcasting to deal with hunger.

Keeping the shield on or not depends on various factors. Large races with high str can afford to leave it on at the start. Good shield apts can afford to leave it on at the start. Trog or Oka worshipers should leave it equipped at the start. Everyone else is better of training shield with it off or downgrading to a buckler which needs only 4 skill levels I think. A +3 buckler is almost just as good as a normal shield.

For casters, it's important to focus on your spells. My regular method is to first max the spellpower of the level 1 attack spell. Next, train spellcasting to make it hunger less. Then train the next spell to 1%. Finally, train spellcasting to reduce hunger. Repeat this method, aim for around 5-6 spellpower before training fighting/dodge/shield/armour. I find doing this often provides just enough MP to survive, typically never needing more spellcasting than your current level.

And that doesn't cover any lucky finds or tactical god choices or perks of race/class combos or useful evocable.

[VeryAngryFelid]

A Caster would usually train the school(s) of its primary attack at the time. Then may train spellcasting to deal with hunger.

For casters, it's important to focus on your spells. My regular method is to first max the spellpower of the level 1 attack spell. Next, train spellcasting to make it hunger less. Then train the next spell to 1%. Finally, train spellcasting to reduce hunger. Repeat this method, aim for around 5-6 spellpower before training fighting/dodge/shield/armour. I find doing this often provides just enough MP to survive, typically never needing more spellcasting than your current level

I agree with just last sentence here. Ignoring spellcasting for MP and ignoring defenses before getting 5-6 spellpower leads to easy death. It's better to have more MP (1 MP for every level in spellcasting) and defense for comfortable play, this way you can do some mistakes or even not curse RNG when you get a miscast or two.

[Magipi]

Autotraining is rather good

Autotraining is awful beyond words. You can get easily get away with it if you are a berserker - it's hard to train a berserker badly, and even though autotraining tries hard, it fails

If you are a mage, autotraining is borderline suicide. You will be very slow to train up new skill (for new spells) because no matter what you do, autotraining will fuck you over.

What is even worse, autotraining is a somewhat hidden feature. I was playing for months, already got a few wins before I realized that it exists. I was absolutely shocked.

[VeryAngryFelid]

Autotraining is awful beyond words. You can get easily get away with it if you are a berserker - it's hard to train a berserker badly, and even though autotraining tries hard, it fails

If you are a mage, autotraining is borderline suicide. You will be very slow to train up new skill (for new spells) because no matter what you do, autotraining will fuck you over.

What is even worse, autotraining is a somewhat hidden feature. I was playing for months, already got a few wins before I realized that it exists. I was absolutely shocked.

Autotraining is bad for branching into new high level spells, I agree (that CBlink will never be castable, for example). Yet it is completely ok for typical casters who continue using magic they started with, my SpVM enjoyed increased training of spellcasting, for example and all those venom and draining bolts were very helpful

[Siegurt]

What is even worse, autotraining is a somewhat hidden feature. I was playing for months, already got a few wins before I realized that it exists. I was absolutely shocked.

It's on by default, you had to have turned it off yourself at some point, or perhaps copied someone else's ini file or something...

[VeryAngryFelid]

It's on by default, you had to have turned it off yourself at some point, or perhaps copied someone else's ini file or something...

True, I played my last online game with autotraining until about Lair before I realized something is wrong
That's why I am saying there is no much difference for Be.

[Magipi]

What is even worse, autotraining is a somewhat hidden feature. I was playing for months, already got a few wins before I realized that it exists. I was absolutely shocked.

It's on by default, you had to have turned it off yourself at some point, or perhaps copied someone else's ini file or something...

No, you misunderstand me. That it is turned on by default was exactly what shocked me. I played for months thinking that my xp goes to the skills I want them to go, but then I learned that no, the program deecides where to put them almost no matter what I do.

[VeryAngryFelid]

I played for months thinking that my xp goes to the skills I want them to go

Well, telepathy has not been proven to exist yet

but then I learned that no, the program deecides where to put them almost no matter what I do.

How is that? As far as I know it depends solely on what you do. It is my experience also. Or do you mean you cannot say in advance what will be trained after killing next monster? That's true, but probabilities are still present and depend on your actions.

[4Hooves2Appendages]

Let's say you decide you want more AC and more EV. The optimal way would be to train armour until you get +1 AC, then switch to dodging until you get +1EV, then switch back. I don't think anyone does that. You simply train them both. And yes, in some periods you are -1 AC, but yeah, that is marginal.

I do that every time, and it seems unlikely that I would be the single unusual person out there.

The last time I trained armor and dodging simultaneously was probably in 0.16 or 0.15 with an extremely strong extended character. I would suggest, somewhat flippantly, that if I really prioritised mild convenience over marginal advantages, I would probably stop playing crawl.

[Magipi]

but then I learned that no, the program deecides where to put them almost no matter what I do.

How is that? As far as I know it depends solely on what you do. It is my experience also. Or do you mean you cannot say in advance what will be trained after killing next monster? That's true, but probabilities are still present and depend on your actions.

That is exactly the problem. Think of the cblink example above: I don't use cblink AT ALL when I am training the skill up. So no matter that I turn everything off, it will get only a fraction of the incoming xp. And the most absurd thing is that most new players don't even know that they are fucked over by that autoshit. This is really mind-blowing.

This is nothing but the relic of the ancient past. When was the whole victory dancing stuff removed from the game, 10 years ago?

[VeryAngryFelid]

That is exactly the problem. Think of the cblink example above: I don't use cblink AT ALL when I am training the skill up. So no matter that I turn everything off, it will get only a fraction of the incoming xp. And the most absurd thing is that most new players don't even know that they are fucked over by that autoshit. This is really mind-blowing.

This is nothing but the relic of the ancient past. When was the whole victory dancing stuff removed from the game, 10 years ago?

Sorry, I am lost here, I was talking about having everything enabled. If you turn everything off except Translocations, Translocations will get 100% XP, just like happens in manual training mode.
I disagree with your claim that ignoring Translocations by autotraining if you never cast any Translocations is bad, it is ideal solution IMHO. Do you really want Troll with Int3 in GDA train Translocations for CBlink? Why should it be different for other characters?

[Magipi]

If you turn everything off except Translocations, Translocations will get 100% XP, just like happens in manual training mode.

Are you sure about this? I think this does not work like that.

As for your example, what does species or int or armour have to do with anything? My point was that not whether cblink is good or bad. My point was that when I allocate how to train, I do that because that is how I want it to be. The program overriding my decisions is clearly absurd.