optimal skill management misconception

[letownia]

I have many times seen this sort of advice in these forums:

UC till 10
Fighting till 5.
X skill till Y.

This guideline is then backed up by the assumption that it's "always better to train one skill at a time".

These seems completely false to me. I know the wiki is wrong at times, but according to http://crawl.chaosforge.org/Skills#Experience_Required, the amount of experience to go from level 0 to level 4 in a skill is about the same as going from level 8 to level 9 of a skill.

Clearly, a fighter with 4 fighting 8 UC, is better than 0 fighting 9 UC. You get more +dmg, +acc, +hp, at the cost of decreasing attack speed by less than .1. This clearly applies to other skill situations.

Do any of you actually follow this guideline literally? Ideally it'd probably be best to do gradual increments like UC till 3, fighting till 1, UC till 4, fighting till 2, UC till 5, etc. But this is a lot of micro-managements and now that we have partial skill bonuses (2.5 dodging is better than 2), even this approach isn't correct.

So IMO it seems reasonable and near optimal to simply use the skill screen and divert 2/3rds experience to UC, while diverting 1/3 to fighting (unless you want to be constantly switching to the skill screen every couple of kills). You can update skill allocation as needed every level or so.

How do you skill your characters? Thoughts on my argument?

[Galefury]

Well, yes. Same for dual school spells. And picking up the cheapest defensive skill levels quite early also is usually a very good idea.

Breakpoints mess with this though. For example when you want to learn a new spell and don't have enough slots for it. You want to train spellcasting either 100% or 0%, because every point of exp you put in that does not get you that final slot you need is almost worthless. All you want to do is get the required number of slots as soon as possible. If there are other skills that are more important, you train those first. If they are less important, you train them after reaching the breakpoint. But you don't split your exp.

Something similar can happen for weapons due to mindelay. The closer you get to min delay, the bigger the return from each skill level gets. This increasing return can sometimes surpass the rising exp cost.

Another thing that makes it good to train skills 100% is finding stuff. If you have no good evocables, you don't train evo. If you find a great rod, training evo may suddenly be by far the best way to spend your exp.

So of course you are actually right, but there are many things that make you often want to train only one skill at a time regardless.

[Sar]

it just werks

[letownia]

@Galefury

Good points - the spellcasting skill can definitely be worth investing in 100% (if you need to memorize a spell asap), and same goes for trying to get some single-school spell to a reasonable success rate (although points in spellcasting can help with this too ). Other skills, depending on the situation might also be optimally trained at 100%.

My main point is that I feel like the "train skill X to Y, followed by W to Z" is misleading and that it isn't true that training 1 skill at a time is always optimal.

[dck]

I train defensive skills by keeping an eye on my armor type, current EV/HP/XL and dex.
Actual skill number doesn't matter for pretty much anything other than silly spells like ozoarmor etc and I don't think one should pay attention to it.

[damiac]

That's a good point. I think advice is given the way you point out due to simplicity, it's easier to say 'train weapon till min-delay, then fighting, etc...' than to say 'train weapon skill until it's reasonably effective, then maybe grab some cheap defensive skills depending on what you have for armor, etc.'

Simple advice is easier to give and follow than more complex advice, but you're right to say it's not completely optimal. I do a lot of what you're saying here, training some early fighting/dodging before I'm even close to min delay, while still keeping most of my training focused toward my main weapon.

Clearly, a fighter with 4 fighting 8 UC, is better than 0 fighting 9 UC. You get more +dmg, +acc, +hp, at the cost of decreasing attack speed by less than .1. This clearly applies to other skill situations.

This is a very good point, and quite well stated.

[KoboldLord]

'Train one skill at a time' is an oversimplification. Often, it is useful to train more than one skill at a time, but the more common new player mistake is to train lots of unrelated skills at the same time, even if one or two of those skills is vastly more important than the others at that particular point in the game. This is particularly likely to be the case in the early game, where the new player is likely to be, because it is at this point in the game where getting your starting spellbook working or getting your starting weapon to be a little bit faster is the most critical.

A more complete heuristic to follow is that you should train skills with a specific short-term goal in mind, and only train skills that immediately contribute towards that goal. If your goal is to get to Sticky Flame, for instance, you need a bunch of fire magic and conjurations, and possibly a level or two of spellcasting but no more than that. A new player will get much worse results in the initial rush to fill out their spellbook if they also train unrelated skills at the same time, but they will get almost exactly the same results by switching back and forth between fire magic and conjurations every time one of them levels up, and it's much easier to explain than fiddling with ratios.

[Sar]

it's easier to say 'train weapon till min-delay

That's a terrible advice and I'm pretty sure no one gives it anymore at least here, thankfully. I know I was guilty of that as well, and I do feel regret for spreading that terrible piece of misinformation.

[XuaXua]

Train weapon till minmay.

[crate]

It is strictly beneficial to allocate every single point of xp to the skill that at that particular time and skill level gives you the most benefit per xp. You can't quite go this far; training only one skill is the closest you can get.

In practice there are almost always several different skills that offer similar benefits, and additionally skill training is much less important than playing well tactically, so no players actually do the ideal thing and adjust skill training after every single fight. Many players go further and train more than one skill at a time (and in fact I often train 4-5 skills at once even though doing so is strictly worse than training them individually one at a time, because I feel that the reduced interface hassle from having to press "m" one-tenth as often is well worth the tradeoff).

[Lasty]

crate's got the right take on the actual point of the "train one thing" advice, but as they point out, no one follows it to the logical conclusion, and as the OP points out, sometimes for your purposes the optimal result may involve training in a less single-minded fashion. The most-practically-correct version of this advice is "do whatever helps you put your XP most efficiently towards your top priority, adjusted for your tolerance for opening the skill allocation screen". For some people that means training a single skill at a time, for others it means training 3+ skills because not looking at the skill screen helps them focus on other aspects of good play.

I personally tend to train one thing at a time, where that thing is what I think the character needs most. I break this rule whenever I think there are a handful of things I want equally much and I know I'm going to get the XP to get all of them to my goal in a reasonable period of time -- e.g. if I have a bunch of low-level utility spells that I want to get to a lower failure rate and my character is otherwise in good shape, I might turn on as many as 5 different spell skills and turn them off each time one of them reaches 3 skill or so (adjusting for how much skill I need to get them to the power/failure level I'm comfortable with). I also tend to aim for somewhat arbitrary round-number breakpoints on skills even though it's not optimal, just because it's easy for me to keep an eye on when skills hit those skill levels, and I only deviate if I find that I have a need to increase a skill I've been neglecting. It would be better to proactively evaluate which skills I need every time I gain a skill level or whatever, but there's a limit to how much time I want to spend micromanaging skills.

[Blink]

Some ballpark figures:

As a melee type, I train my weapon only until I can get the attack speed to 1.0 aut. I do this because otherwise every time I take a swing, my opponent might get in two hits. Once I hit 1.0 speed though, I will start branching out into something else--probably defenses. I'll want to hit minimum delay around the time I am finishing off Dungeon and starting on runes.

As a caster, I train my main attack spell's school(s) until it hits 4-5% failure. I don't much worry about spell hunger and train Casting in anticipation of needing more slots soon (e.g., I have 1 free and I can see I need 2 more for Meph or something).

For defenses, I aim to have about 30 combined AC/EV/SH by end of Lair, and then another 10 for each rune--so 60 by the time I have all three. By end of Depth, I want another 10, so in Zot I'm looking at 70+ combined defenses. What that combination is will depend on my apts and finds.

For hp, I want to be nearing 100 by end or Lair/Orc, 130 by the time I'm going for the 2nd rune, and 160 by Depth. With meleers, I'll want more, but that's usually less of a concern because my meleers tend to be beefy races that just naturally get good hp without too much micromanaging.

These are ballparks that I don't hit every game. They just give me an idea of where I'd like to be.

[goblolo]

It is strictly beneficial to allocate every single point of xp to the skill that at that particular time and skill level gives you the most benefit per xp.

I cannot fully agree with you. A bit of strategy should always be in mind and you should generally move towards your final goal. I remember one of mi first DECj, I trained poison magic (7 or 8 levels) for mephitic cloud and poison arrow somewhere about L1-L3. It was pretty easy to confuse yaks and poison them with penetrating bolt. But a bit later I remembered that I am working towards fire magic and fire spells will do better later on. All that XP I spent on poison are enough to boost my fire magic or conj from 8 to 11 (and even more if you take into account that skill xp costs more when you levelup).
In almost all situations you are better to stay with your main weapon course. Even if you find some good randart +10 dagger, and putting only 10 levels to it will net you a better damage output than your current war axe, But you should bear in mind that when you will train your axes from 15 to 22 and find a branded exec axe, you'll do much more damage anyway, but that 10 skill levels will work bet if they were in shields.

[Hurkyl]

It is strictly beneficial to allocate every single point of xp to the skill that at that particular time and skill level gives you the most benefit per xp.

I cannot fully agree with you. A bit of strategy should always be in mind and you should generally move towards your final goal.

It's maybe a bit nitpicky, but your disagreement just stems from measuring benefit improperly: specifically, reading crate's "the most benefit per xp" as excluding long-term benefits.

[Lasty]

In Crawl, short-term benefits are way more important than long-term planning, because you aren't going to have a future unless you survive the present. Having Venom Bolt up and running when it helps you kill yaks you're otherwise struggling with is way better than having another three levels of Fire Magic because you're hoping to get Bolt of Fire and Firestorm. Likewise, if getting 10 levels of SBl right now helps your character survive, then it's worth doing whether or not you hope to eventually find an executioner's axe.

[WalkerBoh]

I don't know if it's worth stickying this thread, but it should definitely be referenced somewhere. This is probably the best summary of skill training that I've seen.

[savageorange]

In Crawl, short-term benefits are way more important than long-term planning, because you aren't going to have a future unless you survive the present. Having Venom Bolt up and running when it helps you kill yaks you're otherwise struggling with is way better than having another three levels of Fire Magic because you're hoping to get Bolt of Fire and Firestorm. Likewise, if getting 10 levels of SBl right now helps your character survive, then it's worth doing whether or not you hope to eventually find an executioner's axe.

This is very true, but I feel it needs the postscript: 'This does not mean you shouldn't have a long term plan. Long term planning remains important even if you end up not using the plan.'

This is especially relevant on builds with more complex skilling -- ie. magic-focused builds, where it is rather easy to become an ineffective dilettante.