Skill training on automatic

[albondingas]

I think the biggest breakthrough I made early on playing crawl was when I realised i could set skill training to manual and not waste experience on skills I didn't really need. Since then, managing skill training as been something I've thought about in every game but recently I was wondering if it really makes a difference, and is it possible to win with skill training set to automatic?

[Giant Orange Brainbow Dash]

It might not be exactly optimal, but it's certainly possible. One of the best game lessons I ever learned was that ultra-specialization hurts you much more than it helps. Spreading the XP around leads to much better balanced characters.

[Hirsch I]

for begginers, automatic skill training helps you focus on your tactics, that are what really matters in this game. it is perfectly possible, and does not hurt your chances of winning by a lot.

[damiac]

It sort of depends on the character itself too. You'd probably have more trouble than it's worth getting a complex background like transmuter, warper, or arcane marksman off the ground with automatic skill training, unless you were doing something to game the autotraining system (In which case, you should just use manual training).

With a most backgrounds, automatic is generally going to be good enough to get the job done, and it's going to be a lot better than using the manual training option badly.

[Magipi]

for begginers, automatic skill training helps you focus on your tactics, that are what really matters in this game.

This is possibly true, however, learning those two things requires a very different amount of effort. To learn good tactics, you have to play a lot, spectate better players, read forums... for months. However, if you spend 10 minutes learning about skilling, you'll suddenly be much-much better than autotraining.

[Jarlyk]

I can't use automatic skill training, but that's mainly due to a personality disorder that requires me to micromanage when it is possible to do so. It's tricky, though, as crawl tends to downplay long-term strategy, since you don't know up-front what tools you'll have at your disposal in a given game. You might wind up making a significant skill investment into a particular weapon type, only to find a ridiculously awesome weapon of a different type that warrants changing. You might start a game thinking you want to cast Firestorm as your 'end game spell', only to get mostly Hexes books and wind up becoming a stealthy 'stabber'. In these cases, the 'wasted' XP is compensated for by the other resources you found that warranted giving up on the prior investment. Obviously the more you have invested, though, the better the new resource needs to be to justify the loss. This is partly why spreading your training around a bit can be a good idea, since the higher levels are vastly more expensive than the lower levels in a given skill and it's nice to have flexibility.

This isn't to say that strategic choices aren't important, but they're dwarfed by the tactical decisions you make when tough fights happen. Poor strategic choices can mean that you have to make careful tactical choices more often, but mainly crawl is a game about avoiding tactical mistakes. There aren't too many situations in crawl that can't be dealt with (in some cases, this means avoided) using stairs, doors and patience. The hardest part is learning to recognize the tough situations before they get out of hand and your options become more limited. Walking away is probably the single most powerful tool in your kit against the majority of the nasty things you run into.

[damiac]

Walking away works great vs most nasty things, but it doesn't work at all vs ugly things. Whose idea was it to make those jerks tougher, and give them a speed boost!!?!

[Sprucery]

My first about 10 wins were done with automatic training. Granted, those were also the days of victory dancing, so I actually did affect which skills to train after battles, especially magic skills.

[damiac]

Well, you didn't have an option back then, did you?

I never played the old versions with victory dancing...

[duvessa]

You could turn skills on or off; skills that were off would receive less (not zero) experience. Sprucery probably meant that they never turned skills off.

[crate]

and is it possible to win with skill training set to automatic?

yes
http://dobrazupa.org/morgue/autodefe/mo ... 051653.txt

[MrPlanck]

is it possible to win with skill training set to automatic?

I almost always use automatic training (>99.9% of the time). Seems to be pretty effective for me.

http://crawl.akrasiac.org/scoring/players/mrplanck.html

[nicolae]

and is it possible to win with skill training set to automatic?

yes
http://dobrazupa.org/morgue/autodefe/mo ... 051653.txt

lol at the name though

[crate]

the_glow made the account just to do that win, so

[damiac]

That dump is a great example of why you're really making things a lot harder on yourself by using auto skilling.

Level 1.3 Fighting
Level 1.3 Short Blades
Level 1.4 Staves
Level 14.6 Dodging
Level 27 Spellcasting

Obviously it got the job done, but this is just plain bad skilling. With just a little knowledge, you can do much better than this with manual skilling. Usually you obtain that knowledge with really poor use of manual skilling, and/or posting a lot of dumps here and asking for skilling advice.

[dpeg]

That game was playing on a challenge. The point is not that "autoskilling only" should be viable -- you can affect skilling in automatic mode just fine. Rather, the point is that new players should be able to get into the game without having to worry about aptitudes, skills and the m-screen at all. This is really important to me: the entry barrier is high enough, in my opinion.

[and into]

Eh... Fighting was definitely lower than I would have liked, and armor was a bit low too, and a few skills were higher than I would have had them. But putting those aside, that isn't terrible experience allocation for that character. The random 1.3 or 2.3 skills were not needed, but on the other hand that did not really hurt the character, since they represent like one or two stone giants' worth of experience. Branching into new magic can be a bit awkward with automatic skill training, under certain circumstances, and the fighting skill on certain characters might be undertrained, despite the fact that the HP bonus (eventually) becomes worthwhile in its own right. But then, that's something that is kind of awkward about the fighting skill in general. I have definitely seen characters with manually allocated experience that had much worse skill allocation than if the player had done automatic.

Also, you can keep it in automatic mode, but still toggle skills on and off ("semi-automatic" you might say) which can make for a better distribution than what the_glow did, without requiring full micromanagement.

[Magipi]

Eh... Fighting was definitely lower than I would have liked, and armor was a bit low too, and a few skills were higher than I would have had them. But putting those aside, that isn't terrible experience allocation for that character.

Imagine a complete beginner, who knows nothing about Crawl skills, and does not care either. He just guesses about them by the sound of skill names. He assumes that fighting is a skill for fighters, and also assumes that spellcasting is the only important skill for spellcasters. This sounds pretty terrible, doesn't it? Now, this is exactly what automatic training does.

If a blind monkey sits at the keyboard, presses 'm' and then a few random keys... well, this has a decent chance to get a better result than autotraining.

[crate]

If a blind monkey sits at the keyboard, presses 'm' and then a few random keys... well, this has a decent chance to get a better result than autotraining.

I see a lot of humans posting characters in the CIP section that I would say probably have worse skill training than auto training.

If you take a player who has never played a roguelike before, at all, ever, then auto skill training is a very big decrease to the barrier of entry to crawl. This is what dpeg is getting at. It dramatically reduces the number of things a completely new player has to fiddle with to be able to play the game. They are going to die regardless of what they do on the skill screen, and the actually-playing-crawl part of crawl is more fun (ok I'm assuming here that no one plays crawl just to fiddle with the skill screen, but I think that's pretty reasonable!) so that should be the part of crawl a completely-new-to-the-game player gets to experience.

Defauts aren't supposed to be for the people who have reached this forum; you're already more advanced than the player the defaults are designed for. They're for the people who just found the game and are playing it for the first time.

If you have actual improvements to the auto training algorithm (it trains fighting more than in that log I posted already!) then you can suggest to the devteam.

[and into]

If a blind monkey sits at the keyboard, presses 'm' and then a few random keys... well, this has a decent chance to get a better result than autotraining.

This is not even remotely close to being true.

[KoboldLord]

Imagine a complete beginner, who knows nothing about Crawl skills, and does not care either. He just guesses about them by the sound of skill names. He assumes that fighting is a skill for fighters, and also assumes that spellcasting is the only important skill for spellcasters. This sounds pretty terrible, doesn't it? Now, this is exactly what automatic training does.

This isn't remotely like automatic training. Overtraining fighting and spellcasting is a much more typical error for a newbie using manual training than automatic.

There are some major holes in automatic training, like undervaluing defensive skills for ranged characters, but if you think manual training is automatically better you are severely underestimating how badly many players misunderstand how the skill system works.

[Magipi]

If you take a player who has never played a roguelike before, at all, ever, then auto skill training is a very big decrease to the barrier of entry to crawl.

It is a fine goal to make crawling possible for someone who never touches the 'm' key. But does such a new player acually exist? I think not, it is practically impossible. Crawl is a package, and part of the package is skills. Every new player knows that skills exist in Crawl. Imagine a guy walking into a computer shop, and saying he wants to buy a keyboard without vowels, because consonants are more important, and he wants to lower the barrier of entry to the world of computers. This guy does not exist, and no keyboard manufacturer is making keyboards without vowels for this nonexisting customer.

If you have actual improvements to the auto training algorithm (it trains fighting more than in that log I posted already!) then you can suggest to the devteam.

The problems with autotraining are well known. One was fighting, and if this problem was really fixed, that is a huge-huge jump in the quality of autotraining. The next probably should be spellcasting, dividing its weight by 3. Or rather by 10.

[Magipi]

Overtraining fighting and spellcasting is a much more typical error for a newbie using manual training than automatic.

I guess you misunderstood something I was saying, but I can't figure out what. I'm lost.

[basil]

It is a fine goal to make crawling possible for someone who never touches the 'm' key. But does such a new player acually exist?

Sure. I've done it before. I've seen other people do it.

I think not, it is practically impossible. Crawl is a package, and part of the package is skills. Every new player knows that skills exist in Crawl. Imagine a guy walking into a computer shop, and saying he wants to buy a keyboard without vowels, because consonants are more important, and he wants to lower the barrier of entry to the world of computers. This guy does not exist, and no keyboard manufacturer is making keyboards without vowels for this nonexisting customer.

The English language is a package. Imagine a guy not using every word in English. Something like that. Anyway, while analogies have uses, I dislike their use in arguments; speak about things as they are instead.

As for skilling, I think you underestimate the knowledge required to make good decisions, especially if that new player chooses not to participate in the Crawl community. So if they'd rather learn now the tactics in killing things instead of how to distribute experience, I don't see the problem.

[Magipi]

It is a fine goal to make crawling possible for someone who never touches the 'm' key. But does such a new player acually exist?

Sure. I've done it before. I've seen other people do it.

As a self-imposed challenge. We were not talking about that, but about new players.

As for skilling, I think you underestimate the knowledge required to make good decisions, especially if that new player chooses not to participate in the Crawl community. So if they'd rather learn now the tactics in killing things instead of how to distribute experience, I don't see the problem.

Skilling is not hard. Beginners mostly make huge mistakes because Crawl is not transparent, and what skills do is often counterintuitive. They don't realize that fighting gives hp, that spellcasting is 4 times less effective than spell schools, or that low skill levels are much cheaper than high skill levels. But these are easy concepts, easy to learn. Good tactics is 10 times more useful, yes... but it is also 1000 times more difficult to learn.

[Fafhrd]

If you take a player who has never played a roguelike before, at all, ever, then auto skill training is a very big decrease to the barrier of entry to crawl.

It is a fine goal to make crawling possible for someone who never touches the 'm' key. But does such a new player acually exist? I think not, it is practically impossible. Crawl is a package, and part of the package is skills. Every new player knows that skills exist in Crawl. Imagine a guy walking into a computer shop, and saying he wants to buy a keyboard without vowels, because consonants are more important, and he wants to lower the barrier of entry to the world of computers. This guy does not exist, and no keyboard manufacturer is making keyboards without vowels for this nonexisting customer.

I am one of the guys which doesn't press the "m" key during my game sessions. I did before. I realized sometime ago that I spent (approximatively) 25% of my gaming time on the Skill screen (or thinking ahead at what I should train next) rather than actually playing.
I have a very busy RL (home, job, wife and kids) like many of you. I now drive my toons on autoskilling in order to gain some time of actual playing (my gaming sessions consist of 10-15 minutes breaks, one kid on one knee, sweeping the kitchen with one hand and responding a phone call while playing) I really don't have time to lose 25% of this time staring at the skill screen!!

It may seem to you a joke but it's real. And i bet I'm not the only one around there. For us, auto-skilling does wonder regarding the ratio of time efficiency/results in a "not-too-bad skilling"
And more often than not, my chars in auto skilling mode end up having a far more coherent skilling than when I was trying to do things in manual skilling....

[MrPlanck]

Pretty sure I misunderstood autoskilling. I use auto training mode, but turn skills on and off throughout the game. So, for whatever it's worth, I don't think I "autoskill".

[jejorda2]

I'm assuming here that no one plays crawl just to fiddle with the skill screen, but I think that's pretty reasonable!

If there were a version of crawl where the fighting and exploring was all automated, and all I did was choose when to memorize spells, equip items, worship, and train skills, then I would probably play that more than standard crawl. I like to see how decisions play out, rather than do what has the best chance of working.

[damiac]

On a HuGl who never presses 'm', who finds regeneration, how do I get it castable? I guess this player just says "100% fail, I guess Gladiators just can't cast spells".

That is not teaching good crawl skills. I'm not saying I don't think auto-skilling should exist, but I am saying nobody should ever recommend it as a way to get good at crawl.

[dpeg]

It's probably a doomed attempt, but I'll try to point out that people use "autoskilling" in two different ways:

1) Games starting in automatic skilling mode and not using the m-screen at all.
2) Games starting in automatic skilling mode.

Some posters have brought up won games in 1) to prove that it's possible. Other posters point out that there are situations where 1) will fail.

Neither of these gets at the core why 2) is the default, and will stay default, threads like this notwithstanding: a player new to the game has to learn many things.
We are very happy if dealing with skills is not one of these -- that can, and will, come later.

Put another way: I am convinced that manual skilling as the default, and thereby forcing every new player to deal with the m-screen, will lead to fewer newbie -> Crawler conversions. You may disagree with this, but I don't think there's any disagreement in the devteam, and this point hasn't been discussed in the this thread anyway.

[damiac]

I'm not saying it shouldn't be the default, I'm just saying nobody should advise anyone to use it.

Like, go ahead and play your first few games on autoskill, and get an idea of how the game works. But then, if you are trying to win, it's a good idea to stop using autoskilling.

I agree with the point that if your going to die on D2 anyway, there's probably no point in messing around with your skill screen.

[cloud4ge]

Pixel Dungeon has more appealing graphics, and you can play it anywhere on your smart phone.

That's what I have to say about autotraining.