Skill levels, and explaining it to a noob.

[Dhanvanthri]

Guys, often, when I ask for advice, I'm told about how I'm training too many skills at once, or that I'm overtraining certain skills. I don't understand when the magic barrier of skill training is hit.
Example: I'm playing a DeepElf FireElementalist, and I'm still training spellcasting (it's at 20+ right now), and when I ask for advice, someone asks wtf I'm doing and tells me switch it off. So I just want to know how you pro crawlers understand when skills need to be switched off.
Also, I've been playing for a year, and I still don't feel like I understand this game at ALL. So any advice to a newbie would be great.

PS: Never won a game, although I did get 4 runes once.

[dck]

Normally characters want a number of things that are attainable from skill training. You train a given skill until you have enough of what it provides relative to your depth and then switch to whatever gives you the best returns for your skill buck.

So say you're that DEFE and are in D: 3 and just got sticky flame down to good failure %. Sticky flame will demolish pretty much everything for a long long time so your offense is good to go and thus the thing that interests you the most to improve your character are skills that make you harder to kill, in this case these would be dodging until you get a bunch of EV and then likely fighting, weapon skill etc. And you probably will alter your training depending on what you may find (so say you find a +6 dex ring you're only getting 2 EV from, you want to train and get a good bunch more from it since it's cheap and great).

If you're going to be training your skills manually (which while ideal isn't quite required) you need to focus not on skill levels you get but on how you're performing in the affected fields. Your actual EV score is more important than the number next to dodging, your offense with spells and melee (normally binary in terms of "enough" and "not enough") more important than weapon skill/fighting or magic skills.

Sometimes you will train a certain skill quite high, but normally only when you already have a very good base on other skills that are also important (and again pretty much everyone wants a number of basic returns from skill training). In your example though there pretty much will never be a reason to train spellcasting anywhere near that high since spellcasting doesn't do a lot to make your character better.

[Patashu]

It's all about knowing the rules of how skills work. Let's examine them:

1) The most important rule: If it takes 1 arbitrary unit of skill (aus) to train from 0.0 to 1.0...
EDIT: These new calculations are correct. Thanks stickyfingers!

1) The most important rule: If it takes 1 arbitrary unit of skill (aus) to train from 0.0 to 1.0, it takes 3 to train from 1.0 to 2.0, then 6, then 10, 15, 21, 28... or the triangular number (x^2+x)/2. This also means that to go from 0 to x in a skill costs the tetrahedral number, (x^3 + 3x^2 + 2x)/6.

Actually, if it takes 1 arbitrary unit of skill to train from 0.0 to 1.0, it takes 3 to train from 0.0 to 2.0, so just 2 to train from 1.0 to 2.0, and so on.
See the source function and its use.
The examples are thus off (but obviously you're right in the general sense).

How certain are you of this? The table on the wiki page indicates that this changes after 10 skill, then again after 19 skill (and I stopped bothering to analyze beyond that point).

The table on the wiki is correct.

Thanks! So it looks like it's more like this:
-Training a skill from 0 to 12 could be used to train two skills to 9 or three skills to 7
-Training a skill from 12 to 20 could be used to train two skills to 14 or three skills to 12
-Training a skill from 20 to 26 could be used to train one skill to 20 (!!)
-Training a skill from 26 to 27 could be used to train one skill to 9
-In general, if you have a skill at level X and another skill at 2*X, the skill at 2*X will take you 2-3 times as long (2 for low X, 3 for high X) to train one extra level of that skill, before aptitudes

(Spoilered old calculations that are incorrect)

Spoiler: show

it takes 3 to train from 1.0 to 2.0, then 6, then 10, 15, 21, 28... or the triangular number (x^2+x)/2. This also means that to go from 0 to x in a skill costs the tetrahedral number, (x^3 + 3x^2 + 2x)/6.
If you don't know what that means, here are some examples:
If training a skill to the next level costs X, it would be 25% as expensive to get one level in a skill half as expensive.
-For the cost of training a 12 to a 20, you could have trained three skills from 0 to 12.
-For the cost of training a 20 to a 26, you could have trained three skills from 0 to 14.
-For the cost of training a 26 to a 27, you could have trained three skills from 0 to 8.

In general - if you try to train one skill super high, you are denying yourself a lot of skill points you could have filled out lesser skills with.
If one skill is a lot higher than others, and you're not about to reach some milestone, maybe you want to train all those more?

2) Spellcasting is 1/4 as effective as training a spell school is for spells of that spell school, but applies to every school at once.
It also increases max mp, spell slots and reduces spell hunger.
So you want to throw in some spellcasting when it's about a half (because remember - the cost in aus to train a skill goes up by the square!) of the level the spell schools you're training (a bit higher for every extra spell school you need to train), or if you need more spell slots desperately.
Otherwise spellcasting goes on the backburner.
Also, when doing this calculation make sure to only consider spell schools that you actually care about training further (e.g. if you already have haste casteable, you probably aren't going to care about your charms improving)

3) Weapon skills' primary benefit is reducing the delay on a weapon.
If you're not trying to get mindelay or you already have it - train something else instead.
(And also think about, if you're going for a very high weapon skill to get min delay - like 20 for great mace, or 26 for exec axe. Especially 26 for exec axe - Wow - you should think, 'that's a LOT of skill points, maybe I should just suck up that it's not at min delay, or use a good batleaxe?' when thinking about doing this!!)

4) Armour skill gives you a benefit directly (linearly) proportional to how much AC (pre-enchantment) your equipment in total gives. (Armour skill * base AC / 22). So if you can't get a lot of both you don't want to train this yet. For example - if you have only 5 base AC then you need to get to 4.4 armour skill to get even your first AC point from this. If you're a book background, you might not even have 5 base AC! Think about how armoured you are, how much armour skill you can spare (=> how much AC you get) and if this is a worthy tradeoff for how much you'd get in other skills, such as your 'kills things' skills instead, and then you'll know how much of this you want.

5) Dodging skill has a weird interaction with heavy armour. Based on how well you carry yourself in that armour (str and dex affect this I believe*) (*apparently dex doesn't matter), training dodging gives a flat ZERO to your EV until you hit an 'invisible breakpoint', then training it further starts pumping out EV as fast as if you were in robe from that point onwards. Usually this 'invisible breakpoint' is in the teens. So if you want to train dodging in heavy armour - wait until your first or second rune, to the point where getting past the invisible breakpoint and seeing your return on EV makes sense to go for, and before then don't give it even a single aus.

6) Fighting gives a steady boost to your melee prowess (I think it also improves your ranged prowess in 0.15?) but even better, it improves your HP. But, similar to armour skill, it is multplicative, in that your base max hp * your fighting skill come together to determine how many extra HP you will see - which means that as you gain XLs there is an increasing push that makes you want to train fighting more, because you're getting more per aus spent, not just because auses are becoming easier to come across.

7) This post should talk about ranged, evo, invo and stealth too but I don't know enough about them to say what are some interesting properties about them. Anyone else have some tips?

[Dhanvanthri]

Okay, I've been trawling through the wiki as well, in a quest to better my understanding, and I wasn't able to find the page that tells you when you hit mindelay with a weapon, is there a way I can check this ingame? Some stat screen I'm unaware of perhaps?
@Patashu Thanks a lot, although I still don't understand some of the other interactions, like size with EV, or the advantages of being a draconian as opposed to any other class due to the way they gain AC as they level up.

[Magipi]

Okay, I've been trawling through the wiki as well, in a quest to better my understanding, and I wasn't able to find the page that tells you when you hit mindelay with a weapon

http://crawl.chaosforge.org/Weapon_speed

[dck]

3) Weapon skills' primary benefit is reducing the delay on a weapon.

This is wrong, the primary benefit is dealing more damage in the same window of time. You'll likely want a sane swinging time regardless of your effective damage if you're not really judging your fights with all the care in the world but there is nothing intrinsically wonderful about min delay itself nor is there a real reason you should pursue it if your damage proves sufficient and you have better things to train.
A good example of this are great swords. Even though they are excellent weapons you can find pretty early on, you don't want to swing a great sword anywhere near min delay the minute you get one (well you want to, but you can't sanely speaking) and indeed it isn't rare at all you'll spend more than half of the game swinging it over its minimal delay.
There exist a couple of corner cases such as fast weapons with additive brands / aux attacks / slaying where there is good reward in pushing to the minimal delay possible but those are rare (and even then you don't invest as much in the first place).

6) Fighting [..]but even better, it improves your HP. [...] which means that as you gain XLs there is an increasing push that makes you want to train fighting more

It doesn't mean that. It means that Fighting is important relatively early on because it's when more damage and hit points make more of a comparative difference and it means that the multiplicative system is there mostly to stop people to fool themselves into pretending they should just endure early game without the boost from fighting since the hp would be "negible" to the fully developed character; it also exists to provide base materials on which to build a character development and continue providing a challenge curve as the player's XL goes up (this doesn't really work for other reasons but whatever).

Regardless of all of this literally nobody should care about skill points and the concept of skills getting ridiculously expensive at high levels compared to much lower ones is self-evident the moment one actually tries to train something from zero by killing any late game exp pinata.

[Magipi]

3) Weapon skills' primary benefit is reducing the delay on a weapon.

This is wrong, the primary benefit is dealing more damage in the same window of time.

And a huge part of that increase of damage output comes from the shrinking attack delay. When you decrease your delay from 1.0 to 0.9, that boosts your damage output by 10%. After reaching min delay, this effect is lost.

[dck]

It' not really that huge, and as explained in the rest of the omitted post the technicalities of why you deal more damage are as self-evident as they are unimportant and placing unnecessary detail on them only further distracts the player from the important fact "you're dealing more damage" and may cause stupid things like pretty much ignoring defenses until you hit min delay on your weapon that say requires 16 skill sound not crazy because hey, you're after all maximizing your weapon's potential!

when you decrease your delay from 1.0 to 0.9, that boosts your damage output by 10%"

This is also also wrong because it ignores the nontrivial damage weapon skill adds.

[Lasty]

For example - if you have only 5 base AC then you need to get to 4.4 armour skill to get even your first AC point from this. If you're a book background, you might not even have 5 base AC! Think about how armoured you are, how much armour skill you can spare (=> how much AC you get) and if this is a worthy tradeoff for how much you'd get in other skills, such as your 'kills things' skills instead, and then you'll know how much of this you want.

Generally speaking, most characters can benefit from training enough armour to get 1 bonus AC by the time they hit Lair. If you have all your armor slots filled, then even with a robe it only takes a little over 3.5 Armour skill, and as you said, training skills to very low levels is quite cheap. Consider carefully how much to invest in Armour skill based on your total base AC, but keep in mind that at least one or two points of AC are usually worth the price.

5) Dodging skill has a weird interaction with heavy armour. Based on how well you carry yourself in that armour (str and dex affect this I believe), training dodging gives a flat ZERO to your EV until you hit an 'invisible breakpoint', then training it further starts pumping out EV as fast as if you were in robe from that point onwards. Usually this 'invisible breakpoint' is in the teens. So if you want to train dodging in heavy armour - wait until your first or second rune, to the point where getting past the invisible breakpoint and seeing your return on EV makes sense to go for, and before then don't give it even a single aus.

I've never heard that dex affects armour evasion penalty or the "invisible breakpoint", from looking at the code briefly, it doesn't appear to have an effect. Of course, Dex does has an effect on the returns from Dodging skill in general, so it has an impact in that sense.

6) Fighting gives a steady boost to your melee prowess (I think it also improves your ranged prowess in 0.15?) but even better, it improves your HP. But, similar to armour skill, it is multplicative, in that your base max hp * your fighting skill come together to determine how many extra HP you will see - which means that as you gain XLs there is an increasing push that makes you want to train fighting more, because you're getting more per aus spent, not just because auses are becoming easier to come across.

In 0.15, ranged and melee combat operate under all the same rules, including the impact of weapon skill on min delay and the bonus to hit and damage from Fighting skill.

Also, the new formula for the impact of Fighting on max hp is

  Code:

Fighting * (3/2 + XL/14)

which means that Fighting gives a flat bonus in addition to a bonus that scales with XL. All hp bonuses are scaled by hp aptitude. For example, if an XL1 Felid could go from 0 levels of Fighting to 5 levels of Fighting, its hp would go from 7 to 12.

7) This post should talk about ranged, evo, invo and stealth too but I don't know enough about them to say what are some interesting properties about them. Anyone else have some tips?

Ranged now works the same as melee. The other skills you mention don't have much in the way of notable breakpoints, though Invo may depending on the god -- some god powers top out at different Invo levels. Generally if you want to get better at the things these skills cover, invest more XP in them.

[duvessa]

I would recommend ignoring dck's posts in this thread. He seems to have developed a fondness for giving new players deliberately terrible advice.

when you decrease your delay from 1.0 to 0.9, that boosts your damage output by 10%"

This is also also wrong because it ignores the nontrivial damage weapon skill adds.

Example, telling them weapon skill adds damage and then completely ignoring how much damage it adds. A level of weapon skill increases average damage per hit by 2%. A level of fighting increases it by 1.67%, and also increases aux attack (headbutt, offhand punch, etc) damage by 1.25%. Both skills add +0.25 accuracy per level on average.

[dck]

What exactly is bad about telling new players to care about the results their skill investment yields and not the numbers on the skill screen?

[Galefury]

You slightly overstated the damage increase from weapon skill. The most important thing you get from training weapon skill actually is lower delay. Yes, as you said this is because attacking faster means you deal more damage in the same time.

Talking about delay and damage increasing effects separately makes sense, because delay first gets better (per skill level) with every skill level you add, then magically there is no additional effect once you pass a breakpoint. So min delay is quite important to weapon skill training, and the number on the skill screen actually matters in this case.

[dck]

I don't think it does, there are two readily apparent results from training a weapon skill and those are damage dealt and swinging speed. The swinging speed is literally displayed in the game screen and overall weapon performance is very simply judged by hitting things that are relevant to your depth with it.
What I defend is that there is no point in caring about min delay or indeed swinging speed since if you see that your weapon is killing everything quite alright around your depth then it needs no further training and your exp would be better located elsewhere where you can gain another benefit that matters.

[Hurkyl]

It's all about knowing the rules of how skills work. Let's examine them:

1) The most important rule: If it takes 1 arbitrary unit of skill (aus) to train from 0.0 to 1.0, it takes 3 to train from 1.0 to 2.0, then 6, then 10, 15, 21, 28... or the triangular number (x^2+x)/2. This also means that to go from 0 to x in a skill costs the tetrahedral number, (x^3 + 3x^2 + 2x)/6.

How certain are you of this? The table on the wiki page indicates that this changes after 10 skill, then again after 19 skill (and I stopped bothering to analyze beyond that point).

[Hurkyl]

I'm playing a DeepElf FireElementalist, and I'm still training spellcasting (it's at 20+ right now), and when I ask for advice, someone asks wtf I'm doing and tells me switch it off. So I just want to know how you pro crawlers understand when skills need to be switched off.

If you are seeking to increase the effectiveness and miscast rates of your spells, training spellcasting is not as efficient as training your spell skills directly, unless you are trying to simultaneously improve spells using many different schools. It's rather rare that you have four spell schools in the high teens (or twenties), so spellcasting is inefficient for that purpose.

So the other effects of spellcasting are increased MP (which becomes irrelevant because gaining MP beyond 50 is very slow), reduced spell hunger, and more spell levels.

Since many in the forum ascribe little value to having a large number of spells and do not consider spell hunger something worth worrying about, you'll find that training spellcasting to high levels is much maligned. Overly so, in my opinion, but form your own opinion.

---------

The importance of magic skills tends to have rather sharp thresholds: e.g. for a DEFE that has Conjure Flame, training Conjurations and Fire isn't very useful until you get to the point where you can cast Sticky Flame. And once there, training those skills further isn't very useful until you get to the point where you can cast Fireball. If you're training Spellcasting for spell levels, it's not very useful until you get enough levels for the spells you're looking for. And so forth.

So frequently, you want to turn off spell training completely, until you decide it's time to advance to the next threshold, at which point you turn off everything else and focus on your magic training exclusively.

Eventually that becomes untrue, because you want to train magic to increase spell power and you're happy with incremental gains.

---------

Weapon skills are kind of odd, because their most significant effect is how they decrease the delay of your weapon, and the way this works means that higher skill levels give you more effect and the greatest gain happens while training the skill level that will let you reach minimum delay. (minimum delay is displayed in the item details, which can be reached from the inventory screen)

There is another psychological breakpoint in that swinging a weapon at 1.0 aut will prevent you from making most mistakes involving an enemy taking two actions after you attack.

So weapon skills are a sort of thing you want to incrementally improve, until you get to the point where you're nearing the minimum delay of a weapon you're using, at which point you probably want to race to get there. (which means turning off training in other skills)

---------

Regarding training many skills at once, it's a reasonable thing to do when your character is well-balanced. However, characters often have imbalances, such as "increasing my weapon damage output is much more important than other concerns", which imply that you should focus more on correcting the imbalances rather than spreading out your training amongst many skills.

[crate]

It's all about knowing the rules of how skills work. Let's examine them:

1) The most important rule: If it takes 1 arbitrary unit of skill (aus) to train from 0.0 to 1.0, it takes 3 to train from 1.0 to 2.0, then 6, then 10, 15, 21, 28... or the triangular number (x^2+x)/2. This also means that to go from 0 to x in a skill costs the tetrahedral number, (x^3 + 3x^2 + 2x)/6.

How certain are you of this? The table on the wiki page indicates that this changes after 10 skill, then again after 19 skill (and I stopped bothering to analyze beyond that point).

The actual numbers here aren't important. I think mentioning the numbers here is not useful, and as you point out is actually a cause for confusion. The important thing is that skills get expensive quickly. If you really care enough that you think you need the actual numbers, then you will need more numbers than Patashu is providing anyway--for instance, you will also need to know how much more expensive skill B gets when you train skill A (because all skills get more expensive as your total character xp increases) (note I am not intending to use the appropriate technical terms here like "skill points" or whatever because I don't know how this works in the code and I am not interested), because without that information you can't compare directly "I could get axes 20->26 or I could get fighting 0->15 plus armour 0->15 plus dodging 0->15" (I made these numbers up and they are almost certainly wrong because I'm not even trying to be accurate)

You likely also need to know what +1 apt means exactly (since not a single race has every single skill with the same aptitude, though dg comes close) and......

As you can see the numbers provided are not useful in context of this thread and the information provided therein. I suggest that in fact omitting any mention of said numbers would have been better advice.

---

the primary benefit [of weapon skill] is dealing more damage in the same window of time

This is the important point to take away re: weapon skill. It is true that the function is piecewise-continuous; i.e. there is a specific point (when you reach min delay) that the behaviour of the function suddenly changes. But I think that thinking in terms of "damage per turn" is a useful thing to do.

---

Regarding training many skills at once, it's a reasonable thing to do when your character is well-balanced.

As I said elsewhere, it is (very nearly--single large chunks of xp (e.g. !xp) often don't quite fit here) always strictly inferior to training precisely the one skill which at the present moment in time provides your character with the greatest utility per xp gained. However, since it is often true that several different skills have similar returns per xp spent, and furthermore since playing well tactically is dramatically more important than training skills well as long as you are doing something sensible with your xp you do not have to be nearly that precise with your xp allocation. Additionally, very few players, probably none at all, actually open the skill menu and potentially change skill training after every single instance of gaining experience, which is what you would do for the absolute best possible distribution of your gained xp.

---

btw for the record getting to 27 spellcasting is not at all ridiculous for a deep elf. I cannot speak about the character mentioned in the OP (and as a rule I do not like to give skill advice very often) though.

The only general skill training advice I give is this: Here is how you train skills. First you train the skills that you use to kill dudes. Once you are killing dudes well enough, then you train the skills you use to not die to dudes. If you later run into problems killing dudes, you switch back to killdudes skills.

some people will assuredly point out to you that I am purposely simplifying things here to lead you toward a sensible course of action, and this is true, but I think it leads you toward a sensible course of action so that is my reasoning. You might note what I said above about the relative importance of skill training and proper tactics. Of course, teaching good tactics is not easy either, though if you want help in particular situations and don't mind criticism then ##crawl will (probably) help you out there (or possibly the CIP section of Tavern, though the response time there is of course slower).

[Hurkyl]

As you can see the numbers provided are not useful in context of this thread and the information provided therein

I agree it's tangential, but I'm curious for my own purposes unrelated to this thread.

The important thing is that skills get expensive quickly.

But still having a sense of the right order of magnitude "quickly" means is important. Knowing that going from 12 -> 20 is three times as expensive as going from 0 -> 12, as opposed to 2.8 times as expensive is not really important.

But knowing it's three times as expensive rather than ten times more expensive or half as expensive is, IMO, fairly useful.

[stickyfingers]

1) The most important rule: If it takes 1 arbitrary unit of skill (aus) to train from 0.0 to 1.0, it takes 3 to train from 1.0 to 2.0, then 6, then 10, 15, 21, 28... or the triangular number (x^2+x)/2. This also means that to go from 0 to x in a skill costs the tetrahedral number, (x^3 + 3x^2 + 2x)/6.

Actually, if it takes 1 arbitrary unit of skill to train from 0.0 to 1.0, it takes 3 to train from 0.0 to 2.0, so just 2 to train from 1.0 to 2.0, and so on.
See the source function and its use.
The examples are thus off (but obviously you're right in the general sense).

How certain are you of this? The table on the wiki page indicates that this changes after 10 skill, then again after 19 skill (and I stopped bothering to analyze beyond that point).

The table on the wiki is correct.

[Patashu]

1) The most important rule: If it takes 1 arbitrary unit of skill (aus) to train from 0.0 to 1.0, it takes 3 to train from 1.0 to 2.0, then 6, then 10, 15, 21, 28... or the triangular number (x^2+x)/2. This also means that to go from 0 to x in a skill costs the tetrahedral number, (x^3 + 3x^2 + 2x)/6.

Actually, if it takes 1 arbitrary unit of skill to train from 0.0 to 1.0, it takes 3 to train from 0.0 to 2.0, so just 2 to train from 1.0 to 2.0, and so on.
See the source function and its use.
The examples are thus off (but obviously you're right in the general sense).

How certain are you of this? The table on the wiki page indicates that this changes after 10 skill, then again after 19 skill (and I stopped bothering to analyze beyond that point).

The table on the wiki is correct.

Thanks! So it looks like it's more like this:
-Training a skill from 0 to 12 could be used to train two skills to 9 or three skills to 7
-Training a skill from 12 to 20 could be used to train two skills to 14 or three skills to 12
-Training a skill from 20 to 26 could be used to train one skill to 20 (!!)
-Training a skill from 26 to 27 could be used to train one skill to 9
-In general, if you have a skill at level X and another skill at 2*X, the skill at 2*X will take you 2-3 times as long (2 for low X, 3 for high X) to train one extra level of that skill, before aptitudes

[Aule]

... before aptitudes

So, does +1 aptitude take half the XP to train? Does +2 take one-third? What is the relationship, there? What does +3 give? For the aforementioned DE, Spellcasting is +3 while Fighting is -2, so surely this has to be considered when deciding allocations.

[Hurkyl]

... before aptitudes

So, does +1 aptitude take half the XP to train? Does +2 take one-third? What is the relationship, there? What does +3 give? For the aforementioned DE, Spellcasting is +3 while Fighting is -2, so surely this has to be considered when deciding allocations.

Each +1 multiplies the skill points you get by a constant. The normalization is +4 doubles your skill point gain. Negatives divide by the same constant. (this is on that wiki page too)

[Patashu]

So that would make the constant 1.19. right?

[njvack]

Mod note: I've deleted some rather flame-y, content-free posts. Disagree with each other all you like, but please don't get personal.

[Aule]

... before aptitudes

So, does +1 aptitude take half the XP to train? Does +2 take one-third? What is the relationship, there? What does +3 give? For the aforementioned DE, Spellcasting is +3 while Fighting is -2, so surely this has to be considered when deciding allocations.

Each +1 multiplies the skill points you get by a constant. The normalization is +4 doubles your skill point gain. Negatives divide by the same constant. (this is on that wiki page too)

So +3 makes raising a skill about 2/3 more efficient than one of 0 apt (168%). But the difference between -2 and +3 is over twice as great, in all. That means, to answer my question, A DE raises fighting at a rate about 42% of what the same XP can raise in Spellcasting, which for this species is 238% more efficient than Fighting, in raw XP terms. I knew it was a lot. It takes forever to raise fighting. I manual my DE's skills by even percentages, giving double to fighting than to spellcasting, and it still can't keep up. This explains why.

[Sprucery]

Here is a table that shows how many experience points you need to train a skill from one level to the next. The first row is the aptitude and the first column is the next level.

  Code:

   +5   +4   +3   +2   +1   0   -1   -2   -3   -4   -5
1   21   25   30   35   42   50   60   70   85   100   119
2   42   50   60   71   84   100   119   141   169   200   238
3   63   75   89   106   126   150   179   211   254   300   357
4   84   100   119   142   168   200   238   282   339   400   476
5   105   125   149   177   210   250   298   352   424   500   595
6   126   150   179   213   252   300   357   423   508   600   714
7   147   175   208   248   294   350   417   493   593   700   833
8   168   200   238   284   336   400   476   563   678   800   952
9   189   225   268   319   378   450   536   634   763   900   1071
10   231   275   327   390   462   550   655   775   932   1100   1310
11   273   325   387   461   546   650   774   915   1102   1300   1548
12   315   375   446   532   630   750   893   1056   1271   1500   1786
13   357   425   506   603   714   850   1012   1197   1441   1700   2024
14   399   475   565   674   798   950   1131   1338   1610   1900   2262
15   441   525   625   745   882   1050   1250   1479   1780   2100   2500
16   483   575   685   816   966   1150   1369   1620   1949   2300   2738
17   525   625   744   887   1050   1250   1488   1761   2119   2500   2976
18   567   675   804   957   1134   1350   1607   1901   2288   2700   3214
19   630   750   893   1064   1261   1500   1786   2113   2542   3000   3571
20   693   825   982   1170   1387   1650   1964   2324   2797   3300   3929
21   756   900   1071   1277   1513   1800   2143   2535   3051   3600   4286
22   819   975   1161   1383   1639   1950   2321   2746   3305   3900   4643
23   882   1050   1250   1489   1765   2100   2500   2958   3559   4200   5000
24   945   1125   1339   1596   1891   2250   2679   3169   3814   4500   5357
25   1008   1200   1429   1702   2017   2400   2857   3380   4068   4800   5714
26   1071   1275   1518   1809   2143   2550   3036   3592   4322   5100   6071
27   1155   1375   1637   1950   2311   2750   3274   3873   4661   5500   6548


In reality the exp cost goes up as the character gains experience, but the relative differences remain the same.

[Dhanvanthri]

Thanks guys, you've really helped out a LOT! Just managed to get a rune with a TrMo of Oka.

[Galefury]

Emphasis mine.

for instance, you will also need to know how much more expensive skill B gets when you train skill A (because all skills get more expensive as your total character xp increases) (note I am not intending to use the appropriate technical terms here like "skill points" or whatever because I don't know how this works in the code and I am not interested), because without that information you can't compare directly "I could get axes 20->26 or I could get fighting 0->15 plus armour 0->15 plus dodging 0->15"

No you don't. This is actually completely irrelevant to skill training, because skill A also gets more expensive when you train skill A. The order you train skills in does not change the end result (except of course you might be dead ). The conversion rate from exp to skill points changes, but you will always get the same amount of skill points no matter where you put them.

[crate]

Well, then you need to know that that doesn't actually change anything (which was not mentioned in Patashu's post). That was not clear to me (as you can see), though it's pretty obvious from playing that all skills get more expensive as you gain more xp.

[Patashu]

Well, then you need to know that that doesn't actually change anything (which was not mentioned in Patashu's post).

I didn't mention it specifically because it doesn't influence your decisions at all.

[damiac]

Knowing that my level 0 armor will go up 3 levels in the time it takes to get my dodging skill from 4 to 5 helps me make decisions. If I know I will get more AC from 3 levels of armor skill than I will get EV from 1 level of dodging, I would be smart to spend that XP on armor instead of dodging. If I only know that lower levels cost less, but now how much less, I cannot make an intelligent decision on what training will benefit me most.

That skill cost table should be stickied on this forum. It's very useful information, and I don't understand why so many people insist on trying to suppress very useful information like that.

[crate]

Knowing that my level 0 armor will go up 3 levels in the time it takes to get my dodging skill from 4 to 5 helps me make decisions. If I know I will get more AC from 3 levels of armor skill than I will get EV from 1 level of dodging, I would be smart to spend that XP on armor instead of dodging. If I only know that lower levels cost less, but now how much less, I cannot make an intelligent decision on what training will benefit me most.

That skill cost table should be stickied on this forum. It's very useful information, and I don't understand why so many people insist on trying to suppress very useful information like that.

Patashu didn't give you that information though. He provided some numbers, but not enough. (If nothing else, what aptitudes mean was also necessary.) I don't agree with providing all the numbers since I don't think that that actually helps players and it is a lot of math, but providing just some of the numbers is, to me, clearly worse than either option of providing no numbers or providing all numbers.

(also in this case you'd further have to know: how much AC does going from 0->3 armour give you (admittedly not hard to figure out) and how much EV going from 4->5 dodging gets you (harder to figure out) and then if we're going to go that far you might as well also get some fsim numbers to see which actually helps your survivability more, so you actually need even more numbers; you might not agree with my stance but I hope you can at least see where it is coming from.)

[mikee]

I'm all about suppressing the very high cognitive load generated by stuff like that table. I just shudder when I look at it and I don't really know what the application of it would be for me. What would I do, add up the xp value of all the monsters that I kill until I get to 630, and then I know that's how much killing I need to get a skill to 12? I don't even know how much xp an orc gives.

[Sprucery]

Well, I presented the table because some numbers and formulas were mentioned earlier. From there you can get, at a glance, the idea of how much the aptitudes affect training. But I wouldn't expect a normal (or should I say sane) player to pay closer attention to that.

(Personally I use that table all the time because I greatly enjoy deciding what skill to train next to get results as fast as possible. But I would never advice anyone to do that, I know I'm a bit crazy that way. I also have no idea how much exp monsters give, but that doesn't affect the comparison between skills.)

Now that I think of it, it _could_ sometimes be useful for anyone when trying to get for example Deflect Missiles online. Depending on you're aptitudes you could see whether it's most effective to train Charms, Air or Spellcasting in a given situation. Of course many other things affect that decision too.

[damiac]

Seeing the numbers laid out in front of you like that gives you the ability to grasp exactly how much more expensive the skills get as they level up. It gives a good idea of when one might want to switch training around for the best benefit.

I understand someone saying you do not need information like that to win. That's completely true. But what I totally disagree with is the idea that information like that makes you less likely to win. There's a reason when engineers or scientists want to produce a certain result, they get as much information as possible about all the processes involved. No competent engineer or scientist ever said "Whoa whoa, don't give me that information, because it will make this problem harder to solve".

Whether you wish to approach crawl with that mindset is completely up to you. Whether I wish to approach crawl with that mindset is not up to you, so don't tell people to stop giving me information.

[dck]

Indeed that's why when engineers or scientists have to work with π they use the full value only.

[crate]

No competent engineer or scientist ever said "Whoa whoa, don't give me that information, because it will make this problem harder to solve".

Um, this is not true at all. In fact in science you are very nearly always purposely not using all the information you possibly could, precisely because it makes the problem harder to solve! An easy example is that it's very rare to use general relativity to describe gravity; instead people use Newton's law of gravitation, which is incorrect but a good approximation.

[Sprucery]

OK now I have to find a way to take pi and gravity into account when making skilling decisions...

[Hurkyl]

But on the flip side, "stuff falls; just look at it" is not a particularly good substitute for Newtonian gravity.

[dck]

It is when staying alive depends on getting out the crumbling tunnel before the walls cave in on you, and not calculating the exact force with which they'll crush you bloody if you don't.
that's a nice way to get back on track now ain't it

[damiac]

In the early game when XP spending order actually matters, in a game where you have all the time in the world to make a decision, it makes sense to have an idea how to get the most bang for your buck.
Please stop telling people not to give information. It's fine with me if you choose not to use it, but I appreciate someone making it available.

[mikee]

Well if you're going to position yourself as the receiver of advice, then I can tell you that in effective learning environments, there's a reason why the teacher plans the curriculum and not the student. This is not to say that interest doesn't play a role in education or that your interest in the numbers is inherently meaningless, but it's rather something you should pursue as your own hobby. I still wouldn't praise skill point knowledge as useful for playing crawl any more than I would knowing the color of Stalin's mustache for understanding history.

[ackack]

While clearly number overload is a thing (e.g. TOME), I feel like there's a certain degree of anti-number snobbery that goes on sometimes re: Crawl advice. I agree that the exact numbers on the skill level chart are not essential information. But qualitative, rule of thumb estimates extracted from looking at the table for a minute can be very helpful. Yes, if you play for a while you know that it's more expensive to get high levels of skill, but for a long time I had absolutely no feel for just how much. Tidbits like "training a skill from 20 to 26 could be used to train one skill to 20," are, IMO, very valuable and have big implications for how I play. Things like ??skills are probably the most concise summary of how experienced people play, but I feel like they assume too much ability to assess the game. If your judgment about Crawl is good enough that you can answer questions like "am I killing things good enough for now?" reliably you probably don't need skilling advice.

[mikee]

I feel like you're only half saying a lot of things. Am I being an anti-number snob, or is my opinion of the chart anti-number snobbery? Is my ability to give advice impaired by the fact that I no longer remember what it's like to be a new player?

My original comment in this thread is a response to the idea that the chart is very useful information that should be stickied to the top of this forum, and that people who are against that sort of information are suppressing knowledge. That might be easier to see as wrong, but what about the middle ground? What are the implications of the qualitative sense you get from the chart that you wouldn't have otherwise? And is the chart itself the source of that sense? It seems like you would also get that sense if all of the numbers were replaced by X and coefficients of X or with bars or types of fruit.

[ackack]

I feel like you're only half saying a lot of things. Am I being an anti-number snob, or is my opinion of the chart anti-number snobbery?

While you are a good example of somebody I'd say is in the anti-number camp, it is much broader than just you. It seems like the default position among experienced players in ##crawl, for example. As far as the snobbery goes, I just feel like many of the conversations on the subject read something like "I want numbers!" "You don't actually want them and you certainly don't need them."

Is my ability to give advice impaired by the fact that I no longer remember what it's like to be a new player?

No, what I said was that I think that some advice (??skills is a very good example) is not very useful to new players because it relies too much on already having good judgment. I would say that your general theme that truly new players should just leave autoskills on and worry about other things is actually pretty good advice.

What are the implications of the qualitative sense you get from the chart that you wouldn't have otherwise?

After having a look at how much more expensive later levels are compared to earlier ones

--- I spread skills substantially earlier;
--- I am more likely to invest modest amounts in skills to boost the effectiveness of things I find (e.g. a little Throwing for an early stack of javelins, Evo for early wands) because it's not actually all that expensive
--- except in very rare cases, I avoid claymores/exec axes/bardiches;

It is impossible to say that I would never have done these things otherwise. All I can say is that I started doing them after looking at the chart.

And is the chart itself the source of that sense? It seems like you would also get that sense if all of the numbers were replaced by X and coefficients of X or with bars or types of fruit.

Of course you could rescale it or present it as a graph and the things that I'm talking about would still be available from it (provided that the relative scale is preserved.) I don't have a problem with somebody saying there are more effective means of presenting the data. But I do think that something more quantitative than "higher skills cost more" is needed to make some of these assessments.

[Hurkyl]

Like my example before the difference between 2 times and 2.2 times more expensive is rarely a useful thing to know. But the difference between 2 times and 10 times more expensive is a relatively big deal.

And even if one has somehow managed to have an accurate feel for the relative rates of skill gain, they're still muddled feelings; being able to explicitly what you're trading off makes it far easier to reason about it, as ackack describes.

[Lasty]

I love numbers. I like looking up numbers, and I like doing needlessly precise calculations. I do it for fun. I even do it for Crawl. But here's the thing: I don't actually use those numbers when playing Crawl. Ever. I don't fsim things when I have questions. I don't save my game and source dive to figure out what the probable outcomes of my next actions are. I don't even use irc commands to look up monster damage or even consider any fixed numbers for monster max damage. The one formula I sometimes stop and employ is the one for Armour skill's impact on AC, because it's quick and gives some useful info. I could easily get by without it, but I do enjoy doing calculations.

The reason I never look those up is that none of those things are necessary to play Crawl well, and none of them are particularly relevant to success unless you're indulging bad play habits. That's not to say that data isn't important: there are a huge number of things to know about the game that will increase your chances of victory, its just that very few of them involve exact numbers. Knowing that each skill level costs more than the previous one contributes significantly to your success, but knowing the degree to which later skill levels cost more than early ones in general terms (significantly more) is just as useful as knowing it in specific terms. Knowing the amount in specific terms and applying it carefully encourages a level of focus on exact skill point distribution and pre-planning that is more likely to subvert good play than promote it, and so it's valid to say that good advice on the subject should be in general terms rather than specific ones.

[WalkerBoh]

Giving someone all of the specific numbers in Crawl is like giving a terminally ill patient a placebo. It doesn't actually help at all, but it makes them feel better, and then they die.

Disclaimer: I am not a doctor.

[damiac]

You don't need a weapon better than a vorpal spear to win at crawl. You don't need to skill wisely to win at crawl. But a better weapon will make it easier to win. A more efficient expenditure of experience will also make it easier to win.

I agree, trying to say ahead of time "I will train X skill to level Y, then switch to Z skill, etc... is a bad idea. Trying to pre plan a character doesn't really work, because by doing so you let the RNG screw you over, rather than playing it to your advantage.

However, as ackack and Hurkl said, I've already seen improvements in my play from looking at the numbers on that table. I knew it was often said 'lower level skills cost less than higher level skills', but seeing a number lets me understand that we're not talking about 10% more a level or something. It's something I can actually use, unlike vague advice like 'improve your offensive skills until your offense is good enough'.

[Hurkyl]

I love numbers. I like looking up numbers, and I like doing needlessly precise calculations. I do it for fun. I even do it for Crawl. But here's the thing: I don't actually use those numbers when playing Crawl. Ever.

Actually you do. All of those numbers you've looked up and all of those calculations you've done have given you information you've learned from, and you put the resulting knowledge to use when you're playing crawl.

Just like I don't look up the numbers on MP efficiency when I play a conjurer early game: I already did that once and learned that Searing Ray is ridiculously awesome and know to use that whenever I can afford to take my time. And other such facts.

[crate]

honestly in this specific case i'm mostly against the numbers because skills dont matter much in crawl. giving numbers gives the impression they do matter.
learn tactics if you want to win

but again, I would like to point out that my primary objection here was not objecting to Patashu giving numbers (I think people are misunderstanding this point somehow despite the fact I made an entire post about it). My objection was that the numbers he gave are completely useless without further information. In this case it is strictly worse than either giving no numbers OR than giving the additional numbers necessary to make the ones he did give not useless.