Stat Strategy ("stategy" as the Pros call it)

[ZipZipskins]

Okay so the shindig is, in previous versions of Crawl which stat to pick when given a choice on levelling up was either binary or a pretty simple flowchart. Has anything changed recently here? And whether it has or hasn't, what is the optimal "Stategy" for placing points? Or at least what are the big considerations?

[petercordia]

I'm not a pro, but here are some of my considerations:
for non-hex spells, once spellpower reaches 50 (#####) Int becomes less important, because it won't affect spell power as much.
for spell success, strength is more important than int when strength<encumbrance, and less important otherwise.
I usually neglect one of my stats (strength, dex, or int)

I'm also curious what the pro's do.

[Siegurt]

Stats are somewhat clearer choices as compared to prior versions,

Strength adds to melee damage, and decreases ER from body armour (most importantly this lets you wear heavier armour and cast spells with higher success rates)
Int adds to spellpower and spell success rate
Dex adds to accuracy, Evasion and (slightly) to stabbing.
Strength and dex in equal measure impact your odds of scoring an aux attack (so if that's your rationale strength is better)
Shields are also impacted by strength and dex, to an extent that you shouldn't care about.

Dex no longer adds to weapon damage, and strength no longer adds to weapon accuracy.

ER (compared to long ago) is always reduced by more strength (there's no hard breakpoint any longer) but with decreasing effect. It's such that adding to strength as a spellcaster can actually increase spell success by more than increasing int, depending on what your strength is, and what body armour you're wearing (An extremely rough "rule of thumb" is that you probably want the ER
of the body armour you want to wear +10-20% in strength, although there's oodles of caveats and "not strictly accurate in all situations" to that rule of thumb)

Int is pretty much unchanged from old versions (int acts as a multiplier for spells, so doubling your int is roughly equal to doubling your skills)

So mostly it's "raise int or strength depending on what you're doing damage with and possibly depending on what armour your wearing" or "raise dex if you want to increase your defense on an EV-centric build"

I guess technically also raise dex on a non-armour-using-non-spellcasting stabber-who-doesn't-do-much-melee, but that's an awful and dumb build, don't do that to yourself.

[Siegurt]

for non-hex spells, once spellpower reaches 50 (#####) Int becomes less important, because it won't affect spell power as much.

Technically speaking Int always effects spellpower up to the cap, it's just that spellpower itself has decreasing effect, this is true of hex and non-hex spells, but hex spells since they have a spellpower-dependant binary effect, so the result of more spellpower is always visible.

Some nitty gritty details for people who want to know them:

Spoiler: show

Int acts as a multiplier on your *raw* spellpower, you get your (spellpower/4 + skills)*Int (leaving out some details here) and that's subject to a 'stepdown' which turns it from a straight 1:1 scale to a logarithmic one, which in turn is used in the various spell formula for power (damage and accuracy on damage spells, hex power on hexes, duration on duration spells etc.)

The actual #'s themselves represent the "stepped down" spellpower in a further (very approximately) logarithmic way, in that more hashes take a lot more "stepped down" power to display.

The breakpoints for one more # in spellpower display get further and further apart, making the "stepped down" power growth appear even more drastic than it actually is.

  Code:

Value          Description
0-9          #.........
10-14        ##........
15-24        ###.......
25-34        ####......
35-49        #####.....
50-74        ######....
75-99        #######...
100-149      ########..
150-199      #########.
200          ##########


Since "stepped down" spellpower isn't actually displayed anywhere in game, it's not terribly helpful to know what it is (moreover how spellpower is used varies from spell to spell, for example nearly all damage spells have some set "base" damage, and get some bonus based on spellpower to that, meaning their damage doesn't increase 1:1 with an increase in stepped-down power, but rather at some ratio less than that.)

Hexes *are* somewhat unusual, in that for many hex spells, the 'stepped down' power (possibly with a positive or negative multiplier) is used in a direct contest against the monster's magic resistance to determine success

Damage spells get more damage from having more int, duration spells get longer duration, etc. etc. it's just that you *care* a lot less if your damage spell gets a 1% increase in it's average damage.

[Airwolf]

Some nitty gritty details for people who want to know them:

Ah! My eyes!

[Siegurt]

Some nitty gritty details for people who want to know them:

Ah! My eyes!

Clearly you did not want to know.

[stormdragon]

Strength adds to melee damage

I thought it increased attack damage in general (including ranged attacks), not just melee. Am I mistaken?

[tasonir]

My rule of thumb is fairly simple: Start by always worshipping chei, then just dump the rest into int since you already have too much strength and dex. Optionally, if you started as a mage background and somehow already have 35+ int, you can put a few points into strength. Don't add more to dex, since with the +15 you're already way into diminishing returns.

Honestly this isn't too different from most non-chei characters. You probably want int unless your strength is too low to get your armor "comfortable". If you aren't casting or only cast a few low level spells, then just take strength. For some truly spell-less characters (trog), you can probably split a few points into dex just because you don't need +9 strength on top of berserker starting strength.

@stormdragon: ranged damage is basically the same as melee damage, but yeah they mean "weapon damage". It's all one weapon system now (they used to have separate mechanics back in .14 or so).

[stormdragon]

@stormdragon: ranged damage is basically the same as melee damage, but yeah they mean "weapon damage". It's all one weapon system now (they used to have separate mechanics back in .14 or so).

I'm pretty strength increases unarmed as well, not just weapon damage. That's why I say it improves "attack damage"...

[stormdragon]

To answer the OP, I think the rule of thumb is: if you kill things with attacks, pick strength. If you kill things with spells, pick intelligence.

Hypothetically if you don't kill things with attacks or with spells, pick dexterity; but like Siegurt said, it's hard to imagine a good build like that. Maybe something to do with Elyvilon.

[Siegurt]

@stormdragon: ranged damage is basically the same as melee damage, but yeah they mean "weapon damage". It's all one weapon system now (they used to have separate mechanics back in .14 or so).

I'm pretty strength increases unarmed as well, not just weapon damage. That's why I say it improves "attack damage"...

Yes, that's correct, my "melee attack" original was an unintentional misrepresentation, it's any damage done with a melee, ranged or aux attack.

[svendre]

I mainly pump strength. Just about any character needs some defense from armour, and melee or ranged ability helps enormously. You're pretty much guaranteed to find some weapon worth using. You aren't guaranteed to find what you want to build a good spell caster. If you never find good spells or spell casting gear, you will always have a build that works well. If you do find some great spells and/or nice artefacts with +Int on them, all that strength you added will help you cast higher level spells in heavier armour. So, you're covered either way.

If you determine at some point you're never going to cast spells, or only just low level ones, there might come a point where I add a few points of dex to slightly improve accuracy if there is a large enough discrepancy between str and dex.

If you find wonderful body armour you believe you want to continue using for the rest of the game, and you are casting spells (or want to), this can be the point at which you might start adding int when you have enough str+int to overcome the encumbrance of the base armour type for the spells you want to case, then you are adding spell power. Other reasons to add int may be Gozag, or you're having food issues and casting.

The gods which help you populate your gear sooner than later (Oka, Gozag, Trog), actually help you in deciding which stat to bring up. This is one reason why I prefer them over taking Chei to start. Chei's stat boosts are extremely powerful, but not as insightful if you're trying to optimize for a longer game. By the time you get a good set of gear, and if you're going into extended, things like torment wind up being a bigger factor in god selection than having extra stats (surely debatable, Chei is strong).

Some rough examples of proportions to aim for that I find work well (without all the math) from experience:

Hybrid with level 1-8 spells:
str 30
int 30
dex 15

Bruiser with level 1-5 spells:
str 40
int 20
dex 15

Pure bruiser:
Str 40
int 10 or less
dex 25

At around 40 str, you're pretty deadly to the point of more damage not helping that much.
It's good to keep stats above at least 5-10 to be safe from stat draining effects.
Finally, you want to consider your aptitudes in determining stat allocation, to a degree.

[nago]

How the fuck you get those stat in a game without zig raiding or chei?
(ok a DE can reach 40+ int without too much hassle, or a Dg another stat...)

[svendre]

How the fuck you get those stat in a game without zig raiding or chei?
(ok a DE can reach 40+ int without too much hassle, or a Dg another stat...)

As I said, Gozag or Oka. Mostly Gozag. If you save your gold until you've checked all the usual spots where shops happen, and don't pay for consumables much then you can spend your money on armour and jewelry shops and search for permanent end-game quality level gear. I prioritize gear with stats on it. That's not to say I'd get those kind of stats every game and certainly not right away. The point was demonstrating values you could aim for with different focuses, and may be obtained by level 27 and with good gear.

[byrel]

To answer the OP, I think the rule of thumb is: if you kill things with attacks, pick strength. If you kill things with spells, pick intelligence.

Hypothetically if you don't kill things with attacks or with spells, pick dexterity; but like Siegurt said, it's hard to imagine a good build like that. Maybe something to do with Elyvilon.

I believe Dex is highly valued on transmuters, since you aren't wearing armour (so Str is less valuable), but you don't care about spell power much (so Int is less valuable.)

[petercordia]

Dex is also good on felids and octopodes

[stormdragon]

I believe Dex is highly valued on transmuters, since you aren't wearing armour (so Str is less valuable), but you don't care about spell power much (so Int is less valuable.)

Str is almost certainly better than Dex in this case because of the damage bonus.

For example, say you have 13 UC and blade hands. Your base damage is 13 + 22 = 35. Then maybe you have about 10 fighting which will increase this by about 20%, to 42. A single added point of strength increases your base damage by 2.5%, and 42 * 0.025 = 1.05. You get about as much damage from Str as from slaying. This is a typical lair-ish transmuter. As you get more UC & fighting it gets even better. I don't know how to calculate the EV from Dex but in my experience you get between 0 and 1 EV per point of Dex, and even at 1 EV per Dex, Str would probably be better.

[tasonir]

@stormdragon: ranged damage is basically the same as melee damage, but yeah they mean "weapon damage". It's all one weapon system now (they used to have separate mechanics back in .14 or so).

I'm pretty strength increases unarmed as well, not just weapon damage. That's why I say it improves "attack damage"...

Correct. That's what I meant, although I'll fully admit that using the term weapon damage could be misleading. In this context, unarmed is a weapon, so I probably should say "attack damage" or the much longer "strength increases all melee weapon, unarmed, and ranged weapon attacks". I think it also includes aux attacks, as people were mentioning, although I don't know if I've ever checked that. Aux attacks are usually not really that important of a source of damage, unless you're a minotaur or vine stalker.

[Siegurt]

I think it also includes aux attacks.

It does.

[TheMeInTeam]

Dex is good on stabbers also, it contributes to both how much damage you get when stabbing successfully but also how frequently you can stab monsters when they're confused/non-tier 1 stabs. On a character likely to benefit from the EV conferred also it's actually doing a fair amount. For builds swinging a heavy weapon, relying on spells, or wearing heavy armor you have better options.

Most builds should just go STR for more weapon damage or INT for better spell power/spell success rates. If you're going to do extended you might want to push a low stat a bit to avoid stat 0 but this usually doesn't matter otherwise.

[acoolguy]

Not sure if this is suboptimal but I tend to value dex a bit more. In general, I train skills in such a way that I can kill dudes reasonably well and then train skills that help me not die. Similarly, I tend to allocate my stats in a similar manner.

On str vs dex:
Early on, I'll pump strength to ensure I have enough damage to kill dudes and wear the ideal armor. Later, I will generally allocate a few levels in dex and train some dodging to increase my ev even on high ac melee dudes.

On int vs dex:
I'll pump int whenever I need to get key spells castable. Later on, I'll pump either str to wear heavier armor or dex for higher EV gains.

[tabstorm]

Strength on every levelup if you mostly use melee or archery, int on every level up if you mostly use conjurations/summoning/necro/hexes. On lower-strength warrior-mages it can sometimes be better to take strength, I think? Warrior-mages are generally not the best.

[TheMeInTeam]

Warrior mages are perfectly fine and eventually outperform pure warriors by a margin, and arguably outperform mages too since they have mostly the same spells by then.

You want enough STR to wear your armor, then INT for spell success in their case. Fire, ice, pearl dragon scales are all ~11 ER, so 13 str is plenty for those and still castable. Going a bit lighter you have swamp scales, which can still be enchanted to +7 for very respectable AC and only 7 ER.

Going with troll leather or steam allows Ozocubu's armor, which at good power will give more than 10 AC. Bit less GDR but your evasion will be better than characters in heavier armor, a steam dragon scales setup can therefore reach 80+ total across AC/EV/SH just by casting that, doing even better with statue form of course.

Here's an example of a warrior-mage that can handle anything in 3 rune pretty easily and would still make it through extended with a few small tweaks via amnesia scrolls + minimal training: https://crawl.kelbi.org/crawl/morgue/Th ... 121045.txt

In that case I went strength for the extra damage, though were I to do that particular game again I'd probably take dex to proc stabs more often and have the rapier do more damage when stabbing, since this would also have granted more EV.

[damerell]

for spell success, strength is more important than int when strength<encumbrance, and less important otherwise.

Variants on this have been said several times in this thread. It isn't true; there is an important ER^2 term in the formula. No rule that compares ER and strength linearly is correct.

[petercordia]

The formula is roughly

int-penalty = ER^2/strength

taking the derivative you find:

effective-increase-in-intelligence-per-extra-point-of-strength = ER^2/Strength^2

Hence this is greater than 1 if ER>Strength, smaller than 1 otherwise.

[damerell]

The formula is roughly
int-penalty = ER^2/strength
taking the derivative you find:

So the fundamental principle here is sound (and new to me and very interesting); since it is the derivative with respect to strength that matters when comparing the relative utility of Str and Int increases, we do get a Str^2 term.

However, the actual formula is a bit more complicated than that. A point of Int is worth 2% spell failure; that bit is easy.

http://crawl.chaosforge.org/Spell_failure#Armour_and_shield_penalties_.28Under_Construction.29 is presently accurate for body armour. (I'm not saying it's not for shields, I just haven't checked).

Hence for example if I wear armour with an ER of 18 and have an Armour skill of 18 and a Strength of 18, the armour adds 70.4% to spell failure chances. (Beware wizmode testing of this - the extensive post-processing of the number means you can't ever see that 70.4% in game.)

If I were to increase my Strength by one, it would change the penalty to 67.2% - a difference of 3.2%. This is far greater than the 2% I would get by taking a point of Intelligence. The situation favours Strength even more with low Armour skill.