Do we need points based stat generation at all?

[PowerDiver]

You may have meant this, but in my own words:

If a game engine provides for "multiple blows per round" or shots per round, then they should all fire at once and not allow the player to use part of them and then move midway through the round.

So if somebody had 2 shots per round, their bow would automatically fire twice when used. (Or for a different perspective, would notch and fire two arrows with the same string pull.)

What I said is a bit more subtle. I want to eliminate the whole idea of a round. So when fighting with 4 blows against someone fast, rather than 4swings by you then 2 moves by them, you get 2 swings by you then 1 move by them. I am a bit ambivalent about doing the same thing with a bow, since ego ammo is so precious, and at a minimum you'd have to special case when a stack of ammo was depleted anyway. In a world of unenchanted ammo with attributes imparted by the launcher it would be clear.

Also, my system could still abused for hack-and-back if you were fighting a pack of monsters, since most likely you would be disturbed quite often. Something would have to be done about that, not because of the abuse, but because it would be too annoying to get only one swing per keypress.

[Atarlost]

Of course hack'n'back is the only actual tactic for melee that's more advanced than bashing the enemy mindlessly.

[ajps]

I think the answer is only to boost dex [before the con phase] if it gives another blow. That means calc_blows needs to be refactored, passed a weight instead of an object, and also passed str and dex mods. In addition, base str of 18 should be considerd if that would give an extra blow. Perhaps looking for extra blows from stats should be combined with noting required stat improvements for the next extra blow when examining a weapon.

I really cba doing that, so I'll just update it to what you originally intended the algorithm to do for now, which will hopefully be enough to provide my aim which was a reasonably playable character for a given random race/class combo.

Incidentally, http://rephial.org/research/birthstats3.txt shows what would happen if we simply skipped the initial allocation of DEX for pure spellcasters. Is this an improvement?

[Nick]

OK, some extreme thread necromancy. Here is the post that the current birth default stat allocations are based on:

One simple approach would be to have "standard" point distributions for every race/class combination and allow re-distribution from that start point.

There are too many combos for a table, so we need a simple algorithm. Here's a possible approach.

buy base STR 17

if possible buy adj DEX of 18/10

spend up to half remaining points on each of spell-stat and con, but only up to max base of 16 unless a pure class [mage or priest or warrior]

If there are any points left, spend as much as possible in order on DEX, non-spell-stat, CHR.

They have served pretty well for the last 11 years or so, but are now clearly non-optimal for blackguards (who prefer INT to DEX). So do we:

1. Leave it as is and let blackguard players work it out for themselves;
2. Adjust the algorithm somehow or
3. Give each class it's own algorithm?

I'm kind of leaning toward 3, and writing the algorithm into the class.txt datafile, but I still need to work out how to do that, and also could do with opinions on how the algorithm should go for all the classes.

[archolewa]

I'm kind of leaning toward 3, and writing the algorithm into the class.txt datafile, but I still need to work out how to do that, and also could do with opinions on how the algorithm should go for all the classes.

Does it need to be anything remotely fancy? Why not make it data driven. For each stat you specify how many stat points should be spent on it. So warrior could look something like (numbers are arbitrary, and bear no relation to how things actually work in the game):

Str: 5
Dex: 4
Int: 0
Wis: 0
Con: 0

Blackguard maybe might be more like

Str: 5
Dex: 2
Int: 2
Wis: 0
Con: 0

This might be a little bit brittle, and could break (either spend too few or too many points) if we ever tweaked the cost of stat buy. However, I have a sneaking suspicion that changing stat buy isn't exactly high on the priority list, so that's probably not a problem. It also means that your logic can be dead simple, and it's really easy to add additional classes, or tweak a class' starting point allocation.

[DavidMedley]

Zombie thread trying to eat my brain

[Nick]

An extra wrinkle is race dependency, because it's not how many points you put in but where you get to that matters. So maybe each class should have successive targets to reach - STR 18, than INT 17, then DEX 18/10 for blackguards for example (numbers not well thought out).

[archolewa]

An extra wrinkle is race dependency, because it's not how many points you put in but where you get to that matters. So maybe each class should have successive targets to reach - STR 18, than INT 17, then DEX 18/10 for blackguards for example (numbers not well thought out).

That's true. I guess a question to ask ourselves is: "Do we want stat allocation to be optimal, or good enough that a player new to the class isn't inordinately screwed and is able to enjoy the class as intended?"

If we want it to be optimal, then yeah we'd have to somehow take race into account, perhaps by setting targets. Though setting targets is tricky, because what happens if you can't reach the target for say Int, but you can for Dex (since you're playing a Half-Troll)? Should you get as close to Int as possible, even if that means you can't hit the Dex target either, or should you hit the Dex target since you can, even though that hurts Int?

If we just care about "good enough to get started, but not necessarily optimal" then you could probably set the targets based on either humans (i.e. no stat bonuses) or the race that we expect is the most common combination (i.e. Half-Troll for warrriors and maybe Blackguards, hobbits for rogues).

[DavidMedley]

Was hoping to escape the zombies long enough to sleep, but they caught up to me.

The true root issue is that the bonuses are very opaque. 3rd edition D&D cleaned this up, but Angband is stuck with this system seemingly based on 2nd ed.

One of the very first things that I put into my spreadsheet is the bonuses for each of the 5 stats. I reference them frequently when creating a character. The approach Nick outlined is basically what I'm doing, and would be a very smart way to program a default buy. But it's very subjective, and would be complex to specify a good list, given the 8 and 12 cost levels.

I think the very best solution would be to throw out those old tables and make something sensible. The next best would be to let the player see in-game what each level actually does for their character.

Next best (tho not mutually exclusive) is Nick's approach above. Hitting the breakpoints that the player doesn't know about would be a big improvement over blithely slamming 12 points into Dex. I'd be happy to add my 2 cents to that list of breakpoint targets.

A simpler approach would be to spend the 20 points based solely on class, and this could fit nicely into class.txt. This wouldn't hit the breakpoints but would communicate which stats are to be valued.

### Warrior ###
Birth-points:8:0:0:6:6

### Blackguard ###
Birth-points:6:4:0:5:5

(I don't love spending 8 points on one stat but I will for crucial breakpoints, but I never spend 12. Others may differ. Also, many eschew Con more than I do. My personal point buy preferences aren't important, but felt obliged to add this note after giving the Birth-points suggestion above.)

[bughunter]

I'm a relative noob, especially when it comes to the esoterica of point-based stat generation. Can someone please explain it in a little more detail?

I quickly learned I could do better by rolling rather than point buying, and got good at adding up the stats and stopping to inspect stat sets with a high sum (usually >72 doesn't take too long to find) and at least one 18 in prime requisite(s).

Why is a point-based generation scheme deemed superior (or at least a desirable alternative) to munchkin rolling like this?

Also, the AD&D 2e PHB has several stat generation options. Are none of these suitable for Angband?

(I assume a better understanding of the first question would help answer the latter two...)

[DavidMedley]

I'm a relative noob, especially when it comes to the esoterica of point-based stat generation. Can someone please explain it in a little more detail?

Pretty much everything I know about where to put your points is in here:

https://docs.google.com/spreadsheets/d/1nAVk2AodMHloqhLBzZm7p_s4mne3My4Kyo9wODJWAFQ/edit#gid=133942851

The DEX "Blows Index" is probably the most important one. I set this one first at 10, 17 or 18/20. Sometimes I'll put a leftover point to get 18 because you get everything except more blows.

For half-casters, I think the 13 and 15 levels for INT or WIS are crucial. For others, it is often worth a point or two to get an improvement in save or devices imo. For full casters I don't think there's a hard and fast rule. You can put in 8 points or more, or you can spread them around to be more well-rounded. Pretty sure I always end up with 5-8 points in a full caster's prime requisite.

For CON, 14 and 18 are great because they improve recovery times from stunning, poison and cuts. Many players don't care about the extra HP, but going from 14 to 18 gives an extra 5 HP every 4 levels, plus the recovery gain, and that seems worthwhile to me if you can find the points. But going from 9 to 13 and 14 to 17 isn't worth much.

Finally, STR is the most important stat overall. It doesn't have the same break-point benefits as the others, but I put 4-8 points in this no matter the class. If you know the weight of the weapon you'll be using then you can use Blows Coefficient breakpoints. For example, the Blackguard starts with a 10 lb Tulwar, and therefore get extra blows at STR 18, 18/20 and 18/40. I'll put in that 8th point to hit one of these, but not to hit 18/10 or 18/30. The Carry Limit is very useful for all classes.

I don't like spending 12 points on anything. That's 3 extra statgain potions you'll need later.

[whartung]

Develop the mechanism to capture the algorithm in a text file.

Provide a default algorithm.

Allow it to be specialized for class and class/race combos.

Provide one for the Blackguard.

Let the community do the rest, like macros.

It's there any reason the algorithm can't be a simple list of desired stats?

For example, using your default:

STR=17,DEX=18/10,(INT,CON)=16

Which reads:

Allocate points to 17 STR
Allocate points to 18/10 DEX
Allocate 1 point to INT, then CON, one at a time, until both are 16.

I could be that simple, then just tag it for class or class/race. And then let fly.

For you default, you may have to create a specialization for, say, priests to use WIS instead of INT.

But, no need out the gate to do one for everything. Just let folks customize it to their liking, and over time, recipes will start to accumulate.

[Nick]

Some further thoughts:

• I agree with David's point that spending 12 on anything is a bad deal
• I agree with Pete's point (from 2009) that CON is undervalued
• I think the current immediate spend of 8 on STR is probably correct (although maybe not for half-trolls)
• I feel like break-points are really only interesting for DEX and blows, and then only for warriors, paladins, and probably rangers and rogues (maybe priests and druids are slightly interested?)
• I think full casters, especially mages and necros, want to spend big on their spell stat

This is probably getting pretty close to an algorithm.

[DavidMedley]

Seems closer to specifying a default buy for each class, which is easier anyway.

[Werbaer]

I always spend 12 points on the spell stat as a pure caster. Getting the stat up to 18/50 is huge. And with an internal stat of 18, you only need one stat (swap) potion to get real good benefits.