Random Hit Points

[zaimoni]

To zaimoni's comment: I must disagree; it *is* noticeable. It's how you get those really low 8th and 10th level hit point totals.

That's not how I got ~48 hp at CL 8 for a V2.9.3 Dunadan paladin with 17 CON.

When your rolls are below average, even -1 hit point per level REALLY stands out in those cases when the die size is smaller...primarily, any mage, any priest except maybe half-troll, and any hobbit, kobold, or elf ranger, rogue, or paladin.

-1 hp/level is CON 5 (until the stat tables get rebalanced).

In the early game, rolling a bunch of 1's and 2's, versus near-maximum rolls, dominates starting CON. Even that matters before stat-gain only for the optimized melee competent classes (fighter and paladin); the others don't have enough melee attack damage rate to make burning through hp, and then recovering, a viable level-vacuuming early-game strategy.

[Timo Pietilä]

Actually, the hit dice for all levels are rolled at character creation. [ZAngband variants also provide potions for rerolling those hit dice en masse.] Character level fifty hitpoints, ignoring CON bonuses, are constrained to be between 3/8ths and 5/8ths of theoretical maximum.

Potion of life used to re-roll those giving you new "life rating". Not sure if this is still the case.

[Derakon]

No it didn't. That's the Potion of New Life, which, as noted, was added in ZAngband.

[MKula]

Is it really so bad to have a character with low early hitpoints, though? It'll balance out eventually. Some characters have harder starts than others; some have easier.

I don't know if I really agree with this. If it's supposed to balance out eventually, why not make it balance out right from the start?
In any case, the reason it does indeed 'balance out eventually' is that in the end the deciding factor is the huge CON bonus (a fixed value). Seems to me that reducing the random factor (either a bit or a lot) would be a better approach to "correct" hit points.

[Derakon]

I don't know if I really agree with this. If it's supposed to balance out eventually, why not make it balance out right from the start?

Because that means that you never get to experience having to protect a character with unusually low health -- an extra challenge, the kind of unpredictable thing that makes Angband fun. Keeping hitpoints always basically the same (which is what reducing / removing the random factor would do) means that you're never pushed outside your comfort zone, and who wants that?

In any case, the reason it does indeed 'balance out eventually' is that in the end the deciding factor is the huge CON bonus (a fixed value). Seems to me that reducing the random factor (either a bit or a lot) would be a better approach to "correct" hit points.

It is true that CON bonuses tend to overwhelm the influence of hitdice, but that's not what I meant by averaging out. It balances out in the end because the sum of your hitpoint rolls is not allowed to deviate more than a maximal amount from the average (see zaimoni's post earlier in the thread). In theory you could roll max hitdice for your first 25 levels and min hitdice for your second 25 levels and still come out average, or vice versa.

[Timo Pietilä]

No it didn't. That's the Potion of New Life, which, as noted, was added in ZAngband.

Are you absolutely sure about that? I have vague memory getting my life rating changed with potion of life (in the time when mages still had invulnerability). Though I played TY-era Z quite a lot so I'm not entirely sure how accurate my memory is.

[Timo Pietilä]

I don't know if I really agree with this. If it's supposed to balance out eventually, why not make it balance out right from the start?
In any case, the reason it does indeed 'balance out eventually' is that in the end the deciding factor is the huge CON bonus (a fixed value). Seems to me that reducing the random factor (either a bit or a lot) would be a better approach to "correct" hit points.

Wrong use of word "balanced". Game is balanced when there are differences between games, it makes life a bit more interesting. Game is already too "balanced" in a sense that all differences have been tried to eradicate.

Rough start smooths out at the end.

[Derakon]

Are you absolutely sure about that? I have vague memory getting my life rating changed with potion of life (in the time when mages still had invulnerability). Though I played TY-era Z quite a lot so I'm not entirely sure how accurate my memory is.

Not absolutely, but damn near. I certainly would have mentioned rerolling the life rating in the potion of life description in TANG if that had been one of its effects, and I didn't.

[MKula]

Your points are all valid (Dera and Timo). I certainly agree that variability between characters is a good thing, but I don't know if completely randomized hit points is the way to approach it. That seems to me less a "feature"... more of an "unpolished" (not quite a "bug").

The point I'm trying to make is that when you roll a certain new @, you expect certain things. Hit points should be the same. When you roll a gnome mage, you expect to have low HP. This is non-negotiable based on the race/class combo you selected. When you roll a half-troll warrior, you expect (among other things) to have high HP. The fact that low HP as a half-troll warrior is a real possibility seems a bit... I dunno... sloppy?
Again, this is all about the low levels, since CON bonuses at high levels blah blah.

Based on this thread, I like 1d(race) + 1d(class) over 1d(race +class). Actually, I just realized it might even simplify the code since law of averages with this system would not really require a check to make sure HP is between 3/8 and 5/8 of max? Can someone verify this?

[Hariolor]

A popular hack in many RPGs is to simply guarantee max HP at first level...

[Derakon]

Which is already done in Angband.

[Hariolor]

right...point being it's good enough for other games, probably no reason Angband needs to go beyond that. Occasionally you'll get mages with an excess of HP and warriors with too few - I fail to see what needs correcting about that.

[buzzkill]

right...point being it's good enough for other games, probably no reason Angband needs to go beyond that. Occasionally you'll get mages with an excess of HP and warriors with too few - I fail to see what needs correcting about that.

... and a vast majority of the time it will be something between the extremes and extremely appropriate (whatever that is). I'm with Hariolor on this. In a game where almost everything is random, why worry about HP's being a little to high/low. If anything, I'd vote for more variability, not less.

[kaypy]

Assuming for the sake of technical argument that we want to "fix" this...

A technically easy solution is to sort the hit dice rolls in nonstrictly decreasing order.

This seems in the right direction, but giving all the good rolls upfront seems somewhat overkill. Is there a reasonable algorithm to go from an increasing or decreasing sort order to something roughly even?

So from something like
2 3 3 4 5 6 7 7 8 8 9 9
to something like
2 7 6 9 5 4 3 8 8 7 3 9

I got the latter by:
- dividing the 12 items into 4 buckets
- calculating the sum of the differences between the buckets
- randomly picking 2 items
- swapping them if it decreased the total difference
- repeating until I had 5 no-swaps in a row
but surely there's a better way than that (for starters, this can still wind up a no-op if the RNG conspires against you...)

[zaimoni]

Assuming for the sake of technical argument that we want to "fix" this...

A technically easy solution is to sort the hit dice rolls in nonstrictly decreasing order.

This seems in the right direction, but giving all the good rolls upfront seems somewhat overkill.

It is overkill in isolation, yes.

As for reasonable algorithm...ignoring the vagueness of "reasonable", the real problem is that sorting has a lot of research so is technically easy. "almost sorting" is harder to recognize when it's working.