Fighting reform

[crate]

My numbers concerns that I posted earlier are still there.

I will assume that changing the effectiveness of common earlygame weapons at 0/low weapon skill is not an option (because of the fact this would need to be accompanied by changes to basically every early game enemy as well as every low level offensive spell).

In a linear delay system--which is what Crawl currently has--then how fast a weapon gets better is determined by its base delay. Faster weapons get better more quickly than slow weapons. In a linear speed system this is completely different--how fast a weapon gets better is instead determined by its damage. More damaging weapons get better faster than less damaging ones. Roughly, damage and delay swap roles: low base delay -> high base damage, and high base damage -> low base delay. (see footnote 1)

Something is going to have to change pretty dramatically if you want to use a linear speed formula, since right now most onehanders are bad weapons (dam/delay < 1, where delay is in aut ... in fact among common weapons scimitars are the best at 6/7 or about .85) that get better quickly and most twohanders are good weapons (dam/delay ~= 1; exec axe and bardiche are 0.9 but get cleave/reaching) that get better slowly. Since changing early game weapons is not really an option probably the least disruptive course of action is to make twohanders very bad weapons (dam/delay < ~.6) that get better quickly, while leaving onehanded weapons as bad weapons that get better slowly. In doing this you can manage to make onehanders actually better weapons than twohanders at low skill. This will lead to other balance changes though, since to keep twohanders better than onehanders at high skill (which absolutely should be desirable) and to also keep their dam/turn comparable to the current situation near mindelay skill (I assume this is also desirable) probably what happens is they will need a base damage increase as well as a much larger base delay increase (see footnote 2), which then reduces the importance of AC and slaying and other additive/subtractive things that don't get multiplied by weapon skill.

Additionally you make twohanders completely unusable at low skill. Perhaps this is a good change though.

footnote 1:

Spoiler: show

Obviously you can adjust weapon improvement in a linear delay system by changing the weapon's damage. In fact many crawl onehanders have bad enough damage that despite their low base delay they still don't improve quickly enough to ever beat out twohanded weapons (compare hand axes to an exec axe). Similarly you can adjust weapon improvement in a linear speed system by changing the base delay. These numbers don't qualitatively effect how quickly the weapon gets better however, so they're less important.

footnote 2:

Spoiler: show

For example, let's assume we make the base delay of an executioner's axe 400% to get it into the "very bad weapon" category. If we want to keep its damage at skill 26 comparable to the current dam 18/delay 200% exec axe then it needs to have a base damage of 21! A 21/400% exec axe in the proposed linear speed system actually manages to be worse than a quick blade all the way from skill 0 to skill 13, and worse than a scimitar from skill 0 to skill 12. It's also complete suicide to swing at skill 0 against anything dangerous (and you'll be getting double hit sometimes until you get to axe skill 23).

If we instead make the exec base delay only 300% then it needs dam 20 to be comparable to current exec axe at skill 26 and is better than a qblade at skill 11 and a scimitar at skill 7. As you can see you need really huge base delay on the big twohanders to make onehanded weapons ever a good choice for damage.

Additionally all onehanded weapons deal less damage at their mindelay skill in the proposed linear speed system than they do currently in the linear delay system, so probably the high-end onehanders would need stat adjustments as well....

Anyway these adjustments may be possible but you are potentially disrupting melee combat's damage an awful lot with these changes, and even afterward there is going to be that important qualitative difference in how weapons improve.

edit: obviously you can get around this by changing how quickly different weapons improve but as I said earlier I fail to see how that is actually an improvement way over the current linear delay system, since you will basically have to tell players what "C" (in the formula on the devwiki) is for each weapon so you're just replacing one magic number that's reasonably easy to understand with a magic number that's hard to understand.

[Wahaha]

Do I understand correctly that with this change we would need to get 27 skill to get the same attack speed that we currently get with much less skill? If I wanted to use a lajatang I'd have to get 27 skill instead of 14 to have it at 0.7? If this is correct, how is this good in any way?

[CommanderC]

I have played several characters until Lair with the new branch (HuFi, 2xHuGl, and HaGl). This is what I have observed:

• High strength is probably too good on the first levels of the dungeon. Despite this, I like this change a lot. Raising strength is now a very good choice.
• Weapons with high delay and damage (like glaives) are incredibly good with ~10 skill levels. Halberds are excellent as soon as you can get one.
• Short blades are pretty weak. Despite this, the HaGl of Trog cleared all the levels until Lair:6 with a dagger of draining, then I switched to a scimitar of flaming (I had to berserk too much).
• The thing I like the most is that training exclusively your weapon skill is not as interesting or powerful as it is now. You have to think more about training your other skills (fighting, armour, dodging).

[crate]

The thing I like the most is that training exclusively your weapon skill is not as interesting or powerful as it is now.

This is both true and not true. For onehanders it's true, you just plain get less benefit from weapon skill in the linear speed system here. For good weapons (of course these pretty much don't exist in early game unless you're an ogre) you actually get significantly more benefit from the first some number of weapon skill levels (exec axe or bardiche at skill 15 in the linear speed system is the same speed as exec axe at skill ~20 in the current crawl system) so it's actually even better to train just weapon skill early on in the linear speed system. But then later levels of weapon skill both cost more and don't give you more benefit so you probably stop training your weapon earlier.

[njvack]

Would it be reasonable to deploy this to trunk on cdo/cao and see what the sentiment is?

[byrel]

I'd be interested in trying it out if it was available on webtiles.

[Thasero]

I've played a few characters on the branch for testing.

As crate says above, it does change the balance of light vs. heavy weapons quite a bit - unless your character is so young that the accuracy malus still matters, you should always use the heaviest weapon you've found. Light weapons of electricity, pain, and venom are also less interesting, because although they're still faster than heavy weapons at low skill, it's not by as much.

The change solves the hard breakpoint, but it's still the case that exclusively training your weapon skill is always the right decision for young melee characters. Although the benefits of weapon skill are spread out more evenly by the change, they're still extremely *large* benefits. You just no longer have an obvious skill level to stop training your weapon and start training everything else.

I feel the change to the strength formula is positive - strength gives a noticeable boost to melee damage. As CommanderC says, it might be too good in the early game, but high stats are supposed to be good! An ogre berserker with a huge strength score should be powerful in melee, just like a deep elf wizard with a huge intelligence score is powerful in spellcasting.

It does, however, get out of hand for transmuters. Statue Form, Dragon Form, and Blade Hands already give large bonuses to base unarmed combat damage for high strength, and now you can double-dip and get that base damage multiplied by the strength bonus to damage. Dragon Form comes with a built-in +10 strength that now translates to a +37% bonus to melee damage in the strength formula. I would suggest removing the stat-based bonus to base UC damage for these forms and making their flat bonuses larger to compensate if the changes to the strength formula are merged in.

[galehar]

Yes, I agree that the proposed formulae buff heavy weapons too much and reduce weapon differentiation. It would be good to find a function that doesn't have a breakpoint but approximate the current system more closely. So far, my (poor) mathematical skills have failed me.
Or maybe as crate suggested, heavy weapons need to have much bigger base delay. Making them impractical at low skill sounds like a pretty good change to me.

For the stat buff, there is also the option of going with a smaller buff. Maybe tripling it was too much, we can double, I'm not sure.

[CommanderC]

It would be good to find a function that doesn't have a breakpoint but approximate the current system more closely.

A cautious approach would be to keep the current system and randomize the minimum delay so we have a fuzzy breakpoint.

The current model is:

  Code:

min_delay = is_short_blade? 5 : min(base_delay/2, 7);
delay     = max(min_delay, base_delay - skill/2);


For example, in the the second formula min_delay could be changed to min_delay - 1 + 1d3. The breakpoint is still there but hidden. In this example, the allowed fluctuation is only one aut. If we want a greater standard deviation, the mean should be increased.

[some12fat2move]

Maybe delay = baseDelay - (baseDelay - 7)skill/27 would be a good formula. The negative constant in this equation, where I put a 7, is the universal minimum delay that all weapons reach at skill 27. Using this formula a graph of the axes' delays would look like this, with y representing delay and x representing skill. Lighter weapons have a distinct advantage at 0 skill, but are less affected by skill, and eventually at max skill all weapons hit the universal minimum delay. This lets lighter weapons be better at low skill while heavier weapons surpass them at higher skill (the proposed formula does this too but not as much). An overhaul of individual weapon stats would probably be warranted with this formula. Also, 5 or 6 might be better for the minimum delay.

[galehar]

There's also the option of not trying to have a progression of base types at all. It doesn't even work in the current system - I get the impression you're expected to start with a falchion and switch to a great sword when you get enough skill, but in reality the correct skill to switch at is about 4.

I'm aware of that. What I'm saying is that if a side effect of a balanced linear speed system is that heavy weapons are crappy at low skill, then we do have a progression of base types and I see it as a good thing. This seems to be more intuitive than the current system. Let's kill 2 birds with one stone.

Nobody thought of a way to fix this "problem" in linear delay short of giving you negative accuracy, and I can't see it being any easier with the new formula.

Well I just said the opposite (it should be easier with the new formula), and so did crate in a way. I'll try to introduce base delay changes in my proposal and we'll see where it goes.

[galehar]

How about this. The formula is:

  Code:

delay = base_delay / (1 + 2 * skill / 27)

Which means training the weapon skill to 27 divides base delay by 3. The 3 examples shown are quick blade (base delay 7, no change), hand axe (base delay changed from 13 to 15) and exec axe (base delay changed from 20 to 25).

[CommanderC]

It is a good approximation to the current system and it only has one parameter, base_delay. The best proposal so far.

[sanka]

The formula is:

  Code:

delay = base_delay / (1 + 2 * skill / 27)

With this formula the average damage over time of any weapon will be multiplied by the same factor at a given skill, no matter which weapon it is. The ratio of the damage of different weapons won't change with increasing skill based on this formula alone. You simply multiply the outgoing damage with (1+2*skill/27) (regardless of the weapon).

(Of course a more complex damage/accuracy function may change it, but since than that funciont will be the more important, I think it would be better to start the design with that.)

I wrote this on the assumption that the main design goal of these formulas are to differentatie the weapons, since the acutal damage of the weapon can be set arbitrarily by adjusting it's base stats.

[sanka]

Also, I do not really understand what exactly the problem is with the current min delay system. With spells there is usualy a sharp point where you want to stop training the spell school - just it's much more visible on the interface.

In crawl's system it's a fact that some skills are not really worth to train past a certain point. Weapon skills are not the worst in this regard. This seems to me as an integral part of the strategical part of crawl.

I really like that strength has more effect on damage tough. If high strength seems to be too good, then we can simply adjust strength of the races to make reaching high strength harder. I feel that high strength is easier to get than other stats.

[some12fat2move]

It's basically a given that hard breakpoints are a bad thing. That is the problem with the current min delay system. The breakpoints for spells are okay because they're relatively soft and vary according to the character.

[sanka]

Well, "relatively soft" means it's in between 1-2 skill levels. If this is the goal just use the current system and smooth the breakpoints by using splines or whatever.

[njvack]

I wrote this on the assumption that the main design goal of these formulas are to differentatie the weapons

Is this really a design goal? Meaning: should there be a time that a battleaxe is a better weapon than an exec axe?

If so, how will players know when it's time to switch?

[crate]

You switch when you try the weapons and the exec axe is doing better, the same way you choose now. If you can't tell the difference between two weapons when you use them, then there isn't an important difference.

[galehar]

How about this. The formula is:

  Code:

delay = base_delay / (1 + 2 * skill / 27)

Which means training the weapon skill to 27 divides base delay by 3. The 3 examples shown are quick blade (base delay 7, no change), hand axe (base delay changed from 13 to 15) and exec axe (base delay changed from 20 to 25).

I have polished this proposal. The base delay changes go like this:
<= 10 no change
11, 12 and 13: +1
14, 15 and 16: +2
>= 17: +3

The proposal also comes with a handy tool, to easily compare any weapons, with charts of delay and speed. It's also easy to tweak delays and even the formula.
Of course, this will offset the balance, but we can also compensate by also tweaking the damage and accuracy stats of some weapons.
The real question isn't about balance, but about differentiation. I think it succeeds in keeping weapon differentiation but it might be hard to tell without actual playtesting.

[galehar]

So, I ran a few fsim, and it turns out that a very bad consequence of the change is that it makes weapon choice almost completely independent of skill level. But of course, what make the curves cross with the current formula is min delay. So maybe, instead of trying to completely eliminate it, I should try to smooth the breakpoint somehow.

[Leafsnail]

What is the problem with a min delay breakpoint, from a philosophy perspective? I personally think it wouldn't be a huge problem if it were clearly mentioned in a weapon's description ("You need X Maces and Flails skill to use this weapon effectively"). The choice of whether to spend massive of XP to get a great weapon or a modest amount of XP to get a passable weapon is a legitimate one.

I can see that breakpoints are bad if they are opaque and involve multiple variables, but in this case the breakpoint is clear and easy to understand once you are told about it.

e: Actually, since min delay is already listed on a weapon now, it could just outright say what skill you need for it.

[galehar]

What is the problem with a min delay breakpoint, from a philosophy perspective?

They force you to pay more attention to numbers. This one is especially bad, since it's a very important stat and a very sharp breakpoint. Because attack speed increases faster as you get close to the min delay.
I'll try smooth the breakpoint while keeping the overall curve the same. This should be doable by using stepdown, as usual.

[Wahaha]

Looking at the spreadsheets, the effective damage per turn seems to stay approximately the same between old and new values. With the new system, training the weapon skill past the old min delay value keeps increasing the speed and damage, this is good because there's no point where the weapon skill suddenly becomes much less useful. This looks like a potentially very nice weapon reform.

One solution to make skill level affect weapon choice within a weapon type is to increase the effect of accuracy. Specifically to increase the difference in accuracy between small and big weapons, so that higher skill levels are needed to have good accuracy with bigger weapons. This is already the case, but I think that in general accuracy could benefit from being made more significant, because it would actually matter when choosing which weapon to use.

I'm up for play testing this if the branch is hosted on a server.

[BBQsauce]

Attempting to sum up concerns about speed formula

* Weapons should be tiered by effectivity, strength and skill required, allowing some weapons to be midly effective with low skill and some weapons to be very effective, but only after heavy investment
* We would like to get rid of caps and reaching magic numbers, e.g. speed return going to 0 when reaching a certain skill amount

Proposal : switch behavior when you reach 10 delay. Make a linear delay decrease before that, and linear speed increase after it. Reaching that point would take more time for high delay weapons, similarly than it currently is

Relevant graphs
For 14 base delay -> http://imgur.com/gVtUFCa
For 17 base delay -> http://imgur.com/y2PYgeE

This makes the early leveling of high delay weapons return less than early leveling of low delay weapons, as intended, and effectively make rid of magic numbers for the player. Kind of works as a base skill amounts required to get the weapon in an "effective" zone where player gets a base speed return for every skill point. Exact tuning (how much delay decrease per level) being ofc not accurate.

e/
Fooplot, horrible formulas
http://fooplot.com/plot/2wk60n1y11

[BBQsauce]

Also a suggestion about differentiating weapon ability to bypass AC :

If i remember correctly damage formulas, you end up having a huge damage dice that is 1d(weapon damage * skills multipliers). So if you get good rolls you can bypass a lot of armor. If you want to make a weapon less able than others to do this, it should either launch 2 dice, or have a min damage cap, and see it's max damage decreased, but without affecting GDR. Ex : 1d10 to 1d6+2 or 2d5. Vs ac roll of 5, the latter 2 will have way lower average damage. It's a classic trick to avoid introducing a penetration number.

I support what has been said about multipliers multiplying multipliers. With damage and accuracy and speed all increasing of the same skill, you have cubic returns. So it's very rewarding to invest in high weapon skill. If it was not intended, it could use some reworking.

[siprus]

I have to say that I do like how in current system some weapons are better than other with low skill, an other weapons are better with highskill. (generally weapons with high delay are not worth it with low skill in my experience, so people who don't want invest a lot in weapon skill (but want to invest a bit) can take advantage of faster weapons.