attack delay

[galehar]

Yet another attempt at a melee attack delay formula. Details on the wiki.

[Lasty]

The chart looks like a pretty decent set of outcomes.

With the old formula, when people would ask what weapon to choose, the heuristic was generally "wield the biggest thing you can, and train it to minimum delay". That heuristic appears to no longer be true under the new metric; what's the new heuristic?

[and into]

The chart looks like a pretty decent set of outcomes.

With the old formula, when people would ask what weapon to choose, the heuristic was generally "wield the biggest thing you can, and train it to minimum delay". That heuristic appears to no longer be true under the new metric; what's the new heuristic?

Apparently, something like "all else being equal, use best one-handed weapon until skill level 10 or so, unless that one-handed weapon is a demon whip, in which case great mace does not outstrip it until skill level 14."

[Lasty]

Edit: Now that I've looked at the SB data, here's some analysis and feedback:

Dagger: a little better at skill 0, but takes 16 SB skill to get to the point where it matches old daggers at skill 10.
Quickblade: also start better, but doesn't beat old quickblade at min delay (skill 8) until it hits skill 17.
Short sword: same pattern, but doesn't beat old short swords at min delay (skill 12) until skill 19.
Cutlass: same pattern, but doesn't beat old cutlass at min delay (skill 14) until skill 23!

Even a mace, a generally pretty bad weapon, only takes 20 skill levels to beat cutlass in the new system, and a hammer, which is basically a joke weapon, is approximately on par. Flails beat it at 13 skill.

This seems like a huge nerf to short blades, which were already terrible apart from stabbing and quick blades. I think short blades either need to use a modified version of the general formula, or just need a damage boost -- unless making short blades worse than they already were is one of the goals of this change, and I don't think that it is.

[evilmike]

Looks to me like the numbers are fine for 2 handers, but 1 handed weapons take a fairly big hit. The best 1handers (demon/holy) seem to need much more skill under the new formula to reach the same damage numbers. For me the main reason to use a demon whip/blade is the low skill investment needed, a great mace/great sword will hit harder eventually but it takes much more xp to outclass the demon weapons. The charts you posted make it look like there's only a small difference now, so I think the balance would be pushed even more in favour of 2 handed weapons. I don't know what can be changed to fix that though.

[nicolae]

Looks to me like the numbers are fine for 2 handers, but 1 handed weapons take a fairly big hit. The best 1handers (demon/holy) seem to need much more skill under the new formula to reach the same damage numbers. For me the main reason to use a demon whip/blade is the low skill investment needed, a great mace/great sword will hit harder eventually but it takes much more xp to outclass the demon weapons. The charts you posted make it look like there's only a small difference now, so I think the balance would be pushed even more in favour of 2 handed weapons. I don't know what can be changed to fix that though.

What about different parameters for one and two handed weapons?

[galehar]

Dagger: a little better at skill 0, but takes 16 SB skill to get to the point where it matches old daggers at skill 10.
Quickblade: also start better, but doesn't beat old quickblade at min delay (skill 8) until it hits skill 17.
Short sword: same pattern, but doesn't beat old short swords at min delay (skill 12) until skill 19.
Cutlass: same pattern, but doesn't beat old cutlass at min delay (skill 14) until skill 23!

At this point, I was more looking for feedback on the formula rather than the balance of individual weapons. The new base delays are just rough estimates that don't completely break the balance, there is still a lot tuning to do. I haven't looked closely at short blades, so maybe they'll need to be buffed with a smaller base delay or a higher base damage. What is certain is that under the new formula, the fast weapons benefit less from training skill. Not sure if it's much worse than the old formula where they benefited a lot until min delay (which was reached rather quickly) and then almost not at all.

If set the following goals:
* some weapons need more training than others to be efficient
* weapon choice depends on skill level (corollary to the previous)
* no hard breakpoints

Then the consequence is that some weapons benefit much less from skill training.

Looks to me like the numbers are fine for 2 handers, but 1 handed weapons take a fairly big hit. The best 1handers (demon/holy) seem to need much more skill under the new formula to reach the same damage numbers.

I only really tried to balance demon whip and demon blade (the ones on the comparison charts on the wiki). All the others need more fine tuning.
If you compare demon whip to the old formula, you get:
skill < 10 : new > old
10 < skill < 22 : new < old
skill > 22 : new > old

And it's better than a great mace until skill level 13 (instead of 16). I think it's still a good weapon with low skill investment. Of course, it cannot be as good as it was at min delay unless you buff it a lot, because the old curve has the spike there, and the whole point is to remove it. I can try to increase the base damage by a point or two to make it more competitive with 2 handers.

[Lasty]

At this point, I was more looking for feedback on the formula rather than the balance of individual weapons.

I get that, but I guess what I'm saying is that it looks like entire classes of weapons (short blades) work very badly under this formula. As evilmike said, the same issue may apply to most one-handed weapons, but short blades in particular get beat up by this.

Also, and I think this may be a necessary corollary of the goals, low-delay weapons with additive brands end up being significantly weaker. A whip|spear|SB of electrocution|venom|distortion|draining, while still good, will be significantly worse than it is now. In general, all the additive brands will be significantly worse, since all weapons will be swinging slower through the majority of each game, though that may well be a good thing.

The new base delays are just rough estimates that don't completely break the balance, there is still a lot tuning to do. I haven't looked closely at short blades, so maybe they'll need to be buffed with a smaller base delay or a higher base damage. What is certain is that under the new formula, the fast weapons benefit less from training skill. Not sure if it's much worse than the old formula where they benefited a lot until min delay (which was reached rather quickly) and then almost not at all.

If set the following goals:
* some weapons need more training than others to be efficient
* weapon choice depends on skill level (corollary to the previous)
* no hard breakpoints

Then the consequence is that some weapons benefit much less from skill training.

Unfortunately, that implies that there will be weapons that effectively you should not train for, currently including everything that formerly had base delay 11 or lower. I think it might make sense to add a design goal "all weapons should be worth training", in which case it might make sense to have a different formula for short blades that makes them worth using.

As a suggestion, the second formula might leave damage relatively unchanged while increasing speed more dramatically, possibly past current breakpoints. Weapons could be described as "improving both damage and speed with training", or "improving speed with training" in the description menu to help distinguish which formula they use.

[DracheReborn]

I think the point is that with old demon whip you can get fairly decent damage at 12 M&F skill, while with the new formula it would take something like 19-20. And if you're getting 20 M&F anyway, you might as well be using a great mace.

To put it in another way, under the old formula, great mace doesn't become better than demon whip until M&F reaches something like 17 skill, while under the new formula it looks like it becomes better at only 13 skill. So those characters who used to train only exactly 14 weapon skill for their demon weapon, or lajatang, or dire flail, would now be better off using a great mace at that point.

The heuristic under the new formula seems to be, bigger is always better, as long as you have moderate amount of skill. This is even more true for LB, where it looks like claymore outclasses every other LB starting at skill 10.

I do get that numbers for specific weapons can be tweaked, but it seems to me that some sort of breakpoint seems unavoidable if we really want to preserve the current system where some weapons are *much* better than others at moderate skill (and not just low skill).

[ackack]

Other than an imo irrational fear of breakpoints, I don't really get the need for change here. Having a formula like this seems more rather than less likely to encourage "spoilery" knowledge to plot out your weapon choices. We've already moved to putting mindelay in the weapon descriptions --- why not just mention the skill needed to achieve mindelay?

I agree with the critiques thus far that this seems really bad for one handers.

[Leafsnail]

Personally I'd prefer the breakpoints were kept too (although I agree that the weapon should get better at a roughly linear rate before the breakpoint is reached, rather than suddenly getting way better in the last few levels).

The really meaningful tradeoff when it comes to weapons is XP invested vs effectiveness, and breakpoints mean that choice is easy to understand providing they are simple and transparent ("I need to invest this much more XP to use that better weapon"). Under this system as far as I can tell you aren't choosing a weapon - you're choosing a skill level you want to end up at, and then picking whichever weapon happens to be best at that skill level. That seems like a more confusing decision to me - instead of comparing a few different weapon options I'm now trying to work out where I want to be on a continuous curve of XP investment and effectiveness.

[Siegurt]

To make the formulas work the way you want them to, without hard break points, but where smaller faster weapons still get good skill bonus at low training levels, the delay formula needs to be logarithmic (see the new armour EVP formula for an example of one that works)

The base needs to have a higher return for low-investment weapons, with a longer flatter curve of delay past a certain point, where the bigger heavier weapons need to have a smaller:

Something vaguely like this:
Chart

(Blue being a two-hander, red being a one-hander, and black being a short blade)

[TheDefiniteArticle]

With the old formula, when people would ask what weapon to choose, the heuristic was generally "wield the biggest thing you can, and train it to minimum delay". That heuristic appears to no longer be true under the new metric; what's the new heuristic?

This would make it "wield the biggest thing you can".

[galehar]

ok, I think introducing beta in the formula might have been a bad idea. I'll try setting it to 1 and see how it turns out. This should lower the difference between fast and slow weapons. Thanks all for your feedback.

[Wahaha]

As I said in the other thread, the very easy way to fix one handers vs two handers is to give big weapons a bigger accuracy penalty, so that higher skill levels are needed to use them effectively. Suddenly a claymore isn't the best at 10 skill anymore!

instead of comparing a few different weapon options I'm now trying to work out where I want to be on a continuous curve of XP investment and effectiveness.

That's the point of removing break points. You have to decide how effective you want to be with a weapon in the same way that you decide how much armor or dodging to train. Players don't have too much trouble deciding how much armor and dodging they need, weapon skill should be the same.

However it's not as simple as "I have 15 skill, let me look at the charts and get the bestest weapon for 15 skill". In a real game weapons aren't just base types, they have enchantments and brands. Not to mention demon weapons and lajatangs are rare. Often there will be an obvious choice from the weapons that you actually have.

Another thing to note is that with the new system, higher skill levels continue to significantly improve all weapons, even those that would've hit their breakpoint in the old system. So unlike the old system, with the new system even if your weapon isn't the best type for the skill level you have, it still benefits from the extra skill levels and in the end the difference between similar base types is small. The fact that your weapon is +5 of freezing instead of +0 unbranded is what matters.

[duvessa]

As I said in the other thread, the very easy way to fix one handers vs two handers is to give big weapons a bigger accuracy penalty, so that higher skill levels are needed to use them effectively. Suddenly a claymore isn't the best at 10 skill anymore!

Unless you completely change the EV formula, wouldn't this pretty much break the instant a weapon gets a nonzero to-hit enchantment or a player gets nonzero to-hit slaying?

[Wahaha]

Unless you completely change the EV formula, wouldn't this pretty much break the instant a weapon gets a nonzero to-hit enchantment or a player gets nonzero to-hit slaying?

I couldn't answer your question so I did some crappy tests with fsim. The results are below. Graphs would be much much more useful but I don't know how to do those.

great mace vs yak
3 fighting 4 maces: 78%.
10 fighting 14 maces: 82%.
15 fighting 24 maces: 84%.

demon whip vs yak
3 fighting 4 maces: 79%.
10 fighting 14 maces: 82%.
15 fighting 24 maces: 84%.

So there's no difference between +1 acc on the demon whip and -4 on the great mace.

Now let's change the great mace base accuracy to -20 (wearing -16 acc ring) and the demon whip base accuracy to +10 (wearing +9 acc ring). (Is base acc the same as other acc?)

great mace vs yak
3 fighting 4 maces: 63%.
10 fighting 14 maces: 70%.
15 fighting 24 maces: 74%.

demon whip vs yak
3 fighting 4 maces: 83%. (20% more accurate than great mace)
10 fighting 14 maces: 85%. (15% more accurate) (this is the point where the demon whip is supposed to be better. This change in accuracy gives it an extra ~15% advantage compared to the default numbers.)
15 fighting 24 maces: 85%. (11% more accurate) (this is where the great mace should be better. The demon whip is still 11% more accurate which isn't the ideal result but the great mace does more damage at this point)

What if the -20 great mace and +10 demon whip are both enchanted to +6?

great mace vs yak
3 fighting 4 maces: 69%.
10 fighting 14 maces: 73%.
15 fighting 24 maces: 77%.

demon whip vs yak
3 fighting 4 maces: 84% (15% more accurate).
10 fighting 14 maces: 85% (12% more accurate).
15 fighting 24 maces: 86%. (9% more accurate).

Conclusion: This only kind of works because the demon whip quickly hits the point where more accuracy does almost nothing. So if small weapons all had >+5 accuracy, accuracy would basically not matter for them. This might be bad because it makes accuracy a useless stat for fast weapons(even more useless than it is now) but at the same time they're supposed to be accurate. On the other hand it makes accuracy matter more for big weapons.

Adding +acc from enchantment/slaying reduces the difference between small and big weapons, but not by a lot.

[Leafsnail]

That's the point of removing break points. You have to decide how effective you want to be with a weapon in the same way that you decide how much armor or dodging to train. Players don't have too much trouble deciding how much armor and dodging they need, weapon skill should be the same.

I have two answers to this. The first is that the main effects of the dodging and armour (EV and AC) skills are much easier to see than the effects of weapon skill. The second is that I don't regard skill investment in dodging and armour to be that interesting when compared to investment in weapons and spell skills (which also effectively have breakpoints in the form of spell slots and spell success), where you have to make tactical decisions based on what items are available rather than dumping some amount of XP into it at some point.

However it's not as simple as "I have 15 skill, let me look at the charts and get the bestest weapon for 15 skill". In a real game weapons aren't just base types, they have enchantments and brands. Not to mention demon weapons and lajatangs are rare. Often there will be an obvious choice from the weapons that you actually have.

I don't see how this addresses anything I said. The fact that finding which weapon is best for you right now might require you to run a few fight sims of your own doesn't make it any more interesting.

Another thing to note is that with the new system, higher skill levels continue to significantly improve all weapons, even those that would've hit their breakpoint in the old system. So unlike the old system, with the new system even if your weapon isn't the best type for the skill level you have, it still benefits from the extra skill levels and in the end the difference between similar base types is small. The fact that your weapon is +5 of freezing instead of +0 unbranded is what matters.

Is this a good thing? To me it seems to just mean that how much you want to train your weapons skill won't be affected at all by what you find - you'll just use whatever weapon happens to be best for you right now and keep training to whatever number/ in whatever ratio you've decided to go with.

e: and also that you'll have to run more fight sims to decide which weapon to use I guess.

[Wahaha]

I don't see how this addresses anything I said. The fact that finding which weapon is best for you right now might require you to run a few fight sims of your own doesn't make it any more interesting.

You're misunderstanding something or perhaps everything because this was an argument against needing charts or fsims.

You're still assuming that with the new system it's better to pick a skill level and then pick a weapon that suits that skill level. That doesn't make any sense. A good weapon suitable for your skill level won't suddenly appear in front of you. First you find A good weapon. Then you train enough skill to use it. Same as with the current system. The only difference is that instead of an exact spoilery skill level at which to stop, the amount of skill you can train is variable, but you're sure to be fairly rewarded with extra speed and damage for every level.

"But it's impossible to tell which of these two similar weapons is better!"
If you can't tell, it doesn't matter. Coincidentally this is exactly how it works in the current system, but you don't do fsims do you? That's why I find that you saying this

e: and also that you'll have to run more fight sims to decide which weapon to use I guess.

is silly. Do you fsim the exact point when a +3 long sword becomes worse than a +2 scimitar? No, and there's no reason to do it with the new system.

To me it seems to just mean that how much you want to train your weapons skill won't be affected at all by what you find - you'll just use whatever weapon happens to be best for you right now and keep training to whatever number/ in whatever ratio you've decided to go with.

And when the skill gets a lot higher you'll grab a bigger weapon because it's better than the smaller weapon, just like people do with the current system. Wow!

The current system's breakpoints don't make weapon selection easier. They make deciding at which skill level to stop training easier. These are very different things. The new system doesn't make weapon selection harder.

[XuaXua]

Looks to me like the numbers are fine for 2 handers, but 1 handed weapons take a fairly big hit. ... I don't know what can be changed to fix that though.

Improve shields?

[Leafsnail]

I think you misunderstand what I'm saying with my fightsim comments. I'm not saying "This change is bad because it would make you require more fight sims", I'm just saying that "the fact that it's not immediately obvious which of the current weapons available to you is the best isn't a good thing and doesn't address the wider point I'm making". I was just rebutting your points, it's not meant to be a point in my favour.

Ultimately I just prefer having big, easily understood decisions when it comes to my character's offensive capabilities (like with spell success/slots) but if that's not a design goal then I guess it's moot.

[TheDefiniteArticle]

Do you fsim the exact point when a +3 long sword becomes worse than a +2 scimitar? No, and there's no reason to do it with the new system.

Well that's an amusing example since the correct choice is literally always the scimitar in the current system. But yes you're right, doing math for every two vaguely similar weapons is a waste of time, which is part of the problem with this proposal.

You're still assuming that with the new system it's better to pick a skill level and then pick a weapon that suits that skill level. That doesn't make any sense. A good weapon suitable for your skill level won't suddenly appear in front of you. First you find A good weapon. Then you train enough skill to use it. Same as with the current system. The only difference is that instead of an exact spoilery skill level at which to stop, the amount of skill you can train is variable, but you're sure to be fairly rewarded with extra speed and damage for every level.

No. In the current system, an exceptional weapon find can change your plans. "Oh, I had 16 polearm skill for this glaive, but I just found Wyrmbane. Oh well, Wyrmbane will be even better than usual." "Wow, vampiric great mace, maybe I'll get more than 12 maces after all". The difference in the new system is that your plans NEVER change because even at 12 skill, the great mace is better than a demon whip. On the off-chance it's not better, the margin is small enough that no one will know/care because, as you adequately explained, they aren't gonna bother doing the math on every pair of weapons, they'll just assume any weapon without ridiculous enchantments is worse than a big heavy one. This is already partly true in the current system; the only thing preventing it is the fact that there are significant periods where lighter weapons are dramatically better than heavy ones, and the proposal here is to remove those periods.

These "magic numbers" are quite literally the defining characteristic of weapon base types, and the idea of "tweaking numbers to balance weapons in the new system" has the end goal of recreating the magic numbers in a different way (i.e., get 12 skill for demon whip, get 18 skill for great sword, get 20 skill for great mace). If you just want to make the magic numbers fuzzier, that is A- easily done without completely up-ending weapon numbers, and B- kind of unnecessary given that they are already somewhat fuzzy due to random drops etc.

I think [demon whip's] still a good weapon with low skill investment.

But not by enough. Since I am not using fsim every two seconds, why would I bother with it when even at 9 skill I would be hard-pressed to tell the difference, and I know for a fact that the comparison will only change in the great mace's favour?

tl;dr "Stop at 14" isn't a hard rule, it's just a convenient number in a wider range of good numbers, the real reason people don't get 20 skill with a dwhip is that "best weapon at 12 skill" is the motivation for choosing it in the first place

[galehar]

Ultimately I just prefer having big, easily understood decisions when it comes to my character's offensive capabilities (like with spell success/slots) but if that's not a design goal then I guess it's moot.

The weapon choice is not always obvious in the current system, and the new one doesn't really change in that regard. Ultimately, it would be good to have a way to compare weapon effective damage in-game, to help with those decisions, but that's a different issue. It's true that with the current system, there are sometimes very obvious choices because of the min delay spike. If you have access to a demon weapon, then training it to min delay has a great damage / skill training ratio which dominate other potential choices (some might call it no-brainer).
With this proposal, there's the potential for a smoother progression. Train your skill, and upgrade your weapon as better ones become available. The progression isn't balanced well for now, I'm working on it, but that's the idea.

[Lasty]

Galehar, thanks for bringing the formulas you're testing to the community for feedback. I'm very invested in seeing the best outcome from this significant change, and it's clear that others are too. I appreciate that you're letting us contribute to this project through feedback.

[Leafsnail]

It isn't really a no brainer, though. Demon weapons are never going to do anything like as much damage as a good two-hander. So a clear and meaningful choice is presented: should I stick with this good, inexpensive weapon, or should I invest lots of XP to get a great weapon? I'd say this is a good example of items found changing the strategic decisions a player makes, incidentally.

I mean I guess they're a no brainer before you get good enough to use a two hander but I don't see this system changing that.

[galehar]

After some more experimentation, I am not satisfied with the stepdown. Starting from the linear speed formula (speed = 10/delay + k * skill), I replaced skill by stepdown(skill, base_delay * alpha) to give a steeper slope to slow weapons speed (so that they can catch up). It also had the side effect of inverting the direction of the curvature which I think is counter productive. Maybe by going back to a linear speed formula, it will be easier to approximate the old system.
So, let's go back to the linear speed formula (speed = 10/delay + k * skill) and simply make the slope depends on base delay:

speed = 10/base_delay + (alpha + base_delay) * skill / beta

I'll try that and let you know

[sanka]

I'm a little bit confused about the exact design goals (yes I know they were stated in this thread, just I do not realy understand them), but I'd like add some personal opinion:

1. It does not matter what formula you choose: the more you divert from "spikes" the more better a simple linear approximation will be and the more you will loose the current startegic aspect of weapon choice. That is because these spikes are the exact reason that the questions like "quick blade with 8 skill investment, demon whip with 12 skill investment, lajatang with 14 skill investment or great mace with 20 skill investment" can be an interesting one, so that differnt weapons can provide radically different opportunities for skill investment/damage, affecting strategy. A spike in the damage over time per skill level function means an association with a weapon and a skill level - and this is why weapons provide strategical opportunities.

2. One thing I would like to propose: do not try to choose a formula for speed first. Instead first try to find furmula for the planned damage over time against a tipical opponent at a given skill level. This of course will depend on the speed/damage/accuracy formulas, but determine those later. That is because for balance the most important thing is damage over time, so you want to differentiate weapons by that function, and it will be easier to understand which formula will give good results.

3. The facts that some poster has pointed out will hold for any formula where damage over time is roughly linear: the startegy will be more like EV instead of spells. Just remember that there are only one type of EV so there's no differentiation, and armour choice is determined by the fact that armour not only gives AC (all simple give it, just some more) but also gives penalties (unlike current weapons). So for weapons I do not think that linear damage over time function will work very well.
One more thing against linear damage over time increases per skill: at first it seems that for a linear function you have two parameters, and that seems enough. However I think that neither very steep nor very flat lines will work very well.

[Siegurt]

A suggestion:

min(20,<delay>/(<skill>+4)*(14-(<delay>*,<skill>/27/5)))

See here:
Chart for a chart of delays.

See here:
Chart for a chart of attack speeds.

See here:
Chart for a sample of damage times attack speeds for various weapons (quickblade/short sword/longsword/great sword/claymore)

[Leafsnail]

That formula is very much "biggest weapon is literally always better". Having it so that weapons don't benefit at all for low levels of weapon skill also seems really wonky and is basically just an unintuitive reversal of the current system.

[Siegurt]

Well, if we took out the max delay of 2.0 AUT it'd fit the bill, but I felt like having a weapon that took 6 turns to swing would be... awful.

[siprus]

After some more experimentation, I am not satisfied with the stepdown. Starting from the linear speed formula (speed = 10/delay + k * skill), I replaced skill by stepdown(skill, base_delay * alpha) to give a steeper slope to slow weapons speed (so that they can catch up). It also had the side effect of inverting the direction of the curvature which I think is counter productive. Maybe by going back to a linear speed formula, it will be easier to approximate the old system.
So, let's go back to the linear speed formula (speed = 10/delay + k * skill) and simply make the slope depends on base delay:

speed = 10/base_delay + (alpha + base_delay) * skill / beta

I'll try that and let you know

So basically everything will be as it used to be expect that the weapon speed rises steadily with skill? (instead of delay) until it hit's the cap. Sounds reasonable to me.


I don't think having "magic number" with weapon speeds is quite as bad as with, let's say armour and strength, because higher weapon skill will always be better than lower one, because of the damage and accuracy increase. If different weapons are going to be better at different skill levels there will always be some sweet point for each weapon. More we smooth out those sweet points, less relevant those sweet points become (and more and more certain type of weapon is just better than other type).

[galehar]

I've updated the spreadsheet and the wiki with a new formula. Speed is linear, and there is a (delay+2)/delay introduced to make slower weapon gain more speed from skill. Seems like it's getting closer. Demon blade is more competitive, claymore only catches up at level 15. However, damage difference between the two at skill 27 is smaller compared to the min delay system (3.7 instead of 7). It's probably the weak point of the proposal, still not sure how to address it.

[TheDefiniteArticle]

The weak point of the proposal is that every long blade character will always use the claymore and never use anything else because outside of pain/elec brands there is no compelling reason to ever use any other long blade.

The damage difference at 27 skill is totally irrelevant because good players will never get that much skill. Requiring 26 skill instead of 24 was a strict nerf to triple sword/claymore even with the base damage buff it received.

[galehar]

The weak point of the proposal is that every long blade character will always use the claymore and never use anything else because outside of pain/elec brands there is no compelling reason to ever use any other long blade.

Assuming for the sake of simplicity availability of both weapons (which are both rare), up to skill level 15 there is a very good reason not to use a claymore: it deals less damage than a demon blade. You might even want to keep the demon blade for a few more levels if you're using a shield. So "always use the claymore" is non-sense unless you're playing a version of crawl in which you start the game with 15 levels in long blades.

(yeah, I can read through the annoying arrogant snarkiness, I guess your point is that claymore is too good at mid-level).

[DracheReborn]

galehar: On the new formula you posted, it looks like claymore eventually gets a lower delay than a dagger, which seems counterintuitive. You can see that by inspecting by the term (skill * base_delay^2) in the denominator of the delay formula - for large values of skill, big weapons with large base_delay will eventually become faster than little weapons like daggers.

Anyway, I would like to propose yet another formula. I'm not sure if it has been brought up before or not, since it's very simple it may have been. It is very similar to the existing system, but with the piecewise linear delay formula being replaced by an exponential. Old delay formula vs new looks like this:

http://fooplot.com/plot/50r7dnff16

Red is the current delay for mace plotted vs skill, blue is the proposed new formula. The delay formula takes 2 parameters, base_delay and min_delay, which are identical with the current parameters. While min_delay is also something of a magic number, it's at least somewhat derived from base_delay.

new_delay = (base_delay - min_delay)*exp(-skill/(base_delay - min_delay)) + min_delay

I plotted a simplified damage vs skill using this new formula, where damage = base_damage/new_delay. It looks like this:

http://fooplot.com/plot/dq5vxgrv8i

(Blue = mace, red = demon whip, orange = great mace)

As in the current system, dwhip is better than great mace in the mid skill range, but past that great mace is better. Damage increases with skill smoothly, so there are no breakpoints. The greatest returns for training weapon skill actually occurs at the low levels, but there will always be benefit for training more skill. Weapons won't actually reach min_delay now, instead they will approach that point asymptotically. This affects the slower weapons especially (in my example, great mace only reaches a delay of 7.7 aut at 27 skill, compared to 7 aut at 20 skill in the old system), so further compensation like increased base damage is probably needed for the slow weapons.

Thoughts on this idea?

[galehar]

galehar: On the new formula you posted, it looks like claymore eventually gets a lower delay than a dagger, which seems counterintuitive.

That might happen with some weird values of alpha and beta. Or maybe at skill level 892. With the values I'm using, it doesn't happen.
I'll look into your proposal later, thanks.

[Siegurt]

I am heavily in favor of DracheReborn's formula. It mimics the current balance, eliminating it's drawbacks without causing any other problems. The problem with linear speed is that it heavily de-emphasis skill training on lighter, faster weapons in the examples for the current working set, quickblades take 27 levels of skill to improve delay by 1, and there's almost no incentive at all to train those 27 levels, the problem is less dramatic when comparing say demon blades to claymores, but it's still present, heavier weapons always get more benefit from skill, there's no incentive to use less skill and use a lighter weapon, more training will always make the lighter weapon obsolete.

It also avoids having to re-code and rebalance the numbers for every weapon, and avoids things like using a claymore unskilled and it taking four rounds to swing it (Yeilding a lot of uncomprehending newbie deaths) Also it avoids having to redo all the acquirement weightings and fiddle with everything while preserving the players ability to take a lighter, quicker, less damaging weapon and have it be a smaller skill investment to be effective with it

About the only change I might make to it is to change exp() (Which is merely e^) to 4^ or 5^, this slightly sharpens the logarithmic curve and gets weapons with a delay of 20 slightly closer to their current max speed.

[acvar]

At this point, I was more looking for feedback on the formula rather than the balance of individual weapons. The new base delays are just rough estimates that don't completely break the balance, there is still a lot tuning to do. I haven't looked closely at short blades, so maybe they'll need to be buffed with a smaller base delay or a higher base damage. What is certain is that under the new formula, the fast weapons benefit less from training skill. Not sure if it's much worse than the old formula where they benefited a lot until min delay (which was reached rather quickly) and then almost not at all.

If set the following goals:
* some weapons need more training than others to be efficient
* weapon choice depends on skill level (corollary to the previous)
* no hard breakpoints

Then the consequence is that some weapons benefit much less from skill training.

Here is a suggestion. Just make damage more dependent on weapon skill and str/dex, and less dependant on the actual weapon. For example:

1. All weapons except the quick blade have the same min delay but different starting delays.
2. Weapons have a damage multiplier that ranges about 90% to 110% for one handers, and 125% to 130% for two handers.
3. Base damage is something like rand(skill) + rand(stat). This is overly simplified for discussion purpouses.
4. Weapon enchantments add directly to the percentage for the weapon damage modifier.

So how does this effect things. Well lets assume that the min delay for weapons is 7. So a falchion that starts with a delay of 13 hits it min delay at 12 weapon skill, but do you stop training there? Of course not. There is a very big bonus to continue traning long blades since it has such a large effect on damage. The hard break point is gone. It is still a good idea to keep track of your weapon delay to maximize your damage, but now even more so since there is not much difference in damage per hit between a falchion and a tripple sword. This fits all your criteria. It also has the added benefit of making the physical stats more relevent, and making it less relevent if you find the "right" upgrade to your current weapon, all good things.

[SublimeSnake]

At this point, I was more looking for feedback on the formula rather than the balance of individual weapons. The new base delays are just rough estimates that don't completely break the balance, there is still a lot tuning to do. I haven't looked closely at short blades, so maybe they'll need to be buffed with a smaller base delay or a higher base damage. What is certain is that under the new formula, the fast weapons benefit less from training skill. Not sure if it's much worse than the old formula where they benefited a lot until min delay (which was reached rather quickly) and then almost not at all.

If set the following goals:
* some weapons need more training than others to be efficient
* weapon choice depends on skill level (corollary to the previous)
* no hard breakpoints

Then the consequence is that some weapons benefit much less from skill training.

Here is a suggestion. Just make damage more dependent on weapon skill and str/dex, and less dependant on the actual weapon. For example:

1. All weapons except the quick blade have the same min delay but different starting delays.
2. Weapons have a damage multiplier that ranges about 90% to 110% for one handers, and 125% to 130% for two handers.
3. Base damage is something like rand(skill) + rand(stat). This is overly simplified for discussion purpouses.
4. Weapon enchantments add directly to the percentage for the weapon damage modifier.

So how does this effect things. Well lets assume that the min delay for weapons is 7. So a falchion that starts with a delay of 13 hits it min delay at 12 weapon skill, but do you stop training there? Of course not. There is a very big bonus to continue traning long blades since it has such a large effect on damage. The hard break point is gone. It is still a good idea to keep track of your weapon delay to maximize your damage, but now even more so since there is not much difference in damage per hit between a falchion and a tripple sword. This fits all your criteria. It also has the added benefit of making the physical stats more relevent, and making it less relevent if you find the "right" upgrade to your current weapon, all good things.

My feeling is that a change like this would change the meta and balancing so hugely that it's not going to happen regardless of how reasonable it is.

[Siegurt]

At this point, I was more looking for feedback on the formula rather than the balance of individual weapons. The new base delays are just rough estimates that don't completely break the balance, there is still a lot tuning to do. I haven't looked closely at short blades, so maybe they'll need to be buffed with a smaller base delay or a higher base damage. What is certain is that under the new formula, the fast weapons benefit less from training skill. Not sure if it's much worse than the old formula where they benefited a lot until min delay (which was reached rather quickly) and then almost not at all.

If set the following goals:
* some weapons need more training than others to be efficient
* weapon choice depends on skill level (corollary to the previous)
* no hard breakpoints

Then the consequence is that some weapons benefit much less from skill training.

Here is a suggestion. Just make damage more dependent on weapon skill and str/dex, and less dependant on the actual weapon. For example:

1. All weapons except the quick blade have the same min delay but different starting delays.
2. Weapons have a damage multiplier that ranges about 90% to 110% for one handers, and 125% to 130% for two handers.
3. Base damage is something like rand(skill) + rand(stat). This is overly simplified for discussion purpouses.
4. Weapon enchantments add directly to the percentage for the weapon damage modifier.

So how does this effect things. Well lets assume that the min delay for weapons is 7. So a falchion that starts with a delay of 13 hits it min delay at 12 weapon skill, but do you stop training there? Of course not. There is a very big bonus to continue traning long blades since it has such a large effect on damage. The hard break point is gone. It is still a good idea to keep track of your weapon delay to maximize your damage, but now even more so since there is not much difference in damage per hit between a falchion and a tripple sword. This fits all your criteria. It also has the added benefit of making the physical stats more relevent, and making it less relevent if you find the "right" upgrade to your current weapon, all good things.

~ Meant to quote ~

One thing with a system like this is that it *does* de-emphasize the current focus on finding weapon upgrades. Currently when you *don't* find a weapon upgrade in the category you're training now, you have to choose: do you continue to hold out, do you train for a different weapon, do you go for spells instead of trying to become a melee god. All those choices become moot if you can simply pick your starting weapon and use it through the whole game, it makes it a simpler, and non-adaptive sort of game, and IMHO less interesting. One of the things that makes crawl a good game, is you *can't* count on the RNG to give you just the right upgrades, and you have to adapt. To give it a counterpoint, we wouldn't want to make all starting spellbooks with attack spells that scaled up in skill until they were as powerful as level 9 spells. Part of the point of the game is to take the good that you find and make the best you can of it.

[Leafsnail]

I've had an idea for how to differentiate weapons more: make it so that weapons don't mostly end up at the same minimum delay.

Why? Because DPS (that is, base damage per AUT) vs attack speed is actually a sortof interesting tradeoff. With high DPS but low attack speed you can kill things faster, but you can't respond to the developing situation as quickly. The issue with the current system is that you can make this distinction disappear completely by training your weapon skill, simply adjusting the formula and damage/delay values could solve this.

To me this is a lot more appealing than trying to make Accuracy more relevant. The problem with accuracy is that it doesn't actually provide any differentiation other than making different weapons better against different enemy EVs, and enemy EV is a completely hidden and opaque mechanic (I guess it could also provide differentiation in how much training you need to make each weapon viable, but min delay already does that). I'm not sure if it would even be interesting if enemy EV were shown, that would just mean you'd carry around a high accuracy weapon for dodgy enemies.

One thing with a system like this is that it *does* de-emphasize the current focus on finding weapon upgrades. Currently when you *don't* find a weapon upgrade in the category you're training now, you have to choose: do you continue to hold out, do you train for a different weapon, do you go for spells instead of trying to become a melee god. All those choices become moot if you can simply pick your starting weapon and use it through the whole game, it makes it a simpler, and non-adaptive sort of game, and IMHO less interesting. One of the things that makes crawl a good game, is you *can't* count on the RNG to give you just the right upgrades, and you have to adapt. To give it a counterpoint, we wouldn't want to make all starting spellbooks with attack spells that scaled up in skill until they were as powerful as level 9 spells. Part of the point of the game is to take the good that you find and make the best you can of it.

I really agree with this. There are breakpoints everywhere when training magic (getting enough spell slots to learn a spell, getting enough spell success to cast a spell reliably, hitting the spell power ceiling) and I think that makes it a lot more interesting, especially since the breakpoints are different every game.

[acvar]

Siegurt: Note that the numbers I gave were just an example. They could easily be played with to make weapon upgrades more important. Having 1 handers range from 50% to 100% and two handers from 150% to 200% would cetainly retain the need for weapon upgrades while still doing all the things we want. What is really important is to make damage more tied to weapon skill. The reason I gave the small variation in weapon damage is in part a philosphical difference in what the game should be from you, but lets face facts the starting spellbooks for pure casters scale far better then starting weapons for fighters. To balance them out a fighter would have to start with a mace, flail, and morning star, and then hope only to find that great mace he needs. In addition it is stupid that a pure spellcaster can just happen upon a random artifact after totally neglecting his martial skills and suddenly becom a competent fighter with zero training. That would be like a fighter finding a spellbook with firestorm in it and immediatly torching everything in sight with firestorm without any training. I think everybody would agree that the latter is stupid, but for some reason most accept the former. Why?

The long and the short of it is that martial characters should be rewarded more for their skill (and stat) investment. Doing so will incentivise them to actaully invest in those skills which is what we all want isn't it?

[duvessa]

In addition it is stupid that a pure spellcaster can just happen upon a random artifact after totally neglecting his martial skills and suddenly becom a competent fighter with zero training.

Against low AC monsters, with reasonable str/dex values, a +0 falchion with 14 fighting/long blades does more damage per aut than a +5 demon blade of flaming with 0 fighting/long blades. (Also, good artefact weapons are vanishingly rare if you don't worship Trog or use every acquirement scroll on them.)

[Siegurt]

Siegurt: Note that the numbers I gave were just an example. They could easily be played with to make weapon upgrades more important. Having 1 handers range from 50% to 100% and two handers from 150% to 200% would cetainly retain the need for weapon upgrades while still doing all the things we want. What is really important is to make damage more tied to weapon skill. The reason I gave the small variation in weapon damage is in part a philosphical difference in what the game should be from you, but lets face facts the starting spellbooks for pure casters scale far better then starting weapons for fighters.

Well, this is already the current situation, weapon skill already influences damage, you're just inverting the formula, so that rather than x times y, you're doing y times x. Increasing the damage that skill gives could be just as easily accomplished in the current paradigm by increasing the percentage bonus that weapon skills have, and dropping the base damage by the same margin.

One of the problems I have with this is under your proposition, having 0 skill means you do no damage with all weapons, and having 1 skill means all weapons are the same (Yes it's easy enough to add some constants to this, but one of the reasons for the speed proposal in the first place is to remove spoilery constants, besides, we'd start to want big weapons to have larger constants than small ones, and it starts to look like a complicated version of what we've already got)

To balance them out a fighter would have to start with a mace, flail, and morning star, and then hope only to find that great mace he needs. In addition it is stupid that a pure spellcaster can just happen upon a random artifact after totally neglecting his martial skills and suddenly becom a competent fighter with zero training. That would be like a fighter finding a spellbook with firestorm in it and immediatly torching everything in sight with firestorm without any training. I think everybody would agree that the latter is stupid, but for some reason most accept the former. Why?

The long and the short of it is that martial characters should be rewarded more for their skill (and stat) investment. Doing so will incentivise them to actaully invest in those skills which is what we all want isn't it?

I would argue the following:
1. Spells have additional constraints and additional rewards aside from doing damage, comparing one to one isn't logical as they don't follow the same rules (for example, if there was a melee artifact which let you do 50-100 points of damage to everything in an 8 square circle at range, it wouldn't fly)
2. Spells aren't drop-in replacements or direct upgrades for earlier versions, many are better versions of attack spells, but they come with additional costs and deficits. A fighter that started with a mace, flail, and morning star can simply drop the mace and flail and use the morningstar at level 1, You can't memorize a 3rd or 4th level attack spell at level 1 and attack with it.
3. Someone who has trained 0 fighting and 0 weapon skills in exchange for large proficiencies in spellcasting, can pick up a weapon and use it, but can't be nearly as effective with it as someone who has trained for it (And certianly not as effective as they are with spells). Someone who hasn't trained any spellcasting skills can pick up a book, memorize a first level conjuration and cast it, and will not be nearly as effective as they are with melee (or as someone who has trained for spellcasting skills)
4. Spells and spellcasting have more functionality than simply being damaging. In many ways it can act as a compliment or augmentation *to* melee, and as such they should be harder to obtain. Being able to whack things with a weapon doesn't make you any better at blasting things with spells, being good at blasting things with spells can make you better at whacking things with a weapon.

[Psiweapon]

You're such a bunch of bioroids in conference <3 <3 <3

[Leafsnail]

In addition it is stupid that a pure spellcaster can just happen upon a random artifact after totally neglecting his martial skills and suddenly becom a competent fighter with zero training. That would be like a fighter finding a spellbook with firestorm in it and immediatly torching everything in sight with firestorm without any training. I think everybody would agree that the latter is stupid, but for some reason most accept the former. Why?

This is a seriously invalid comparison. If you want to find a melee analogue for Fire Storm then you should surely look to high-end weapons such as Triple Swords, and it's pretty obvious that a primarily spellcasting character can't just pick up one of those and use it well for zero investment. They could certainly pick up something like a Demon Blade and invest a bunch of XP into it to become a fairly good melee fighter, but non-spellcasting characters can equally pick up some mid-level conjurations.

[acvar]

Well, this is already the current situation, weapon skill already influences damage, you're just inverting the formula, so that rather than x times y, you're doing y times x. Increasing the damage that skill gives could be just as easily accomplished in the current paradigm by increasing the percentage bonus that weapon skills have, and dropping the base damage by the same margin.

Obviously. All I ever said was to make skills/stats more important and weapons less important. I also think that the current equations are much too complex and could be simplifed at the same time to make things less spoilery.

One of the problems I have with this is under your proposition, having 0 skill means you do no damage with all weapons, and having 1 skill means all weapons are the same

Not when you consider a characters stats also. If we use something similar to (rand(skill) + rand(stat)) X (weapon modifier) then you still do damage with a 0 skill, and the weapon modifer still diferentiats damage. The main difference at such a low skill will be attack speed however which means that a dagger is far better at low level then a great sword which is one of the goals isn't it?

I would argue the following:
1. Spells have additional constraints and additional rewards aside from doing damage, comparing one to one isn't logical as they don't follow the same rules (for example, if there was a melee artifact which let you do 50-100 points of damage to everything in an 8 square circle at range, it wouldn't fly)
2. Spells aren't drop-in replacements or direct upgrades for earlier versions, many are better versions of attack spells, but they come with additional costs and deficits. A fighter that started with a mace, flail, and morning star can simply drop the mace and flail and use the morningstar at level 1, You can't memorize a 3rd or 4th level attack spell at level 1 and attack with it.
3. Someone who has trained 0 fighting and 0 weapon skills in exchange for large proficiencies in spellcasting, can pick up a weapon and use it, but can't be nearly as effective with it as someone who has trained for it (And certianly not as effective as they are with spells). Someone who hasn't trained any spellcasting skills can pick up a book, memorize a first level conjuration and cast it, and will not be nearly as effective as they are with melee (or as someone who has trained for spellcasting skills)
4. Spells and spellcasting have more functionality than simply being damaging. In many ways it can act as a compliment or augmentation *to* melee, and as such they should be harder to obtain. Being able to whack things with a weapon doesn't make you any better at blasting things with spells, being good at blasting things with spells can make you better at whacking things with a weapon.

All true but totally irelevent to the point being made. I was not comparing the balance of melee with spells. I was pointing out the difference between the two in how far their starting packages take them. In other words how long they can go before not finding the right upgrades kills them. The longer you can hang with your starting package the more things you find and the less likely it is to be a problem. So based on what you said spellcasters gain a great deal of versatility, a bigger safety net, and easier diversification. What do melee based characters get? A more powerful start that is more likely to lead to a dead end.

This is a seriously invalid comparison. If you want to find a melee analogue for Fire Storm then you should surely look to high-end weapons such as Triple Swords, and it's pretty obvious that a primarily spellcasting character can't just pick up one of those and use it well for zero investment. They could certainly pick up something like a Demon Blade and invest a bunch of XP into it to become a fairly good melee fighter, but non-spellcasting characters can equally pick up some mid-level conjurations.

My experiences directly contradict you. I just recently played a HEAE that found a freezing triple sword in the middle of lair. He was able to pick it up and use it to kill most of the opponents he faced in the middle of the lair quite effectively without any skill invested (obviously I started investing in long swords once I found it). The effectivness of a weapon without skill is > 0. The effecitveness of a spell without training < 0. What justifies the difference?

[Leafsnail]

Well yeah, I guess you can use a Triple Sword at 0 skill. It's just an incredibly dumb idea because you could swing faster and do more damage with almost any other long blade (not to mention the potential for using a shield) So the opportunity cost of using a Triple Sword at zero skill makes it effectively just as worthless as Fire Storm at zero skill.

[duvessa]

Well yeah, I guess you can use a Triple Sword at 0 skill. It's just an incredibly dumb idea because you could swing faster and do more damage with almost any other long blade

this isn't actually true

[Leafsnail]

I guess maybe it could do with a minor nerf at low levels then, I dunno it doesn't really seem like much of a problem to me if some rare items change how you play when you find them.