Make >mindelay weapon skill more important

[chequers]

For meleedudes, the general 'best practice' recommendation is to train weapon skill to mindelay and then ignore training it further until lategame. This is because weapon skill gives massive dps benefit below mindelay but only minor benefit above. I think there are several problems in this area:

1. Mindelay is difficult for newbies to grasp. It's not really visible ingame (only in '@')
2. It relies on remembering cutoff points for each weapon or looking at '@' every skill up to disable skill training.
3. If you find a bigger weapon it's almost always better to use it regardless of your current skill (similar issue with armour, but not shields)
4. The benefit from weapon skill is not linear. You get a massive benefit below mindelay and minor benefit above. Skills should give constant benefit per level.
5. Fighting skill overlaps with weapon skill in improving acc/dmg.

Here's my initial proposal, I'd like to see what other people think and if anyone can see problems/has better ideas:

My suggestions:
1. Weapons should have a skill requirement (like armour's ER) which suggests a goal for weapon skill. When below this point, acc/dmg/delay are penalised. Above this point, only delay is reduced (train fighting for acc/dmg).
2. Mindelay should never be perfectly reached. Instead, the further above the threshold your skill is, the greater the chance of a mindelay attack (so demon whips attack at mindelay more than GSCs)
3. These changes significantly reduce dps compared to current, so would need some buffing of weapons to compensate.

[duvessa]

3. If you find a bigger weapon it's almost always better to use it regardless of your current skill (similar issue with armour

this isn't true for armour at all once you have something that makes armour penalties actually exist (dodging or spells)

5. Fighting skill overlaps with weapon skill in improving acc/dmg.

...how is this even related? So do weapons themselves, enchantment, slaying, forms, strength, dexterity, haste...

1. Weapons should have a skill requirement (like armour's ER) which suggests a goal for weapon skill. When below this point, acc/dmg/delay are penalised. Above this point, only delay is reduced (train fighting for acc/dmg).
2. Mindelay should never be perfectly reached. Instead, the further above the threshold your skill is, the greater the chance of a mindelay attack (so demon whips attack at mindelay more than GSCs)
3. These changes significantly reduce dps compared to current, so would need some buffing of weapons to compensate.

I really don't understand your motivation here. You want to get rid of current minimum delay because it adds a breakpoint...then you immediately suggest replacing it with another breakpoint that does exactly the same thing?
The usual suggestion is to make all weapons reach minimum delay at skill 27, replacing linear delay with skill with linear speed with skill. This thread provides a good overview of the problems with that, provided you ignore all the posts that aren't by crate (otherwise it will just look like another bad thread which it is). This post is a good summary.

[and into]

For meleedudes, the general 'best practice' recommendation is to train weapon skill to mindelay and then ignore training it further until lategame. This is because weapon skill gives massive dps benefit below mindelay but only minor benefit above. I think there are several problems in this area:

1. Mindelay is difficult for newbies to grasp. It's not really visible ingame (only in '@')
2. It relies on remembering cutoff points for each weapon or looking at '@' every skill up to disable skill training.
3. If you find a bigger weapon it's almost always better to use it regardless of your current skill (similar issue with armour, but not shields)
4. The benefit from weapon skill is not linear. You get a massive benefit below mindelay and minor benefit above. Skills should give constant benefit per level.
5. Fighting skill overlaps with weapon skill in improving acc/dmg.

Whether items 3 and 5 above are very problematic is debatable, I think.

But in general, yes, the existence of all these break-points in weapon skill training is widely felt to be a problem (or at least not ideal), including by at least a few developers. There have been quite a few threads in the Tavern about it (though not very recently, and there would be so many hits in any search, I don't blame you for not sifting through them).

My suggestions:
1. Weapons should have a skill requirement (like armour's ER) which suggests a goal for weapon skill. When below this point, acc/dmg/delay are penalised. Above this point, only delay is reduced (train fighting for acc/dmg).
2. Mindelay should never be perfectly reached. Instead, the further above the threshold your skill is, the greater the chance of a mindelay attack (so demon whips attack at mindelay more than GSCs)
3. These changes significantly reduce dps compared to current, so would need some buffing of weapons to compensate.

Well, wouldn't the first one just introduce *more* break-points and thus retain non-linear progression? I'm also skeptical of whether combat in the game would become more fun if the time cost of an attack were so swing-y, and I don't see how the new role of fighting vs. weapon skill that you describe would be any more intuitive than the current situation.

EDIT: yeah duvessa brought up most of these points, beat me to the punch.

[chequers]

Good points. You're right that my proposal is still non-linear. The beakpoints are less severe than the current design, but it's not much better.

Perhaps the right place to start is by describing what dps curves are desired and reverse engineering a formula. Here's a basic approach, based on a few simple principles: simple weapons do good damage in early game but don't scale well whereas advanced weapons are the opposite, and 2h should scale faster but still be usable in the early game. Here's an illustration of that:


(the numbers are made up, here's the source spreadsheet.)

Is this what weapon scaling should look like in DCSS?

[duvessa]

Good points. You're right that my proposal is still non-linear. The beakpoints are less severe than the current design, but it's not much better.

Perhaps the right place to start is by describing what dps curves are desired and reverse engineering a formula. Here's a basic approach, based on a few simple principles: simple weapons do good damage in early game but don't scale well whereas advanced weapons are the opposite, and 2h should scale faster but still be usable in the early game. Here's an illustration of that:
[big image]
(the numbers are made up, here's the source spreadsheet.)

Is this what weapon scaling should look like in DCSS?

This is literally the exact design proposed in the thread I linked (well, except for totally changing the relationships between dire flail and mace, morningstar and eveningstar, etc).

[mps]

Isn't it generally true that there are uneven marginal benefits to training most skills? Learning the contours of the skill system is a major part of the game strategically. It's also worth noting that among the key differences between different classes of weapons, of which there are already arguably too few, is exactly this issue of mindelay breakpoints. For example, the fact that there's a trade-off in going for mindelay for a double or triple sword vs. a great sword is a good thing and has nontrivial ramifications.

re: Newbie play, if you're talking like MiBe first 3 rune win type games, you can easily max all of your melee and defensive skills anyway and it's probably worthwhile to do so, so it's not a big deal. For building hybrid types, the option to take the pressure off melee skills by using lighter weapons at certain stages of the games adds a lot of flexibility.

[and into]

Variations in the value of training skills at certain levels is definitely good (otherwise there isn't much strategic depth). Really sharp breakpoints are not very good, IMO, because they actually limit this variation. So, for instance, you almost always have very little incentive to train beyond min-delay for your current best weapon, unless you run out of other stuff to train.

It's also worth noting that among the key differences between different classes of weapons, of which there are already arguably too few, is exactly this issue of mindelay breakpoints. For example, the fact that there's a trade-off in going for mindelay for a double or triple sword vs. a great sword is a good thing and has nontrivial ramifications.

This pinpoints main issue, as I see it: Current situation is pretty good for having some degree of differentiation among categories of weapons (something that Crawl struggles with compared to some other games), but at the cost of strategic depth, and also transparency. In the thread that was linked to above, Galehar basically was attempting to enhance strategic depth (in that there would be value to training more than just to min-delay), while keeping weapon differentiation, and also make the numbers work out reasonably well for balance—but it seems that is a pretty tall order.