Proposal: Display relative cost of skill levels

[gerbias]

I propose that a display be added to the skills screen showing the cost of gaining skill levels relative to each other (ie 1.1, 2.0 etc). This would appear in the same location as where the training percentage is displayed, and could be toggled between the two displays. This is intended to allow players to make more informed decisions regarding skill point allocation.

Reasons:
1. The formula for skill level cost is not very intuitive
2. The aptitude factor is not particularly clear
3. Getting a spell castable potentially involves training four skills; this makes the fastest route to do so clear to the player

I have put together a working, but relatively untested version of this functionality and would like to know if it has any chance of being accepted before working on it further.
If anyone has any concerns or suggestions, please feel free to point them out. Thank you for reading!

Summary: Display the cost of gaining skill levels relative to each other on the training screen

[wheals]

How exactly would the interface work? Would some skill randomly be selected as 1.0? Would it use some arbitrary unit system (perhaps based on what the code uses)? Would the player select a skill to compare others to? How many skill levels of gain would it show (important because everything is so nonlinear)?

I think this is a good idea, but ironing out the right interface is the biggest issue.

[jejorda2]

If you pick the most expensive skill as the basis for comparison, then every other skill is less expensive than that, and could be expressed as a percentage between 1 and 99%. We already know those numbers fit on the screen where the allocation numbers are now.

Picking the highest leveled skill as the baseline makes some sense, too, and depending on aptitudes, the relative cost of other skills could go a little over 100%. But never over 999%, so it wouldn't take more room on the display.

The relative cost of the next level is what makes sense to display, to me.

[byrel]

Personally, I'd prefer to see an absolute number based on raising a +0 skill from 9->10 being 1. Then a Troll raising Fighting from 26->27 will show 7.07, and a Merfolk raising Polearms from 0->1 will show 0.13.

[archaeo]

Maybe one of those ideas is workable, but where do you actually explain the ratio in question to the player? It looks like you have two lines of text to work with (where "The percentage of the experience used..." bit lives); otherwise, you have to put it in the help menu, where, if you wanted to, you could just provide the XP table instead of explaining this "relative cost" number.

Of course, if you can a) find a good numerical way to explain the relative cost and b) explain those numbers in 200 characters or so, then this would be a cool and good thing to have.

[jejorda2]

SH is displayed on the main screen without explaining that bolts can't be blocked or that the second attack in the same turn isn't blocked as well as the first. It isn't unprecedented to show a useful number without completely explaining it.

[archaeo]

I mean, the game doesn't give you the formulas for how SH works, but it gets covered in the manual (which is available in-game) twice; once in the Attributes and Stats section to explain what that SH number is, and once in the Items section, where it tells you that it only works on "some ranged attacks."

In comparison, if you just hit a toggle and the "Train" column switched to the "Ratio" column, it would be hard to figure it out. The game could explain that number below the skills or in the skill help menu, but the former is a pretty small space, while the latter may as well just tell you the XP table if we're going to be giving the ratios anyway.

[jejorda2]

The ratio is more valuable to me than the table. I usually want to know: Of these skills that can benefit my character, which is the cheapest?

I don't want to try to remember the levels of 6 skills and my aptitudes while I cross reference a table.

I had a spreadsheet that did this calculation for me at one time, but that was when aptitudes were expressed as percentages instead of exponents. I haven't updated it in a while.

If the column is labeled "Skill Cost," then it is clear that lower numbers are better, because lower costs are better. And that's really all that's important.

[duvessa]

I mean, if you don't do something goofy like aptitudes, I think most players will assume a skill cost of 2 is twice as expensive as a skill cost of 1.

[archaeo]

I don't want to try to remember the levels of 6 skills and my aptitudes while I cross reference a table.

I guess. I would personally be just as happy with a nice line graph showing the XP curve, as knowing exactly which skill will be cheapest to train seems like it would just drive me to do a bunch of unpleasant optimal stuff.

That said, if we're going to include these ratios, "lower numbers are better" seems like it's setting the bar too low. If we want the player to have this information, we should probably make sure it's a usable amount of information. Otherwise, if "lower numbers are better" is enough for the ratios, why is "higher numbers are worse" not already doing the trick?

(I will note that I can't find where in the manual it tells you that each level of a skill/XL requires an increasingly higher amount of XP; if it's not in the manual, it probably should be.)

[jejorda2]

Without pulling up the XP table and looking up what the definition of an aptitude is, I can't tell you whether a Hill Orc finds it cheaper to train conjurations from 12 to 13 or fire magic from 14 to 16. Even looking at the skill menu and seeing the aptitudes, I really don't know which is cheaper.

Right now, I turn on both when I need more spellpower or spell success. If I knew that one was 80% of the other, or that one was 320% of a human raising a skill from 9 to 10 and the other was 400%, I would only train one until the costs were equal. Then I'd train both.

On a mummy or a troll, or something else with bad aptitudes, I've actually added up how much XP it would take to get a spell online, and looked at how much HP I could get for the same XP in fighting, and decided whether to go for it. On most chars, I just kind of think about what I have available that doesn't work well yet and train towards it the cheapest thing that can help me survive something I can't currently survive.

[byrel]

That said, if we're going to include these ratios, "lower numbers are better" seems like it's setting the bar too low. If we want the player to have this information, we should probably make sure it's a usable amount of information. Otherwise, if "lower numbers are better" is enough for the ratios, why is "higher numbers are worse" not already doing the trick?

They key feature of both suggested schemes is not that 'lower is better' but that 'the derivative of skill levels with respect to investment' is actually comparable across a given character, or across all characters depending on whether you go relative or absolute.

Right now, I have no way of knowing (without a decent amount of calculation) the relative point at which I should train spellcasting instead of Cj+PM+AM to get Mephitic cloud online. Or spellcasting vs Cj+AM+IM for freezing cloud. I know how the calculations work, but I don't want to cross-reference a graph when the program can just hand me numbers and I can say 'Oh, spc's number is less than 3/4s the others. I should train it.'

[gerbias]

Okay, some very good questions!

Here's an example screenshot of the interface:
https://github.com/smitnich/crawl/blob/ ... xample.png

Which skill is 1.0: Currently I'm using the median of all skills trained >= 1. I feel like this provides a representative value even in extreme cases such as a player having level 26 in one skill and 2 in the rest. However, once you account for aptitudes, there might not necessarily be a skill with exactly 1.0.

Amount of level gain: Currently I'm using 1/10 of a level as the amount to show; that is 12.0 and 12.9 will display the same cost. This has the advantage of always being accurate (ie if I used a full level when the player had a manual, it might run out partway through). I had considered using one full level as the baseline, ie (12.5 would be half the cost of 12 and half the cost of 13), and am still undecided about the best method.

Choosing the baseline skill: Honestly this is probably a good idea to include somehow, but I have not thought about it.

Explaining the cost: "A skill with 2.0 here costs twice as much to gain as one with 1.0. This takes aptitudes as well as manuals into account."

Thank you all for the feedback!

[duvessa]

Honestly the display in that screenshot is already intuitive to me.

Amount of level gain: Currently I'm using 1/10 of a level as the amount to show; that is 12.0 and 12.9 will display the same cost. This has the advantage of always being accurate (ie if I used a full level when the player had a manual, it might run out partway through). I had considered using one full level as the baseline, ie (12.5 would be half the cost of 12 and half the cost of 13), and am still undecided about the best method.

This approach has the advantage that it would help dispel that dumb "whole numbers aren't breakpoints for skills" myth that keeps getting paraded around.

[archaeo]

Honestly the display in that screenshot is already intuitive to me.

It's intuitive insofar as "smaller numbers are cheaper," sure. I still like gerbias' little blurb about what the numbers mean in context.

I have reservations as to how much this would encourage a lot of extremely tedious skill micromanagement, something that's already super annoying even without this, but lord knows I beat the anti-numbers drum enough that people probably expect I want it to say "cost: choko."

[jejorda2]

The advantage of byrel's suggestion for a human 9 to 10 as the baseline cost is that the same number means the same thing (presumably about the same number of enemies to defeat) in different games across different character builds. It makes both the XP aptitude and the skill aptitudes clearer. If there is room, this could replace aptitude displays on the skill screen by default.

[johlstei]

I would like another sig fig, but that looks really good to me, gerbias.

[reaver]

I think this is a very good idea.

Why not just make raising a +0 skill from level 0 -> level 1 defined as 1.0? It makes way more sense then having "1.0" change over time or based on what skill the character has trained.

[tasonir]

I assume the reason they didn't suggest 0->1 is because then getting high level skills will display something like 304, which is a "large" number. I'm not sure if it would go over 999, but there's already "18.3" which is 4 characters, so I don't know if that's a problem. You'd just have to drop the fractional cost, which wouldn't matter much when the cost is >999.

But choosing 9->10 gives it a more "friendly" scale. Either way it's probably a minor issue; overall I love the idea and I agree that a simple Cost column is intuitive and useful.

[gerbias]

Alright, so it seems like the biggest issue is what to use as the baseline value.
I took both byrels and reaver's suggestions and provided an example picture for 0-1 as base and 9-10 as base.

0
https://github.com/smitnich/crawl/blob/ ... 1-base.png

9
https://github.com/smitnich/crawl/blob/ ... 9-base.png

The issue I see with 1 as base is that the number can get very large (55 for +0 aptitude, 130.8 for -5) with 26 skill.

One possible issue I am worried about with 9 as base is that if I see two skills compared, with a cost of 0.13 and 0.26, my first instinct is not that the second is twice as expensive, but that they're both really cheap and which I pick doesn't matter that much. I'm not sure how much of an issue this would be for other players though.

Anyways, I would appreciate it if some more people could provide feedback/suggestions on which base they prefer. Thank you!

[byrel]

I think the lesson that training fighting from 0-1 is a trivial investment is actually needed for a lot of blaster players.

[duvessa]

I like the first one better.

[WalrusKing]

I concur with Duvessa here, the first one is better in my opinion. Its more fine grained and I like the feeling of seeing big numbers, even if big numbers are worse in this case. Perhaps if we could constrain the numbers to four digits it might be even better though, as the last digit probably doesn't matter in that case and it could serve to reduce the amount of visual clutter.

[Siegurt]

I actually like the original that uses the median better, personally, but all of them seem decent enough.

[chequers]

0
https://github.com/smitnich/crawl/blob/ ... 1-base.png

This is my favourite.

For precision: show 1.0 for <10, and then whole numbers above.

[jejorda2]

I like for all numbers to be right-aligned with the same number of decimal places. It makes it easier to scan down a list and find bigger or smaller numbers, because they are wider or narrower.

Which column is quicker for you to scan?

  Code:

50.7     50.7
 1.2     1.24
12.3     12.3
 0.3     0.29


[neil]

I like for all numbers to be right-aligned with the same number of decimal places. It makes it easier to scan down a list and find bigger or smaller numbers, because they are wider or narrower.

Which column is quicker for you to scan?

  Code:

50.7     50.7
 1.2     1.24
12.3     12.3
 0.3     0.29


It takes one more column, but you can also show a fixed number of significant figures while keeping the alignment:

  Code:

51
 1.2
12
 0.29


[chequers]

I find that really hard to read though. Maybe harder than with unaligned dots

[Sprucery]

I am really looking forward for this feature to be implemented. I micromanage skills all the time* by using a skill-aptitude-skill points table and having this information on screen would save a lot of time for me.

I'd go for 0-1 as base, and either "aligned dots" or "fixed amoung of significant figures" display.


*That doesn't mean that I would train optimally, though...

[archaeo]

Am I crazy for not exactly getting what those numbers represent? It's fine to say "relative cost," but relative to what? All I get is that the bigger number is more expensive, but it isn't clear to me how much more expensive, especially if the bar just says "costs."

The "am I crazy" question isn't rhetorical, it's possible I'm just woefully math illiterate and should let the grown ups talk, but if the goal is increasing the skill menu's clarity, I guess I'd like it if it were clear for morons like myself too. At the very least, the text at the bottom there needs to include something like, "Costs reflect the amount of XP necessary to gain the next level, where 1.0 represents taking a +0 skill from x to y," or however that's best written.

[jejorda2]

Yeah, I guess you don't know that it's the cost of the next level, as opposed to the cost already paid to get where you are now.

But I don't think defining the baseline on that screen is important. That's what the knowledge bots and the manual are for.

Knowing the cost of one skill relative to the cost of another skill right now is useful.

[bel]

If the number of the digits is an issue for display, why not just display at most 3 significant digits? so 1.2 or 20 are displayed as it is (right aligned or whatever), but 310.9 is displayed as 311 or just 310 if you don't want to bother with rounding.

(Edit: oh this was already mentioned above by neil, nvm)

[archaeo]

But I don't think defining the baseline on that screen is important. That's what the knowledge bots and the manual are for.

Isn't that exactly the same defense one would use for not showing the cost of skill levels at all?

I'm not asking for a full math lecture or anything (if I want that, I can log onto tileschat), just some context for the numbers I'm seeing.

[jejorda2]

Looking at the current skill display, you know the level of the skills and your aptitudes. To find out which of two skills costs more, and how much, you also have to have a table of how much xp (or skill points- I use the terms interchangeably) each level costs. Then you have to use the aptitudes, levels, and the table to calculate the cost of the next level of each skill. Then you can compare the results of the calculation to find out which skll is cheaper.

With the proposed display, the skill with the lower cost number is cheaper. If it's half as much as the higher skill, it's half as expensive.

From the proposed numbers, I don't know how many ancient liches or bats I need to defeat to get that skill, but I know that I need more armour and dodging, and right now dodging is cheaper. Or I know I can get bolt of fire's accuracy improved more by training conjurations instead of fire. Because the cost is a lower number.

I'm not trying to use these numbers to figure out if I can get some skill from 6 to 18 before I get to Depths, or whether it costs more to use a large shield with no penalty and a one-hander at minimum delay vs. a two-hander at minimum delay. I'm trying to figure out which of two skills is cheaper right now to train on the character I'm playing right now.

And the lower number is the cheaper cost. Much lower is much cheaper.

[gerbias]

Alrighty!

I've uploaded a patch here:
https://github.com/smitnich/crawl/blob/ ... play.patch

I have to concur with those who said that 0->1 should be the baseline: I like this a lot after actually playing a decent length game with it. So thanks for the suggestion!

For formatting, I'm just following the standard set by the skill level display (ie 1.0, 26.1) aligned to the right with the exception of values >= 100 losing the decimal place.

Regarding archaeo's concern about knowing the baseline, I think players will quickly realize what the value is relative to the first time they raise a skill from 0->1.

[archaeo]

Regarding archaeo's concern about knowing the baseline, I think players will quickly realize what the value is relative to the first time they raise a skill from 0->1.

I think you vastly overestimate the cognitive abilities of people as dim as myself, but I can see I'm the only one waving this flag, so no harm done. Thanks for the patch!

[johlstei]

But I don't think defining the baseline on that screen is important. That's what the knowledge bots and the manual are for.

Isn't that exactly the same defense one would use for not showing the cost of skill levels at all?

I'm not asking for a full math lecture or anything (if I want that, I can log onto tileschat), just some context for the numbers I'm seeing.

The absolute value is irrelevant, it's the ratio that matters. So if conjurations says 6 and charms says 2, you know it would take 3 times as much XP to raise conjurations by one level than charms. One particular case where this is obviously valuable is when you want to cast a dual-school spell. Assuming you don't care about other spells in each school (work with me), you should train whichever school has the lower XP cost if you want to cast the spell ASAP. This isn't useful if you've trained the skills the same amount and you're a human so the aptitudes are the same, but that's usually not the case.

There are other situations when you want to do things like weigh the cost of a level of dodging vs a level of fighting where it's more complicated, but this at least provides some baseline for comparison as opposed to nothing. It also more easily communicates in-game that skill costs are non-linear.

[archaeo]

I get it now, johlstei, I just think that the game should explain things as clearly as you guys have here. I doubt I'll be the only person confused by these numbers at first.

[wheals]

Thanks for the patch gerbias, this is now in trunk!

[byrel]

Awesome sauce! I look forward to using this.

[le_nerd]

Quick feedback: it is indeed as awesome as it seemed it would be. Thanks a bunch to everyone!

[gerbias]

Thank you to everyone who showed their support for this idea!
(Also thanks to wheals for cleaning it up; I had left the code necessary to support changing the baseline in in case it was needed later, but that was probably pointless.)

[emikaela]

nice work! sadly, my game crashed after i trained a single skill to 27 and got prompted to choose a new one (crashed upon closing the prompt), so that probably needs looking at.

since the game state apparently reverts to my last staircase, i can replicate it if needed. or i could just get around it by training another skill at the same time and finish the game. i'll hold on for now.

[gerbias]

Thank you for the bug report! As a quick update, I have narrowed down the cause quite a bit and am testing a fix right now. I'll get back to you as soon as possible.

Edit: Also you should be able to continue playing just fine if you switch off of the cost menu and back to training percentage display.

Edit2: Alright, I've uploaded a patch to Mantis; if any devs are reading this, could you either apply my patch (or just remove the call to fixup_skills in set_cost and commit it yourself). Thanks!

[WalkerBoh]

This is the best GDD thread I have ever seen. Great idea, glad to see it being pushed.

[nago]

I've just noticed the introduction of skill cost in the game - great thing btw - but I can't find a commit which describes it and the git's patch page is gone: shall I assume that the cost displayed is normalized on the value which cost for going from 0 to 1 skill? And in that case, the apt considered is 0 or the best\worst of the race?

[vible]

I've just noticed the introduction of skill cost in the game - great thing btw - but I can't find a commit which describes it and the git's patch page is gone: shall I assume that the cost displayed is normalized on the value which cost for going from 0 to 1 skill? And in that case, the apt considered is 0 or the best\worst of the race?

its from 0 to 1 with apt 0

[Pollen_Golem]

The issue I see with 1 as base is that the number can get very large (55 for +0 aptitude, 130.8 for -5) with 26 skill.

According to
http://crawl.chaosforge.org/Skill_point
the ratio in cost between 0->1 and 26->27 is 1:265, rather than 1:55. Is the wiki wrong/outdated?

[neil]

According to
http://crawl.chaosforge.org/Skill_point
the ratio in cost between 0->1 and 26->27 is 1:265, rather than 1:55. Is the wiki wrong/outdated?

Skill point cost is something else entirely: it affects all skills across the board, regardless of aptitude or the skill's level, and is based solely on the total amount of XP the player has gained. So it does not affect the relative cost, and is not shown on the m screen. It exists to prevent high-level characters from taking a skill from level 0 to level 9 by killing a single titan (or from 0 to 1.8 by killing a single orc warrior), while still allowing low-level characters to get a skill up to level 1 by killing D:1 monsters.

This is the table you are looking for.