XP and learning - opinions?

Game Talk The Bunker
1

After realizing that the way skill comprehension works means that values above 100% don’t matter much, I decided to play around with the system and see if I could improve it. I think I may be on to something, although I’d be interested to hear how other people play, as well.

To that end, I figure I’ll share my experience with the current system, and with some variants that I’ve tried out. These were all with 14 int, Fast Learner, and no skill rust.

b Current playstyle[/b]:
When I have 0 XP in my pool, I am annoyed. So, I find a book, any skill book, then read it until I’ve got a nice amount of XP in my pool (say, 200+). I then engage in a mixture of training activities (welding, butchering, etc.) and normal combat/exploration until my XP pool is empty. Once the pool is empty, I often fall back to a safe area, and continue reading. If I can’t get new books, I end up being rather annoyed. Not enough to stop playing, obviously, but enough that I seek out libraries more than I possibly should.

b Skill comprehension boosts learning rate[/b]:
The first experiment I tried was making skill comprehension boost your learning rate (per point of XP) when above 100%. So, 150% skill comprehension would result in learning skills 1.5x as fast, for the same XP cost. My observation here was that certain slow-to-raise skills ended up being trained up fairly fast - construction ended up being higher than I’m used to, for instance. Overall, I think that’s probably bad for game balance, particularly since I’m already thinking of applying a learning rate bonus in another context - stacking could get a little silly. Plus, there are some things that seem to be intended as hard to raise, and I imagine that having a huge between-character disparity in potential learning rate will make that hard to adjust properly.

b Skill comprehension affects XP gain[/b]:
The next experiment I tried was having skill comprehension act as a multiplier to the amount of XP you gain. This had the effect of filling up my XP pool a LOT faster on my first character, which actually got me out into the streets. I think I’d still follow roughly the same pattern as before, reading until I’ve got XP in the pool then heading out, but the reading would proportionately be much less of the day for me. Plus, I can see myself having enough to do that I don’t always bother with reading - cookbooks, for instance, may actually get skipped over now (I don’t usually have a need to train cooking from books, I just want the XP). In terms of balance, I think this will work out a lot better. Yes, there will be huge differences in the rate at which characters accrue XP, but they’ll still have to spend the same amount of time practicing - it’s just that for some characters, practicing can happen a lot more often.

I’m going to play a bit more with what I’ve outlined in (2), especially because the first character didn’t last long. However, I can only really get a feeling for how it impacts my playstyle, and I’d be curious to know what others think.

2

Further testing of mechanic (2)! Having both int 14 and Fast Learner is a little ridiculous (+92% XP gain). However, it’s also self-limiting in a way: if you’re not actively training your skills, you’ll fill up your XP pool and end up wasting potential points. That in turn causes me to head out into the streets, as I said in the original post, and apparently I play too aggressively. Basically, I’m finding that XP expenditure ends up being limited by exploration and resource acquisition, and that in turn ends up being limited by rate of healing. Although the XP-optimized build has potential to improve very quickly, I’m finding that practically, it has issues being able to spend the XP.

Dropping Fast Learner (+28% XP gain), it ends up being a nice buff, but not game-changing.

Lowering to int 11 and taking Fast Learner again (+56% XP gain) seems to be a good balance, for my personal taste. I don’t end up doing stupid things in a vain attempt to empty my XP pool, but that +56% boost is also noticeable.

I’m happy with this change, so I’ll submit it via Github.

3

Righty-ho.

I’m glad for the parameter-exploration in the OP; option 1 was what I’d hoped for as higher-level skills seem to require more XP, and I’d had times (especially before disabling skill rust) where leveling up required more than 800 XP. Getting Cooking 8/9 was a major project.

I recall GlyphGryph doing something with the XP capacity, but wasn’t sure if xe actually pushed it or not.

Any possibility of, say, 24 hours for notice & comment next time? I don’t have any particular objections to your submission as I understand it, but it would have been interested in how you proposed to boost the XP-efficiency, and how that would have interacted.

4

The concern I had with option 1 is that it seemed to make the early levels a bit too easy to get. I agree that later levels are hard to get, but I personally think that should be handled with a separate balance adjustment - maybe make the scaling a bit less extreme, or something.

Another concern I had is that if I add other things which increase learning rate (e.g., what I’m considering with regards to professions), then the stacking could get out of hand.

Good point, should probably keep that in mind next time. I don’t object to further discussion, of course. Part of why I submitted the pull request was to make it easier for people to check it out, although you’re right, more time to discuss could be good.

5

OK, this isn’t final yet, but is a solid draft. Good to have that insight; I’m still at the point where finding the time to sit down with a learn-to-C++ book and a laptop is an issue. Some of the code I can read and understand, but reworking the Mutations is a long way off.

Thanks much.

As regards the OP: I can accept that #2 may be the better fit in terms of balancing/code economy/etc. No problem there. Calling a modifier to the XP-pool’s fill rate “skill comprehension” seems confusing to me, as “skill comprehension” sounds like it should affect skill-gain and not XP-gain. I don’t have a better term right now, though–“Learning rate” hits the same issue. Will think about it some more.

Your point about no longer “needing” skill books is interesting. I’ve played from the perspective of maximizing book use, because early game XP-pool is a finite resource: in the beginning of the game, XP spent leveling a book-eligible skill is XP that could have gone to Dodge or a combat skill. Reducing the time value of XP would help reduce the problem of people just hunkering in with books.

6

Yeah, I fully welcome continued debate, but seeing as I’ve got something which is fun, seems reasonably balanced, and is better than the current system, I figured I’d stick it up on Github anyway. It’s also a system that feels right to me (see my comment about reduced book usage), so I figure there’s little point in just sitting on the code at this point.

I agree that “skill comprehension” is a poor name for my proposal in option #2. I think now I’m going to call it “XP gain percentage”, since that’s fairly unambiguous.

And that is exactly why option #2 felt right to me. To some extent, I like the concept of a study/training cycle, but the way it worked out in practice, it seemed a bit too heavy on the study side. Personally, I like the idea that a build optimized for skill learning would have reason to regularly head out and get some practical experience, rather than just hunkering in a house reading.

7

With Method 2, if I’m reading it right, is that the more intelligent your character is the less they’ll depend on books. So a total nerd can go outside to gain hands-on experience, because they’ve got the big XP pool; but a stupid person will run out of XP by noon and pretty much need to read books to keep up. (Basically, you’re sending a dumb character back to your Method 0.) That’s sort of the opposite of how I would expect things to work. I would think that a high INT character would profit the most from a library, while a low INT character would more or less ignore all that fancy book-learnin’.

8

Yeah, a quirk about method 2 is that it ends up mixes street smarts and book smarts in the same stat. Well, all the methods, except that the current system doesn’t have much variation in street smarts, so it’s less noticeable. If we went with an Angband/D&D-style 6 stat system, with Int and Wis separate, then I’d probably put the street smarts/XP-affecting aspect under Wis, and the book smarts/reading speed aspect under Int.

9

Some alternative ideas:
A)
Rather than changing the speed at which your xp pool fills up, why not increase the size of the xp pool?
8 intelligence means an 800 points xp pool. 14 means an 1,100 point exp pool. (100 points per level for the first 8 levels, 50 per level afterwards) But in both cases, they take the same amount of time to fill, at a given morale level. This means unlike in your current system where morale is less important, it’s still vital, but having a higher intelligence means you can learn more per session.
Technically a bit difficult to achieve, but there are a few ways to do it (and all of them are a serious improvement to the back end that would probably be appreciated in their own right). Probably end up providing a percentage modifier of some sort, here being a boost of about 40%.

B)
Have skill books pull from the experience pool, but at a reduced rate. Especially when combined with the above system, this would mean intelligent characters would have more reason to read (they can do it more often) while stupider characters should take up other hobbies that are less of a resource drain.

C)
Just have the bonus apply to skill points as they are spent like people already think it does. If it’s giving too big of an advantage, simply make the bonus smaller. How does it work now? I was figuring that when a point of xp was spent with 142% skill comprehension, it was 100% chance of the first point going through and a 42% chance of a free point. I think this mechanic is pretty solid and intuitive, and if it’s too powerful, it’s easy enough to reduce the bonus, right?

D)
Rather than outright give you skill, have books give you a temporary skill boost. As long as your current skill is below that level, it will learn much faster, but that level rusts much faster than ‘real’ levels. This could be displayed in Green, for an elevated (as opposed to rusted) skill level. So reviewing a book could give you a quick boost to the given skill, but unless you apply it while the knowledge is still fresh in your mind, you’ll lose it. Obviously, again, this could be combined with any of the other ideas.

10

Are you looking at just how Int (and thereby “Skill Comprehension”) affects skill gain, or the entire skill system in general?

I’ve been in a love-hate relationship with the skill system since I started playing a few weeks ago. I’m a big fan of skill-based systems as opposed to just standard “I got XP, now I level up” setups, but Cata’ frustrates me some.

The current mechanism of racking up XP while “safe” and/or inactive, and burning through it quickly took some time to get used to. I think, naturally, “I do things to earn XP”, not sit around reading a book.

I had to view my XP total more as “Concentration” - my character’s calmed state of mind that enables him/her to focus enough to actually learn something as opposed to just survive. A well rested, happy self might notice that spear s/he threw arched better this throw, while an “OMG Z’s!” screaming, panicked self may not. It somewhat explains about the “burst then empty” aspect of XP too … you concentrate well at first, but can only soak up so much before you get fatigued.

This is how I play currently and it feels too “gamey” to me. After day 1, I’m not surviving and fighting for my life. Read - train - repeat, becomes the mantra.

[b](1) Skill comprehension boosts learning rate[/b]:

b Skill comprehension affects XP gain[/b]:

I like your option 2 better so far, simply because it gets the character out in the world more instead of goofing around with books and throwing cans around. Does #2 make skills ramp up that much faster though? Are you powering some of them so fast it’ll imbalance things? With the current system I focused on throwing for a couple days in my last game, and went from “which end do I aim at them?” to an Olympic javelin thrower in basically a weekend. At that point everything felt trivial.

I'm going to play a bit more with what I've outlined in (2), especially because the first character didn't last long. However, I can only really get a feeling for how it impacts [i]my[/i] playstyle, and I'd be curious to know what others think.

Balancing it all is tough I’m sure. You want character progress at a good rate. Too fast and I’m too powerful too early (or so powerful the game becomes trivial - see Olympian above) … too slow and I’m sick of “grinding”, especially in a rogue-like, where death happens often and that marks the end of the character.

In Cata’, I feel like I can quickly (day 2 even if I properly balance risk-v-reward, plan escape routes while scavenging, and get lucky with drops, especially a working vehicle) establish a bunker and basically be set up forever. I go from “omg I could die” to “grind grind grind” safely. There currently isn’t something that keeps me pushing or reaching or risking my life.

Dwarf Fortress was good at finding a middle ground. I always wanted to refine things … push a little deeper … look around that next corner. If I did “turtle” and get totally defensive, it’d throw some invasions at me to keep me struggling … I could never get too comfortable.

On that note - when NPCs get back in, I wouldn’t mind if you had to specialize your character in a craft/trade … and NPCs would make up for whatever you didn’t have. If I focused on being a carpenter to fix up a base somewhere, cannot learn everything (like I can now), but my car is crap, I am forced to venture out in the world. I need to track down and recruit/save a mechanic, keep that person alive to get them home, keep them happy (collect food & water for them, clothing, whatever else they want/need). That then opens up the ability to design upgrades for our vehicle and while I go collect the parts they stay put & work on it, etc.

It evolves from “sole survival” into “I’m trying to keep my rag-tag group alive & sane”. If you want to LMOE it, you can, of course, but I’d enjoy the complexity of dealing with your NPCs at home base and all that entails.

Sorry … brain wandered off there … more back on topic …

What do you think about book smarts vs street smarts – Intelligence vs Perception maybe?

I feel that Int is basically a required stat to bump at this point. Don’t take Int? Well then you need to buy talents that make up for your int deficit.

Frankly, I’d almost see Perception as a variable in practice-based skill gain. I have to pay attention to my golf ball in flight, where it lands, and my body especially, to help me realize what I need to do to get better next swing.

11

Couple quick thoughts:

  1. Careful about adding/splitting stats. One Mines finale can stunlock (thus, effectively one-shot) someone with 14 IN, which I thought was a pretty good score. Splitting scores also implies moving the “saving throw” mechanics around, etc.

  2. Is 8 or 10 considered the human norm for stats? I was thinking 8 but would appreciate official word on that.

  3. I like the idea of NPCs handling skill-specialization but would worry about forcing all characters to specialize, pretty much for the LMOE reason. Not everyone wants to manage a team. (I’d like for a craft-y character to be able to equip troops–forming NPCs into a militia or somesuch to patrol the region, etc, but that’s rather down the timeline at best.)

Re GlyphGryph: I don’t mind options A or C. Would want to hear more about the reduced rate of B, and D seems like it might break require more tedious micro-skilling to effectively use high-level books, since higher level skills require more XP spent for the same gain. Skill rust already causes problems around level 7 or so, so having that happen even faster would demand grindy micromanagement.

12

It’s sounding more and more like the skill system could do with an overhaul sooner or later. I don’t know whether we want to try incremental improvements, or a massive all-in-one change. Personally, I think what I’ve currently got is at least worth testing, but I can see the merits of a large overhaul as well.

why not increase the size of the xp pool?

That’s a thought, although if anything, that would just make me more inclined to hunker down and read books until it’s full. Combining it with additional tweaks might make sense, though.

This means unlike in your current system where morale is less important, it's still vital, but having a higher intelligence means you can learn more per session.

I’ve actually never found much use in morale. Although, I don’t know if that’s because I’m playing suboptimally, if it’s because significant XP increases are tough to get, if morale itself doesn’t have enough effect, or what. One thing I notice about morale is that each additional 20 points of morale has a higher and higher effect on XP gain (the difference between 1 in 10 turns and 1 in 9 is much less than the difference between 1 in 2 turns and every turn). Maybe we should tweak morale’s effect a bit, as well? Make it more obvious, somehow? More fine-grained?

Also, I didn’t actually remove morale’s effect - which, come to think of it, could mean very high XP gain for high-morale, high-Int characters.

Have skill books pull from the experience pool, but at a reduced rate.

I like this suggestion, it would de-emphasize the current perceived need to learn from skill books before you learn from practice. Maybe have Int affect the XP ratio, with higher Int == less XP spent during book-learning?

On the other hand… the amount of XP spent would often be less than what you get while you’re reading. I wonder if we want to instead reduce XP gain while reading skill books?

How does it work now? I was figuring that when a point of xp was spent with 142% skill comprehension, it was 100% chance of the first point going through and a 42% chance of a free point.

That’s how I thought it would work, too. That’s also how my variation #1 handles it. Instead, the current system is that skill comprehension gives the probability that XP will be transferred from the pool to the skill. Numbers above 100% don’t do much (except for high-level skills, there’s a skill comprehension penalty once it gets higher than half your Int). And failed rolls don’t actually penalize you, they just mean that XP wasn’t transferred. That’s a major annoyance with stuff like electronics, sure, but for combat skills and cooking and such? You may not even notice a failed roll, given how fast XP gets spent.

Rather than outright give you skill, have books give you a temporary skill boost.

I’ve heard DarklingWolf talk about changing how skill books work, possibly removing them entirely. I agree that changing how they work might be good for a full overhaul of the system, although I’m not sure at the moment what the best solution overall is regarding skill books.

Are you looking at just how Int (and thereby "Skill Comprehension") affects skill gain, or the entire skill system in general?

Just Int and skill comprehension for now, since I consider the current implementation to be functionally broken (it’s Fast Learner as a 3-point trait that really bugs me). I’m not opposed to a full skill system overhaul, though - it would just take more time and thought to do well.

[quote="Soron, post:1, topic:703"][b](0) Current playstyle[/b]:[/quote]

This is how I play currently and it feels too “gamey” to me. After day 1, I’m not surviving and fighting for my life. Read - train - repeat, becomes the mantra.

Cool, so I’m not the only one who has that issue.

Does #2 make skills ramp up that much faster though? Are you powering some of them so fast it'll imbalance things?

Option #2 does have the effect of making easily-trained things even faster to train. Which… now that I think about it, is probably a negative. I think it’s less of a negative than the current system (and the numbers could be tweaked), but we should definitely watch out for stuff like that.

Frankly, I'd almost see Perception as a variable in practice-based skill gain. I have to pay attention to my golf ball in flight, where it lands, and my body especially, to help me realize what I need to do to get better next swing.

That seems like a good way to look at things, actually. I don’t know if we want to split stuff like that, but it’s definitely good to think about.

On that note - when NPCs get back in, I wouldn't mind if you had to specialize your character in a craft/trade .. and NPCs would make up for whatever you didn't have.

…and…

  1. I like the idea of NPCs handling skill-specialization but would worry about forcing all characters to specialize, pretty much for the LMOE reason. Not everyone wants to manage a team. (I’d like for a craft-y character to be able to equip troops–forming NPCs into a militia or somesuch to patrol the region, etc, but that’s rather down the timeline at best.)

I like the idea of encouraging specialization, but I don’t think we’d want to force it. It would also have to be clearly explained somewhere in the UI or help screen. Maybe reduce learning rate by some amount based on the sum of all skill levels? Not enough to cause excessive frustration, but enough to make people think twice about skilling up Cooking to 8 without a reason to. Actually, that might address some (not all) of the issues with certain skills getting way out of hand.

Now, with all that replied to… that’s given me some ideas for a more significant overhaul. For instance, maybe we want to introduce more ways to affect morale, and replace the “hunker down and read” phase with activities meant to maintain your sanity? And in line with that, it may be sensible to smooth out the XP bonus from morale a bit, like maybe use morale the way I’m using the reworked skill comprehension-thingy in variation #2 from the OP. In that sort of situation, I’m not sure whether or not we want to have stats affecting skill learning or not.

Also, after reflection… it seems like the numbers involved are too small (and too coarse) for long-term play. They work well for someone surviving a few days, sure, but not so much for long-term survivors. That’s possibly something that should be tweaked, as well, if/when we overhaul the system.

13

Personally I think skill rust is terrible and either it should be taken out Or at least greatly slowed and capped, and recovering from skill rust should be much faster than gaining the skill to begin with and not cost any experience at all. (You’re not relearning, after all! You’re just shaking off the rust.)

Also, Soron, I think your style is missing out on the awesomeness of good morale. How long do you usually read for? With morale boosts, I can easily fill up my experience pool three or four times a day, while still doing stuff in between. XP pool increases are based solely on your morale, after all. If you’re happy enough (usually through drugs), you’ll actually gain xp faster than you can realistically spend it.

14

No need to agree with me quite so thoroughly, GlyphGryph. From a hardcore-realism standpoint it makes sense that top/line skills (I’d define those as 8+, just for a starting number) would require fairly consistent use. Having rust set in at 3 or so like it does now, yeah, that makes skilling tricky and I turned it off.

I’m not Soron but I’ve got some experience with alcohol-fueled melee monsters. Yeah, 1XP/turn is neato but not when I’m expecting to have the character hit -100 morale in about 10 hours or so because he’s in alcohol withdrawal. Repeated use could help, but for the thirst problem–IIRC about three or four shots is enough to drop a Thirst category, and that does debuff. I was doing some experiments with the mp3 player and that was useful; seemed fairly battery-effective if you remember to look for the morale capping and turn it off.

15

Morale has it’s own serious problem right now, in that the benefit function is… It’s pretty absurd. High stuff is absurdly good, and bad absurdly low.

Example:
Item A gives 5 morale
Item B gives 10 morale

You’d expect item B to be about twice as good, no? But, unfortunately, duration is (to my understanding, I could be wrong) coupled tightly with quantity.

So Item A gives 5+4+3+2+1=15 total, while Item B gives 10+9+…+2+1=50 total. Item B is actually more than three times as good as item A. And this gets even worse! A 60 morale event gives around 1,800 morale!

Obviously, these numbers aren’t the exact numbers used in game, but it describes the growth rate as I understand. And it is… pretty terrible.

I’m gonna have to dig in soon to see if it really works this way and if it does, what can be done to make it less terrible…

16

Just to save people heading to the wiki, XP gain rate is +/-1 turn for every 20 morale. Baseline is 1 XP per 10 turns=1XP/minute. So the mp3 player at full effect would have the player gain one XP every 8 turns.

180 morale is achievable with about 4 shots each of three different liquors, IIRC. (Last I used was Tequila, Gin, and Vodka, IIRC.) And with Ethanol Burner, Blood Filter, and an Integrated Toolset, you’ll probably gain battery charge in the bargain even after accounting for boiling replacement Clean Water. Trouble is that after about three rounds of that sort of alcohol use, you’ll have a stiff withdrawal hit once you stop.

Not sure if the withdrawal/side effects might be a reasonable balance to the effects of morale-boosting items: you’re Up but then you’re Down. If it simply that the effects are too outsized at higher levels, I can understand that.

17

You don’t actually lose experience for low morale, and haven’t for a while.

18

Hmm, I wonder: are there any effects of morale other than the change to XP gain? I mean, when I hear the term “morale” in a zombie survival scenario, I think of something that should be maintained at a reasonable level constantly, rather than something that only matters if I’m training up my skills. And I certainly don’t think of something that I should be min-maxing with sudden spikes when I want to train a lot of skills, followed by seeing my morale level drop tremendously when it no longer matters (what “no longer matters”? That’s how you get suicides!).

I’m starting to think that we should consider shifting the focus of morale.

19

It isn’t so much “negative XP, drain the pool” as it is “abysmally slow gain, if any at all”. (Draining would be pretty rough. Thanks for that small mercy.) I learned about addiction levels the hard way* and once I’d gone through the 700-odd XP I’d had, that was pretty much it for the next several hours. Maybe I got 5-10 XP over that time; morale was hovering in the -90 to -110 range.

*I’d read about them, but presumed that Blood Filter removed that problem as part of the whole filtration thing. It didn’t.

If XP gain can’t drop below 1 XP/minute, somebody needs to fix the wiki.

Soron: Ja. Moderately bad morale ( < -50, IIRC) locks out crafting/reading, and significantly good/bad morale (100 points or more) gives stat bonuses/penalties. (Primarily to IN, but enough morale will reach everything.)

20

My main problem with int influencing xp gain is thematic.
Like Dominae my understanding of xp was that it is the “giving a shit” score, and it isn’t something that would be influenced by any of the existing stats. Perhaps a better way to describe it is “mental fatigue”. It’s difficult because it’s a fairly novel concept in gaming, so there isn’t a term I’m aware of that represents it.

Your description of the current way it works is exactly as I envision things typically working, you head out and get shit done, and get physically, emotionally, and mentally beat up, then you retreat to whatever refuge you’ve secured to recover until you’re ready to attack the world again. The “fun” activities you can partake in aren’t all that fun for you as a player, and that’s something to look at, but it sounds like it’s working as intended.

That’s not to say that things are working perfectly right now, but I’m not sure what your criteria for “it’s working better now” is. From what I can tell you’re frustrated by your character needing downtime, but… people need downtime.