Is animate skeleton strong or not? (again)

[dowan]

Or, evenplayers trying their damnedest to win realize that animate skeleton isn't that great.

Or is there something special about Ne that just makes people play poorly? I mean, you can accuse berder of reaching with his statistics, but I find it even more unbelievable that necros do poorly because their starting book is just so great. The tavern seems to think every spell in the necro book is awesome, although I would strongly disagree, as do the statistics, apparently.

[Sandman25]

Or, evenplayers trying their damnedest to win realize that animate skeleton isn't that great.

Or is there something special about Ne that just makes people play poorly? I mean, you can accuse berder of reaching with his statistics, but I find it even more unbelievable that necros do poorly because their starting book is just so great. The tavern seems to think every spell in the necro book is awesome, although I would strongly disagree, as do the statistics, apparently.

Have you read the thread?

This. Ne is one of the last backgrounds I won with, exactly because I didn't enjoy animating every possible corpse, luring every monster upstairs to be attacked by zombies from previous floor, exploring manually to make sure I am always in the best position to take advantage of my zombies, using portal projectile instead of simple shooting etc.

Good players trying to streak does not mean that they are playing optimally or are doing other things that they hate.

[Berder]

Most of those deaths were caused by players playing Ne as fighter (ignoring AS)

Sandman25, isn't it interesting that most of your "No AS" games appear earlier in your list of links? I provided the links in descending order of streak length. It appears that the stronger the streak player, the less they are inclined to use AS.

or by player mistake which had nothing to do with being a Ne.

Almost ALL deaths are ultimately caused by player mistake, and having a stronger character helps protect you from your own mistakes. The correct reading here is: the player made a mistake, which AS was insufficiently powerful to save them from.

[dowan]

Unlike the old as, where when you made a mistake you just backed into the cloud of skellies, and problem solved most of the time. I think the OPness of AS was already fixed when the skellies became temporary, however, I did make an alternative suggestion in the GDD thread on the same subject, which is barely a nerf for spellcasters, but prevents the 0 skill AS generating a titan problem.

[Sandman25]

Sandman25, isn't it interesting that most of your "No AS" games appear earlier in your list of links? I provided the links in descending order of streak length. It appears that the stronger the streak player, the less they are inclined to use AS.

Do you really want me to accuse those players in bad play? It's ok, I don't have problems with that.
Hyperelliptic was playing badly in that game (both tactically and strategically), jeanjacques was playing badly in that game (both tactically and strategically) and so on. Please let me know if you want to know details, they are quite obvious for everyone who analyzes the morgues without piety.
You really should not assume that being a good player means that player does not do any mistakes, it is not scientific approach, closer to religious dogmas which don't work in crawl.

Almost ALL deaths are ultimately caused by player mistake, and having a stronger character helps protect you from your own mistakes. The correct reading here is: the player made a mistake, which AS was insufficiently powerful to save them from.

Yes, I got your point. Forgetting about scrolls of blinking or getting 127 damage from A.Lich in Zot 5 proves that Ne is not strong and Animate Skeleton is not OP. Sorry, I don't know what else I can say here. We can only agree to disagree I guess.

[archaeo]

The correct reading here is: the player made a mistake, which AS was insufficiently powerful to save them from.

I'm pretty sure the correct reading is that everybody should just stop arguing with you about your use of these statistics, since it's abundantly clear that no matter how many people bring up their reservations with your problematic analyses, you're completely unwilling to bend.

[Berder]

Yes, I got your point. Forgetting about scrolls of blinking or getting 127 damage from A.Lich in Zot 5 proves that Ne is not strong and Animate Skeleton is not OP. Sorry, I don't know what else I can say here. We can only agree to disagree I guess.

Quite seriously - if taking 127 damage from an ALich on Zot:5 is the kind of thing that kills you at that stage of the game, and your skeletons weren't strong enough to kill the lich for you, then the skeletons aren't protecting you against the things that kill you. Same with waiting too long to blink vs a hill giant. If the skeletons had killed the hill giant, there wouldn't have been a problem.

[Sandman25]

Quite seriously - if taking 127 damage from an ALich on Zot:5 is the kind of thing that kills you at that stage of the game, and your skeletons weren't strong enough to kill the lich for you, then the skeletons aren't protecting you against the things that kill you. Same with waiting too long to blink vs a hill giant. If the skeletons had killed the hill giant, there wouldn't have been a problem.

Yes, I remember your point, we have discussed it already. If something protects you in 99% situations instead of 100%, that "something" is useless and it should not be used. If something kills 99% monsters instead of 100% and then you die to that 1% monster, that "something" is useless and should not be used.
Please don't argue about 99%, this was an exaggeration to make my point clear.
I agree with archaeo, this discussion is pointless

[Zooty]

I can solve this "which background/race/combo is great" problem for you. All I ask of you is 20 newborn babies, 20 million dollars and 20 years time.

[Berder]

Yes, I remember your point, we have discussed it already. If something protects you in 99% situations instead of 100%, that "something" is useless and it should not be used. If something kills 99% monsters instead of 100% and then you die to that 1% monster, that "something" is useless and should not be used.

If you could have handled the 99% of situations on your own (e.g. draconian packs), and the 1% situation is the one that could actually kill you (e.g. ancient lich with LCS), and the skeletons don't protect you from it - then no, they aren't really helping.

[Sandman25]

If you could have handled the 99% of situations on your own (e.g. draconian packs), and the 1% situation is the one that could actually kill you (e.g. ancient lich with LCS), and the skeletons don't protect you from it - then no, they aren't really helping.

How interesting... Are you training armour/dodging/weapon?

[archaeo]

FWIW, I think Neil proposed (in ##crawl-dev) the best solution to this "problem" if it's actually a problem: make Animate Skeleton always give you a single kind of "skeleton" monster instead of creating a skeleton version of the monster. Maybe call it Animate Bone Golem instead, or Animate Bones.

I still don't think any change is necessary -- I think any argument against AS is an argument against all low-level spells that remain good for the entire game, which is kind of dumb imo -- but if you were going to make one, that seems like the best solution.

Also, Sandman25, either Berder isn't trolling, or he is the most talented and convincing troll I have ever seen on the Internet. I have every reason to believe it's the former: Berder is arguing with you in good faith, even if you disagree with him.

[Lasty]

As someone mentioned twice in the list of morgues, let me be really clear: 1) Until I nerfed animate skeleton I mostly didn't use it, because it was so strong it felt like cheating; 2) animate skeleton is sometimes more obnoxious to use than I feel like dealing with, though the nerf made using it well much more interesting for me; 3) even when I'm on a streak, I clearly don't play optimally, and the reason is that I play crawl for fun, not to prove something to someone. When I was a worse player I would attempt to keep streaks going, and I found that I didn't enjoy the process at all; I'd much rather just play for fun and then get excited when that ends up generating a streak, because it shows that I'm actually improving rather than just playing so deliberately that it's painful in order to look good.

As for this thread, I strongly recommend that if you find someone to be completely unable to exchange ideas, you simply add them to your enemies list and thereby stop reading their posts. If enough of you do that, then maybe I'll stop having to read those posts in quotes. :p

Edit: that hill giant morgue was so painful. I considered the possibility of a double move + fatal damage really unlikely, so I was being greedy with consumables. Really stupid play.

[Berder]

Mighty neighborly.

Anyway, Sandman, you do have a point about the necromancers that never used Animate Skeleton, though you don't have a point about the player mistakes. I suppose a better comparison would be to look at streak morgues that used Animate Skeleton, vs those that did not, and compare the winrate.

[Sandman25]

Berder,
I apologize, I edited my post.

[Berder]

The idea here is to track whether AS usage reduces the risk of player death immediately afterward. For this, I can use the Action table, which is split into divisions by XL. Here is an example of an Action table (irrelevant parts cropped):

  Code:

Action                   |  1- 3 |  4- 6 |  7- 9 | 10-12 | 13-15 | 16-18 | 19-21 | 22-24 | 25-27 || total
-------------------------+-------+-------+-------+-------+-------+-------+-------+-------+-------++-------
Melee: Unarmed           |     5 |       |       |       |     1 |       |       |       |       ||     6
       Great mace        |       |       |       |   146 |   159 |    24 |       |       |       ||   329
...
 Cast: Pain              |    73 |    79 |    52 |    26 |    46 |    23 |       |    17 |     2 ||   318
       Vampiric Draining |     7 |    25 |    80 |   203 |   552 |   516 |   352 |   378 |   337 ||  2450
       Animate Skeleton  |       |     1 |     2 |     1 |       |       |       |       |       ||     4
...


If a player casts AS during a given XL division, and does not die in that XL division, then that would speak to the power of AS.
If a player casts AS during a given XL division, but dies anyway in that XL division, then that would speak for the weakness of AS.
If a player does not cast AS during a given XL division, and does not die in that XL division, then that would speak to the power of not using AS
If a player does not cast AS during a given XL division, and dies in that XL division, then that would speak for the weakness of not using AS

I used the following method:

1. Download the morgues for the relevant streak games
2. Look only at games that have the potential to be 4th in streak or later, discarding the griefer games
3. Parse the morgues looking for the Action table. Morgues without an Action table are ignored.
4. Look for morgues with a Cast: line in the Action table. Morgues without a Cast: line are also ignored, since I'm not interested in comparing characters that cast Animate Skeleton to Troglodytes. I only want to compare those casting AS to those who might potentially be casting AS, but aren't.
5. Look for the Animate Skeleton part of the Action table.
6. For each XL division of the Action table, I tallied up the following
   
   1. number of games that survived that XL division without casting AS
   2. number of games that died in that XL division without casting AS
   3. number of games that survived that XL division and cast AS in that XL division
   4. number of games that died in that XL division and cast AS in that XL division
7. calculate sum(b)/(sum(a)+sum(b)), the death ratio without AS. sum(b) means the sum of the b entries for each possible XL division
8. calculate sum(d)/(sum(c)+sum(d)), the death ratio with AS

The death ratio without AS is 9% greater. This difference is not statistically significant.

My a, b, c, d:

  Code:

a = [983, 908, 844, 774, 729, 698, 651, 628, 575]
b = [57, 67, 48, 60, 28, 31, 33, 15, 21]
c = [33, 40, 55, 61, 76, 69, 77, 83, 62]
d = [1, 1, 1, 4, 2, 7, 5, 1, 5]
sum(a) = 6790
sum(b) = 360
sum(c) = 556
sum(d) = 27
sum(b) / (sum(a) + sum(b)) = 0.0503
sum(d) / (sum(c) + sum(d)) = 0.0463


It may come to light that there is a bug in my code. It's complicated enough to have one. Anyway, these are my results for now. Based on this I would conclude the benefit of AS is relatively small.

A note on the interpretation of a, b, c, d:
Pr(you die from xl 4-6 given that you reach xl4 and don't use AS from xl4-6) ≈ b[1] / (a[1] + b[1]) (zero indexing)

[dowan]

9% seems somewhat significant, to be honest. Of course, I wonder how that compares to butterflies, and also whether we're just confirming that better players think AS is good, therefore players with AS are on average slightly better than those without, and therefore they win more. That is to say, if there were a well believed hoax within the crawl community that engraving your weapon with "Elbereth" makes the RNG treat you better, then we'd probably see that people who engraved "Elbereth" win more often than those who do not. That's not because of the engraving though, obviously, it's because those are the players that try to maximize their chance to win. Of course, AS is more useful than a useless engraving, but I hope you see my point.

[Sandman25]

There is a caveat. Some high XL characters cast AS only because it is faster than butchering a corpse, they use Animate Dead otherwise. So late game it can be a comparison of characters without AS with characters with AD. AD is much more powerful than AS of course.
I like first numbers (for low XL) in Berder's research, they ("d = [1, 1, 1, 4, 2") show that it is hard to die with AS.
Despite I like the results I still disagree with the method of course, for example, it does not take into account deaths in Slime Pits where there are no corpses (unless Kiku of course).

[Rast]

As someone mentioned twice in the list of morgues, let me be really clear: 1) Until I nerfed animate skeleton I mostly didn't use it, because it was so strong it felt like cheating;

I don't play trunk so correct me if I'm wrong: The "nerfed" AS produces monsters that are still just as strong, but they time out faster?

[Berder]

9% seems somewhat significant, to be honest.

It's not. sum(c) = 556, sum(d) = 27. Treating it as a binomial random variable with n = 556 + 27, and p = 27 / (556 + 27), the standard deviation is 5.07. If the number of deaths with skeletons swung by 5 in either direction then that 9% would be wiped out. That isn't a full difference-of-means significance test, but it's enough to tell.

I've tried this same method with search terms other than "Animate Skeleton." If I search for Repel Missiles, for example, I find that not casting RMSL is associated with a 36% higher chance of death than casting it, among characters that ever cast anything, which is statistically significant. For Deflect Missiles, it's 79% higher (not significant). For Phase Shift, it's 52% (not significant). So that 9% doesn't look very good by comparison to these other examples.

Of course, I wonder how that compares to butterflies, and also whether we're just confirming that better players think AS is good, therefore players with AS are on average slightly better than those without, and therefore they win more. That is to say, if there were a well believed hoax within the crawl community that engraving your weapon with "Elbereth" makes the RNG treat you better, then we'd probably see that people who engraved "Elbereth" win more often than those who do not. That's not because of the engraving though, obviously, it's because those are the players that try to maximize their chance to win. Of course, AS is more useful than a useless engraving, but I hope you see my point.

I actually think the reverse might be true: that stronger players use it less because they don't want to bother. Sandman's examination of the Ne losses seemed to indicate that. I should probably see what happens if I control for player strength. It's a little tricky to figure out how to best to do that.

There is a caveat. Some high XL characters cast AS only because it is faster than butchering a corpse, they use Animate Dead otherwise. So late game it can be a comparison of characters without AS with characters with AD. AD is much more powerful than AS of course.

If I look at the death rate for characters that cast AS, and compare it to the death rate for characters that don't cast either AS or AD, I find the death rate for those who don't cast AS or AD is only 2% higher than those who cast AS.

I like first numbers (for low XL) in Berder's research, they ("d = [1, 1, 1, 4, 2") show that it is hard to die with AS.

The numbers are too small to conclude anything about XL-dependence.

Despite I like the results I still disagree with the method of course, for example, it does not take into account deaths in Slime Pits where there are no corpses (unless Kiku of course).

That's probably not a large enough factor to worry about. It would affect the a and b terms, not the c and d terms.

[ydeve]

The correct reading here is: the player made a mistake, which AS was insufficiently powerful to save them from.

I'm pretty sure the correct reading is that everybody should just stop arguing with you about your use of these statistics, since it's abundantly clear that no matter how many people bring up their reservations with your problematic analyses, you're completely unwilling to bend.

This thread reminds me of another involving chickens.

[Lasty]

As someone mentioned twice in the list of morgues, let me be really clear: 1) Until I nerfed animate skeleton I mostly didn't use it, because it was so strong it felt like cheating;

I don't play trunk so correct me if I'm wrong: The "nerfed" AS produces monsters that are still just as strong, but they time out faster?

Yup!

[Aethrus]

From what I see with this debate, a lot of the arguments is based around two different looks at the spell animate skeleton.

For players using the spell from the beginning of the game (ie Necromancers, Kiku worshipers, those that find early spellbooks) the spell gives reasonable returns for reasonable investment... most characters that early need at least some spellcasting / necromancy to make the failure rate tolerable, especially if they're not a background that emphasizes Int. Hypothetical Optimal Players might posit that you can ignore any investment in the spell, cast at 58% failure with 0 spellcasting / 0 necromancy and just summon them at the end of fights, but I think that that's a waste of the spell's usefulness anyways, especially early on, as characters can benefit from getting a skeleton mid-fight as well. From what I understand, most are not arguing that the spell is a problem until around Lair.

In the middle-late game, there are a lot more creatures that are far stronger than hobgoblins and the occasional gnoll... elephants, death yaks, stone giants, etc. are the ones often pointed out to showcase that the spell gives too much reward for too little investment at that point in the game. I do think that the actual level of reward is somewhat overstated... animate skeleton requires you to still kill the monster tougher than the skeleton you get out of it, without that skeleton to help you. And late game, more and more enemies are going to have special features as living beings they lose as skeletons... a Titan skeleton isn't the same level of threat as a Titan, it's slower and lacks its threatening spells. It's still a useful cold body to soak damage, soften up opponents, and maybe even take some popcorn out...

I'm not convinced that animate skeleton gives too much power late in the game; while it does scale with enemy strength, it does not in fact scale as quickly as the enemies themselves do. Many highly threatening monsters don't drop corpses at all to use the spell on, but most importantly, while derived undead let you turn corpses of foes into strong allies, they won't help you actually get those initial kills (barring worshipping Kiku, in which case we're not really talking about using Skeleton anymore anyway). And that's where I hesitate. If players had to invest similar amounts of experience into necromancy to get hydra skeletons that summoners have to invest to summon Hydras out of nothing, the player who went into summoning has made summoning his primary source of killdudes. The player who went into necromancy isn't going to killdudes as well, is limited to whatever corpses he scrounges from the dungeon as allies, etc. Scaling animate skeleton on the basis of the golden dragon and titan skeletons players can get in the end-game I feel will likely make necromancy too much of an investment to justify.

My separate issue with the idea of raising the level of Animate Skeleton is, admittedly, one of flavor, but I feel like the idea of choosing a background called Necromancer, going into the dungeon, and not being able to make any dead rise until level 2 or 3 would be the wrong direction. Necromancers should be able to make spooky skeletons out of the gate, 'fixing' Animate Skeleton and making it a higher level spell, even assuming it is more balanced for the game as a whole, is a solution that takes away some of the character and fun of that background.

As far as what solutions, if any, might need to be implemented, I'm not terribly sure. Making them expire over time was a necessary change to limit the ability to snowball, and actually puts some time pressure on necromancers to make it to the next fight with as much of their undead horde as they can. Solutions that come to mind might be splitting animate skeleton into multiple spells, with higher level spells for higher hit dice monsters, but that seems a solution far worse than the cure.

One idea I particularly like is making animate skeleton destroy corpses if it is miscast. It would at least solve the issue of it being a spell that players can cast 'with any failure %' and mean that characters would need either significant spellcasting and Int or some level of necromancy investment to be able to cast it. Some of the complaints about Animate Skeleton seem to revolve around the fact that its only opportunity cost is a spell slot for any character who even considers casting and has a god that allows it. This would give it some (small) experience requirement for it to have significant usefulness.

Final thought I've had is making skeleton duration scale as a function of spellpower versus hit dice. If you've only got 2 pips in Animate Skeleton, a Titan Skeleton wouldn't last you the time to rest off the fight you encountered it.

[ZipZipskins]

For players using the spell from the beginning of the game (ie Necromancers, Kiku worshipers, those that find early spellbooks) the spell gives reasonable returns for reasonable investment... most characters that early need at least some spellcasting / necromancy to make the failure rate tolerable, especially if they're not a background that emphasizes Int.

a level 1 ddne casts animate skeleton at 4%, a mfgl that finds the book in leather armour after raising int at level 6 (15str 10int 14dex) casts AS at 47% with 0 spellcasting and 0 necromancy. 3 levels of necromancy (0 spellcasting) brings that down to 14% fail which is a perfectly tolerable failure rate for most spells, let alone a more relaxed, 1 mp support-type like AS. even without raising int and instead raising dex, that same mfgl casts AS at 17%.

3 levels of necromancy is not an actual investment by any stretch of the imagination.

(a 19 str 5 int migl in plate needs 8 spellcasting 10 necro to cast animate skeleton at 17%. this is an extreme example for contrast only.)

[Aethrus]

For players using the spell from the beginning of the game (ie Necromancers, Kiku worshipers, those that find early spellbooks) the spell gives reasonable returns for reasonable investment... most characters that early need at least some spellcasting / necromancy to make the failure rate tolerable, especially if they're not a background that emphasizes Int.

a level 1 ddne casts animate skeleton at 4%, a mfgl that finds the book in leather armour after raising int at level 6 (15str 10int 14dex) casts AS at 47% with 0 spellcasting and 0 necromancy. 3 levels of necromancy (0 spellcasting) brings that down to 14% fail which is a perfectly tolerable failure rate for most spells, let alone a more relaxed, 1 mp support-type like AS. even without raising int and instead raising dex, that same mfgl casts AS at 17%.

3 levels of necromancy is not an actual investment by any stretch of the imagination.

(a 19 str 5 int migl in plate needs 8 spellcasting 10 necro to cast animate skeleton at 17%. this is an extreme example for contrast only.)

A level 6 MfGl who raises Int, memorizes a level 1 support spell and puts experience in a skill they have a poor aptitude for that could be going into getting mindelay on a trident deserves some gnoll and hound skeletons, I'd say. 3 levels isnt an unnoticeable amount that early in the game with a negative aptitude. I'm not claiming its an onerous burdern, but its a noticeable divergement. It's better than just going all in melee... but support magic should be, otherwise its pointless.

The place where you and I are concerned about the spell isn't a level 6 MfGl picking this spell up. It's about the level 17 MiFi or so who has it barely castable but getting powerful skeleton support through the rune branches and depths. And I mentioned what I thought might be other considerations for that point, but I'm not sure what if any are better than the existing situation.

[bcadren]

Animate Skeleton is good early. In late it can be pretty meh (mostly because the creatures are weaker than normal variants and only having melee and being slow means...they are pretty much meat shields late. [take hits but don't do much damage]); and in extended...well you won't have enough corpses to get much use unless you are of Kiku and if you are of Kiku, you'll be using a better slave spell. So...I'd say generally it's great up until...late lair. Good for meat shields for most the rest of standard, but no longer able to really do much damage. Which is...fair for a level 1. I think. It's more powerful than Summon Small Mammal and weaker (in late game) than Summon Butterflies so...pretty fair.

[duvessa]

level 3 and 4 spells often fail to be "great up until late lair", i think it is pretty reasonable to say that a level 1 spell definitely shouldn't be

[Berder]

level 3 and 4 spells often fail to be "great up until late lair", i think it is pretty reasonable to say that a level 1 spell definitely shouldn't be

Level 3 and 4 spells often succeed at being great through extended.