Brands: protection -> evasion

[ololoev]

Why do we have protection brand while evasion is much more cool? (despite the fact both are boring)
Most tabbers (who usually benefit from the brand) rely on AC rather than EV and this lead to defensive stats like 35AC/8EV. Thus having +7 EV will matter more than +7 AC (I don't sure it will, but 35/15 looks cooler that 42/8).
Also, this brand can be reworked to give +3AC and +3EV.
Or removed.

[bel]

Sounds like a good idea, especially the second last point (giving EV and AC both). Keep in mind that protection weapons no longer give a fixed value of AC in trunk; it depends on the number of times you hit. (edit: nvm, it has been changed again)

Tangentially related: I did some tests using a modified version of fsim and the utility of per point AC or EV depends on the GDR. With high GDR, one point of AC is more useful than one point of EV, while with low GDR, it's the reverse (evasion is even more important for an octopode or draconian). There are many caveats to this finding: you are putting all your eggs into one basket, so undodgeable attacks (for instance) are a bit more dangerous. Also, a point of AC and a point of EV don't have the same opportunity cost (though this is mitigated by EV being almost twice as good as AC in my tests - for low GDR).

[Sar]

http://s-z.org/neil/git/?p=crawl.git;a=commit;h=9d3f7082a8e28e3f5d69b42a311a389155c2454a

  Code:

Instead of giving an increasing bonus, give a fixed 7 AC bonus after striking
an enemy. The bonus is slightly higher than the old protection bonus to
make it a bit more appealing than old protection and because it's harder to
trigger. This is mostly a buff from the incremental version, but it offers
less press-your-luck/pie-in-the-sky potential.

The motivation for this change is that the incremental version is more
complex and could potentially lead to scummy behavior.

[stoneychips]

This is a rare case where I have to admit I've pretty much given up on the brand myself... I stopped using Protection mostly, even before you had to hit with the things.

I'm not sure I feel strongly about it. Others might have tactics that get some mileage out of it? I generally like to play more offensive lately. I'll often use shields because so strong/so many resistances, but otherwise I want to get through the dungeon a little faster and take less "special" attacks while slugging where possible.

Again, it's just my two cents. I wouldn't go so far as to insist Protection "should" be removed, but (for once) I probably wouldn't personally consider arguing for it to remain, either.

[Siegurt]

Polearms with protection are slightly better than non polearms with protection...

[prozacelf]

I use new protection more often than I ever did old protection.

[stoneychips]

Actually, with a decent enchantment, I'm finding one pretty handy in the early dungeon.

Still not sure I'd bother using it in late.

[Gozigzag]

plot twist: we used to have evasion brand and it was removed due to being good. Increases to AC are worthless if you dont have gdr, while gdr only comes for heavy armor which already have decent AC, making the bonus AC trivial, so using a protection brand instead of a offensive brand is always a bad choice.

[VeryAngryFelid]

Probably you don't know how gdr works. I don't see how EV+5 can be better than AC+7 for many characters provided extra EV does exactly nothing in most cases.

[Gozigzag]

Im not comparing AC and EV, im comparing current protection with any other offensive brand that works against the enemy you are currently fighting. The EV brand was a joke, it only spawned on ranged weapons and it was AT BEST a mage stick. But if people prefer reducing the damage they take by 1d7 or increasing their chance to evade an attack by less than 10% at exchange of around 30% effective weapon damage... oh well.

[stoneychips]

Im not comparing AC and EV, im comparing current protection with any other offensive brand that works against the enemy you are currently fighting. The EV brand was a joke, it only spawned on ranged weapons and it was AT BEST a mage stick. But if people prefer reducing the damage rthey take by 1d7 or increasing their chance to evade an attack by less than 10% at exchange of around 30% effective weapon damage... oh well.

Maybe we're playing different games, but I never found evasion to be a joke. 5 points of EV is nothing to laugh at. To be fair, again it was more impressive in early game when you didn't have as many offensive choices sometimes.

Still, I really found it helpful for yes ranged combat - would you really expect it not to be good for ranged?? - especially, centaurs (whose barding is pretty random as to drop timing). Presumably someone will say forget it and just spend more levels training Dodging early on or something (or maybe totally fighting if you're that sort of fanatic), but I hate being single-tracked that way. And it was really nice when you had Evasion along with other bonuses on the same item.

I don't see how EV+5 can be better than AC+7 for many characters provided extra EV does exactly nothing in most cases.

I don't see how I just won with any deep elf focusing on EV if this is true. What do you mean "extra"? From what to what? That might help.

---
Somehow I have a feeling we need to specify where in the game people are talking about, before we get the sweeping statements about things being useless.

[Gozigzag]

Maybe we're playing different games, but I never found evasion to be a joke. 5 points of EV is nothing to laugh at. To be fair, again it was more impressive in early game when you didn't have as many offensive choices sometimes.

Its a joke when you are expecting a brand on a weapon that you are going to use to kill things from a position where they cannot touch you. Again, it was a mage stick.

I don't see how I just won with any deep elf focusing on EV

Did the enemies gave up on hitting your elf and comminted "sudoku"? I dont see how i won a deep elf without training any EV, other than the game has other means to achieve victory.

Somehow I have a feeling we need to specify where in the game people are talking about, before we get the sweeping statements about things being useless.

I would agree that prot/evasion is better than nothing, or ok back when you didnt had to hit things to get their effect and they only lasted for 2 turns.

[stoneychips]

[
Its a joke when you are expecting a brand on a weapon that you are going to use to kill things from a position where they cannot touch you. Again, it was a mage stick.

I wasn't expecting quite that, although I might have expected it a bit stronger. (Were you expecting that? Where did this begin, again?) I wouldn't have expected too much stronger from any brand that would be lumped alongside others on many artifacts though, now that you mention it.

Did the enemies gave up on hitting your elf and comminted "sudoku"? I dont see how i won a deep elf without training any EV, other than the game has other means to achieve victory.

Well sure I fire stormed a lot of stuff too and the Servitor must have finished some fights... Even blocked some boulders and bolts with a buckler +0 at around 5-6 Shields, to be complete. But I did spend more than one turn at a time with a few things in melee. Which says to me EV is doing something right. I honestly expected things in Vaults and Depths to slam me a lot more often than they did.

I would agree that prot/evasion is better than nothing, or ok back when you didnt had to hit things to get their effect and they only lasted for 2 turns.

I did find Protection much easier to use in early game when you didn't have to smack stuff to get the effect. Having to train it makes it much more iffy in my opinion. And on this particular run, I picked it up with a Gargoyle Wanderer (natural armor) who did start with a weapon of the same class (and thus was biased toward training it at least slightly already). So I'm not sure how much it should count in the large scheme of things.

At least with evasion you didn't have to train just to get the benefit.

[Sprucery]

Did the enemies gave up on hitting your elf and comminted "sudoku"?

Impossible; there are no pens or pencils in the dungeon.

[Siegurt]

On a slightly serious note.

I think it is ok to have some brands be weaker, i used a +3 protection glaive on a recent mf because my other option was a vanilla +3 glaive, it was ok, i enachanted it up to +5 and i was already in lighter body armor (ring mail) so the extra ac was welcome.

When given the ability to switch to a +4 glaive of chopping i ultimately decided against it, mostly because going from 17 ac to 24 ac felt substantial, i did pay with both for a while, but went with protection, and when i replaced it with a bardiche of flaming much later i felt like the ac loss was noticeable.

I think as a brand, current protection is probably on par with or perhaps slightly weaker than, vorpal, which is ok imho.

I would personally rather it be a steady bonus, the "you get the bonus after you hit something" i find annoying, i do understand why it is like it is, but i would rather it be a straight bonus worth, like, weapon skill/5+1 or something like that, but as is, it has a niche.

[VeryAngryFelid]

I don't see how I just won with any deep elf focusing on EV if this is true. What do you mean "extra"? From what to what? That might help.

You know how hit or miss check works, right? Basically d{to_hit} is compared to d{EV}, or to d{EV+5} for evasion brand. In most cases d{to_hit}-d{EV} and d{to_hit}-d{EV+5} are either both positive or both negative.
I did some fsim tests a long time ago and I remember, for example, that +8 EV changed chance of being hit for some Felid from 30% to 20%, that means that extra EV is important in just 10% cases and irrelevant in 90% cases.

[tasonir]

I'm not sure I'd call that irrelevant in 90% of cases - you reduced your incoming damage by 33%, which is pretty significant. Now that might not be as reliable on a felid when 2 hits can kill you, but for a normal hp/ac race, taking off 1/3 of hits is going to give you a lot more time to kill things or escape from them. This is basically how I try to tab the whole game - reduce incoming damage, stack some regeneration, watch your hp stay high all the time (in theory, until I forget to pay attention and die).

[VeryAngryFelid]

It is really irrelevant in 90% cases, the formula is linear. You get attacked 10 times and 9 of those attacks don't care about your extra EV.

[Siegurt]

It is really irrelevant in 90% cases, the formula is linear. You get attacked 10 times and 9 of those attacks don't care about your extra EV.

Yes, but 7 of them already were misses, so 1 out of the remaining 3 is still a good deal.

The equivalent in ac land would be to say that something that reduced your damage per hit by 33% would really only be reducing your damage taken by 10% because armour doesn't reduce damage from misses

[stoneychips]

It is really irrelevant in 90% cases, the formula is linear. You get attacked 10 times and 9 of those attacks don't care about your extra EV.

Okay, disclaimer I'm not a math whiz at all. I'm terrible about making calculation errors when I try to do longhand work myself. And no, I don't find it natural to muck around in all of Crawl's formulae. (The format makes me dizzy to begin with. I'd have to change the fonts and size to start with.) Heck, I couldn't even get an online file to dump to offline fsim the other day, if you recall.

But I'm going to try picking at this anyway, because I honestly do not believe I am conceptually that horrible at it... And if someone can show me it's a totally silly effort, then I'll simply shrug I tried to explore it to my satisfaction and probably laugh and walk away. Here goes. It depends of course what you want some EV to do, and you didn't talk about particular Crawl threshold points so far that I noticed (15 or 25 EV versus 9-12 or 30-40 for that matter). I have no idea what "some Felid" was working with to begin with... But taking what there was, that 1 in 10 attacks business sounds way too simplistic a figure for me.

From what little I recall of my high school probability lectures, I somehow suspected Tasonir has a bit of the reality there. In simple probability (or is that "in theory" where games go on for ages and nothing ever happens?) sure any one roll has those odds. And maybe that works out neatly in 100 or 1,000 trials but... We're actually talking about cumulative probability here because it's over the course of multiple trials. And btw I'd venture that except for oh Abyss/Pandemonium (maybe?), we're also probably going to decide life or death more by "Did I avoid a few extra hits while repositioning or did this push me over a nice Crawl EV threshold I didn't have otherwise" than "Did this item really prevent all sorts of ranged attacks all day?" Well, I am anyway.

Back to the point. So if I recall, to actually simulate the likely outcomes, some multiplicative and counterfactual math (odds of event actually not happening over multiple runs) are also involved. I'm not sure about the neat formula - if I could even write the neat thing out once I found it somewhere - but... After quite a bit of googling, I think if it might be something like whatever powers this binomial calculator.

Here are the parameters that make me think so. In particular:

What is a binomial experiment?
A binomial experiment has the following characteristics:
The experiment involves repeated trials.
Each trial has only two possible outcomes - a success or a failure.
The probability that a particular outcome will occur on any given trial is constant.
All of the trials in the experiment are independent.

Now, if I'm right (or in the ballpark at least, I hope):
This calculator looks neat because it will tell you a range of probabilities over cumulative trials: You can see the chances of getting not/less, equal, or more than a given number of hits over however many rolls. I'm not sure what its boundaries are; at some point things break down with larger numbers and you get Error figures, but that's any calculator I guess.

If this is indeed an appropriate place to look, then you can plug in some numbers and see:

0.3 probability for hits. The probability of less than 3 hits given 10 shots? Around 38.3 % And of less than 2 hits? 14.9%
Probability for less than 3 hits in 10 attacks with probability of hit at 0.2 ? About 67.8% For less than 2 hits? 37.6%

That's some absolute shift in percentage terms, moving the dial for success by 20 (edit: 20-some to 30?) %.
(See, maybe I can copy from calculators but I'm pretty lousy at consistent subtraction. )
Assuming my goal is not to be absolutely untouchable but rather to avoid taking damage a notable chunk of the time. Unless one finds a totally awesome robe/skin, that's before we even get to a touch of armor help even on many elves, resistances and damage randomness (and see also Deflect Missiles). I think that's fairly substantial.

Of course, you can play this any way you like on interpretation. You can also plug in other numbers of trials, other numbers of hits you'd like to look at. But where my concern is, does adding EV matter (at least adding 8 EV by whatever combination of training and items), then it looks to me like I could say, well if my goal is to at least noticeably reduce the chance of being hit 2 or 3 times in as many as ten attacks (which I also wonder is perhaps rather a lot)... Then in that case, that added EV actually changed my chances of pulling it off by much more than 1 in 10 per attack. At least, that's the case within the limited data emphasized so far. For all I know, that Felid could be sitting just 2 points right or left of some benchmark where the curve is completely skewed within Crawl's system.

That is, if I'm doing and interpreting this right. I don't think it's a simple linear calculation, in any case.

[VeryAngryFelid]

It is really irrelevant in 90% cases, the formula is linear. You get attacked 10 times and 9 of those attacks don't care about your extra EV.

Yes, but 7 of them already were misses, so 1 out of the remaining 3 is still a good deal.

No. Another example to make it more obvious hopefully. 12% without extra EV, 6% with extra EV. It is not 50% improvement, it is just 6% improvement because I almost don't care about attacks anyway. I would rather use another item instead.

[bel]

The 10% comment doesn't make sense. To see this, consider what happens when you go from 10% to zero.

Again, this is modulo things like undodgeable attacks and so on.

[VeryAngryFelid]

stoneychips


I don't know how to use online file for offline wizmode, I didn't even know it was possible before your post

I am not sure what's your point. Of course 10% in my example was for one attack and the more attacks we have the greater that 10% becomes. But the same happens to other items, you might use ring of fire and enjoy greater spell power in more attacks, or ring of wizardry and enjoy less falure rate for spells, or ring of protection from fire and reduce more damage from fireball etc. I mean greater number of attacks does not change relative usefulness of items (at least in most cases except those when you stair-dance V:5 and get hit by 8 vault guards) so it can be easily ignored.

[VeryAngryFelid]

The 10% comment doesn't make sense. To see this, consider what happens when you go from 10% to zero.

It's impossible. You always have at least 5% chance to be hit (and miss too, by the way) in crawl.

Do you see difference between going from 95% to 94% and from 6% to 5%? According to you the latter is really great because you decrease chance to be hit by 16%! According to me it is not a big deal in both cases, 1% is almost irrelevant.

[bel]

It was a thought experiment to make the point that the reduction isn't linear. The exact mechanics are irrelevant to the point.

Yes going from 6% to 3% is much better than going from 93% to 90% - the former doubles your effective HP for the particular attack, while the latter does almost nothing. The reason why the former case isn't usually important has little to do with the reduction from 6% to 3%, it has to do with the fact that monsters usually don't have only one attack; if one attack has only 6% chance of landing, other attacks become more important, especially hexes and other undodgeable attacks.

To see this in another way, consider MR, which works roughly in the same way as EV (again, not exactly, but that's not relevant to my point). Do you think that going from 90% to 80% banishment rate is same as going from 20% to 10%. Or 10% to 1%?

[VeryAngryFelid]

Probably I see where we have disagreement. You say going from 95% to 5% gives you 1800% extra HP, I say it just allows you to ignore extra 90% attacks. It's the same thing, if you think about it.

[VeryAngryFelid]

To see this in another way, consider MR, which works roughly in the same way as EV (again, not exactly, but that's not relevant to my point). Do you think that going from 90% to 80% banishment rate is same as going from 20% to 10%. Or 10% to 1%?

Of course. In hellcrawl I often had to fight monsters while having 30+% chance to be paralyzed/banished. That's after putting a ring of MR on. I still used the ring. Would you use the ring?

[bel]

No, it's not a matter of semantics. The main point is that the "10%" number is totally the wrong thing to look at. The effect isn't "linear" or "irrelevant", it depends on the starting point of reduction. Here's the quote which started the discussion:

Probably you don't know how gdr works. I don't see how EV+5 can be better than AC+7 for many characters provided extra EV does exactly nothing in most cases.

That is somewhat fine if you are talking about very high GDR. But you were asked, what is meant by "extra", and you gave the example of felid, which has zero GDR!

All this is very strange because I remember an old thread where you were arguing that an evasion ring on a felid is better than a protection ring (you were right, modulo undodgeable attacks). You still have, in your signature, that "EV is underrated while AC is overrated". So, I have no idea if you changed your mind since then, or what on Earth you're talking about in this thread.

[VeryAngryFelid]

My example with Felid had nothing to do with AC, its only purpose was to demonstrate that extra EV can be almost meaningless. Extra AC can be almost meaningless too so I tried to show that "meaningless +5" cannot be much better than "meaningless +7".

Edit. I am not comparing AC to EV at this point. All I am saying is that EV can be as useless as AC (in one attack!) so we should not overestimate either (OP overestimates EV IMHO)

[VeryAngryFelid]

Maybe I should clarify my position better. AC and EV do roughly the same thing, they allow to avoid some damage. Players should use fsim to decide which ring (+AC or +EV) or which brand (evasion or protection) are better for specific character in specific fight, everyone who says "evasion was great and protection sucks" is wrong, presense/lack of GDR does not change that

[stoneychips]

I don't know how to use online file for offline wizmode, I didn't even know it was possible before your post

Not saying it is. The fact is I just don't know if it is or not. I was just trying to say, I'm a complete novice as far as mechanical and technical stuff here. I mostly play by feel (along with reading advice and wikis). If I keep finding that when I add a few points of EV, I don't get hit as often then I tend to think it works for something. Whether that's enough "something" to satisfy you, might be another matter. But yeah, I'm not claiming to be anything awesome about picking at the internals of the game.

I am not sure what's your point. Of course 10% in my example was for one attack and the more attacks we have the greater that 10% becomes. But the same happens to other items, you might use ring of fire and enjoy greater spell power in more attacks, or ring of wizardry and enjoy less falure rate for spells, or ring of protection from fire and reduce more damage from fireball etc. I mean greater number of attacks does not change relative usefulness of items (at least in most cases except those when you stair-dance V:5 and get hit by 8 vault guards) so it can be easily ignored.

I'm just not buying your application of the fractions to understand whether (or precisely how much) EV works. As I understand probability, it isn't that simple and I think the pages I linked to would show that all else being equal, EV should have made a much bigger difference in the bare bones numbers you gave than you concluded it did. The difference between 30% and 20% is a lot greater than 1 hit in 10, because for any set of only 10 trials we're also thinking about "assuming whatever happened during all the rest of those 10" and that changes the probability of actually getting a string of events (not just one roll). Yes, I assume it's different for large numbers of trials, but I don't think we generally want to sit around and wait for the value to come down that long during any single battle. Now if you want to argue EV benefits relative to other brands and tactics, well that's an endless field and there are oodles of threads.

[Siegurt]

Maybe I should clarify my position better. AC and EV do roughly the same thing, they allow to avoid some damage. Players should use fsim to decide which ring (+AC or +EV) or which brand (evasion or protection) are better for specific character in specific fight, everyone who says "evasion was great and protection sucks" is wrong, presense/lack of GDR does not change that

So then your point is really not "additional protection is not significant" it is "once you are sufficiently well protected, additional levels of protection are probably not as important as something other than protection"

That certainly makes sense, particularly once you look at it from the extreme end of things where you are so well protected from that you have no chance of dying to them at all, adding more ac or ev to your ac/ev 30/30 character will not be important in a combat against any number of rats.

Of course, personally, i simply ignore those types of combats altogether when deciding what to use, i only consider combats where i might lose a significant amount of health, and hence i am biased towards adding more defenses. (This is quite possibly wrong for ev, since monster's to hit doesn't scale upwards indefinitely)

Note the difference between important and significant here, going from 90 to 95% protection (whether ac or ev) is *significant*, because it halves the amount of damage taken, but may not be *important*, because either way you aren't taking much damage to begin with (unless it really is the case that 10% of the damage is a large chunk of your life)

[VeryAngryFelid]

I see I failed to describe my point. I disagree with "Why do we have protection brand while evasion is much more cool?" and "Thus having +7 EV will matter more than +7 AC"

[stoneychips]

Whether EV is considered important, also depends on how you like to deal with damage. If you're playing aggressively and taking on huge waves of stuff (continually charging vaults to test your limits, or maybe pushing it into deeper Zig), then it doesn't matter as much to "save" a lot of HP over time because you can't guarantee prompt recuperation. If you have huge regeneration or Makhleb with fast kills, or at least a moment to sit back and eat chunks with a ghoul, then sure a big cumulative "savings" on HP may make more difference. You only have so many HP to save from, but your overall style and restoration tendencies matter too.

On the EV side, I've been told to really emphasize getting either 15 or 25 EV in the 3-5 rune game at least. Which leads me to believe the curve inside the game isn't 1 for 1 benefits with each increase. But actually playing, I still tend to feel that every few points matter somewhat even within that range.

[stoneychips]

Okay, after crunching the gargoyle through a couple Lair branches, I have to admit protection can be pretty awesome early on --

at least for somewhat armored characters and against things that appreciate it, like hydrae. (I did have a trident enchanted up to +8, though)