suspicious of EV averaging and step-downs

[HardboiledGargoyle]

Hit-or-miss is determined, roughly speaking, by rolling EV twice, averaging the results, and comparing it to the to-hit roll. In practice, there is very little difference between doing this and simply comparing 1d(EV) vs 1d(to-hit)... unless EV is considerably higher than to-hit. Try it out on http://anydice.com. So Crawl incommensurately rewards EV considerably higher than monster to-hit. But monster to-hit is pretty high - it starts at +18 and goes up from there, using HD. For player characters evading monster attacks, this is hard to achieve. First off, you have to be at least in mid-game to get EV considerably higher than the to-hit of monsters that pose any threat. Second, the game makes it hard to get EV high enough for this to matter by implementing step-downs past 24 DEX and 30 EV. It's like the game creates a mechanic to make EV especially good, but only for really high EV, and it turns out to be too good, so the game creates step-downs to compensate.

Looks to me like the step-downs are a hack to keep EV roll averaging, an unnecessary mechanic, from becoming ridiculously effective. Why not get rid of both? I don't think much will change. Invisibility and repel/deflect missiles lower monster to-hit by a big proportion, and some other stuff like umbra also lower monster to-hit somewhat. These let you take advantage of EV roll averaging without actually having incredibly high EV. But I don't think that the mechanics of these special status effects are a good enough reason to keep step-downs and the mostly irrelevant EV roll averaging mechanic. Surely you can invent some other method of making invis/Rmsl/Dmsl especially good with great EV, if that is even desirable.

I presume the double-rolling EV also works for monsters, making high-EV monsters particularly hard to hit when your to-hit is lower. I doubt this is pertinent to many situations outside of bolt of inaccuracy, squashing early adders while wielding a GSC or wearing D:1 plate, and the ocasional monster with invis/DMsl.

Meanwhile, rather anecdotally, high AC is extremely effective. AC doesn't get any step-downs though. What AC gets instead is a nifty bonus in the form of GDR, and works on almost everything that can be dodged (doesn't do anything against some melee brands) and works on many things that can't be dodged. ???

[advil]

I'm having a bit of trouble following this. Have you played a high EV character through a big chunk of the game? Does this connect to some practical problem you're having with EV?

I mean, having done this myself I have my own complaints about the way high EV players play, but it's more about flat distributions for damage rolls leading to really swingy damage for EV-focused characters than anything to do with stepdowns. Also, if you think that EV can't be really overpowered in some parts of crawl as-is, try doing spider with no rpois with a high EV character.

[WingedEspeon]

I have played a lot of high EV characters and the EV stepdowns really hurt. A lot. (Especialy my HeGl with 44 dex in steam dragon Armour )

[chequers]

AC doesn't get any step-downs though.

Well, high AC gives spellcasting penalties.

[HardboiledGargoyle]

TL;DR version is that EV has two distinct mechanisms that cancel each other out, so why do we even have them.

AC doesn't get any step-downs though.

Well, high AC gives spellcasting penalties.

No it doesn't.

[GBudee]

It seems to me that the averaging effect was specifically intended to prevent EV from being too spiky (make it fairly consistent). Your analysis implies that this effect is not as strong as one might expect, but it's something of a happy medium I suppose, as stronger gaussianizing might make it very all-or-nothing.

[tabstorm]

TL;DR version is that EV has two distinct mechanisms that cancel each other out, so why do we even have them.

Crawl combat formulas: ?????????
As far as I can tell, changes to make the combat formulas simpler would require more work than is desired, and might lead to paralysis by analysis due to thinking about the formulae. I mean, if showing enemy damage or player damage is too much information, just imagine what it'd be like if people could actually figure out how the combat formulae worked in practice.

[duvessa]

making the combat formulas simpler makes it easier to balance them though, in the event that crawl wants to balance them at some point in the future (this is a part of why ranged combat mechanics were changed)

[Wahaha]

I find that EV is consistent damage reduction and does not really have "swingy" damage taken. Basically same as AC aside from gdr.
Edit: this is not addressed to the OP, in part because I read the OP after posting

[byrel]

Honestly, I doubt the stepdowns are related to the EV averaging. Undoubtedly the goal of the EV averaging was to smooth out the damage, but whoever did that screwed up. Double rolling and averaging generally has a big effect on the distribution of the result. But when you immediately threshold it, the difference between the probabilities is fairly minimal. The actual impact of this is more to provide a monster-specific pseudo-breakpoint in EV, after which EV has very high gains vs that monster. Note that this theoretically makes draining more useful for EV characters.

I don't think that's ideal, but in any event, the differences are small enough to not matter much. I'm generally in favor of removing stepdowns, because I think they're rarely the most efficient way of solving the problem at hand, tend to be spoilery, and tend to frustrate/mislead spoiled players. I don't think it would be bad to remove EV stepdowns. I also don't think it would be bad to remove the EV roll averaging.

[WingedEspeon]

The EV averaging might get out of hand without the EV stepdowns. A HEGl (possibly^Chei) with 70-80 EV would almost never get hit. That said I have been wanting to make a DCSS fork to try out just removing EV stepdowns (including the dex-EV stepdowns).

[byrel]

The EV averaging might get out of hand without the EV stepdowns. A HEGl (possibly^Chei) with 70-80 EV would almost never get hit. That said I have been wanting to make a DCSS fork to try out just removing EV stepdowns (including the dex-EV stepdowns).

And if you have 80 AC, you'll almost never take meaningful amounts of damage.

Ignoring stepdowns, EV scales as EV~=Dodging*Dex/20. Even with lucking into all points in dex, and choosing dex at all upgrades an HEEn has 30 dex at level 27. With maxed dodging, that works out to ~40 EV. If you worship Chei, that goes up to 60 EV. 60 EV is a bit absurd, but it's still not really OP for level 27 with max Chei piety, and not wearing plate (as Chei characters are prone to.) There are still threats, and you'll still get hit occasionally.

Also, to clarify how test_hit() works: 2d(2*EV-1)/2-1d(to_land). So the EV roll max is 2*EV-1, and averages EV-0.5. The OP stated it was 2d(EV)/2-1d(to_land) which isn't correct.

Sample values of to_land:

  Code:

Monster        to_land      60EV     40EV  40EV No Avg
------         -------      ----     ----  -----------
Jackal          19.5        2.0%     4.7%     13.3%
Hydra           37.5        7.2%    15.5%     24.7%
Orc Warlord     55.5       15.4%    32.9%     36.0%
Caustic Shrike  45.0       10.0%    22.6%     29.0%
Juggernaut      68.0       22.3%    43.1%     43.7%
Orb Guardian    55.5       15.4%    32.9%     36.0%
Antaeus         73.0       25.5%    46.9%     46.8%

All percentages ignore the small chance to hit and miss regardless of roll. I also included a column of chances calculated with simply 1d(2*EV-1)-1d(to_land) (OPs suggestion). In retrospect, I appreciate the way it actually trivializes the damage from trivial monsters. A Hydra shouldn't be a threat to a level 27, dodging 27 character.

Basically, 60EV still leaves you with a 22% chance of getting jugged. That's (IMHO) quite reasonable for a character who is a Chei full dodge-tank.

[tedric]

Well, high AC gives spellcasting penalties.

No it doesn't.

To make this point clearer: Spellcasting penalties come only from body armour and shields. AC by itself is unrelated to spellcasting.

[sweetrabies]

I mean, if showing enemy damage or player damage is too much information, just imagine what it'd be like if people could actually figure out how the combat formulae worked in practice.

The game already does show you enemy damage, just in a way that forces you to keep track of your change in HP between turns instead of simply showing a number. Player damage is hidden because I guess trying to figure out what's better between ##### sandblast or ### stone arrow or a long sword of electrocution is *~*~fUn~*~*

[HardboiledGargoyle]

In retrospect, I appreciate the way it actually trivializes the damage from trivial monsters. A Hydra shouldn't be a threat to a level 27, dodging 27 character.

The thing about that is, you don't care how good your defenses are against the harmless monsters, because they're harmless anyway and you don't care if the proverbial rat lacks an extra-strong penalty to its accuracy.

I suggested removing not only EV averaging but also stepdowns, so you'd likely be comparing 40 EV with averaging not to 40 EV without averaging but to something like 45 EV without averaging.