Could someone edit in the 0.7 AC change, so we can bring the discussion up to date with CRD? I've heard both 5/2 and 3/2 for the formula, despite what the commit said. On another note, I was wondering if we intend there to be any guaranteed damage reduction for special cases such as ozocubu's armour, statue form, etc. Some of that I'll take to the transmutations page, but I thought it worthwhile to toss that question out here since the mechanics are so drastically different. — lordsloth 2010-04-08 17:34
Unfortunately, I can't. — dpeg 2010-05-26 17:37
Here are some recent ideas from irc about AC. One problem is that heavy armours got hit a bit hard by the nerf (but probably not as much as some complainers think). Another problem is that you can still get absurdly high AC in a robe. — dpeg 2010-05-26 17:37
An OpenOffice spreadsheet file with the old and proposed new formulas, courtesy of syllogism (many thanks!): armour.ods Excel: armour.xlsx
For the sake of quick discussion, the original discussion is still on this page, at the bottom. Here, I reorganised everything. The title “AC nerf” doesn't do the modification justice: we will be changing all of AC, EV and SH. The top part is by dpeg 2010-01-16 23:34. Comment!
Okay, first, I am *double not* suggesting any change to SH, which I believe is already revised somehow (but have not checked.) I am only suggesting a change to the *penalty applied by shields to other things*. — drpraetor 2010-01-17 04:33
I think AEVP should be used exclusively wherever Evasion matters. (I.e. always refer to AEVP, never to EVP or Dex directly.)
syllogism points out that the Old formula cannot be valid as is — there might be other things missing?
drpraetor looks right to me! See the post far, far below. — drpraetor 2010-01-17 04:33
Old: AEVP = EVP * MAX(30, 60-Armour * Str) / 60 New: AEVP = (EVP + MAX(0, 3*EVP - Str) ) * (45 - Armour) /45 New Again: AEVP = MIN ( (EVP + MAX(0, 3*EVP - Str) - Armour / 4, EVP * 2 / 5)
OKAY WHY AM I PROPOSING THIS CHANGE? Two reasons - one, this means that relatively low levels of armour skill will still reduce the EVP of relatively light armour like chain mail. Under “New” you need to get your Armour skill up to 8 before the EVP of chain mail is reduced by even 1 point. The *effect* of this is to reward medium armor like chain mail, especially for characters with relatively low armor skill. Meanwhile, “New:” and “New Again:” will give roughly the same effect on crystal plate mail. — drpraetor 2010-02-14 20:24
It is intended that some Strength is needed to overcome the evasion penalties for the heavier armours.
Armor should cap the effectiveness of the Dodging skill to some extent as well. Playing with the spreadsheet, a Mountain Dwarf with 15 Dodging wearing Banded Mail gets only 1 more EV at 10 Armor Skill than at 0. I realize this was true with the old formula as well, but it seems like if you aren't used to wearing heavy armor your dodging would be hampered significantly, even if you're good at dodging to begin with. I think I'd like to see a low Armor skill directly affect the EV gains of Dodging if possible, rather than just partially factoring in as part of the whole. It should at least hamper your Dodging abilities up to Armor-EV-Penalty = Armor Skill, I'd say. — twinge 2010-01-21 15:19
TEV is the number shown on the main screen as EV (Evasion).
Old: EV = 10 + Size + MIN[0 or (MIN[ (7+Dodge*Dex)/(20-size) AND Dodge*(7+size)/9) ] - (30*EVP+15)/Str [if heavy armour]] - EVP - Shield Mod + Bonuses New: EV = 10 + Size + (7 + Dodge * Dex)/(20 + AEVP - Size) - AEVP - ASP + Bonuses
The New formula is by drpraetor. It now has centralised penalties. Most importantly, the effect of Dex EV is now uncapped! This is a huge boost for Dex and may need nerfing soon — but I suggest not for 0.6. The term which could run away is (7 + Dodge * Dex)/(20 + AEVP - Size) — this could be a stepdown (better than a cap, probably).
ASP is the value subtracted from EV by virtue of you using a shield. It is *not* the SH value, which rises (instead of falling) with shield skill. — drpraetor 2010-01-17 04:33
Old: ASP,SP = XOM's choice New: ASP = SP - 1d(Shields) / (5 + Size Factor) SP: 1 for Buckler, 3 for Shield, 5 for Large Shield
This is the number shown on the main screen as AC (Armour Class). As you can see, the “AC nerf” is just a numerical nerf at this point.
Old: TAC = sumAC * (15 + Armour) /15 + enchantments + modifiers New: TAC = sumAC * (23 + Armour) /23 + enchantments + modifiers
Analysis: Obviously, you now need 23 Armour skill to double your AC from armoury. At the high end, you can still easily get TAC of over 50.
GDR is only relevant for melee damage. Should be only a function of body_base_AC and Armour skill (i.e. not of enchantments or other modifiers).
Old: GDR = MIN(50, 2*(body_base_AC + Armour)) * max_dam/100 NewR: GDR = body_base_AC*(13+Armour)/17 * max_dam/100 NewRRR: GDR = MIN [max_dam / 2, MAX [stepdown (5, (body_base_AC - 2)*(13+Armour)/17), body_base_AC*(13+Armour)/17 * max_dam/100] ] stepdown (X): halve the pts > X, then halve the pts > X*2, until pts >= X.
WHY AM I PROPOSING THIS CHANGE? Low level fighter survivability. The old formula provided an early fighter with 20% GDR, roughly. The new formula provides a starting fighter with 5% GDR. The upshot is that we want a system that provides the Armour 8/Chain Mail regime with 3 pts of GDR against a hobgoblin (dam 10), but doesn't provide 26 pts of GDR against an axe-wielding frost giants (dam 52). So, my target values for that formula where: robes and leather give negligible GDR, scale mail with starting armour skill gives *some* GDR against any damage amount, and GDR against said frost giants won't exceed 14*40/34 ~= 16 pts. The stepdown is fairly steep: 12*40/17→28→17→13 pts. So around 40 pts of max damage the % component will be bigger than the non-proportional component. — drpraetor 2010-02-14 20:24
The two new formulas are still competing.
Old Spell Mod: -20 * EVP + 15 + Armour * Str / 15 + Shield + Race Mod Old Shield: Buckler is -5, Shield is -15, Large Shield is -30 (-20 for large races) Old Race Mod: 20 for elven, -10 for dwarven, 10 for race match New Spell Mod: -25 * (AEVP + ASP) + 20 + Race Mod New Race Mod: 25 for elven, -15 for dwarven, 15 for race match (plus an indirect effect if you get virtual armour skill)
Note that the new formulas have AEVP and ASP instead of EVP and the fixed Shield values. (Here, the random effect of ASP could be problematic — we could use MASP, Mean Adjusted Shield Penalty instead.)
The importance of Strength seems a little too high in some cases. Dropping from Good with the old formula to useless with the new is rather extreme; the formula seems to get too overzealous once you hit any armor with a -5 EV penalty or higher. I agree that casting should be harder in these heavy armors than it is now, but it should tier down more slowly than the current proposal. Str 10 in Storm Dragon armor can't cast anything Level6+ worth a crap even with pretty good Spellcasting, Spell Skill, and Armor levels - yet at only 15 Strength, the success rate is actually *higher* than it was with the old formula. It should be hampered at low Str, but not gimped. — twinge 2010-01-21 15:19This sounds to me like “working as designed”. — lemuel 2010-01-21 18:37
These are used verbatim from drpraetor's proposal below. This is all New, as the old formulas are too complicated to write down. Someone with a grasp on the combat mechanics please double check!
To-hit = 1d(AEVP) + 1d(ASP) To-hit for 1.5 handers = 1d(AEVP) + 2d(ASP) To-dam = none To-dam for 1.5 handers = -1d(ASP) [reducing base damage] Delay1 base_delay = MAX [base_delay, 1d10 + 1d(AEVP)] base_delay for unarmed = MAX [base_delay, 1d10 + 2d(AEVP)] Delay2 final_delay += MIN [1d(AEVP), 1d(AEVP)] final_delay for 1.5 handers = MIN [1d(AEVP) + 1d(ASP), 1d(AEVP) + 1d(ASP)]
The newest version of Samuel's proposal: Now updated to my full proposal. — drpraetor 2010-01-16 14:52
Old: AEVP = EVP * MAX(30, 60-Armour * Str) / 60 New: AEVP = (EVP + MIN(0, 3*EVP - Str) ) * (45 - Armour) /45
(Adjusted Evasion Penalty)
Old: ASP,SP = XOM's choice New: ASP = SP - 1D(Shields) / (5 + Size Factor) SP: 1 for Buckler, 3 for Shield, 5 for Tower Shield
(Adjusted Shield Penalty)
Old: totalAC = sumAC * (15 + Armour) /15 + enchantments + modifiers New: totalAC = (sumAC + enchantments) * (36+Armour) /36 + modifiers NewR: totalAC = (sumAC) * (23+Armour) / 23 + enchantments + modifiers
I support the NewR formula. — dpeg 2010-01-08 01:37
Old: GDR = MIN(50, 2*(body_base_AC + Armour))*max_dam/100 New: GDR = MIN(50, (body_base_AC+enchantments)*(36+Armour)/36/2)*max_dam/100 NewR: GDR = body_base_AC* (13+Armour)/17 * max_dam/100 NewRR: GDR = MAX [body_base_AC squared/100 * (3 + Armour) * max_dam/200 ; MIN [max_dam/2; (body_base_AC - 3) * (5 + Armour) / 10] ]
The NewR formula is by me. GDR should be just a function of base AC, in my opinion. With crystal plate (14 base AC) at maximal skill, you get a fraction of 33%. — dpeg
I think GDR should be a trifle more generous than that, especially for high armour skill. In fact, if we just double the armour skill component… — drpraetor 2010-01-16 14:52
I think it'd be good if CPM was better than GDA for strict AC and GDR purposes. — dpeg 2010-01-16 16:32
NewRR is somewhat more complicated, so I should explain. In the late game, it means GDR starts to matter for plate mail (which maxes at GDR 15%), and that crystal plate maxes at GDR 25%. In the early to mid game, fighters get a GDR of 2-4 against most things, which also matters. So GDR isn't huge but it matters all the time. — drpraetor 2010-01-16 16:35
I don't understand why the NewR formula doesn't suffice to achieve this: with plate mail and a skill of 4, you'd get 10% reduction already. Which should be noticeable. — dpeg 2010-01-16 23:34
Old: EV = 10 + Size + MIN[0 or (MIN[ (7+Dodge*Dex)/(20-size) AND Dodge*(7+size)/9) ] - (30*EVP+15)/Str [if heavy armour]] - EVP - Shield Mod + Bonuses New: EV = 10 + size + **(7 + Dodge * Dex)/(20 + AEVP - Size)** - AEVP - ASP + Bonuses
Couple of changes here. First, I've centralized all the penalties. Second, I've uncapped EV as a function of dex. The bold part of the formula may require stepdown or something. — drpraetor 2010-01-16 14:52
Old Spell Mod: -20 * EVP + 15 + **Armour * Str / 15** + Shield + Race Mod Old Shield: Buckler is -5, Shield is -15, Large Shield is -30 (-20 for large races) Old Race Mod: 20 for elven, -10 for dwarven, 10 for race match New Spell Mod: -25 * (**AEVP** + ASP) + 20 + Race Mod New Race Mod: 25 for elven, -15 for dwarven, 15 for race match (plus an indirect effect if you get virtual armour skill)
These formulae deal *only* with the penalty for armour and shields, I'm not reproducing the rest of the spellcasting math here, but suggest leaving it alone. The major change is to compare the bold text in the two versions. Armour skill used to have an incredibly minor (1EVP equivalent, at ~20 skill, 2EVP for a very strong character at 27 skill) effect. Since we're now using the main EVP math instead of a site-specific modifier, this is no longer the case. A second change - in the previous system, you could get a “free” buckler *and” a “free” leather armour. Under this system, it's either-or. Also, especially at low shields skill, the spellcasting penalty for shields is somewhat larger all around; we could go (-15 * ASP) instead if we wanted shields to have a lesser effect on spellcasting than on evade, but this never made a lick of sense to me. — drpraetor 2010-01-16 14:52
This seems very good to me. — dpeg 2010-01-16 23:34
And for the combat penalties, I'm not giving the "old" versions because they're all over the map: To-hit - 1DAEVP + 1DASP for 1.5 handers - 1DAEVP + 2DASP Damage - none for 1.5 handers - base_damage -= 1DASP Delay1 - base_delay = MAX [base_delay, 1D10 + 1DAEVP] for unarmed - MAX [base delay, 1D10 + 2DAEVP] Delay2 - final_delay += MIN [1DAEVP, 1DAEVP] for 1.5 handers - MIN [1DAEVP + 1DASP, 1DAEVP + 1DASP]
This is actually not so different from the status quo, but is a massive simplification (the status quo involves a huge host of coin-flips, especially for large shields.) One fundamental change - the base_delay modifier is intended to discourage plate mail + quick blade more than is currently done. — drpraetor 2010-01-16 14:52
On the current AEVP system: In function player_evasion() in player.cc, under “Handle main body armour penalty”: first, we get the EVP. then, we modify by size (which I haven't mentioned above, possibly should?)
…then, we *reduce the EVP by*: (Armour * Str) / 60 but, this is *capped* at the base EVP / 2. So, since EVP / 2 = EVP * 30 / 60, we get: AEVP = EVP * MAX[30, 60 - (Armour * Str)] / 60. No? Do I need to check the current trunk for changes to this? — drpraetor 2010-01-17 04:33
Samuel's comments:
So, +8 GDA *old* gives 21 AC + 13/15 AC per point of armour skill.
+8 plate mail *new* gives 21 AC + 21/36 AC per point of armour skill.
That's a nerf of ~7 AC at high skill, which is quite significant.
OTOH, +8 SDA *old* gives 15 AC + 7/15 AC per point of armour skill.
+8 SDA *new* gives 15 AC + 15/36 AC per point of armour skill.
This *sometimes* costs you a whole point of AC, depending on armour skill.
Now, I agree that medium armour currently *sucks* and you never use it, but I think this is a bit too strong, so I agree with other posters that enchantment bonuses should be moved back out of the denominator.
Solving for the same proportional nerf, we get:
This is an exactly-equal nerf on +8 GDA, and a proportional nerf on SDA, but remember that the EV penalty for -2 EV heavy armour is *way* less severe now, so I think this is better.
kilobyte's comments:
Having enchantment inside the parenthesis means that once you train Armour, get boots+helmet+gloves+cloak and grab a bunch of EA scrolls, you can drop heavy armour altogether and still enjoy 75% of AC, with full dodging and full spellcasting. Thus, only the NewR formula is good. Same applies to GDR if light and heavy armour are to be equalized.
dpeg:
I agree. The idea to get enchantments inside the brackets was an attempt to treat base AC and pluses on an equal footing. As you figured out, this is not a good idea. Also, that was more for consistency and did not make up the “AC nerf” or anything.
What about making GDR come only from the base AC of your armor while adding some guaranteed Dodge- or Dex-related damage reduction (from rolling with blows), and allowing these two to stack? A heavy armor user could enjoy both sources, but an EV character would receive only one. This would also provide an incentive not to wear the heaviest armor you can as soon as you find it, and with the right formula, this would make “medium armor” roles feasible.
dpeg:
Apart from getting enchantments out of the GDR formula again or at least outside of the brackets (as proposed by kilobyte), I believe that GDR (which itself is generally overrated) should be a function of only base AC. I think that's simple and exactly what one would expect. If you wear light armour, there is no guaranteed damage reduction, no matter how evasive you are.
GDR from base armor only seems exactly right to me, for whatever it's worth. Simple and transparent is always good. Note that the new EV penalty formula already creates a role for medium armors, by making a high Str essential for the heaviest armors.
I've adjusted the text above inline to reflect the full proposal as best as I can.
Lemuel, I'm not sure needing str vs AEVP will matter much at all. As far as I can tell, AEVP is only used here, in melee accuracy (where it gets drowned by a million other factors), and unarmed delay.
I disagree with the presented AC figures: All AC should be created equal. Someone in a +8 robe should have the same AC as someone in a suit of +0 plate mail with the same armor skill. AC from mutations should also count the same way. This would make well enchanted or artefact versions of the lesser armours worth using, rather than pointless curiosities. The formula for total AC could be something as simple as total_AC = nominal_AC_from_all_sources * (40 + armour_skill) / 40. You would never double your AC, but it would be made up for by counting enchantments and ac from other sources. 50 AC becomes the realm of maximally enchanted crystal plate mail, scaly creatures, and people giving up ring slots or weapon brands.
GDR could be handled similarly: GDR = stepdown_value(AC, 15, 15, 60, -1) * max_dam / 100. This causes GDR to equal 15% at 15 AC, 30% at 45 AC, 45% at 105 AC and 60% in the unlikely event someone reaches 195 AC, with no hard cap.
I'll let others comment on the philosophy (although I think this would have the effect of making plate mail irrelevant,
while I play in robes all the time.) But that is *not* the only place AEVP is used!
AEVP is used for spellcasting and of course for EV itself.
Yes, the combat penalties associated with wearing armour are not that severe. Never have been.
But under this system, there *is* a reason to wear “lesser” armours. Lesser armours have lower EV penalties, and under these
rules it is possible (even highly desirable) to use both Dodge skill and Armour skill at the same time.
Irrelevant. EV is useless unless you have quite a lot of it. Take, for example, a suit of chain mail. With no armor skill and 12 str you have an EV of 6. Your armor skill effectively stops training at 12, giving you an AEVP of 2 (only barely, if AEVP were a float it would be practically 3) and 8 ev (your average end game melee attacker will hit 84.1% of the time). A HuCr might wish to make use of this: They start with all the str they'll need and both cast as well as melee. If the rng is nice about their stats from level and they split their elective points evenly between int and dex, a HuCr will have 13 str, 19 int, 18 dex. In a 3 rune game, they are unlikely to get much over 15 dodging, due to how slow it trains. This yields 20 EV, which has a 64.1% chance of getting hit by your average endgame melee enemy. An electric golem slinging lightning bolts (fairly inaccurately) has a 77.5% chance of hitting, vs 92.4% at ev 8. All this skill and stat dilution has gained is ~20% less chance of getting hit. Should the game go on long enough to hit 27 dodging, an EV of 30 can be obtained, getting hit 47.5% of the time vs melee and 65.1% of the time vs beams, while having (under AC new and GDR newR) 32 ac at maximum enchantments (nearly half of this comes from enchantments) and 8% GDR. You dodge a few more hits, but you take much more damage from them than if you had used a heavier piece of armor: crystal plate with 21 armor skill (not difficult to have during a 3 rune game in armor this heavy) yields 52 AC with 28% GDR.
Skill dilution isn't that big a deal considering the small XP investment. Furthermore, it's hardly fair to compare a plain chain mail to a crystal plate; even an okawaru worshipper won't usually see one while better pieces of combo armor are readily available. Obviously a piece of regular chain mail will not remain useful very long, but the new formulas do make combo more viable, even discounting spell casting. GDR has very little effect overall.
dpeg 2010-01-18 13:03
I am not in a position right now to compare numbers. Just two remarks: Training will be changed, and I guess this is for Armour and for Dodging, not sure about Shields. And the other one, the first new formulas (by me) tried to put armour enchantments on the same level as base AC. This proved very tricky, essentially because enchantments can be very large relative to your base AC (especially for robe wearers). This is a bit like in Nethack. It seemed best to move enchantments back to “other boni” (a la rings or “of protection”).
danei 2010-06-15 15:55
A thought about AEVP, combat penalties, and Armour skill; currently (in 0.6 and in 0.7 trunk), the reduction in AEVP from Armour skill functions as a percentage of your armour's base EVP. This leads to a situation where, for example, a medium sized character (with adequate strength) requires 11 armour skill to reduce AEVP by 1 in scale mail, and only 6 to reduce AEVP by 1 in plate mail. This may (or may not) make sense in terms of actual EV reduction, but when AEVP is also used in combat penalty calculations, it conceptually seems rather odd that one should have to train almost twice as much to improve combat abilities by one increment in scale mail than for the same increment in plate, and in terms of game effects, players in the early and early-mid games in medium armours see zero or very little effect from armour skill on anything (melee combat, AC, or EV) besides possibly spellcasting. Perhaps something like an extra roll for AEVP to be cut in half for the purpose of each armor-related combat penalty calculation would be appropriate, or something similar to that. The chances of it occurring could rise with increasing armour skill and fall with increasing base EVP on armour. Something like that would retain the multiplicative nature of AEVP while providing more immediate benefits of training armour skill to users of medium armour.
My thought is that we may be going a bit too in depth to formulas without having an idea of how methods of combat should work in gameplay. Here's how I see these stats:
HP: This is all about being big and bulky and tough. It's not affected by your equipment, but it affects the other defenses regardless because of how much buffer it provides.
AC: This defense should excel against large numbers of weaker foes, but be good most of the time. It should only be weak against attacks where it makes you easier to hit without providing any further protection (such as ghostly attacks that pass through your armor). It should always reduce the damage received.
EV: This defense should excel against single opponents, particularly ones with slow or inaccurate attacks. It should have the disadvantage that it requires mobility - it should absolutely tank when you are paralyzed, netted, constricted, or so on. It should include glancing blows - you may be able to move somewhat, but not completely, out of the way.
SH: This defense should excel against single opponents and ranged attacks, but restrict your damage output/spellcasting. It should be weak against stronger or larger attacks (such as giant clubs or large rocks). It should also be weak against being surrounded (not strictly multiple attackers, but attacks coming from more than one side). This is a defense that should be either be all-or-nothing: you interpose the shield in time, or you don't.
We should examine how the player gets their defense score, for each defense. EV represents chance to not be hit, but it can come from something like Phase Shift as well as skill in Dodging. The “guaranteed AC” proposals that consider base armor value most important are one part of this - but we should also look at, for example, Dodging having a chance for glancing blows which doesn't depend on EV directly but instead on your amount of Dodging skill. (This gives an incentive to train both Dodging and Armor, for instance: you cannot ever jump out of the way, but you can roll with blows and so on.) Shields have their own base shield value, as well, which could be put to similar uses. Bulky shields could provide better protection against larger attacks, even if the SH value is the same between two characters, just as a factor of protecting a larger area.
As for armor affecting spellcasting: it may be beneficial to move to a low independent failure chance, which is reduced by STR and armor skill. This way, you can't get super good at spellcasting and then hop into armor once you don't need any more success rate. Doing so would tank your success rates - still usable, but a flat penalty making success unreliable.
doy's formula:
11:56 <+doy> use something like... (11 - min(8, armour/6 + min(4, (str - 12) / 6))) * evp 11:56 <+doy> maybe?
eronarn's formula:
f(X) = 5 * EVP-(X/3) Where X is either STR or SKILL. The actual failure rate is: Failure Chance = max(0, f(STR)) + max(0, f(SKILL)) + (min(0, f(STR)) + min(0, f(SKILL)) / 2) In other words, your chances of failure are a net of 10% per EVP with no STR/SKILL. If you have enough STR or SKILL that you have a 'negative' penalty, half of it is applied as a bonus to the other factor. This way, two points of STR makes up for one point of skill, or vice versa.
I think a combination of these two with a bit more tweaking is what we actually want. From my formula, I think the hard lower limit (so you're never going to get better than a certain amount at casting in plate mail) is important, and I think scaling strength bonuses from some value greater than 0 would also work better. The idea of sharing between strength and skill at a penalty like in Eronarn's formula is a good one, I think. In addition to these, I think that some sort of stepdown would be useful here, so that the first few points of strength and skill are worth more - we don't necessarily want it to be difficult for an early character to be able to pick up leather armour or ring mail without being crippled, we just want to make it a lot more challenging to be able to cast things in the heavier armours without a serious training commitment. This should help make mid-range armours more viable for certain playstyles. — doy 2010-10-04 19:03
I love these ideas. How would a system like this interact with the way miscast severity is chosen? — danei 2010-10-04 19:08Personally I'd say that miscast severity depends on your skill with magic, rather than your armor. We can use a message similar to what we currently do with missing an attack due to wearing armor - that your armor got in the way. — eronarn 2010-10-04 19:25