Fail rates for magic devices

[Marble Dice]

I don't see how exclusive 0% fail on devices for pure casters is much of an imbalance. By the time mages and priests can get 0% fail on spells (18/200 INT), they almost certainly already have rBlind. Depending on how devices are implemented, it might not be practical to get 0% fail on devices any earlier, which would make the spell-like effects while blind point moot.

Casters supposedly excel in the tactical/information meta-game, to make up for their frailty. This could be in-line with that perspective, so long as it's balanced properly.

[Pete Mack]

I'm really not a fan of 0-fail devices. In my opinion magic devices are not broken; they do need better UI for fail rate (as Eddie has written.)

EDIT: Modest tweaks (like improving behavior when confused or enforcing a max fail rate) might be in order. I think the basic asymptotic behavior at somewhere around 3% for good device users is about right. In any case, devices should definitely work better than spells for half-casters. 5% floor is ng.

[bebo]

Giving only the pure casters 0% fail devices creates a big imbalance

yeah, since mages and priests are so much better than all the other classes right now...

[buzzkill]

I only been half following this thread and don't feel like re-reading the whole thing so forgive me if I'm a little off, but here's my 2 cents...

0% failure SHOULD NOT be an option. I'd cap the the low end of the failure rate at 5%. However if anyone should be capable 0% failure, if should be mages. They have the skills and knowledge to use these devices better than any other class. I would like to see these failure rates be item-type dependent. Priests should have a better chance of using healing items, mages a better chance of destructive items. Perhaps rangers and druids nature attuned devices, and so on. I know it seems counter-productive since these classes are the ones who need those particular items the least, but it's precisely that familiarity with the spell-effects that should make them better at using such items.

[Atarlost]

nonzero fails on critical abilities are an example of fake difficulty. 95% survival in a situation is about as good as 0% survival in a long game with permadeath. I don't like situations where you can do everything right and still get killed. There should allways be some point in preparation where you made a decision wrong. Since using consumable monster detection isn't very viable that means there should in as many cases as possible be some escape that could have saved you if you'd bought enough of it. Failing to activate a staff of teleport while blind is just random and shouldn't happen if you level conservatively and don't play half-troll warriors.

[buzzkill]

nonzero fails on critical abilities are an example of fake difficulty.

Angband's very heart is the RNG. It giveth and it taketh away. It's the nature of the beast. Unlike 'classic' role playing games where you have to follow a storyline, Angband is completely random. I don't think the fake difficulty argument is a good one (though admittedly I just browsed through it).

95% survival in a situation is about as good as 0% survival in a long game with permadeath.

So the key would be to minimize the situations where you have to face those 1 in 20 odds, with scrolls, spells or other fail proof methods which already exist.

I don't like situations where you can do everything right and still get killed.

Then maybe this isn't the game for you. Even you do play perfectly you aren't guaranteed success. Thank you RNG.

There should always be some point in preparation where you made a decision wrong (that causes death).

And I'm sure there is. Nobody ever played a perfect game of Angband. Nobody ever will. I'm not saying the RNG isn't unfair sometimes. It is. Furthermore, it's unavoidable due to the random nature of the game. Sometime it's (randomly) excessively generous as well. Any serious attempt to satisfy this point would result in the game becoming much easier.

Since using consumable monster detection isn't very viable, that means there should, in as many cases as possible, be some escape that could have saved you if you'd bought enough of it.

Why isn't consumable monster detection (or anything else) viable? Aren't staffs consumable as well? I don't quite understand this (I cleaned it up a little with commas)... seems like a contradiction. I think I agree.

Failing to activate a staff of teleport while blind is just random and shouldn't happen if you level conservatively and don't play half-troll warriors.

Yes it should, rarely, but it should be possible. We'll just have to agree to disagree on that.

Perhaps we need to discuss what physical, mental, and verbal components are involved in the activation process of various objects.

[Magnate]

nonzero fails on critical abilities are an example of fake difficulty. 95% survival in a situation is about as good as 0% survival in a long game with permadeath. I don't like situations where you can do everything right and still get killed. There should allways be some point in preparation where you made a decision wrong. Since using consumable monster detection isn't very viable that means there should in as many cases as possible be some escape that could have saved you if you'd bought enough of it. Failing to activate a staff of teleport while blind is just random and shouldn't happen if you level conservatively and don't play half-troll warriors.

I totally agree with this. I don't see any advantage in either gameplay sophistication or enjoyability to preventing 0% fails, at least for the easier devices and the better race/class combos.

Having a look at device skills, they range from 46 (L50 HT Warrior with 14 INT) to 140 (L50 HE Mage with 18/220 INT - please note that there is currently no way to boost device skill other than by raising INT - there is no pval or other flag that can boost it - yet).

If we keep device difficulty on the 3-to-100 scale of depth (there is no need to do this, but it doesn't hurt as a first approximation), the difference between skill and difficulty ranges from 137 down to -54 for L50 chars (obviously much lower for L1 chars).

I propose that where skill minus difficulty > 100 you have 0% fail. This would allow HE mages to get 0% fail on devices up to level 39 (which currently includes wands of telOther but not rods of same). A dwarf priest (with max INT) would get 0% fail on devices up to level 22 (which still includes wands of telOther). It would mean that half-troll paladins couldn't get 0% fail with anything, and nor could half-orc or half-troll warriors. But other warriors could get 0% fail with the easiest devices.

This needs more thought, especially because a straight linear scale doesn't work, but I don't see any point in preventing 0% fail altogether, or capping success at 95%.

[Nick]

nonzero fails on critical abilities are an example of fake difficulty. 95% survival in a situation is about as good as 0% survival in a long game with permadeath. I don't like situations where you can do everything right and still get killed.

Nice link, but I don't think the argument quite holds here. The current position is that you should not rely on a device in a critical situation - if you need -tOther to survive, you have already done something wrong. In fact, I'm rethinking my original opinion that 100% success should be possible. I'll be happy either way.

[Zikke]

Having non-zero success rate isn't "fake difficulty" if part of the game makes you choose which area to focus on with gear. If you can spike INT to reduce your focus to zero, that's your choice and you will be rewarded with reliable escapes.

If you focus on spiking CON or DEX or anything else, or if you focus on balancing multiple stats, you make the sacrifice of not having guaranteed escapes.

[Magnate]

Hmmm. The tide of opinion seems to have turned against allowing 100% success - which is a shame because the more I think about it the more important I think it is. Well actually it's a linear scale that I think is important, and you can't have that without allowing 100% success (except by simply capping success at 95% or 99% or whatever - I suppose I could live with that).

Anyway, I've done a spreadsheet with a linear chance of success based on

(a * Skill - b * Difficulty + c) * d

with a = 1, b = 0.9, c = 80 and d = 0.7

This gives a 1st level HT warrior a 61% chance of activating a wand of magic missile, and a 50th level max-INT HE mage a 91% chance of activating PDSM. These are in the right ballpark that we've discussed, but it does mean that mages get quite a big chunk of the 0% fail territory (because other classes have lower INT and/or lower per-level skill increments). I'm ok with that, but I know that others aren't.

The spreadsheet is here: http://www.terminalarrogance.com/devices.xls - feel free to tinker with it and have a look at the results. Please remember one thing: we can change the difficulties of devices to suit the new algorithm, so it doesn't have to be perfect! (A rod of speed is the hardest non-DSM at a current difficulty of 95 - I'd recommend bringing the difficulty down a bit while leaving the depth/rarity alone.) The success chances are on sheet 3 - the devices are listed on sheet 2 and the race/class skill values on sheet 1.

Also bear in mind that being confused will result in a penalty to skill (at least -10 points and possibly as harsh as -50%) - and I am also proposing small penalties for several other states - so chances over 100% have some merit in that they still allow reasonable success while debilitated.

If you really don't like a linear approach, let me have alternative suggestions which (i) are not too complex (so preferably nothing more than squares or square roots) and (ii) do not introduce discontinuities (breakpoints are bad, ladies and gents).

[Magnate]

Having non-zero success rate isn't "fake difficulty" if part of the game makes you choose which area to focus on with gear. If you can spike INT to reduce your focus to zero, that's your choice and you will be rewarded with reliable escapes.

If you focus on spiking CON or DEX or anything else, or if you focus on balancing multiple stats, you make the sacrifice of not having guaranteed escapes.

I like this, but I think we would need to increase the effect of INT to make that happen.

Another possibility is to introduce a pval flag for "boosts device skill". Sang has this (at 5 skill points per unit pval) and it works well - it provides yet more difficult choices of kit, which suits your argument well.

[ClaytonAguiar]

Some thoughts

1) Yes, I think 100% should be allowed. How? Read "2" below.
2) The activability should be calculated by a formula weighting the character INT, class (professional afinity) and the complexity of the device. The weights for each factor? I leave this to you, who are more experienced than me. The complexity would be intrinsic to each device, to balance their usage by mid and lower level characters. Higher players with low INT would use better, but not the best items. Using the base INT (and not the modified) for all calculations would help balance the formula, as smarter players would use better items first, but everybody would use everything near the end of the game (when the INT is usually maxed out). It would be also interesting to find (rarely) some "less complex, but powerful" staves and wands (the complexity would get a random reducer every - just to give a number - 100 items of the same class).
3) Yes, and tied to the complexity of the device and the INT of the character, to let lesser wands even more useful to higher players. It's not usual, today, to find a level 40 warrior with a "wand of magic missiles". If they, critically activated, aimed three or more times the usual damage, it would be nice.
4) This would be the "art" of filling the weights of the calculation I talked about above.
5) They could be ranked, from the "simpler" to the "most problematic" handicap, to work as a reducer in the above formula. It would be a divisor or a subtraction, depending on the formula itself.

6) I think you missed this: magic devices usage calculation should predict "critical failures" or fumbles. These could be a simple charge reduction (you used the charge, but didn't see the result) or even more creative lists of possible consequences. I played MERP and Rolemaster for a while, and spellcasting there caused, even without failures, occasional "problems" (ambushes, for example).

[Pete Mack]

nonzero fails on critical abilities are an example of fake difficulty. 95% survival in a situation is about as good as 0% survival in a long game with permadeath. I don't like situations where you can do everything right and still get killed. There should allways be some point in preparation where you made a decision wrong. Since using consumable monster detection isn't very viable that means there should in as many cases as possible be some escape that could have saved you if you'd bought enough of it. Failing to activate a staff of teleport while blind is just random and shouldn't happen if you level conservatively and don't play half-troll warriors.

I read the article, and I disagree. The fake difficulty things are non-optional steps where you have a large chance of failing. Facing any one monster is almost always optional. (There are a few exceptions, like going down stairs, or recalling, or getting teleported.) Avoiding the situation of many low-risk events adding up to death is makes Angband much more interesting.

In the current model, if you want to rely on devices for evasion (never mind escapes), you have to play fairly aggressively, so you just won't use them too often in the game. If you let everyone get 0-fail with a wand of teleportation, everyone can play like a mage. If you give 0-fail to a Mage, it has minimal effect on game-play, as he already likely has 0-fail from the spell by the time he reaches cl 32. (The only difference is he has 0-fail when blind--this is a second-order effect, as it's already pretty straightforward to play so that you will never need 0-fail when blind!)

[bebo]

Nice link, but I don't think the argument quite holds here. The current position is that you should not rely on a device in a critical situation - if you need -tOther to survive, you have already done something wrong. In fact, I'm rethinking my original opinion that 100% success should be possible. I'll be happy either way.

Exactly. We already have 100% escapes - scrolls.

Sure, they require you not to be blinded, but detection and anticipation is an essential part of angband - for example if you see/detect agroup of dark hounds and don't have rBlind and are really low on hp you should escape right away, not wait until you are blinded with terminal hp so that if you fail to activate the staff of teleport you're dead. (lame example, i know)

The way i see it if you are blinded and are down at 5hp and are surrounded by monsters, you already made a mistake - you should have tried to use _teleport at least also the turn before to give some leeway (and come on, how many times are you going to fail two times in a row in critical situations?)

Another possibility is to introduce a pval flag for "boosts device skill". Sang has this (at 5 skill points per unit pval) and it works well - it provides yet more difficult choices of kit, which suits your argument well.

I really like this.

Might as well also suggest to reduce a bit the rarity of rods of speed/detection/mapping. As things are now, even if you find one, chances are you are not going to waste an inventory slot to keep it, mainly because it's very unlikely you will find another and recharge time is very high - thus not providing a viable alternative to the less powerful but plenty available comparable lower level objects (ie staff/potion of speed, spells/staff detect evil/invisible etc, staff/scroll of mapping, etc).

[Magnate]

Hmmm. The tide of opinion seems to have turned against allowing 100% success - which is a shame because the more I think about it the more important I think it is. Well actually it's a linear scale that I think is important, and you can't have that without allowing 100% success (except by simply capping success at 95% or 99% or whatever - I suppose I could live with that).

Anyway, I've done a spreadsheet with a linear chance of success based on

(a * Skill - b * Difficulty + c) * d

with a = 1, b = 0.9, c = 80 and d = 0.7

This gives a 1st level HT warrior a 61% chance of activating a wand of magic missile, and a 50th level max-INT HE mage a 91% chance of activating PDSM. These are in the right ballpark that we've discussed, but it does mean that mages get quite a big chunk of the 0% fail territory (because other classes have lower INT and/or lower per-level skill increments). I'm ok with that, but I know that others aren't.

The spreadsheet is here: http://www.terminalarrogance.com/devices.xls - feel free to tinker with it and have a look at the results. Please remember one thing: we can change the difficulties of devices to suit the new algorithm, so it doesn't have to be perfect! (A rod of speed is the hardest non-DSM at a current difficulty of 95 - I'd recommend bringing the difficulty down a bit while leaving the depth/rarity alone.) The success chances are on sheet 3 - the devices are listed on sheet 2 and the race/class skill values on sheet 1.

Also bear in mind that being confused will result in a penalty to skill (at least -10 points and possibly as harsh as -50%) - and I am also proposing small penalties for several other states - so chances over 100% have some merit in that they still allow reasonable success while debilitated.

If you really don't like a linear approach, let me have alternative suggestions which (i) are not too complex (so preferably nothing more than squares or square roots) and (ii) do not introduce discontinuities (breakpoints are bad, ladies and gents).

In case anyone is still following this issue, I've had a chat with Takkaria about it and he's relaxed about modifying the base and per-level device skill for different classes. This will enable the creation of a linear success scale without mages getting too much of an advantage. They will obviously still be the best, but not by the massive 30% margin that they are at the moment, and it shouldn't prevent warriors being reasonably proficient with useful detection devices etc., while allowing mages to excel with the endgame attack devices. I'm reworking the spreadsheet to this effect, and I'll post here when it's ready.

Having thought about it, it's very easy to put in a line that says

if (chance > 99) chance = 99;

... so if the consensus is that we should prevent 0% fail I'll put it in.