Item Identification

Recent changes to item identification made obsolete most of the content on this page. I will delete all old content at a later point. — dpeg 2014-05-05 00:58

Instead, I will start a general and abstract discussion about the id minigame altogether. Note that dtsund's Crawl Light fork has completely dispensed with this (all items identify right away) — this is always an option, but I am trying to argue that with some thought, we might be able to devise a system that's better than that and than status quo.

General analysis

We assume that items are unidentified when you find them (as in DCSS 0.14). You will want to know what they are as early as possible, for the following reasons:

  • Correct (tactical) decisions. This one is crucial: using unknown potions/scrolls in a fight is a desperate measure. But if you could have use-id'd and prevented death, it's your fault.
  • Avoid losing useful consumables (because of use-id). Typical examples are high value tactical items, such as scrolls of blinking/fog/fear, or rare strategical items, such as potions of cure mutation.
  • Avoid harmful effects (from use-id). The main examples are the potions of mutation/degeneration/decay.

(Introducing more reasons might be possible, and could make the id minigame richer.)

Here are my desiderata for the id minigame:

  • The optimal approach should depend on the specific situation of each game. This includes the following decisions:
    1. Use-id, scroll-id or hold out?
    2. If you are about to use-id or scroll-id, what to use or what to apply the scroll on?
  • The fact that items are unknown makes the early game more interesting, by providing decisions at a time where you have few tools.
Identification methods
  • Use-id. Use an item. This used to not guarantee identification; since 0.14, it does.
  • Scroll-id. Apply scroll of identification. Obviously, you can only do this if you possess these scrolls and have them identified, but that's generally easy (they're common).
  • Buy-id. Buying unidentified items from shops identifies them. Much less relevant than use-id and scroll-id, but a parameter we could change. (For example, would the presence of a shop selling consumables in D:3-6 affect your id strategy?)
  • Monster-id. Monsters using an item. Only applies to few items.

The main tension is between use-id and scroll-id.

Use identification

Use-id did not guarantee to identify the item prior to 0.14. Now it always does, which is a good change (otherwise, we're always prone to having players jump through hoops in order to increase identification chances). Also, there were cases of harmful item use with preventable drawback, as with the old scroll of immolation. Of course, this is equally out of the question.

Reasons in favour of use-id (and hence against scroll-id):

  • immediate information (this is the main draw)
  • item destruction (minor) — this is gone from Crawl by now. — dpeg 2014-06-03 14:04

Since your new knowledge is only useful if you have another item, using singletons right away seems questionable.

Scroll identification

Reasons in favour of scroll-id:

  • avoiding bad effects
  • keep item

It is interesting to observe that there are quite different identification schemes employed by players (also among very good players). That does not imply some particular approach is best, but I think it means that there is enough scope for keeping the id minigame, and trying to improve upon it. These differences seem to come from:

  • Which item category to scroll-id first? (Most players go for potions.) Note that as of 0.14, jewellery is not a valid target category anymore.
  • When to start use-id? (E.g. right away. Or only for items where you have three or more.)
  • Which items to scroll-id? (E.g. only singletons.)

What are some parameters that can affect your approach to use-id vs scroll-id early in a game?

  • threat/safety: the riskier your character's life, the more important is immediate survival, i.e. information gain (hence use-id)
  • amount of identify scrolls: if you find many, you may be best off to use them early (hence more scroll-id)
  • presence of early shops (allowing you to bypass use-id and scroll-id)

Currently, there seems to be bias for scroll-if of potions and use-id of scrolls. This is certainly not for the lack of good and rare scrolls (use-id'ing your only scroll of blinking is a painful loss, in my opinion) but rather the absence of harmful scrolls. Adding such scroll(s) would be a way to change that approach, if we want to.

Some suggestions

These are not necessarily good, and in any case they're independent of each other. For now, I just want to collect ideas that can affect decision making.

If we want to make holding out more attractive (i.e. if use-id is the correctly dominant strategy and we want to change that), here are some ideas:

  • certain monsters (e.g. uniques) could drop identified consumables
  • vaults (e.g. closets) with identified consumables
  • more early shops, perhaps also bad items shop
  • monsters use more items

If we want to create more choices regarding scroll targets:

  • Can use the scroll of identify to reveal the top card of decks. [This is a non-suggestion because it's already in the game. I am mentioning this because that was my first attempt to make scroll-id broader.]
  • Evoking unidentified wands uses up more charges; say 1d3 for a start. Players do already identify their top tier wands (healing, hasting), but with this change it might actually become a meaningful option to identify a wand of fire/cold/etc. you found early. For this, it is probably best to have the Evocations-based charge-identification test come before the charge deduction. This is an attempt to make Id-scrolls more relevant by having unidentified items perform worse — this idea can be extended to other items which can be used without being fully identified (currently, that's only potions, scrolls, wands, decks I think).
  • Change how randarts work (for simplicity, I only deal with jewellery here). That's a larger change (and an attempt to make Id-scroll use less trivial past the early game):
    1. Putting on randart jewellery will immediately reveal all negative properties (and these function as now).
    2. Positive properties are hidden and will reveal themselves after a while, one by one (based on experience and/or duration and/or depth of the item; the last is so that D:3 randarts are much quicker to self-id than D:26 ones).
    3. Announce that the ring/amulet is fully identified once the last positive property has been revealed.
    4. There are very many parameters to this, for example: the xp cost; should it be indicated (“this ring will take long to reveal its properties” etc.); give away number of positive traits in advance and so on. All of these are about having the player make a more or less informed decision on whether they should blow a scroll on the item or not.
  • A double-edged scroll (to make blinding reading more risky): Shaft self. This can be worked around by players reading on D:1, so we have to be more radical: the scroll can put you any unvisited level that's 1, 2 or 3 levels away from a visited one. (Problem: now this is indeed risky, but can also be life saving.)

I am interested in feedback in any of the statements I make above, and also about potential changes. Preferably comment here, but any other way is also fine. — dpeg 2014-05-05 02:43

My greatest complaint against the current system is that Scroll ID is a little lame. Getting perfect information kills the ability to make meaningful decisions. What if it was replaced by a system which encouraged more use ID? For example, instead of an ID scroll telling you a potion is type is definitely !Haste, it tells you the potion is either !Haste or !Mutation, and further ID scrolls do not give more information. This would either make the player wait until a shop formed, or until they were in a situation they have enough bad mutations that they're going to quaff !CureMut soon anyway. Another good example would be a potion which could either be !BeneMut or !Degeneration. I imagine most characters would think carefully about whether such a potion is worth quaffing. A third example would a potion which is either !Berserk or !Lignification - Do you wait until there is a enemy either could handle or do you quaff immediately?
I imagine the “pairs” of unIDd potions/scrolls would be randomized each game. — Reaver 2014-05-05 05:27
I really like that idea - but to take it a little bit further, how about 'classes' of items? So for example:
- benemut, curemut, mut could go in the 'mutagenic' class
- heal wounds, curing, poison, strong poison could go in the 'medicinal' class
- haste, slowing, and paralysis could go in the 'speed' class etc.
ScrollID-ing a potion would identify the class (This is a potion that changes your speed! [haste or slowing or paralysis]), so you'd have an idea of what it would do, but not exactly. — Bodrick 2014-05-06 09:09
I definitely see what you two mean, but your proposals would weaken scroll-id — isn't the current concensus that use-id is stronger? Whatever the answer, I once tried something in the opposite direction: make ?id more interesting by giving it more uses. (That use was identifying the top card of a deck.) This could also be expanded upon: wands might lose more than one change when evoked (say 1d2 or 1d3) unless identified.
I am little worried about one aspect of the partial id suggestion: it would lead to a lot of special casing. So perhaps best to only apply it to some interesting groupings, not force everything into a group. — dpeg 2014-05-06 16:52
Well, then invert the situation so that use-id gives a random result from a possible set, but once scroll-id'd the potion “stabilizes” and always gives consistent results. Uncertainty Principle in action! Note this means that different potions might resolve into the same potion type, e.g. a blue potion and a grey potion might both turn out to be haste. More generally, if this principle of multiple potions per type is allowed, that would certainly provide more opportunities for scroll id. — DracheReborn 2014-05-23 14:17

Bodrick: Could you explain the logic behind your “classes” proposal? I don't see any advantage at all to doing that.

dpeg: Use-ID is stronger for scrolls, from what I've seen, while Scroll-ID for potions. It's a nerf to Scroll-ID but that's because I think Scroll-ID is normally boring. — Reaver 2014-05-06 18:59


Often the player can deduce that some piece of jewellery is of some type. There's been a move to identify such items automatically, e.g. rings of regeneration. See also 1459 for a request to autoidentify rings of poison resistance under certain quite circumstances.

Speaking of rings of regeneration, they could cause Regen status when they activate (I'm still not sure if they consume food when at full HP). Then you could make rings of regen identify once Regen status kicks in. — b0rsuk 2010-09-09 21:58

Alternative approach

Here's an idea for moving in the opposite direction. It's a bit radical. It could apply to jewellery only.

  • Items (not item types) get randomized appearance.
  • Use only ever identifies a specific item, not the whole type.
  • Scrolls of identify identify the item and the type, i.e., they give the player enough knowledge about the item to recognize the type again in the future.

Tavern (new shop type)

It doesn't have to be a tavern, but in general a new kind of shop which contains no items at all. Instead, you can pay a small fee to learn the appearance of items you haven't seen. So you might learn how a scroll of vulnerability looks like, but you might not find one in a game.

I'm getting off-topic, but another kind of shop - Enchanter - could enchant items for you, including repair (-1 and lower), vorpalise, ewI, ewII, recharging, and possibly even other fixable brands like flaming, freezing (but these would be about as rare as randarts). What's the point ? Well, Enchanter could only rarely appear only at bazaars, so you either vorpalise that trident or miss the opportunity completely.

b0rsuk 2010-09-09 22:00

Possible new use-id cases

  • rings of poison resistance (some are 100%, others not really)
    • drinking potions of poison
    • eating poisonous corpses
    • being hit by poisonous attacks, both ranged and melee
    • resisting mephitic or poisonous clouds

Weapons' plusses

In midgame, learning weapons' plusses safely is usually a matter of hitting enough plants in Lair, which is tedious. Since weapon skill training is supposed to correspond to real training with a weapon, I propose a small chance of learning a wieldable weapon's plusses when the corresponding skill is trained. — ortoslon 2011-12-15 20:09

What about this: items already have a random value associated with them (currently used for tile choice). We could reuse it (or rather, a hash to prevent an info leak), to give the item an use-id difficulty – let's say, 0..10.0. If your skill is X, you instantly recognize the plusses of weapons you wield if their difficulty is lesser than X. The chance to recognize plusses on hit goes away. — KiloByte 2011-12-16 13:09
Using a weapon for an extended period of time and not identifying the enchantment is bad: the player will have a good estimate of the enchantment just seeing how accurate/damaging it is in combat. Maybe just have the enchantment always ID when you wield the weapon? Then again, I'm in support of *all* items always fully IDing when wielded/worn/read. — minmay 2011-12-17 04:20

As of 0.11, weapon pluses identify based on your weapon skill and a hidden item-specific number, but this still leaves the problem minmay mentions. I guess, perhaps we indeed should make weapons consistent with armour and fully id them on wield? Because you can write a script to calculate probabilities, and give you the enchantment after a small number of whacks against some foe. — KiloByte 2012-11-10 20:17

How about reverting to the old system, but just making it so that zero-XP monsters can't be used to identify weapons? This would make it so that you'd have to actually use the weapon to identify it: there really aren't any nonzero-XP monsters you wouldn't normally kill, barring monsters that are too difficult to kill, which you wouldn't use an un-ID'd weapon on anyway. If you expand zero-XP monsters to include summons, then there'd be no scummy way to identify weapons. — IonFrigate 2012-12-11 02:14
What would THAT be good for? You'd still be able to identify the weapon on any plant or summon, it'd just not show in-game. — KiloByte 2012-11-12 22:08
That's the case now. You can still write a script to get the weapon's plusses based on attacking plants or fungi, and this will work for a weapon you have zero skill with. There's no way to get around the fact that you can deduce plusses from doing damage a few times, and as long as Crawl is going to have non-threat monsters or spammable monsters, there's no way to ensure that players are “really” using the weapon while doing so. However, most players *really* don't care enough to do tedious crap like that: it's not like Alter Self, where a ton of tedious effort could leave you with a supermutant character. Early identification of weapon plusses is never going to be game-breaking.
And I think the old system was better because it actually felt like you identified the weapon by using it. It was very fixable by removing identification via zero-XP monsters. -IonFrigate 2012-11-12 16:26

Anti-spoiler countermeasures

This section in spired by the statement that “items with fixed appearances shouldn't pretend to be unidentified, it's pure spoiler” (OG17). This was in response to the fact that a potion of water is always described as a clear potion, a potion of porridge is always described as “gluggy” (though the color can change), and that potions of (coagulated) blood are always red (red or brown). These descriptions are obviously in accordance with the physical properties of water, porridge, and blood, but indeed a seasoned player would know that a clear potion is nothing to waste a scroll of identify on.

My observation is that there are other instances involving identification that might qualify as spoilers. For instance, by about D:4 or :5, the most numerous kind of scroll in your inventory is almost certainly of identify and the most numerous potion is almost certainly of healing. Unspoiled characters may not realize the relative safety of trying these items compared to the others in their bag. Further, if an unidentified scroll asks you to choose an item for its target, a seasoned player knows it's one of identify, enchant armour, or recharging, and so will try to use it on an unidentified wand or piece of armour, if possible. The spoiler potential becomes even greater once one or two of identify, enchant armour, and recharging are known.

Eliminating these spoilers completely would take some fairly extreme changes:

  1. To resolve “the biggest stack of scrolls in your inventory is of identify,” etc.
    1. Normalize the frequency of all scrolls/potions so that there are no “common” items. (May make the game significantly harder.)
    2. Start characters off with knowledge of the various common scrolls/potions. (May make the game significantly easier.)
    3. Auto-identify common scrolls/potions when lots of them have been sitting in your inventory for long enough. (A compromise, but potentially abusable.)
  2. To resolve “a scroll that asks for an item is one of identify, enchant armour, or recharging”
    1. ID the scroll *before* you choose the item to use it on. (I don't see any huge problems with this; you still lose the scroll if you can't use it effectively, which is likely if you're reading scrolls out of the blue.)
    2. Every unidentified scroll asks for an item! Behavior is otherwise unchanged. (Makes it a lot harder to distinguish item-buffing scrolls from other scrolls that don't ID on use. Seasoned players will still prefer to target armour/wands. Mostly, this would just be annoying.)
  3. Resolving “gluggy potions are probably porridge, red potions might be blood,” etc.
    1. Add something like “It looks like porridge” or “It looks like blood” to the description. (This is a gentle hint to unspoiled players, but said players probably don't even know about porridge or blood or what they're good for.)
    2. Outright identify porridge/blood/water on sight. (These items aren't all that common or game-changing anyway, so why not?)

Also, the first listed issue could be resolved with the “alternative approach” above, though this would tend to clog up starting character's inventories with repeat, though differently named unidentified scrolls/potions. Perhaps this isn't a terrible thing, as it encourages slightly more risky potion experimentation. I would rather that scroll/potion types still be identified on use, potentially leading to the combining of stacks. –rriegs

Knowing items rarity helps for identification. This is something you can learn from experience or from spoilers, but I don't think we should change it. We want to make crawl winnable without spoilers, and avoid features which give a big advantage to spoiled players. The semi-randomised potion is ok, it's very minor. And we'll have to live with the spoiled rarity, because there's no acceptable solution. And you only need a few games to learn it anyway. — galehar 2011-05-24 10:10
Logged in as: Anonymous (VIEWER)
dcss/brainstorm/item/identification.txt · Last modified: 2014-06-03 14:14 by dpeg
Recent changes RSS feed Donate Powered by PHP Valid XHTML 1.0 Valid CSS Driven by DokuWiki