Designing Secret / Hidden Features

Here I will define a secret as any feature or area of a level that is hidden from the player. This includes all secret doors, as well as any room that can only be accessed by teleporting or digging. Some secrets have hints, others do not.

I believe that some vaults make acceptable use of secrets, and others do not. It is hard to draw the line here: some people would argue that all secrets constitute spoilers (and are therefore bad), while other people would argue that secrets can be acceptable under certain conditions. I will use this page describe I think constitutes acceptable usage of secrets.

I will also propose that we should modify vaults which go against the principles I describe here. The goal would be to recuce (and ideally, eliminate) the need to read .des files (or use ashenzari) in order to find loot (I emphasize need - some players may want to read the level map anyway, but this should never be necessary in order for an intelligent player to find everything).

This page also contains a lot of my opinions on level design, and could be useful reading for other vault makers.

For a secret to be unacceptable, I believe that it must satisfy two conditions.

(1.) The first condition is that it must have no clear hint (I will go over what counts as a “clear hint” a bit later). An example of this would be 10×5 box full of loot, with no windows, no doors, no stairs, and no indication that this box is made of anything but solid rock. The only way for a player to discover the secret would be to luckily teleport into it, to luckily dig into it, or to study the .des files. Ashenzari also works, but that is a special case which does not apply to most characters (and does not help the issue of spoilers).

(2.) The second condition is of being important. The term “important” has a specific meaning here: a secret is important if it conceals something that could help the player in a significant way. In almost all cases, this is loot. In at least one, it includes a rune. Thus, a secret is not important if it is just an empty room. It is also not important if it is only a shortcut between two non-secret areas, with no strong advantage gained by taking the shortcut.

If this definition isn't good enough, then another way to look at it is this: if a normal player might be encouraged to look at the .des file to find the secret (having been told there is one, and what is in it, but not knowing where exactly it is located), then it is an important secret. On the other hand, if the player knows there is a secret but does not think it's worth looking it up in the .des file (for example, the player knows it is just an empty room), the secret probably isn't important.

If both (1) and (2) apply, a secret is unacceptable and should be edited.

With the above in mind, I can identify three cases of acceptable usage of secrets:

(I.) If both conditions (1) and (2) are false, then it is acceptable: it has a clear hint, and it leads to nowhere particularly important anyway.

(II.) If (1) is true and (2) is false, then it is acceptable: While the secret lacks a hint, the player would not want to (or need to) find - it is unimportant.

(III.) If (1) is false and (2) is true, then it is acceptable: it has a hint, and leads somewhere important. Even though the secret is important, a clever player will get the hint and find the secret. This is the trickiest of the three cases, since it hinges on what counts as a clear hint.

Here things get a bit muddier. I do not think it is easy to precicely define what a “clear hint” is. I think the best we can do is look at examples. I will be using vault syntax here. Please read syntax.txt if you do not understand the notation. Note that all of these examples are valid vaults, so you could try them out in-game if you want.

NAME:   example_1
DEPTH:  D:1-27
MAP
xxxxxx.
xx***x.
xx*8*x@
xx***x.
xxxxxx.
ENDMAP

Here we have a box with loot (*) and a nasty monster (8), no door, and no hint at all. The only way the player would be able to guess something is here is if the vault happened to generate in the open. In that case, the 6×5 box might appear a bit conspicuous. But, even then, the player would have to be spoiled to know that “6×5 box in the open means loot”.

NAME:   example_2
DEPTH:  D:1-27
COLOUR: x = blue
TILE:   x = wall_zot_blue
MAP
xxxxxx.
xx***x.
xx*8*x@
xx***x.
xxxxxx.
ENDMAP

Now we have something a bit better. The box is the same, but it is coloured blue. Since this appears in the dungeon, and there are otherwise no natural blue walls, our vault gives a hint. A player who sees this will know there is something special, and will try to find a way inside. When they can't find a way in, they might try teleporting or digging, and will find the loot. However, a player might still autoexplore past this, and never see the blue of the vault (particularly with “explore_delay = -1”). Note that in console mode, all out-of-view squares are dark grey. Therefore, this is a hint, but it probably isn't a clear one, since it is quite possible to miss it.

NAME:   example_3
DEPTH:  D:1-27
MAP
xxxxxx.
xx***xG
xx*8*=@
xx***xG
xxxxxx.
ENDMAP

Now our vault has a secret door flanked by statues. Even if a player autoexplores past this, they can see the two statues when examining the map. A smart player will find these two statues suspicious (particularly if there are no other statues nearby), and will go to investiage. This is a clear hint. Although, it is not perfect: an unobservant player might still glance over the map and miss it. This is partially their fault, but the vault could still be improved.

NAME:   example_4
DEPTH:  D:1-27
MARKER: G = lua:props_marker { stop_explore = "a statue."}
KFEAT:  H = statue
MAP
xxxxxx.
xx***xG
xx*8*=@
xx***xH
xxxxxx.
ENDMAP

The marker property used here will stop autoexplore when the statue 'G' comes into view. 'H' is also defined as a statue, but we only need one marker here. Now, our vault not only has a hint, but also stops autoexplore to ensure that the player can't miss it. This hint is very clear. Note that there are other ways of stopping autoexplore: you can place a weak monster, or an item, or some other feature that causes autoexplore to be interrupted. All of these are equally valid.

NAME:   example_5
DEPTH:  D:1-27
MAP
.ccccc.
.c***c.
@n*8*=@
.c***c.
.ccccc.
ENDMAP

In this example, the vault is made of stone (undiggable) instead of rock (diggable). The only reliable way in by finding the secret door (a player may also force their way in by teleporting). However, this vault has a window (n) on the side opposite of the secret door. As long as the player explores the whole level, they will see the window, and and suspect there is a way inside the vault. Note that the window also serves to stop autoexplore: the monster (8) will be seen by the player (new monsters will stop autoexplore when discovered, even when safely behind glass). This vault has a clear hint, albeit an extremely overt one.

NAME:   example_6
DEPTH:  Zot:5
NSUBST: ! = 1:=cccc / *:c
MAP
.............
...c!!!!!!!!c
............@
ENDMAP

Here we have a somewhat special example. Assume that this is part of the zot:5 vault, dividing the lower and upper part of one of the “lungs”. There is a 1-in-5 chance that in the middle of this wall, there is a secret door somewhere (placed randomly). Finding this door would give the player a little shortcut to the upper section of the vault… enough of an advantage (thanks to how extremely dangerous the level is) that this secret qualifies as important. There is also a “hint” of sorts: since the player is likely to take the vault slowly, hug the walls, and to search for traps, they are likely to stumble across this secret door. It also helps that you are very likely to magic map the level.

I am being somewhat flexible with the definition of “hint” here, but I think this case is acceptable. Note that the actual Zot:5 vault does something very similar to this example.

NAME:   example_7
DEPTH:  D:1-27
MAP
......
.====.
.=**=.
.=*8=.
.====.
......
ENDMAP

This is another weird example. In this one, the player will originally see an inconspicuous 4×4 box. However, the entire box is made of secret doors. Because of this, the chance of discovering it is quite high for any character with T&D skill. Also, the box is guaranteed to stand by itself, is square, and is surrounded by floor: this further increases the chances of the player spotting it. Thus, the fact that there is a huge chance of detecting the secret doors means that this secret is probably acceptable. A real example of this in crawl would be minivault_9 (the “door vault”).

This concludes my analysis of what constitutes an “acceptable secret”. This isn't the end of the discussion, however. There is still the question of why to use secrets in the first place. This is especially pertinent to example_4 and example_5, where the secrets are very obvious.

I will give five reasons.

One reason is increasing the sense of reward a player feels when they find a room full of loot. Loot is rewarding on its own, but if a player has to find their way into the room, it gives a sense that they found something special. Results are usually more rewarding if some amount of work is required to achieve them (also, this applies to more than just video games). Even if the hint is a very overt one, the fact that it is a secret still lends it a feel of being “hidden”. Finding it isn't an entirely automatic thing. For some players, this makes the vault more fun.

This reason may seem trivial, or even stupid: shouldn't gameplay trump everything else? My own opinion is that the psychological element of the game is very important to how fun it is. You need to reward players for them to enjoy the game, and part of what determines how they feel rewarded is not just what they find, but how they find it. Presentation is important. Letting players find secrets is just one of many ways of giving them a sense of accomplishment.

Of course, there are two sides to this coin: if the vault does not provide a clear hint, a player might just feel ripped off if they discover that they missed loot. In that case the psychological element is a negative one rather than a positive one. Thus, it is important for there to be a hint: it should be the player's fault for missing the loot, and not the vault maker's fault.

As an aside, I think the XP percentage on the main view is one example of how psychological elements can make the game more fun. Players simply enjoy seeing that percentage go up as they kill monsters, even though the number does not really have any impact on gameplay. It's just fun to see numbers go higher - having the number there doesn't do much to make the game play more smoothly or the interface cleaner, but it does make the game more fun to play. That percentage a measure of progress and achivement. Almost every RPG has an “XP indicator” (often in the form of a bar) because of this.

By its very nature, this reason applies to all secret areas that I have defined as “important”.

Another reason is to prevent the player from wandering into a dangerous room via autoexplore (an “autoexplore trap”). By using a secret door (or requiring digging, or teleportation) we force the player to intentionally stop and look for a way in. This way, they won't bumble into the vault by accident. There are many ways of preventing accidents like this; secret areas are simply one of them. This method is particularly useful if there is a dangerous monster guarding loot. If a secret door is used, it is often a good idea to use a lua:restrict_door() property for it - this way, only the player may open the door.

Another is to slow the player down. If the player has to search for a secret, bad things might happen to them in the mean time. This is the case in the dis_mu Dis ending, where the player needs to find their way into the middle of the “pyramid” by finding a series of secret doors. While doing this, the player is hit with hell effects. Another example of this is timed portal vaults where the player knows there is a door, but has to find it before the portal vanishes.

Secrets can be used for theme/flavour purposes. A lot of vaults try to follow some kind of theme, and sometimes this theme is enchanced by using secrets. For instance, you would probably expect an ancient pyramid to have a few hidden areas. This sort of thing makes the vault more flavourful and more interesting (of course, the secrets still need to be used in a way that I've defined as “acceptable”).

Some players (particularly those who have seen everything the game has to offer) do not care much for vaults that focus on flavour. But, for other players (and, I suspect, most) this enchances their enjoyment of the game. One example which I've heard a lot of positive feedback on is lexackson_kobold_supper in mini_monsters.des (the one with the hungry kobolds). I've seen a lot of people comment on how cool this vault is, even though it doesn't really offer anything in the way of interesting gameplay - it's all flavour. Overall I'd say this is a very good vault.

Another example (and one that actually uses secrets) is minivault_9 (the “door vault”). For new players, this vault is weird and interesting (and a nasty surprise, when they find out what's inside it). For experienced players, the vault loses this element, but gains a rather ominous feel to it. This vault is more abstract than lexackson_kobold_supper, but it still has a definite “flavour” aspect (more emotional than visual) that makes it enjoyable.

Obviously, a vault can have a theme without using any secrets at all. But, secrets can somtimes be used to enchance it.

Secrets can be used to provide the player with a way out of an area they might become trapped in. Take the following example:

NAME:   example_8
DEPTH:  D:1-26
MAP
xxxxxxx
@.l*=>x
xxxxxxx
ENDMAP

Here we have a piece of loot across a square of lava. In theory, a player could levitate over to the loot, stop levitating, and get stuck. This is very bad if there's no way out! Luckily, there is actually a secret door behind the item, concealing an escape hatch. Of course, a door doesn't need to be used here - the main reason for hiding the escape hatch is just to control how the vault looks, and to de-emphasize the area with the hatch. A similar technique to this one is to use a guaranteed teleport trap (which, again, will be hidden initially).

I believe all five of these are good reasons, and can be seen in several vaults in Crawl.

There may be other reasons than these five. If so, I believe that the vault is fine as long as the vault meets the criteria for acceptability that I have laid out.

There is also a bad reason, which I think can also be seen in several vaults: to hide loot which can only be obtained by the usage of spoilers or ashenzari (or, in ancient versions, divination spells). Vaults which do this have already been defined as unacceptable, however.

So far, everything here has been my opinions. There are people who would disagree with me, however, and it is only fair that I acknowledge that. Thus, I will give a counter-argument to the usage of secrets:

All secrets are spoilers, even when they give hints. Even an obvious secret like example_4 might be missed by an inexperienced player. Thus, in order for crawl to be spoiler-free, secrets should be eliminated. In cases where the secret door serves to block autoexplore, the secret door can simply be replaced by a normal door with a (y/n) prompt. For rooms which require digging (even ones that have an obvious vault-shape), transparent walls should be provided to let the player know there is something inside. For secret doors that serve to slow down the player, some other obstruction should be provided (such as a bush that needs to be hacked through). For vaults which use secrets for flavour purposes, they can simply just lose the flavour (being free of spoilers is more important).

I think this argument misses the psychological elements that I described in the above section. Removing all secrets would make Crawl more spoiler-free, yes, but it would also lessen the sense of reward that players get when they find a loot room (particularly when it is for the first time). I also think that flavour is a very strong element for what makes a vault fun. When I see people talk about which vaults they like the most, they usually talk about ones that have a clear theme to them.

I also think spoilers are acceptable to a small degree. My stance is not a radical one where crawl must have no spoilers at all. I think a spoiler can be allowed as long as an intelligent player can be expected discover everything in-game, without having to resort to “meta” resources (such as reading .des files). I would call this sort of thing a “soft spoiler”.

I agree. Sometimes, there is confusion about what “spoiler” means (it's not a well-defined concept, I guess). As I see it, if information can be obtained by an attentive player, even at the cost of a death or two, then it is fine. We will never get around the fact that a player armed with the source code (or the data base etc.) will have better chances at survival. That's okay! We don't want to hose the other players, that is all.
So I speak of a “spoiler” if the mechanic is (a) opaque enough so that a player cannot be expected to figure it out on his own and (b) relevant enough so that a player should know it! For example, knowing the precise damage formula is not spoiler material, in my opinion: knowing it will improve your performance, but just going with common sense and being attentive leads to wins. On the other hand, I am not sure if a player can figure out on his own that banishment chances (of elf monsters, say) is affected by MR. [Insert other example once we get rid of this one.] — dpeg 2011-12-23 00:06

This extends to more than just vaults: a newbie will not know that Margery is one of the most dangerous enemies in the game if you don't resist fire (the game tells you that margery is dangerous, but does not tell you how much damage her fireballs do, or that she casts fireball). This seems fine to me. There is in-game information that clearly tells you she is dangerous (her fireballs hurt, especially when you die to them). “Die and learn”, as I like to say.

Likewise, a newbie might miss the secret in example_4. But, it places an obvious secret door between statues. Once a player knows secret doors exist, they will be inclined to look for them in suspicious spots. Even moreso when the player figures out that statues usually are parts of vaults.

Now that I have gone over these principles, I can give examples of vaults that I think should be changed. I should mention, I like these vaults. My issue is only with their usage of secret areas. While I might sound like I'm strongly in favour of secret areas and such, I actually am quite against the way they are used in many places. Also, this list is far from exhaustive. There are many vaults which have problems.

Crossed out entries are ones that have been fixed already. They are being left here as examples.

tar_old (hells.des): This tartarus ending places the boss behind a semi-obvious secret door, and the rune behind one that isn't obvious at all (a monster is likely to open it, but this doesn't always happen). A stronger hint should be given - right now the only hint is the fact that the boss is nearby. I would probably use a window to where the rune is, but leave the secret doors. This map also has a lot of other loot in non-obvious secret areas.

church_of_pain_bobbens (crypt.des): There are many secret rooms here which have loot, and little to no hint that there is a door. The only indication is the empty space in the vault… but this is not good enough, since the vault has plenty of actual empty space in it! It also doesn't help that the rooms and doors are always in the same spots. Very spoilery.

box_level_dp (encompass.des): This one is a big offender. There is a huge area in the middle of this level which can only be accessed via teleportation, and contains a great deal of loot. Magic mappping doesn't even reveal it. To get the central loot, you essentially have to know it is there. This vault probably was designed with divination spells in mind.

evilmike_wizard_prison (large_themed.des). I'm not entirely innocent here. This vault has secret loot (that you have to dig to), and a hint, but it's likely that the hint isn't good enough. The hint is a wall of rock (brown) where the rest of the vault is stone (grey). I think the concept behind this hint is good, and it works in other vaults (such as in minmay_expanding_circles, in large-abstract.des), but its execution in evilmike_wizard_prison is poor. The rock part should be made more prominent.

These are example vaults that have been repaired since appearing on this page:

wormcave (lair.des): There is a dragon behind a secret door, guarding gold and loot. This vault does a give hint (one of four randomly chosen ones, actually). However, most of these hints are insufficient. Most people never know the loot is there unless it is pointed out to them (either by another player or by Ashenzari). My solution here would be to get rid of the randomness and use only one hint, and to make it a stronger one.

due_jungle_book (lair.des): The main loot (guarded by the anaconda) is behind a secret door. The only hint is that there is a dead end. Most people know it's there just because it's such a popular and distinctive ending, but the door is still non-obvious. This one is less problematic than some others, but could still be improved. Note that the door itself is good: being secret prevents the player from bumbling into a very dangerous area. I would just make it more obvious, perhaps by having it flanked by elephant statues.

evil_forest (lair.des): There is a secret room where the door is between statues. The room is empty and leads nowhere. The intuitive thing to do would be to search for a second secret door in the empty room. This will prove fruitless, however. The actual secret entrance is outside: a second door on an inconspicuous corner of stone. It leads to the crazed wizard and an entrance to hell. I would recommend moving the secret door to the empty room. The way it is now is very spoilery.

vgs_many_chambers (volcano.des)… and many others in volcano.des. There are secret doors here, with loot, and no hint. Particularly bad is that this subvault has loot elsewhere in it, leading the player to think that they found everything (when, in fact, the real prize is completely hidden).

I'd like what I have written here to become a policy for new vaults in crawl. It is certainly something I will stand by when making new vaults, and when evaluating ones that other people submit. I think it would be good for other vault makers to adopt this principle too.

If other devs agree, I would like to apply these principles to already existing vaults. My goal will be, whenever possible, to give the secret a hint (or strengthen a currently too-weak hint) rather than remove the secret.

Many of the problematic vaults I have identified are by currently active devs, so I will not make any changes unless there is agreement. Thus, if there is anyone who disagrees with me, please discuss it on this page. If there is enough agreement (or a notable lack of disagreement) I will go ahead and make changes to old vaults.

I am particularly interested in hearing the opinions of vault makers, but other people's opinions are welcome.

I agree with your principles. Feel free to address my vaults.
There is one, somehow metagamey reason to use secret doors: this makes the T&D skill more relevant. For a while I championed a level design where hidden closets with minor loot (think of normal consumables) are in very inconspicious places (e.g. at the boundary of the map). The drawback is that a player without T&D could still press 5 very often, and our clocks do probably not punish this behaviour enough. (Said several times but here I go once more: we briefly discussed scrapping traps, secret doors and T&D at some point, but didn't… we already scrapped so much and traps seemed to have potential. By now, I am absolutely against scrapping traps as a whole, with shafts being the strongest reason. Therefore, I am also happy to keep secret doors. I believe they add flavour.)
I would suggest to just change vaults if you want to. If the vault maker is still around, she can have a look at the new version and talk about it later, if there is some kind of disagreement. — dpeg 2011-12-23 00:14

Rather than setting repaired vaults as deleted using strikeout, I moved them to a separate section. I suggest updating their descriptions with whatever was done to “repair” them. — XuaXua 2011-12-23 16:33