Wednesday, September 7, 2011

The Shell Game in the Sandbox

I see some places I frequent (Hack & Slash, Monsters and Manuals) are speaking of sandboxes and player agency, and it's got me thinking again about something I've been mulling related to the problem of Illusionism.  The question is whether player agency in a game is an objective condition or subjective; is it defined by player expectation?  To get there, let's first look at The Shell Game in the Sandbox (a problem of palette shifting).

You probably know the Shell Game - there are 3 cups on the table, underneath one of the cups is a little ball, and the skilled operator shuffles the cups around, trying to mix up the location of the ball so the player can't guess the location.  The real shell game on the street is a fraud; the operator palms the ball with sleight of hand, so there's no way to guess the correct shell; the mark always loses his money.  The metaphor holds up.

Let's imagine a gaming scenario where the group must search a series of woods looking for the MacGuffin - Woods A, B, and C.  The MacGuffin could be in any of the woods.  Pick a wood, pick any wood!  It could be the classic Shell Game in D&D!

One DM - we'll call him Scripto-DM - scripts the content for all 3 woods in advance, and locks the MacGuffin into Wood B.  The other DM - Improv-DM - makes a detailed encounter with an Ogre, and keeps that game content unassigned.  Regardless of which woods the players choose first, he'd like the party to have the opportunity to encounter the ogre.  The MacGuffin will be somewhere else.

Most folks will say that Scripto-DM has enabled player agency and free choice; Improv-DM is setting up a railroad.  Let's first take a closer look at Improv-DM.

When the party boldly announces they will head out to Wood C first, looking for the MacGuffin, they run into Improv-DM's (supposedly excellent) Ogre encounter.  He reasons that he could have improvised the woods with random encounter tables, but instead developed an encounter in advance.  By deciding at game time that the MacGuffin is not in Wood C, and the Ogre is there instead, has he *actually* violated player agency?  Player will or choice has not been thwarted.  They wanted to go to the woods, and Lo! - they are in the woods.  And yet objectively he has preordained a game result.

Edit: redacted a section where I discussed how Scripto DM's standard sandbox could create the impression of a railroad through poor design - will put that in a succeeding article.

This isn't a new topic for me; I worked through similar issues with Illusionism and the Sandbox back in March in parts 1, parts 2 and parts 3; ultimately I came to the following conclusion for myself:

Player choices should be meaningful and have consequences.  Player choices are based on partial information.  The DM is obligated to administer the setting in a way that ensures player choice is meaningful, in accordance with the previously established facts.

Barring a change in temperament, that puts me on the side of the fence that says Player Agency is a Subjective Standard.  Improv-DM is probably okay with his approach to palette-shifting the ogre encounter, since the party's will isn't thwarted and they had no expectations for or against meeting an ogre.  If the players in Improv-DM's game asked the villagers where the ogre lives, and they're told he lives in Woods A, now Improv-DM is constrained from palette shifting, because he has introduced a fact about the game world that will be used by the players in decision making.

Edit:  While I continue to maintain that player agency is subjective (what the players don't know doesn't hurt them), I ultimately came around to the position that any kind of illusionism is a bad practice; rather than test it from the angle of player knowledge, its best that the DM doesn't predetermine outcomes in any manner; it's a simpler "test":  Fixing the quantum ogre.

So here's why this is so topical for me:  I've been reading the Red Tide setting, from Sine Nomine Games, and it provides some off-beat sandbox design advice (a review is forthcoming).  Much of the advice involves developing pre-built towns, ruins and encounters that can be moved around the sandbox - a clear case of palette shifting.  (As I mention in the comments, the Red Tide approach is either brilliant - or brilliantly manipulative!)  But in a sprawling sandbox with lots of undefined white space, who cares if the serial numbers are switched around on two remote villages that the players never visited?

It's also a potential answer to the problem of the Sandbox Triangle.  I'm a project manager by trade (although these days I do more IT functional management) so I'm very familiar with the Iron Triangle; you see the same concept in product development.  For the sandbox, the constraining factors are Freedom, Detail, and Effort.  The group can have total freedom with lots of detail, but the DM needs to put in a lot of effort if they're a Scripto-DM.

The "harmless" form of palette shifting perpetrated by Improv-DM provides a way to cheat the Law of the Triangle.  The DM puts in a lot of detail into smaller areas (so smaller effort); however, according to the Law of the Triangle, player freedom has to be constrained.  The provocative question is this:  If the player's don't know they're taking part in the Shell Game, does it matter?

I know there are sandbox purists that avoid Improvisation, Reflection, and changing content on the fly, so I'm eager to see the pros and cons regarding what I'm calling "harmless palette shifting".  Convince me I'm wrong!


  1. That Ogre example weakens your point. So much so that your point is essentially unfounded.

    I don't think players would truly feel like they had no agency in Scripto-DMs game. Their decision is predicated on their knowledge that many ogres exist ("they wish to avoid Ogres" plural). If it was really an at-all-costs thing, they would have asked more questions, but they didn't. Plus it's not unreasonable for a singular ogre to roam around a bit ... possibly to another woods.

    Spin the situation again, but replace Ogre with something more implicitly unique ... maybe a location with a name given to the players that clearly reflects its uniqueness: The Ruins of Duke Bubba's Keep, or something. You need a more unique – or at least immobile – thing to support your point.

    (But if you do simply swap in a unique/immobile thing in place of ogres, the example becomes about Scripto-DM *actually* pallet shifting.)

  2. The example -C raised yesterday over on Hack & Slash regarding pallet shifting involved the group learning about bandits on the west road; when the group decides to go east to avoid bandits, and the DM shifts the encounter anyway, he claims that is a refutation of player agency.

    Splitting hairs over the use of the term 'an Ogre lives in the Wood A' versus 'Ogres live in Wood A' dodges the question - whether any pallet shifting is distasteful even when players don't know better.

    In your example, the Ruins of Bubba Keep, there are different states of player knowledge vis a vis the sandbox. If they choose to go to one of the woods, and they don't anything about potential ruins in the woods, is it fair to shift Bubba Keep around so they find it in the first wood they enter? Is it a different scenario if they learn in town there could be an undefined old ruin in Wood B, they choose to go to Wood C, and the DM puts the specific Bubba Keep in Wood C?

    My point is that player agency is subjective and is related to how much information the players have for the basis of their decisions; the "test of agency" under these circumstances shifts with player knowledge and the DM's duty to portray a consistent game world.

  3. I don't necessarily disagree with your intended point, and I do think that's worthy of further exploration.

    But my point isn't about your point. It's about the essay/article as a whole. I apologize for the lack of clarity in my other comment. Trying again:

    The comparison between Scripto-DM and Improv-DM is flawed because the details of the Scripto-DM scenario don't actually support the stated player perception. And the article is structured so as to present the comparison as substantiation for the conclusion. (Otherwise why spend so many words deconstructing Scripto-DM's scenario?) But because the comparison is flawed, the conclusion is implicitly undermined, regardless of whether the intended point is valid.

    The comparison presents two ideas:

    1) Improv-DM isn't really hindering agency, despite some peoples' beliefs that this is a non-agency playing style, because the players had no information on which to base their sense of agency. That is, do they still have agency even if the popular notion is that they don't have agency.

    2) Scripto-DM is hindering agency, despite some peoples' beliefs that this is a correct agency playing style, because the players perceived their agency to be violated. That is, do they lack agency even though the popular notion is that they have agency.

    And from this you attempt to draw the universal conclusion that agency is subjective, regardless of whether agency methods are being applied. But the failure in the comparison means you haven't really substantiated idea #2, and so the conclusion falters.

    But it seems that your *real* point actually has nothing to do with Scripto-DM's scenario. It has only to do with validating Improv-DM's scenario. As you say, determining "whether any pallet shifting is distasteful even when players don't know better." And – if i understand correctly – only in this sense do you really care whether agency is subjective.

    I think you should delve deeper into Improv-DM's scenario, but without bringing Scripto-DM's scenario into the discussion. Fodder for thought:
    - Does the DM's knowledge of this encounter-to-be do anything to hurt agency, perhaps subconsciously?
    - How much DM decision is acceptable when player agency is a goal?
    - What are some tools a DM can/should use if they want to knowingly have some of these encounters-to-be in their back pocket?
    - Are these encounters-to-be different from using wandering monster tables, and if so, how? Is the answer different if the DM pre-rolls a bunch of wandering encounters prior to the start of exploration?
    - What is the true test of agency? Is it something players even need to be able to test for, or is something that can only be tested with all the knowledge on the DM's side of the screen?

  4. Yep, your Points 1 and 2 hit on what I wanted to get after - thanks for clarifying thrust of the disagreement. People in this space (rightly) fetishize player agency as the highest good; the measure of a game that supports agency is that players feel empowered because their choices lead to distinct outcomes; right or wrong, an Illusionist DM can get away with shell game tricks and not undermine player agency based on the current state of player knowledge; likewise, the traditional sandbox can (perceptually) fail based on the current state of player knowledge. (I'm stating these like absolute truths, but the point is that they really are debateable based on a subjective or objective definition of player agency).

    Furthermore, as I ponder the Red Tide approach to the sandbox, I vacillate between declaring it brilliant, or just brilliantly manipulative! Clearly I need help.

    As for refocusing the thesis of the article - I'm torn between rewriting it, or leaving the flawed one in place and putting up a second.

  5. Note: After Guy's feedback ("Um, Beedo, your article is a confused mess..") I got rid of the Scripto-DM discussion - how the scripted sandbox can fail the agency test, if the test is a subjective one. I'll put that up as a stand alone.

  6. Your post and C's are jiving with some observations I have about the evolution of the Hill Cantons campaign over the last three years--which started as a militant sandbox a la West Marches experiment and has evolved into something more flexible over time.

    Pallet shifting in particular is an interesting question as it was a road I went down fairly quickly in my first year when I found the players wandering off the reservation--then wholly rejected, only to swing again to finding a workable meeting point (which is still pretty much against).

    I know there are sandbox purists that avoid improvisation, reflection, and changing content on the fly, so I'm eager to see the pros and cons regarding what I'm calling "harmless pallet shifting".
    I will have to unpack the experience a little more (perhaps something for a post), but I have definitely seen and felt both the pros and cons of "borderline" behavior, for the lack of a better descriptor of the grey area between pure player agency and what actually happens at the table to make a satisfying game.

    I agree that it's worth exploring, because I would hazard a guess that given enough time at the table players will eventually move into an area you haven't adequately prepped and when they do it is highly likely that you will be involved in some kind of diminishing of player agency.

    Take for instance the highly-recommended use of random tables. If the party is say in a clearing at the edge of the map with unprepped forests in three directions and they choose to move into one of them. The GM consults the same random “forest stocking” chart in either case. Is this classic railroading or pallet shifting? No, but it's an illusion of agency at some level as their actual choice means little to nothing. It's just a few shades of gray less dark.

    One thing though, when you ask "if the player's don't know they're taking part in the Shell Game, does it matter?"

    As a blanket statement, I do have to say "yes". Leaving aside questions about false agency, there is something to be said about it playing false for you the GM too.

    One of the most attractive features of the sandbox style of play is your ability to be surprised. Take away player agency, even if they don't know it, and you greatly reduce the chance of this happening. (This too I can attest to.)

  7. I wouldn't say player agency is "subjective" I'd say it's objective and on a scale.

    For example, the improv dm probably has decided semi-consciously, how many of the millions of resources available to a PC that the players could use to detect the ogre in advance would actually work.

    Ask villagers?
    Climb a tree?
    Get a ranger to look for tracks?
    Have a dog sniff for anything unusual?

    If the GM decides that "whichever one of these resources that I think will work--if the PCs use it, they'll detect an ogre long before they meet one", this is -objectively- creating a situation where player choice has mattered.

    Almost all games ever have "new" setting elements introduced by GMs during play at some point. These could not have been detected and prepared for because the GM hadn't thought of them/prepared them yet.

    As "unexpected, undetectable facts" these things are less sandboxy and more railroady.

    What makes something more sandboxy and less railroady is the ability to get information about them.

    (Like it's laid out in your triangle.)

    The less possible it is to get information about a game element before it does it's thing to the PCs, the more "railroady" it is, the more possible, the less railroady.

    This extends to abstract information, like: if the PCs head into a sparsely-inhabited polar area and meet winter wolves, this is not railroady even if the PCs never heard the word "winter wolf" before because, abstractly, the words "sparsely inhabited wilderness" should let any thinking PC or player know they are likely to encounter wild animals. The further from civilization, the bigger.

    So, yeah, "railroadiness" vs. "sandboxyiness" is always on a scale, but each element can be judged on the scale relatively objectively: How much of a chance to anticipate this did the PCs have?

    In your ogre example if the PCs had, really, no chance to predict the ogre, that encounter is railroady, HOWEVER, it is merely one encounter on the far extreme of railroady embedded in a complex web of encounters of various high or low degrees of railroadiness, all of which the GM can control.

  8. Why is it that it's ok to use dice for combat resolution but so many folks harsh on you when you are using them for events or encounters? Is is taking away player agency if they roll a 2 to hit? No. So why is it taking away player agency when they meet [2d4 forest creatures] on the road to Nottingham? If they can choose to run away or stay and fight (or fast talk, seduce or bribe...) then they have agency.
    Typically I don't move most encounters around because I feel player choices should have consequences but I do think that random events and encounters should be used to let fate decide what they meet in the forest or if they go the night without being attacked by wolves. I'll weight the random encounters by area and such but that's just tuning the machine.

    I'd say railroading begins when you limit player's actions, not how you generate the encounters.

  9. @ckutalik:

    "Take for instance the highly-recommended use of random tables. If the party is say in a clearing at the edge of the map with unprepped forests in three directions and they choose to move into one of them. The GM consults the same random “forest stocking” chart in either case. Is this classic railroading or pallet shifting? No, but it's an illusion of agency at some level as their actual choice means little to nothing."

    I don't think that's an illusion of agency. It's simply the DM discovering the contents of a particular forest at the same time the PCs do.

    If the PCs are going to investigate what the three forests contain by actually entering them (as opposed to seeking clues via other methods), they are still afforded the ability to leave one if they find it unpalatable early on. And then move onto another one, which stands the chance of having different contents produced by the random stocking table.

  10. As a counterpoint to my last comment, perhaps one advantage to pre-stocking areas is that it better allows the DM to reveal significant (though perhaps subtle) teasers about those areas when the PCs first learn about the existence of those areas. Here's a clumsy example of what I mean:

    By not pre-stocking those forests, a DM might easily present them as simply "three forests, one each to the north, west, and east".

    But by pre-stocking the forests, the DM has advance knowledge about whether, for example, one happened to contain a disproportionate amount of undead, another had mostly wildlife, and the last had a typical mix of humanoids and wildlife. This knowledge might spur the DM to initially present them to the players as the "haunted forest" (and won't have any outlying farms near it as the characters approach), the "Duke's hunting grounds" (and perhaps also require the Duke's permission to enter), and the "Southwood" (a pretty bland name to go along with the "typical" random stocking results).

    Perhaps the fact that the DM's imaginative juices started flowing earlier in the process is a good thing in terms of player agency?

  11. The shell game provides an illusion of choice, which is kind of objectionable. You don't really choose and get Wood A or Wood B, you just choose where the GM is going to stick Ogre Encounter C.

    But on the other hand, if this is bad, why is letting them choose A or B and then making up the details on the fly any better? Or letting them choose A or B and then randomly rolling for monsters and getting an ogre? From a player perspective, is this really different?

    I'm not settled on this, because I've used both - choose A, get A, choose B, get B, and C shows up in either one no matter what. Or C is in Wood B and if you choose A you miss out on it.

    It's a tough one to puzzle about, because we don't have unlimited resources to create detail just in case it's needed and then see where the players go, but actual choice does feel better than the illusion of choice.

  12. Peter, you've hit exactly on why the shell game is attractive; precisely because no DM has unlimited resources. I started considering the technique more seriously after the read-through of Sine Nomine's Red Tide, which advocates whole scale repurposing of content and moving things around as the players choose the focus of the sandbox; here are a few sample quotes:

    "One of the most vital skills for a sandbox referee is the ability to repurpose content that would otherwise go unused. You want to waste nothing of your work...

    Your time and creative energy is precious, and you should never waste the work you do just because it’s not labeled correctly."

    The idea being, the DM built a forest village down the east road; when the group goes down the west road instead, and visits a new forest village, go ahead and use the (never visited) east village instead. Because no information has been spoiled, the players don't know the difference, and the DM doesn't waste any work. It's compelling if pulled off well, but changes many of my ideas about prepping for the sandbox.

  13. Your time and creative energy is precious, and you should never waste the work you do just because it’s not labeled correctly."

    That makes sense. But as I said, I've done it both ways. You don't investigate the corpses that have the Magic Whatzit of Whosit (major plot item)? Okay, then it's lost to you. But I save the ambush and traps from that area for another day, to re-use in another form (either straight up or only lightly modified).

    On the other hand, sometimes all roads lead to Rome, because I only wrote up Rome.

  14. @guies and everyone.

    Goddamn it Zak is smart.

    Also: Holy crap, need to post in order to respond.

  15. Ok, Replied.

  16. Your time and creative energy is precious
    ...and time at the game table is even more precious - hijacking your example, it's far too precious to be wasted scouring three sets of woods for that ogre the players need to meet just because they didn't get the hint back in the ruined village about which one it was supposed to be in.

    In all these cases, though, detectability seems like the determining factor: the question of whether an undetectable DM shift is a bad thing strikes me as a mystical one (except for its potentially spoiling the DM's fun) and the examples as written already show plenty of adventure-possibility restrictions (or as I like to call it, "design"). Why only 3 woods? Why an ogre? Because the DM had some idea of what they wanted to happen, at some stage. How long should it take to establish that the ogre is not in wood A? In practical terms, the correct answer is always "not so long that it stops being fun," and there goes your self-running reality simulation straight out the window.

  17. Spring-boarding off richard's comment about the detectability factor being necessary for players to even know there has been a shell game switch led me to post this:

  18. Player agency is possible with random encounters. The trick is to build tables for signs of encounters. The same thing can be said for the shell game. Players choose woods A B or C. You put your pre-designed ogre encounter wherever they ultimately choose, but the 1st thing they encounter are signs of the ogre. So the players aren't forced to deal with your ogre. But your effort in preparing this encounter is not lost. They could decide later to deal with it, or find out that by not dealing with it some other consequence happened and they can decide to deal with that.

  19. IMO, a DM that goes out of his way to move an encounter when PCs have tried to avoid it is out of bounds. I don't care if the bandit leader is Robin Hood, that's too much. Unless there are bandits on _both_ roads, _& the PCs don't know it_ , you're cheating them. If I learned a DM had done that, I'd never play in a game with that DM in charge ever again.

    The "3 woods", OTOH, is a free choice. If PCs haven't tried to avoid (or find!) an ogre, it doesn't matter if the DM "plants" an ogre. It amounts to a variety of "wandering monster" encounter.