Tuesday, 18 October 2016

On failure, its outcomes and its implications: psychology

I feel like it's been ages since I was able to put together anything substantial for this blog. To be fair, nothing I write now is ever likely to compare to the insane (in relative terms) popularity of my post about animal companions...

And also to be fair, it's been a very busy few months and I'm ill. But still. I do enjoy writing for this blog and feeling like someone appreciated it.

This is going to be a miniseries about failure in RPGs, or at least in some RPGs. I fear it may be a bit dry and very rambly. Still, I present it for your delectation, or at least to keep you mildly diverted on the bus.

So a while ago I wrote some responses to a Walking Eye episode about Numenera. Very little of that is relevant right now, so let me pull out the bit which, randomly, sparked this week's post-game conversation. It is is in fact talking about Dungeon World, for some reason.

Most rolls give a failure or a qualified success, which often ends up as a downer and sometimes pretty much undermines the success; it also means the DM (or group) have to keep on thinking of complications that fit the situation, don’t just negate the success and are somewhat interesting. It tends to dwell on the failures and focus attention on them. While D&D can be similarly whiffy, the failures are generally quick and simple, often without specific consequences for failure, which tends to downplay the sense of incompetence. Failing at (say) an attack roll gives the enemy that much longer to fight back, but doesn't damage you or disarm you, but also it doesn't make a big deal of your failure. I don’t often want to play games that are all about failing, and for some reason it feels preferable to me to take damage because an enemy succeeded on a roll than because I did middlingly on mine. But also, I feel like the constant runs of complications get in the way. The game is arguably more story-focused than D&D, but the kind of stories it produces are stories of failing all the time, and the fixed difficulties make for failures that often feel arbitrary.

We ended up having some long conversations (both in person and by email) about failure in games, touching on both its mechanical impacts and its narrative treatment and repercussions. What follows is a cobbled-together mixture of my thoughts, other people's thoughts, my thoughts about their thoughts, and nonsense. Assume anything particularly insightful was someone else's idea.


There are basically three topics in this miniseries. Thread one is about (mechanical, die-rolling) failure per se and how it feels to me as a player. Thread two is about failures, interestingness and the Fail Forward/Roll Dice or Say Yes sort of concept. Thread three is about the interaction between failure, narrative and the coherence (in the non-Forge-jargon sense) of the play experience.

To just get my disclaimer out of the way beforehand: most, but not all, of this miniseries will consist of looking askance at Dungeon World and comparing it unfavourably to Dungeons & Dragons, but as usual, this is a matter of personal taste. Fundamentally it delivers an experience that, I think, I don't particularly want. Similarly, I will be sucking in my breath while gently prodding Fail Forward, Roll Dice or Say Yes, and spontaneous gaming where players bear a lot of responsibility for generating or altering the world during play. I am aware that this does not mean these things are inherently bad.

Feeling Failure

When I first heard about Dungeon World, and listened to a bunch of actual plays, one of the things which most strongly struck me was the way success rolls work. You roll 2d6 and add your stat modifier. If you get a 10+, you succeed completely. If you get a 7-9, you partially succeed at a cost; in the case of combat, this typically means taking a counterattack and therefore, usually, damage. If you roll less, the DM decides what happens, but it's bad.

One of the interesting things was how salient the failure felt. It seemed like the groups were failing a lot. So much that my take-home impression of the game was, broadly, "man, this is a downer, it would get me down to keep failing like this."

But is this fair? Let's find out!

We can do this by comparing two or more systems. Luckily, finding other systems that are similar to Dungeon World is easy, because it's based on the oldest RPG around and so are lots of other things. I'm going to begin with combat, because it's both an obvious thing to test, and most easily demonstrates the impact of failed rolls (getting stabbed is always important).

In the red corner: Dungeon World, rolling 2d6+stat, with failure on 1-6, mixed success on a 7-9, and flat success on a 10+. Our heroine: a human fighter who put a 10 in Strength for some reason. She is, for reasons best known to herself, fighting a horse. Or an Apocalypse Dragon. Or a small, elderly bluebottle.*

* As far as I can tell, it makes no difference whatsoever what monster you are attacking, so why not?

In the blue corner: Dungeons and Dragons, rolling 1d20+stat, with a second roll by our opponent. Our adventurer is a human fighter with a small club, in which she is proficient, who has decided to fight a draft horse for some reason. It's a slightly scrawny draft horse.

In both cases, the adventurer has roughly a 42% chance to hit the creature and then to be hit back in return, and a 16% chance to hit the creature while evading its attacks.*

Only because I specifically set it up that way for testing purposes. In reality, it turns out more complicated - see below. But that's where the D&D character's description comes from. Also the horse. Draft horses have AC 10 and a typical +5 To Hit, so it needs to be a little scrawnier than usual. I couldn't find any other creature that's close.

We run a round of combat, and in both cases we get the same result: the 42% where our adventurer and their opponent both hit.

But they feel different to me. In the D&D sequence, I basically feel like I succeeded. In the Dungeon World sequence, I kinda feel like I failed.

In the case of D&D, my adventurer takes her turn and lands a triumphant blow on... okay, yes, on the unfortunate horse. My adventurer is a horrible person. But, I have done it, and I grin and say something that isn't as pithy as I hoped. On its turn, the aggrieved horse (rightly, we might think) kicks my adventurer in the face, and I wince and curse at the blasted beast.

In the case of Dungeon World, my adventurer swings at the dragon/bluebottle and manages to wound it, but is unable to evade the mighty blow/faint stroke that staggers her back/slightly tickles her. I roll my eyes and shrug it off as a necessary price.

What's going on?

I think the main difference is in the emphasis created by the separation of turns, and thus by the separation of actions, and the psychological implications that has for how I parse the events and what they say about the game reality.

In the case of D&D, the adventurer attacks and succeeds at her action. There is no compromise. When the horse's turn comes around, perhaps several minutes later, it makes a counterattack which also succeeds. Perhaps because I'm not rolling dice but using a static AC, the adventurer's inability to defend herself doesn't feel like my failure, but like the enemy's success. A feisty one, this damned horse! But dice aside, it's also a very clear sequence of actions. Each action is tied to one character and, fundamentally, I mentally associate the success or failure of that action with that character. It's not that I did badly, it's that my character's AC is low (a property of the universe) or the horse rolled very well (which is out of my hands).

In Dungeon World, the adventurer attacks and partially succeeds. She isn't good enough to land a blow without being buffeted by the enemy. The enemy hasn't done anything; they rolled no dice. I took damage because I just didn't roll well enough to not take damage.

I think basically, the issue is that my goal here is not "injure this creature regardless of whether you get hurt", but "injure this creature and don't take damage". It's never going to be the former; I always care whether I get hurt. So when my roll results in taking damage, I have fundamentally failed in my objective. I took the action, I rolled the dice, and psychologically I feel that I failed. I clawed back something, but I still failed. There are a handful of situations in games where I decide to brave probable injury to attempt an action, but those are normally rare exceptions. In Dungeon World, though, it's every single attack.

As well as the ownership of failure, there's also a sense of consequence.

The D&D combat round is one in which everyone acts. My success or otherwise at bludgeoning a horse is independent of the horse's success at kicking me in the nadgers. When I attack, I decide that I wish to bludgeon the horse, and if I roll adequately, I succeed. I can celebrate that success. Whether the horse later kicks me in the nadgers is another matter entirely, and one over which I have little control.

The other important point is that in D&D, danger to my character is largely independent of whether I attack the horse. My character is not in danger because she attacks, but because she stands where a hostile creature can attack her on its own turn.

In Dungeon World, because a single roll is used to resolve both my and the horse's attacks, they are interdependent. Me getting kicked in the nadgers is a result of my attack being inadequate. It is not a mysterious working-out of the universe, but (pyschologically) "my fault".

What's worse is that mechanically, the only reason I get injured is I took the attack action (or whichever named Move you wish to invoke), which makes it feel doubly my fault. Yes, officially this is not how it works: the DM is supposed to run things so that the adventurer who stands around gawping instead makes a Defy Danger roll to avoid being trampled by the horse. However, there's still a clear chain of mechanical consequence: I take action > I make roll > I suffer damage.

Salience stings

Moreover, there's a difference in salience between the two conditions. In the case of D&D, I fail a roll and nothing happens; we generally shrug and move on. Or, in the case of the counterattack, the horse succeeds at a roll against my flat AC, and I incur damage. Of course, it might be a dramatic failure at a key moment, but most of the time that's not true, just as most successes are fairly routine.

In the case of Dungeon World, any failure is highlighted by the fact that consequences attach to it: it takes physically longer to resolve, more is said about it, and everyone (me included) notices it more. Essentially its psychological size is increased. This attention makes me more aware of the failure, and so influences my perception of the game.

In a pair of games where exactly the same series of successes and failures happen, I believe I will notice the Dungeon World things that go wrong more, simply because of the mechanical and narrative emphasis that is placed on them by the interaction of "only players roll" and mechanical consequences. I admit that's speculative because I haven't played Dungeon World (yet, I would); but my impression of the Actual Play recordings I've listened to was very much one where failure (where failure means "bad things happen to you because of a die roll") was to be expected.

You can try to describe them the same way, of course - the [monster] sees a gap in my defences after my attack, and slips through a counterblow. But you really can't shake off the fact that the D&D horse is making its own attack roll which determines the effectiveness of its attack and which is completely independent of my attack, and whose result I can anxiously anticipate and watch. It might even decide to attack a completely different person, or to run away, drink a healing potion, or cast a spell! I can't not know that the Dungeon World horse is definitely going to hit me because I know I rolled a 9, however much exposition the DM throws in.

The only way to really reconcile this would be for all die rolls to be visible only to the GM, who narrates their results, which is just not going to happen. In fact, that still wouldn't stop me knowing that the horse kicked me because of my roll. What we'd need would be for the DM to use an unspecified resolution system secretly while I simply specify actions. Of course, that would cut off all the bonus-offering and roll-modifying that games tend to allow, and make for a game so different that it's not really worth talking about here.

I appreciate this is a fine distinction, and one that may completely bamboozle some people. Perhaps no such distinction has ever occurred to you. If you're playing a particularly narrative-heavy game of Dungeon World, maybe it does feel like the monsters are responsible when you get hurt. Or maybe your brain just thinks differently.

Extensive tangent on combat stats

Read this only if you are interested in how the two systems match up statistically.

To make this comparison fair, bearing in mind that it's mostly about psychology, I thought I should set things up to ensure matching odds. That is, the odds of the D&D character hitting and being hit match those of the Dungeon World characters, because the point of this exercise is emotional impact rather than mechanical balance.

This was surprisingly (I might say, embarrassingly) difficult to work out. The fact that D&D enemies have highly variable AC doesn't help. I decided to stick with basic 1st-level adventurers with only the basic +2 proficiency bonus, since I'm assuming my Dungeon World adventurer has no stat bonuses also for simplicity's sake.

To summarise: okay, the odds of the adventurer hitting are, overall, 21/36, or 58%... for which the closest match in a d20+2 is, I think, an AC of 10 (you always miss on a roll of 1, after all). We now need the odds of the enemy also hitting to be 42%. I think this occurs when the enemy only needs to roll a 5 to hit, which means we're probably talking about an unarmoured, unDextrous human with a sword (AC 10, proficiency +2) fighting... riffles through Monster Manual... I dunno, a particularly weak draft horse?

In practice, this is probably an inaccurate of how the games really run. And that's important, because what's the use of speculating about how they feel to play if they don't play how I say they do?

Feeling seems to be that in 5e D&D enemy AC generally starts at 12-14 and a character will typically have at least a +2 proficiency bonus, therefore hitting about 50% of the time. Monster attack bonuses are typically +3 (the DMG certainly suggests this) and an average starting character has an AC of about 13 (either high Dex, or lowish Dex + armour), so monsters hit on a 50% also.

This means that on average, in a D&D combat between one 1st-level adventurer and one generic level-appropriate monster:

  • 25% of the time, both adventurer and monster will miss
  • 25% of the time, the adventurer misses but is struck by the monster
  • 25% of the time, the adventurer and monster strike each other in the exchange of blows
  • 25% of the time, the adventurer hits and evades the monster's blow

That is to say:

  • The odds of a Dungeon Worldy 6-or-less failure are 25% rather than 42%
  • The odds of a Dungeon Worldy partial success are 25% rather than 42%
  • The odds of a Dungeon Worldy total success are 25% rather than 16%
  • There is a 25% chance of a null outcome, which does not seem to exist in Dungeon World.

In other words, D&D 5e is either less or more failure-ridden than Dungeon World, depending on whether you consider "this round is a wash" to be a failure. And that's very context-dependent.

But wait!

It gets weirder when you actually factor in stat modifiers.

In Dungeon World, you get stats of 16, 15, 13, 12, 9, 8. These translate into modifiers of +2, +1, +1, 0, 0, and -1. Your mean modifier on a roll is therefore +0.5. We can be charitable and reasonably assume that in combat you're probably using the stat with a +2. In this case, you have an average 17% chance of failure, a 42% chance of succeeding with a cost, and a 15/36 = 42% chance of succeeding outright.

If you instead took your +2 in a different stat - and I'm not sure why, because each class has one clear combat stat and it's generally going to be best to put your +2 there - you'll have a 28% chance of failure, a 44% chance of qualified success, and a 28% chance of outright success.

I don't entirely know why they dispensed with the classic "modifier is (stat-10)/2" scheme. They use one based on probabilities of getting certain distributions of 3-18 on 3d6, although since you don't roll stats anyway I'm not sure why this is the case. There's probably a reason and it might be a perfectly good one, I just find it a bit confusing.

In Dungeons and Dragons, you get default stats of 15, 14, 13, 12, 10, 8. These translate into +2, +2, +1, +1, 0 and -1. You also typically receive a total +2 modifier to stats based on your chosen race, which complicates matters. Your mean modifier on a roll is therefore +1.2, which is twice as good, and following the same logic as above, we can charitably assume that as a 1st-level character you have a +3 on your combat stat of choice. Rolling 1d20 +2 (minimum proficiency) +3 (stat) means you will on average hit a starting monster 65% of the time. This means you'll miss and take damage on a 17% (same as Dungeon World), exchange damage at 32% (instead of 42%), strike without injury at 32% (instead of 42%), and nothing happens the last 17%.

In other words, because Dungeon World rejects the idea that "nothing happens" might be an outcome, an unmodified Dungeon World character succeeds at causing damage 72-84% of the time, and avoids damage 28-42% of the time when they are attacking rather than just defending. The corresponding D&D character succeeds at causing damage only 64% of the time, and avoids damage 50% of the time. The D&D character is arguably less successful than the Dungeon World one! Even allowing that we might subconsciously ignore turns where nothing happens, they effectively succeed only 80% of those rounds, which isn't significantly more success than in Dungeon World.

And yet, I feel that because the compromise of taking damage (or forgetting a spell, or losing an item, or whatever other consequence is applied to a roll) undermines the feeling of success from achieving the main objective, Dungeon World still feels more faily.

Failure in Other Games

Let's think about some other game systems and their failures.

Two of my favourite game systems are, in fact, high-failure. Call of Cthulhu uses a percentile system where skills rarely exceed 70% and are often 25% or less, and where many GMs throw in penalties aplenty but few bonuses. Deathwatch use another percentile system with similar probabilities, though its combat system offers plenty of bonuses that make it easier to fight effectively.

Call of Cthulhu

As an investigative game, Call of Cthulhu's failiness has prompted not only endless discussion, but the creation of an entire alternative game, Trail of Cthulhu. This is because a sequence of failed rolls can leave players with no idea how to proceed. Since you're primarily looking for information rather than wandering around fighting things, you need clues to keep the story moving.

You typically make relatively few rolls in Call of Cthulhu compared to more combat-heavy games. This has two implications for failure. The first is that, because for much of the game you aren't rolling at all, the proportion of the game during which you might be failing is relatively low. In my experience, the investigative nature lends it a fair amount of player-level successiness due to the gratification of working things out and directing characters to do as you want. A good DM also typically gives out some clues based purely on roleplaying and good decisions, rather than insisting on a roll to see whether you spot the item hidden under the pillow you decided to lift.

The second implication, somewhat unintuitively, is that failure matters more. As a skills-based game where you rarely roll, it's common that you only roll a particular skill once per session, and often even less. If you therefore fail - and, with "professional" set at 50% and bonuses rare, that's likely - it's a big deal. Because only what happens in the game narrative feels real, it doesn't matter whether you have a Geology of 70%. If you fail the only Geology roll that comes up, the reality of the story is that your character isn't good enough at Geology. In many ways, their Geology skill does not exist at all.

It's been pointed out to me that some of the changes to opposed rolls in Call of Cthulhu 7e seem designed to address this problem in NPCs, but that's another topic.

My experience is that good GMs work to mitigate this. They provide some information even on a failed roll, particularly where it's important to keeping the story moving, so that a successful roll is a matter of better information or more complete understanding. They describe things and manage the story in such a way that a high-skill character's skill feels salient even if they fail the only roll they make. For example, our Geologist might be given a few scraps of information just by looking around a room. But all of this requires DM skill and consideration, and is not built into the game itself.

Of course, the same issue can apply in any game where some skills/abilities/whatevers are just not rolled very often. It can affect that one spell you've been saving for exactly this circumstance, or a particular power, or even a bit of personal background that makes the Dungeon World GM decide to allow you a roll just this once. It's just relatively common in Call of Cthulhu, partly because there are a substantial number of skills which are relevant for ordinary everyday PC-types, but not necessarily needed in an investigation.

This doesn't make Call of Cthulhu any less faily, though. Sometimes this is indeed frustrating. Sometimes, when it works well, it contributes to an overall atmosphere of mystery and difficulty where the characters have to work hard to make progress, sometimes taking several alternative approaches before realising how the latest puzzle piece fits in. It can also enhance the sense of being an ordinary person caught up in something vast and strange, though that depends very heavily on the scenario. On the whole, though, I absolutely understand why some people find it frustrating. For me, the sheer simplicity and get-out-of-my-wayfulness of the system is enough of a counterbalance.

It's also perhaps worth noting that direct consequences of a failed roll are very variable. You don't injure yourself by unsuccessful punching*, only when you fail a roll to avoid damage. On the other hand, negative social consequences are common and you can (depending on the GM) infer wrong information from a theoretical roll.

* I very nobly resisted the urge to say "unsuccessful Fisting", not least because that joke no longer works in 7e.

Deathwatch and other Warhammer 40K games

Deathwatch is rather different because of its setting. It is a setting where secrecy, obfuscation and ignorance are highly-prized virtues, and where it's expected that most people can't do most things. Specifically, science and education are so fragmented, rote-based and unsystematic that it's entirely plausible for a character to perfectly identify and use one item, while being utterly ignorant of a very similar item.

This for me ameliorates the failiness of the skills system. If my technical character cannot identify or operate a machine, it's because that machine was not part of his training - perhaps knowlege of it is even expressly and arbitrarily forbidden. If a piece of lore is beyond my grasp, it's because knowledge is passed on in whispered rumour and patchy indoctrination, and truth is hard to unpick from gossip, lies and party-lines. Whereas I'd be annoyed if my Call of Cthulhu engineer is baffled by some human technology, I'd be much more okay with it in Deathwatch.

It's less good when it comes to physical skills, such as noticing things, leaping over things and so on. Here it feels like these should be pretty reliable. Even here, though, the latest iterations of the gameline have become reasonably generous at providing modifiers: +10 to vault obstacles when wearing armour only on the chest, or +30 when unarmoured; +30 to do basic maintenance on simple machinery with parts available. I think I'm okay with a fully-armoured soldier with full kit (and we're often talking reinforced armour plate here, not cloth) struggling to vault a wall. Similarly, the setting is one where some characters barely need to roll social skills - few people dare to argue with the most holy Adeptus Astartes - but otherwise self-interest, paranoia and suspicion are rampant, so it's not surprising that convincing people of things is difficult.

What really flips things around for Deathwatch and other 40K games is that where it counts, you tend to succeed. Most rolls are for combat, and in combat there is a whole slew of modifiers to help you. My latest character is a good marksman with a 48% in Ballistic Skill for shooting things: but I can use an Accurate weapon for +10%, take the Aim action for up to +20%, and make a standard attack (normal attack with no special conditions) for another +10%, giving me a 98% chance of hitting. I can therefore reasonably expect to hit things quite a high proportion of the time if I'm playing to my strengths. Even without that, I'll often end up with a 78% or so.

When you do fail at combat, it matters. Combat in this system is extremely deadly, and so the psychological impact of a failed roll at a key moment is significant.

I do think the 40K system struggles a bit with skills, and have written extensively about it. On the whole, though, especially in the latest iterations, I think I'm okay with it. It's a relatively gritty game, and it actually seems reasonable that, say, persuading someone to do something for you is relatively unlikely to succeed unless you are practiced at it.

Again, 40K doesn't mandate consequences for failed rolls. Normally failing the roll is enough of a problem; because of the tactical nature and deadliness of the game, even missing a bit of information can be a severe penalty. The more combat-heavy games tend to shrug off failed skill rolls quietly, because you're mostly supposed to be competent and dwelling on your inadequacy makes little sense; as I said, it's completely expected that you don't understand technology, for example, while the Space Marines have no real place caring about geopolitics. The more subterfuge-heavy games draw on setting and fiction that's rich with mystery, shades of grey, betrayal, forbidden lore and the incomprehensible, so even when failure occurs it mostly feels genre-appropriate.

Summing up

I think there's a significant psychological impact to failure in games, which varies with the tone and premise. I also think that what counts as "failure" must vary from person to person.

For me personally, the "succeed at a cost" format of Dungeon World-style conflict resolution systems compromises the success enough that it feels more like a failure. I think it's particularly significant that the cost comes as a direct result of your roll ("you failed to roll well enough") and the roll comes as a direct result of your action ("you chose poorly"). The result is that these systems feel failure-heavy, and this meshes particularly badly with the action-adventure tone of the game.

In other failure-heavy systems, the failure is a result of low success chances rather than a compromise mechanic. Ironically I find this less frustrating, perhaps because it's rare for your own actions to have directly negative impacts on you even when you fail. Moreover, in the case of Call of Cthulhu you don't necessarily have to do things you're bad at, while in Deathwatch et al. the system ensures that in the arena you are forced to engage in, combat, you can quite reliably succeed regardless of low scores, leaving success more about tactical play than simple stats and rolls.

In my next post, I'll be talking about consequences of failure, and whether interesting consequences are more interesting than uninteresting consequences or even no consequences at all.


  1. To throw in some GURPS to the mix (do say if you don't want me doing this; I'm catching up with the last few months of posts):

    - Generally GURPS is about succeediness more than failiness. Skills you're pretty bad at may be rated at a 9-10 (38-50%), and it's not unusual to have a few main skills at 15 (95%) even as a low-powered character. Yes, there are lots of modifiers to stack on top of that, but the unmodified roll is generally for doing the thing in some haste in an adventuring sort of situation.

    - GURPS introduces the defence roll. So the process of hitting someone is (you make your attack roll, and succeed, or stop here), (they make their defence roll, and fail, or stop here), (you do damage, which may be stopped by armour). That has its own psychological effects: it's annoying to think "aha, I hit… oh, no, that's being taken away from me". And when you're being hit, it's definitely because you failed at something. But at the same time you retain the possibility of the nobody-does-anything null outcome.

    Having both attacks and defences lets you do tradeoffs. There are three basic approaches: normal, all-out attack (I really care about hitting or damaging, and I'm going to give up my defences to do it), and all-out defence (I give up my attacks to make myself as hard to damage as possible). Then you can get cunning with things like feints, which penalise your attack but lower the other guy's defences if it works. All of these things work on a psychological level as a sort of mini-game, where winning or losing a fight is a result of your choices as well as your die rolls

    1. No no, please do, I always like comments and to know people are reading.

      That was certainly the impression I got in our 2015 GURPS game. I feel like between the 3-dice pool and the relatively high skills, most of the time you can have a bash at things and expect to do okay. It's got a reasonable balance of realism and cinematic play, from what I can see.

      I think on balance I slightly prefer opposed combat rolls for most styles of game, and my main reservation is more about slowing down play (and slowing down combat) than the mechanic itself. And for games where you're broadly aiming for occasional small-scale combats, making the combat complex and interesting seems one of the more sensible ways to go.