Yes, yes, yes, in D&D you kill things and take their stuff, and as a result you get better at killing things and taking their stuff. Not what we're talking about here.

Tony Lower-Basch got our local ad hoc gaming gang together to run, within a week of each other, one session each of Vincent Baker's morally agonizing Western game Dogs in the Vineyard (Actual Play posts here) and Tony's own relatively lighthearted superhero game Capes (Actual Play posts here) -- two very different games, but each built around an incentive system that strongly drives its particular style of play. What's particularly interesting is that each system rewards certain forms of defeat as much as, or more than, victory: Dogs tempts players to escalate conflicts and gives them new, double-edged character traits as they suffer ("Fall-Out"); Capes allows players to invest one kind of resource (Debt) in improving their chance of winning a conflict, but converts the winner's investment into a reward for the loser in the form of a more powerful resource (Story Tokens).

I think there's a lot of potential in rewarding defeat, for two big reasons:

1) Game & Story: No interesting story has one guy win all the time; there are ups and downs, triumphs and failures, moments of doubt and moments of certainity. Positive feedback loops as in D&D -- where the more you win now, the more likely you are to win later; the more you lose now, the more likely you are to lose later -- are less interesting a gameplay challenge than negative feedback loops as in Capes --where the more you win now, the more likely you are to lose later, and the more you lose now, the more likely you are to win later.

2) The whole point of roleplaying in the first place: If you have an amazing story you want to tell... go write a novel. The whole point of roleplaying is that you get to enjoy the imagination of other people, your fellow players, and what you come up with together is better than what any of you might have produced alone. But here's the thing: If you really let everyone have influence over the story (i.e. if you reject "the impossible thing before breakfast" that somehow the GM is the author of the story, yet the players control the protagonists in that story), then some things you want to see happen definitely won't. Defeat is inevitable. So why not embrace defeat and make it as much fun as victory?

Two questions, then:

Generally, does my theory of the Joy of Defeat make any sense to people?

Specifically, what other systems have people seen, played, or invented that reward failure? Trollbabe's "the loser gets to narrate" rule comes to mind, but alongside Dogs and Capes, that's only three examples in a vast sea of systems where "success is good, failure is bad."

No big surprise that I'm a fan of negative feedback loops. But I'll add another item, which is not specifically related. More of a "two great tastes that taste great together" thing:

• If you reward victory and punish defeat in equal degrees, you have created a zero sum game. At that point, any player who wins (counting the GM as a player) is doing so at the expense of one or more other players. Even the GM, who has supposedly infinite resources, will often find individual situations where player victory would mean stinging GM defeat ("They can't kill my bad guy NOW! It ruins the whole story!") Setting aside the social ramifications, a zero-sum game encourages players to take risks only when their odds are as assured as possible. In short, it encourages tactical caution. D&D is often played as a zero-sum game.
• If you punish defeat MORE than you reward victory then you have a negative sum game. Any confrontation, even if it benefits one player, worsens the overall situation of the group. Good strategy is to avoid any possible type of confrontation, no matter how much weaselling you have to do in order to do it. Caution becomes cowardice, and then paranoia. Amber and V:tM are often played as negative-sum games.
• If you punish defeat LESS (or less often) than you reward victory then... I don't know what that is, and therefore I think it's very worthy of discussion. I think that InSpectres and My Life with Master fall into this camp, but I'm not at all sure.
• But if you reward victory AND you reward defeat then you have a win-win game. Picking a fight with another player is always a benefit in such a game, and always welcome (assuming rational participants). High-risk, and even self-destructive strategies are logical and inevitable. Capes, Dogs and Universalis are designed in this way. The only way to lose is not to play.

And... yeah, I got one more thing to say, more directly in response to Sydney's point. Positive feedback loops lead to death-spirals (where once you start losing you have, statistically, lost). Negative feedback loops are self-maintaining. That's their most recognizable characteristic. You can push them as far and as hard as you want, and they'll always bounce back to somewhere near the middle ground, usually with a lot of velocity/chaos on the return trip.

This is something that I first started thinking about after playing Trollbabe for the first time. I think the key may be less of players being rewarded for not getting what they want than players getting to have some say over when they fail.

What I found with Trollbabe was that players started seeing the possibilities in failure, since they got some choice as to whether they failed or kept going. Instead of, "Oh shit, I failed", it becomes, "Failed again, hmm, if I stop here I could say that X happened and that would be cool".

I think the very fact that there is something more for the player to contribute if the character fails is reward enough by itself.

Nobilis's disadvantage system is something like what you described, Sydney. Instead of getting extra character points for taking advantages when you start the game, you get miracle points to spend when they cause you hindrance in play (or, for some constant handicaps, you have your maximum miracle points permanently increased).

Hiya,

Sydney, you know Pace, right? Check it out; there's a forum here at the Forge too. Big-time failure-driven play.

Best,
Ron

Of course, by failure driven we mean the PC's failing. The player is succeeding in their goal through that.

Which brings me back to D&D 3.x, where an encounter equal to your party level uses up about a fifth of your resources. You loose the thing that matters in D&D (resources), but that grants you XP, which makes you win latter (leveling).

I think were simply talking about how different CA would use the 'loose to win' technique, here.

On D&D: Uh... no. You're still saying they won. Heroes live, monsters die, virtue saves the day. But if they fail to kill any monsters then they get (at least in the old-school D&D I'm most familiar with) nothing. The characters lose and therefore the players also lose.

We're talking about systems where, for instance, a party would choose to go into a dungeon and come out ragged, battered, bloody, having lost all their magic items, killed nothing and been humiliated by the goblin king... because the players succeed in their goal through that.

You may well be able to craft a "characters lose" scenario that's profitable in D&D (in fact I'm curious to see how "off the beaten track" you need to go to do so) but I don't see it in what you wrote so far.

Oh yeah - Elfs. I think it's the most extreme example of winning through character failure available in a Gamist context.

Best,
Ron

I'm assuming that what is "won" in these games is "player satisfaction" and not an in-game resource--is that right? 'Cause if it's an in-game resource then I think Noon is correct.

Also: Does the "Rocky" scenario count? You go in and give it a heroic try 'cause the challenge was way over your head anyway? Or does this have to be the "Of Mice and Men" scenario where the defeat is complete and ultimate and play stops there (as opposed to there maybe being a Rocky-2 where you might win something).

'Cause I've seen a lot of the "Rocky" mode of play -- but I haven't seen so much of the "Of Mice and Men" style play.

I once ran a game that was designed to be an operatic tragedy. The situation was stacked against the players to begin with and the players were trying to rescue people who were already dead (they didn't know it) and the game was factored (although not required to kill off all their NPC friends along the way).

Unfortunatley we did not finish due to external factors but the game was very dark and the players appreciated it going in (although they were not told it was going to be a tragedy).

I had one game end in a mutual suicide (unforseen on my part) by the PC's after it became clear they were up against a foe they weren't going to beat, had been betrayed by their familes, and were mutating. The suicide was a *kind* of victory: they were denying the forces that wanted them a chance to have them.

But it was pretty dark too.

I don't consider those far off the beaten path (one was run with GURPS. One was run with something like GURPS). Both were considered functional games.

[Note: I saw a lot of this in my CoC games, now that I think about it ... ]

-Marco

Marco: "Won" meaning when the characters win... or lose, or... Uh...

Okay, I'm confused. Your first sentence can (expanded) read as either of the following:

• "I'm assuming that what is "won" (by the players, in the act of letting their characters lose) in these games is "player satisfaction" (e.g. a sense that the story is going as it should be) and not an in-game resource (e.g. Story Tokens to buy later control over the story) -- is that right?"• "I'm assuming that what is "won" (by the players, in the act of making their characters win) in these games is "player satisfaction" (e.g. they beat the bad guys and therefore feel bad-ass) and not an in-game resource (e.g. they beat the bad guys and therefore get lots of experience points) -- is that right?"

If you can clarify what you mean then I can respond. Then, with the common understanding, I might be able to tackle your later points.

I mean it in the first sense (1). And, yeah, I guess I was applying it to story-direction. If a D&D group goes up against a foe they can't "beat" and gets humiliated and beat up--but escapes with Xp to make them more effective later, would that count?

I mean, I think it's a pretty blunt, very basic sense of having the charcters lose but the players think their goal was accomplished--but I sort of think it would count.

-Marco

You're talking about players getting their characters into situations where they will probably have no power to achieve victory.

We are talking about situations where the players have the power to give their characters victory, and choose defeat instead.

For instance, in Dogs in the Vineyard, a fifteen year old girl slaps a Dog in anger during a conversation, escalating to Fighting in order to get the dice to continue the conflict. The Dog's player now often has a choice: Escalate to fighting as well (slap her back), take terrible Fallout (curse her out in a way he'll be ashamed of later) or Give. If they Give, it is not because they were defeated beyond their ability to respond (as you say "goes up against a foe they can't beat") it is because the player chose to lose for the benefits it can garner (in this case, being able to face themselves in the mirror).

Am I reading you correctly? If so, does this help explain why I've been finding it so hard to answer your questions from my viewpoint?

Heya,

Another application of this observation is possible when games provide different advancement rewards for success and failure. Say, for instance, success leads to social status advancement while failure leads to exp rewards. Both improve the character, but in different ways. I woulnd't mind seeing a game designed with that principle in mind.

Peace,

-Troy

Sorry, didn't answer this one. Might as well be explicit....

Marco wrote: If a D&D group goes up against a foe they can't "beat" and gets humiliated and beat up--but escapes with Xp to make them more effective later, would that count?

Okay: I can see that's a difference--I was looking at it as a choice of game level decision in my cases rather than a during-play level decision (i.e. if I choose to play CoC and say "make it dark" I'm choosing to 'lose' for personal satisfaction). As a player looking at two game books (CoC and D&D) I have "the power to achieve victory" based on which one I choose to play (and, really, how I choose to play them--either could be doomladen or victorious in spirit).

I have seen players do things exactly like the "slap" scenario (the player's bad-ass character gets chased out of the bar by the bar-maid) for reasons not related to mechanics. In fact, I don't think it's all that rare unless the stakes are high (continued play in the game).

But I've rarely seen an *entire* game hinge on losing by the intent of the players in a game where it wasn't 'lost' to make a statement (i.e. we play through the whole game and, during the last play session, when it is clear there will not be any more, we are battered, bruised, and defeated but consider the game a success for some reason not related to the story generated by play).*

It's a good point though.

-Marco
* In the two games I cited the players and I discussed the fact that the games would be "very dark" so it wasn't a surprise and the players were, IMO, complicit in the theme of the game (in fact, in the case of the suicide, while they couldn't *beat* their opponent, they could've "gone along with" their opponent and had a much better outcome down the line--it was a fairly informed and passionate choice to 'end it there.')

I suspect that, in this context, "winning" is more about player narrative control than about Maximum Game Fun. The following with a massive dose of IMHO:

One thing that Capes and Dogs (and others, but these were the original examples) do very well is that they give players a great deal of narrative power before the outcome of a conflict is resolved.

In Dogs, every single trait on a character's sheet is relevant to their story, and the player gets to choose what these traits are, more or less at will. The players get to decide how their contribution to the conflict will play out, by choosing which traits to engage, and whether or not to Escalate.

Even so, the system and setting virtually demand that the players go flat out to win every single conflict - I think that the "say yes or roll" really contributes to this, because every conflict is important. believe that the Dogs win conflicts far more often than they lose them, but only because they are willing to pay for it in Fallout.

In Capes the players get to declare their own Goals and Events, and can do anything they want as long as it doesn't resolve a Goal or Event to early (or violate the Code.) That's a helluva lot of narrative control already. The difference from Dogs is that the players are encouraged to include conflicts that they don't care about (but that another player does care very much about!) in order to harvest enough resources to introduce - and win - a conflict that they do care about.

(I think this difference is key to understanding why Dogs has a GM, and Capes doesn't, by the way.)

I think this is an entirely different means of enabling narrative contribution, from systems where the player is allowed to narrate their own character's failure (Monologue of Defeat). This is still empowering but in a different way.

I guess that the difference is between:

"winning even when you lose" - games with a Monologue of Defeat

"winning because you lost before" - Capes, Pace

"winning because you paid the price for winning before" - Dogs

Sydney, does this help with the Joy of Defeat?

PS Crossposted somewhat, I think this is still relevant if you're interested in taxonomy.

I mostly agree with what you're saying, but would propose a slight rewording. Capes isn't about "winning because you lost before". It isn't really about winning at all. In-character victory is just the carrot I use to lure unsuspecting players into using the system enough to realize the freedom it gives them.

The fun of a choose-to-lose game is that in-character defeat is a legitimate option. Many games give you only one permitted goal: In-character victory. With only one goal you have no choices. You achieve that, or die trying. With two permitted goals (win or lose, either is valid) you have a choice.

Your job as a player isn't to win, it is to choose.

You can never fail at that. You can make happy choices, or dreadful choices, or stupid choices, or choose to delegate the choice to someone else, but you have succeeded in making a choice. You win!

At this risk of tangenting, what about games that reward the player for giving up power/narrative control at one point in time by giving them tokens that enhance their power/narrative control later?

Drama Points in Buffy are the example I thought of immediately, but I'm kind of playing Dogs in the Vinyard that way currently - "dice farming" by deliberately courting Fallout so as to gain experience dice.

Is that the kind of "economy of failure" you have in mind?

Best,

Mark

Heh. Mark, does that mean you're Gaming Dogs?

TonyLB wrote: On D&D: Uh... no. You're still saying they won. Heroes live, monsters die, virtue saves the day. But if they fail to kill any monsters then they get (at least in the old-school D&D I'm most familiar with) nothing. The characters lose and therefore the players also lose.

We're talking about systems where, for instance, a party would choose to go into a dungeon and come out ragged, battered, bloody, having lost all their magic items, killed nothing and been humiliated by the goblin king... because the players succeed in their goal through that.

TonyLB wrote:
No. If they went up against a foe that they could beat, but somehow found a situation where they could get more XP for losing, and lost, then that would count.

As I said, I think you might have to go a bit off the beaten path to find that in D&D.

Callan,

I believe that Tony is specifically referring to mechanical metagame rewards here. We're talking about more than the players accomplishing their CA here, we're talking about the mechanics rewarding the accomplishment of that CA.

Tony's not asking, "What games can you have fun losing in?" he's asking "What games mechanically encourage you to lose?" Capes does this through the awarding of Story Tokens. Dogs does this by making the cost of success high (you'll have to escalate beyond what you're comfortable with).

This does not happen in D&D, the mechanical incentives are all for winning: XP, phat loot, whatever. While you may be socially rewarded for losing, the mechanics do not encourage you to lose.

There's also an interesting middle ground where the mechanics don't say one way or the other. Games like The Burning Wheel and The Riddle of Steel where you are rewarded, not for victory, but for trying. Assuming that you survive the results of an encounter then you are likely stronger.

Thomas

Yes, and I'm noting that mechanically their basically the same. Tragically lose in capes and you get some story token resources. Get hit a few times and burn off some spells in D&D and you get some more resources.

Same thing, mechanically. It's just that the descriptions of what's happening that differs.

Besides, I could go all monty python and say "What, in capes your loosing tragically but your PC stays alive...and you don't think that's like D&D? Your characters alive and he's gotten more of these story point things...just the same as the D&D PC living on and prospering! In my day our characters died tragically straight away...and we liked it!. Tell that to kids today and they wont believe you!"

Noon wrote: Yes, and I'm noting that mechanically their basically the same. Tragically lose in capes and you get some story token resources. Get hit a few times and burn off some spells in D&D and you get some more resources.

I think we need to make a more rigorous definition of "in-character victory" in order to continue this discussion. For my purposes, I define thusly:

In Character Victory (for character A): An end-state in which the goals of character A (e.g. to save the town, to humiliate their opponents, etc.) have been accomplished at the expense of the goals of character B (e.g. to burn the town to the ground, to save face).

In Character Defeat (for character A): An end-state in which the goals of character A have been prevented, to the benefit of the goals of character B.

The party which departs a dungeon having been beaten and battered has probably not achieved in-character victory. However, they may have if, for instance, they leave beaten and battered but in possession of the relic they went in to find.

Additional definitions:

Marginal Benefit/Cost of Victory: The amount of resources gained by In-character victory minus the amount of resources gained by In-character Defeat. Benefit when positive, Cost when negative.

Marginal Benefit/Cost of Defeat: The amount of resources gained by In-character defeat minus the amount of resources gained by In-character Victory. Benefit when positive, Cost when negative.


When I talk about "Choose-to-lose" game-play I am discussing situations in which the average Marginal Benefit of Defeat of a game-as-played is zero or positive.

I have never seen D&D fulfill that in any consistent way. Assume that GP and XP are of equivalent value (because it makes the calculation simple).

• Defeat: You kill ten gribblings (10x300XP) and two stinkhogs (2x500XP) but are routed by the Slime Wizard (0x4000XP) and barely escape with your lives. Total benefit: 4000XP.
• Victory: You kill ten gribblings (10x300XP) and two stinkhogs (2x500XP) and spend a two-thousand GP diamond for a spell to defeat one Slime Wizard (1x4000XP). Total benefit: 8000XP - 2000GP = 6000XP.
• Marginal Cost of Defeat: 4000XP - 6000XP = -2000XP.

Now it is possible to rewrite this example in a way that creates a marginal benefit of defeat: If the diamond is five thousand, rather than two thousand, GP then the marginal benefit of defeat is 1000GP. Which is a fancy way of saying that it's not worth sacrificing the diamond to clip the wizard (in the short term... I recognize the compounding earning-potential of XP, but am leaving it out so that the example can be less convoluted).

Callan: Are you suggesting that players in D&D regularly encounter situations where the Marginal Benefit of Defeat encourages them to deliberately lose a conflict they could otherwise have won?

Edited to clarify and fix a little math.

Let me give examples of how two different games might work:

Game A: You get 50 XP if you kill the monster

Game B: You get 50 XP if the monster beats the crap out of you and runs off laughing.


Both are mechanically identical. There's no point in trying to say their different mechanically, by stipulating the character is loosing. They are exactly the same process of accumulation. There is a thematic difference in that your described as loosing to do this...it's certainly different from the majority of the market. But it isn't mechanically different.

Thomas wrote: I don't think these are actually the same thing here... As Tony (I believe) mentioned, there is a difference here. There is a point at which mechanically you still lose. If you expend those resources, but fail utterly then you don't get any other resources back.

This is not true in Capes. There is not point at which you will spend resources and get nothing back. Ever. I'm not sure about Dogs as shamefully I have not read or played it.

Okay, how about we posit Game C: You can get 50 XP and an ally by defeating the monster, or get 500 XP and the potential ally holds you in contempt if you lose.

Where does that stand in your estimation... still just the same mechanics of resource management?

The same. Because if a contemptuous ally is unpleasant for me, then it'll outweigh the 500 XP. So I'll go with the D&D advancement scheme.

However, if as a player I enjoy the idea of a contemptuos ally, then I'll go the D&D advancement scheme and collect both of these rewards. Because either way is D&D like advancement.

The contemptuous ally not being a reward to the PC, doesn't mean it isn't a reward to me, the player. Once you ditch the PC's perspective and see it from where it matters, the player perspective, you see that your collecting two rewards in both cases. In other words, both options are identical, in that they both just present rewards.

Your only case could be that the contemptuous ally might be considered a 'loss' because it might take me out of my comfort zone latter as I squirm in answering some nar question. Fact is though that D&D gamism can equally take me out of my comfort zone as I squirm over tactical descisions. Such squirming isn't a loss in either case, even though it's uncomfortable.


I'll note a clever design mechanic that Ralph described awhile ago. In the thread he described how players will squirm and argue when their PC disadvantages are applied. Because if they can argue their way out of it, they avoid the penalty. Thus arguing is rewarded by the system. So he suggested they actually get a reward when these disads are applied. Soon, he noted, you would find players trying to insert their disads as often as possible. The players themselves would want them there! Because they wouldn't be loosing, even though their PC is.

They would want to use these 'you lose' powers just like they would want to use their 'you win' powers. They'd want to in the same way, because they are the same thing.

Okay... so, continuing this logic: If I want to create a game situation that tells me whether the respect of your ally is worth more or less than 450 XP to you, as a player, Game C provides it, yes?

I give you a choice, you make it based on your personal (and subjective) standards and then we know. Maybe you think the contemptuous ally is neat, so you gladly take it, plus the extra 450 XP. Maybe you value the support you can get from an ally, and think it's worth more than 450 XP, so you win. Maybe you value the ally, but not 450 XP worth, so you reluctantly lose.

Incentive systems provide an objective currency. It is then possible to place two items in the balance: a subjective value ("The value of this ally") and an objective one ("450 XP") and judge how the player compares the two. To do that you will very often (but not always) be assigning the subjective value alongside character victory.

You're offering the players the opportunity to buy a subjective item in the SIS for the Marginal Cost of Victory (which is the same as the Marginal Benefit of Defeat). If you want that to be a sensible exchange that they can actually make a choice about then you need the Marginal Benefit of Defeat to be zero or positive, as I've said.


I agree with you that the mechanics of optimizing your resources are largely identical between Capes and D&D. Offer people the chance to earn a meta-game resource and they will try to do so.

But in concentrating on that you seem to be ignoring how having a Marginal Benefit of Defeat lets you use the system to negotiate in ways that are not available when you have a Marginal Cost of Defeat. It's not mechanically different, but the social interactions it engenders are nothing alike. Does that make more sense of what we've been saying?

Very interesting thread!

Let me see if I've followed correctly:

Basically, what is important for a designer is to consider whether it is the player via PC success or the player through a variation of PC success and defeat, who is rewarded (there's also the design that is supposed to support coherence of the PC's actions according to the game world, which is at first view independant of PC victory or success, and this can also be favored by player reward).

In fact, what we're discussing is what kind of CA we want to support.
I'm not saying that Gamism absolutely requires a positive feedback loop, or that Narrativism necessarily demands a negative feedback loop. The fact is that it changes style of play, and we must consider that from that point of view.


Side note: Since we're talking a lot about player choice here, we must avoid situations where a player will be forced to make a certain choice.
For example by awarding exclusively one kind of ressource for success and another for defeat. If both ressources are necessary, the player will have to make the other choice to continue play.

Artemis wrote: Side note: Since we're talking a lot about player choice here, we must avoid situations where a player will be forced to make a certain choice.
For example by awarding exclusively one kind of ressource for success and another for defeat. If both ressources are necessary, the player will have to make the other choice to continue play.

TonyLB wrote: Okay... so, continuing this logic: If I want to create a game situation that tells me whether the respect of your ally is worth more or less than 450 XP to you, as a player, Game C provides it, yes?

I give you a choice, you make it based on your personal (and subjective) standards and then we know. Maybe you think the contemptuous ally is neat, so you gladly take it, plus the extra 450 XP. Maybe you value the support you can get from an ally, and think it's worth more than 450 XP, so you win. Maybe you value the ally, but not 450 XP worth, so you reluctantly lose.

I agree with you that the mechanics of optimizing your resources are largely identical between Capes and D&D. Offer people the chance to earn a meta-game resource and they will try to do so.

But in concentrating on that you seem to be ignoring how having a Marginal Benefit of Defeat lets you use the system to negotiate in ways that are not available when you have a Marginal Cost of Defeat. It's not mechanically different, but the social interactions it engenders are nothing alike. Does that make more sense of what we've been saying?

Noon wrote: I am ignoring those benefits. Because your evaluating them from the wrong level. Gamism is about me winning. Nar is about me deciding what is winning. But you keep pitching it in terms of the characters defeat, when both nar and gamism are clearly about what the player is doing. In neither case is the player choosing defeat for himself.

TonyLB wrote:

Noon wrote: I am ignoring those benefits. Because your evaluating them from the wrong level. Gamism is about me winning. Nar is about me deciding what is winning. But you keep pitching it in terms of the characters defeat, when both nar and gamism are clearly about what the player is doing. In neither case is the player choosing defeat for himself.

So you're saying that we should recognize that there's a difference between character-defeat and player-defeat, yes?

I agree. Shall we move on to discuss the ways in which rewarding character-defeat can open up negotiations about the objective value of subjective SIS elements?

I think were simply talking about how different CA would use the 'loose to win' technique, here.

I think I'm pretty much in agreement there.

The one caveat I'll put in is that Narrativist choices don't have to be difficult to be Narr. A lot of time in Narr games I hear people say "Wow... the moment that question came up I knew what the answer had to be... but before it had been asked I hadn't even thought about it."

I think your "not anything where 99% of the population would choose the same way" criteria is better than the "difficult choice" thing. That seems to correlate well with the much more slippery (and less profitable) notion I have that the choices should convey a "message".

EDIT: Better yet, perhaps, is the notion that the player should be answering the question in play. That can't happen if the answer is already known.

John Kirk wrote: (...)
1) The rewards must be exactly equal for both winning and losing.

2) The rewards for winning and losing are of a different nature, both of which are needed at various points in game play.

3) The "reward" for winning is something that is desirable sometimes but undesirable at others.

Note that (1) strikes me as being rather boring. (...)

If rewards for winning and losing are of the same type, but are unequal, then the player will always have an incentive to take one option over the other. (...)

Note that criteria 2 and 3 don't really give players the option of always winning or always losing. They do, however, give players the option of deciding when to seek a win and when to seek a loss.

TonyLB wrote: I think I'm pretty much in agreement there.

The one caveat I'll put in is that Narrativist choices don't have to be difficult to be Narr. A lot of time in Narr games I hear people say "Wow... the moment that question came up I knew what the answer had to be... but before it had been asked I hadn't even thought about it."

I think your "not anything where 99% of the population would choose the same way" criteria is better than the "difficult choice" thing. That seems to correlate well with the much more slippery (and less profitable) notion I have that the choices should convey a "message".

EDIT: Better yet, perhaps, is the notion that the player should be answering the question in play. That can't happen if the answer is already known.