The combination of three statistics for every idea and path dependence for the process should ensure that players have to make important in-character decisions, and that those decisions will make a difference. However, the results of the die rolls should also make a difference. The players should not be able to map out their whole strategy before they pick the dice up for the first time. After all, if we don’t have the unexpected, what’s the point of having randomisers?

So, let’s start with a very simple mechanic for the core. Roll something (to be determined later), and the result is the current score for the idea in one of the statistics. Half of that score is noted as the difficulty for each of the other two statistics. In order to improve either of the others, you must beat the difficulty with your die roll, and the quality becomes the amount by which you beat it.

An example will help. Let’s call the statistics A, B, and C. First, we roll for A, and get 10. A now has a quality of 10, while B and C both have difficulties of 5. Next, we roll for B, and get a 10 again. B now has a quality of 5, while A has a difficulty of 3 and C has a difficulty of 8. Finally, roll for C, getting another 10. C has a quality of 2, while A has a difficulty of 5 and B a difficulty of 7.

Obviously, the statistic you roll for first tends to have the highest quality. On the other hand, you might want to save your better scores for later, to ensure some quality for the other statistics. Already, we have the need to make significant decisions. Similarly, a surprisingly high roll for one statistic creates problems, because it makes the difficulty for the other statistics higher than expected.

This is a good start, I think, but there are only three rolls per idea, which is a bit limiting if there are four or more players. The choices are also a bit limited, and there’s nothing you can do about a low roll for one of the qualities. These problems both point in the same direction: more types of action.

At this point, I need to clarify what I mean by “path dependence”. There are two forms of this. In one, which we might call “weak path dependence”, the current state depends on the path you used to get there, but only the current state matters when taking further actions. In the alternative, “strong path dependence”, two situations with identical current states still vary in how they are influenced by further actions, depending on the details of the path. In a role-playing game, you want weak path dependence, because otherwise there is jut too much to keep track of. In fact, you can always recast strong path dependence as weak path dependence, by making more of the history part of the current state. The number of statistics that would be needed to do that is a measure of the complexity of the situation.

Although I do want path dependence, the only statistics I want to have for an idea that is being created are a quality and a difficulty for each statistic. This means that the additional actions I allow must act directly on those statistics.

The first thing to do is allow attempts to improve the quality of statistics that have already been generated. The obvious choice for the difficulty is sum of the difficulty of that statistic (there’s a naming problem that needs addressing here) and the quality. The simplest rule would be to add quality based on the amount by which the new roll exceeds that difficulty, but that would make it possible to simply re-roll the very first attempt. So, instead, let’s say that you only get to add half of the excess. Half of the increase in quality should be added to the difficulties of the other statistics, as before.

Going back to the example, C has a quality of 2 and a difficulty of 8. The difficulty is therefore 10, and if you roll a 13 you get to add 2 to C’s quality (3 divided by 2, rounded up), for a total quality of 4. You also add 1 to the difficulties of the other two statistics. Suppose you now want to boost B. It has a quality of 5, and now a difficulty of 8, so the difficulty for the roll is 13. If you now roll 17, B’s quality increases by 2, to 7, and the difficulties of A and C increase by 1 each. On the other hand, if you’d rolled that 17 first time round, when B had a difficulty of 5, B would have had a quality of 12.

Obviously, if this is the whole of the rule, it is only going to be useful if you roll low the first time round. That shouldn’t be ignored, as it gives you a chance to do something about an unlucky roll, but it’s not enough. I’ll come back to that in a moment, because there’s another complication to look at first.

The relatively high difficulties raise the question of what should happen when you fail to beat the difficulty. There should be some cost for failure, beyond time taken, to discourage repeated rolls in the hope of a really lucky result. Fortunately, there’s an easy option. If your roll result is less than the difficulty, the amount by which you missed is added to the difficulty for that statistic. You’ve started up a blind alley, and it’s going to be difficult to get back to a productive track.

What happens, then, if you fail on the initial roll to generate a statistic? Obviously, the difficulty goes up, but I’d say that the statistic itself stays at 0, and thus you would still add the full amount by which a roll succeeded. This way, it is easy to track whether you have to halve your success total or not: if the statistic is over 0, you have to halve, while if it’s still 0, you don’t.

We still need additional actions. There are two obvious results that they can have. First, they can reduce the difficulty of a statistic. Second, they can provide additional dice to roll.

An action to reduce the difficulty of a statistic is rolled against that statistic’s difficulty. The amount by which it succeeds is subtracted from the difficulty of that statistic, but added to the difficulties of the others. I suspect it is probably best to do this without halving, to make it hard to just repeat this, cycling round the statistics, until all of the statistics are up around the maximum result of the dice you are rolling. This is something for playtest, however.

Actions that grant more dice are a bit harder to design, because granting as many dice as the level of success is obviously excessive. They should probably be a bit more specific, and grant one, two, or three dice. This will depend partly on the difficulty, but also on the resources that can be brought to bear.

The end point of this sequence comes when the characters (and players) decide that the idea is good enough. There will come a point when the characters cannot realistically expect to improve it, and there may come a point where the difficulties for all statistics are higher than they can roll, and the players will also know that, so even if there is no in-game time limit it can’t go on for ever.

I think this works as a first pass; a quick test has it doing the right sort of thing. The end result depends on the order in which you approach the statistics, but the situation at any point is entirely described by six numbers: the quality and difficulty of each statistic. One interesting feature is that it doesn’t require a GM; I suspect that won’t be sustainable for the game as a whole, but we’ll see.

However, this mechanical structure is still not a role-playing game. There’s no indication of what the rolls represent, and so no way to role-play what you are doing. That’s something to start discussing next time.