Here’s a fun dice pool system. I’ve tested it and it works fine. There was a discussion about it at one point on Michael Prescott’s G+ feed, I think, but now that that’s going away I want to preserve it somewhere. Someone else (maybe Michael) started calling it Milton dice so now that’s what it’s called.

- Your attributes and your skills give you a number of d6s to roll.
- The GM sets a target number to hit.
- Roll those dice.
- Add up the results on all the dice,
**but 4s, 5s, and 6s count as 0s. **(So for example, you roll 4 dice and get 1, 1, 3, 4. Your total is 5 because the 4 counts as 0, and you just add up the 1, 1, and 3.)
- If your total is equal or greater than the target number, then you succeed.

What’s cool about this system?

- By only counting 1s, 2s, and 3s, each die has an average result of 1. This is because 1+2+3=6 and there are six sides on a die. It makes things really easy to calculate. If I’m rolling 5 dice, then my average result is going to be 5. Likewise, setting target numbers is easy. A TN of 5 will probably be hit by a character rolling that many dice. It has a pleasing transparency that a lot of dice pool systems lack, as Luka pointed out. The odds are a little funky in the 1 or 2 dice range, but it evens out nicely after that.
- You only have to add up 1s, 2s, and 3s, which makes the addition really fast.
- You can distribute 6 points around a d6 in other ways, like make 1 and 2 be worth 0, 3 and 4 be worth 1, and 5 and 6 be worth 2, but that gets confusing without custom dice. This uses the pips the dice already have.

*Related*

## Sean H

February 5, 2019 at 4:14 AMIt has been a minute, but I remember Jovian Chronicles (and presumably Heavy Gear) uses a system where a +1 to a roll or result, was a big deal. With small TN, you can still have a wide range of results, which could be useful depending on how you wanted to spread out results. Seems like a solid idea.

## Yolande d’Bar

February 5, 2019 at 4:10 PMYes but could you provide an example of what I’d use this for and how?

## Kevin

February 6, 2019 at 4:27 AMIt’s a cool mechanic, but I don’t really understand how it is considered transparent. The gist of transparency seemed to be the ability to easily understand the odds of a roll. Without looking at the AnyDice output, I have no idea what the odds are of rolling 10 or higher on 7 dice. How is this mechanic more transparent than a TN vs 2d6 which is considered opaque?

PS: An easier way to get the result on AnyDice is:

loop DICE over {1..10} {

output DICEd{1,2,3,0,0,0} named “[DICE]d”

}

## Ben

February 6, 2019 at 6:40 AMIt’s more transparent in the sense that you know your average result right away. If the TN is 4 then I’m going to need at least 4 dice if I want to have a good chance of success. It doesn’t give you exact odds, so it is less transparent than say, a percentile system.

## Kevin

February 6, 2019 at 3:52 PMSure, I get that, and it is a nice property of the mechanic. However, I know the average result of a 2d6, too, but it is considered opaque. That’s why I’m a bit confused on what exactly constitutes an opaque or transparent mechanic.

It seems to me that on the continuum between opaque and transparent, this mechanic would be considered more opaque than 2d6, but more transparent then a dice pool system where both the target number and the number of dice change.

The mechanic is similar to one I’ve played around with where three sides of a d6 have the value of one and the other three have a value of zero. The average result of a die is 1/2. Instead of adding a constant modifier to a d20 roll, I added a number of these dice to the roll.

## Luke Hawksbee

February 8, 2019 at 1:11 PMLike Kevin, I’d question the use of ‘transparency’, but I do like some of the properties of this system. The method of rolling seems like a pain to me, though, and I think if I were going for a dice pool like this I’d want it to be a bit less swingy.

Here’s my modified/simplified version: every die is functionally 1d3-1 (range: 0–2, mean: 1). This would be easier to implement in virtual dice rollers, and when playing with physical dice I’d use a ‘count the lines’ method with Fate/FUDGE dice: a blank face is 0, a – is 1, and a + is 2. I think that would probably be even faster than having to remove 4+s and count remaining pips.

This would clip off the long tail of very unlikely high results that you get with higher numbers of dice, as well as reducing the chances of rolling absolutely zero—I think that would be desirable for the kind of ‘adventure games’ I would like to play, but may not work for everyone. I’m the type who doesn’t generally have much interest in extreme fluke results, crits, fumbles, etc so it suits me fine. However, it since the mean remains 1, it preserves the rather aesthetic and practical relationship between dice and target numbers, as well as the probability curve of rolling ≥X on Xd.*

Another nice thing about both systems is that they still give you fairly easy methods for integrating other sub-systems or optional rules or whatever: for instance, ‘passive’ scores or ‘taking ten’ can just be based off of the number of dice you have: 10d in arcane lore is a passive arcane lore of 10, and you can use choose to use that rather than roll (when circumstances are suitable, as with the old ‘take 10’ rules), or whatever.

I think I might have found the perfect dice system for a game I’ve been thinking about that I would ideally like to be diceless, but which might have to have a low level of randomness built into it (or at least offered as an optional system) without seriously screwing up the parts that do function perfectly without randomness (like passive scores)…!!! I’d been thinking about using dF already, but it wasn’t until this post that using them as 1d3-1 by counting the lines occurred to me and all the rest more or less fell into place, so thanks for the inspiration!

* Which, while we’re on the subject, is a little bit weird in both dice systems because that probability falls off as X rises: it’s harder to roll at ≥10 on 10d than it is to roll ≥2 on 2d (57.58% vs. 66.67%). (Although weirdly the odds of rolling ≥1 on 1d and ≥2 on 2d are the same, which I assume is what you mean by odds in that range being ‘funky’) The implication of this is that the game will get harder over time if dice pools and target numbers rise in tandem, so the power curve is the reverse of traditional D&D (where most people seem to consider higher level characters more difficult to effectively challenge and lower level characters very fragile—the famous housecat issue, etc).

Of course this could be juiced a little by fiddling with average dice and target progressions, etc. Alternatively you might decide this power curve makes more sense, depending on the tone of your game, etc. After all, it makes sense that everyday people are generally competent at performing everyday tasks, whereas even world-renowned experts can quite easily make mistakes and fail at very difficult tasks in their area of expertise; it’s much more likely that the world’s best F1 driver would crash and injure themselves while racing than it is that the average person would seriously screw up while making a sandwich.

## IVIaskerade

July 9, 2019 at 4:42 PMI’d combine this with the GoblinPunch system of magic – so you choose how many dice to roll from your pool, and any that roll 4+ are refunded to the pool after the spell.

This doesn’t change your maths at all, so all the work you’ve done about TNs is still valid.

What’s I like about synthesising the two systems is that dice that don’t contribute still hit that dopamine release for “doing something” because you can use them again, while still introducing a resource management aspect to the game instead of always having your full power available – and the fact that it’s a dice pool gives you a nice physical representation of your reserves.