I've been mulling around some ideas in my head about die systems recently.

(Not that I have an actual game in mind to use them for at the moment, and it could be argued that the system is something that should be considered AFTER the setting, so shoot me, I just like die systems).

Anyway, I've been thinking about the idea of a system where the player always rolls 3 dice (d10's) and, depending on the situation, takes either the high, middle or low die (middle = the higher value in a double e.g. 1,3,3 = 3 and 7,7,9 = 9, and the value itself in a triple e.g. 4,4,4 = 4).

This would then be added to an attribute (say) for a final value, which would be compared against a difficulty. Doubles and triples might be important for extra special successes or failures, etc.

The idea was that the high, medium, low mechanic might be determined by skill level at a task, so if I'm trying to do shoot someone untrained, I roll and take the low number, add it to my dex (or whatever) and see if I hit. If I'm somewhat trained, I might roll medium, and if I'm a marksman, I roll high.

I wrote a small program to determine percentages for me. They look something like this:

High
% chance of getting at least a...

1... 100
2... 99.9
3... 99.2
4... 97.3
5... 93.6
6... 87.5
7... 78.4
8... 65.7
9... 48.8
10... 27.1

Medium
% chance of getting at least a...

1... 100
2... 99.9
3... 94.4
4... 84.7
5... 72
6... 57.5
7... 42.4
8... 27.9
9... 15.2
10... 5.5

Low
% chance of getting at least a...

1... 100
2... 72.9
3... 51.2
4... 34.3
5... 21.6
6... 12.5
7... 6.4
8... 2.7
9... 0.8
10... 0.1

Any thoughts? Has this been done before? Stupid? The first problem I can see is that with such a difference between reasonably expected results (50
% of the time you get a 3 with low but 50% of the time you get a 9 with high) it makes determining reasonable difficulties hard, doesn't it? Can someone with a better grasp of mathematics see pitfalls or good ideas for this kind of system?

How about bonuses and penalties? Simple additions to or subtractions from the final result? Extra dice (somehow), etc?

Just after thoughts really. I don't know if the system is viable but I kind of liked the idea of it (it may not even be original, but I have not come across it before), so I'm just trolling for peoples thoughts.

Brian.

Brian,

Just a heads up, first, if this seems to dissapear, check for it in the Theory Forum. The mandate in the forum sticky notice is that only threads about actual games in development should be here. If this is to be part of an actual system, say so quickly, so it can stay here.

As to the mechanic itself, actually, something somewhat similar was proposed here not too long ago, though I can't rememeber exactly where. Clinton's game Dunjon uses something similar for the "Standard" method of CharGen. In his game, you roll three dice, and take the median value, and that becomes the value for the statistic.

But in any case, I think it's definitely worth further investigation.

Looking at the "medium" mechanic, I wonder why you don't use the median? IOW, if you roll doubles on a medium roll, use the double value. It's easier to process mentally, and will produce a more naturalistic result (the one y0ou have is skewed high).

That said, I think that the differences in the results produced could be pretty realistic. A lot depends on your range of abilities to which you are adding these values. But that can be adjusted by changing the die type. Given a somwhat standard range, however, it should work fine.

Let's look at some examples:

If Dex is added to shooting, frex, and Dex is rated from 1 to 20, then the roll will be relatively meaningless. If a character has a Target Number (TN) of 18, then a Dex of 18 or higher makes the shot certain. But with a Dex of 7, the shot is impossible. So we need a narrower range than that.

So let's consider 1 to 9 (wich is nice because it's limited to single digits). Here there are no reasonably selectable TNs that are impossible and one can pick ones that are also always non-certain. For example, a TN of 11 can be achieved by anyone, but is not automatic for anyone either. But then the skill level is critical. A character with a 1 Dex (obviously awful on this scale) and a high skill still has a 27% chance of success. And just going up to slightly less terrible, say 3, means that one now has a 65% chance of success. On the other end, an untrained person of Dex 5 (average, assuming this scale is a bell) would have only a 22% chance of success. Even the 9 Dex peak human would only have a 72% chance of success. A Dex 4 human marksman does better.

This all seems very cool to me. I believe that skill does have these sorts of advantages. That even a "Klutz" can learn to do things reasonably well. And that even the most highly talented individual will have a good chance of failure if unskilled.

But there are some problems. I'd say that in fact, the level of ability associated with a high stat and low skill is too high. No matter how Intelligent or Educated you are, if you are untrained in Surgery, you shouldn't even think of attempting it. Now, that can be represented by some sort of difficulty factor, but that leads to the next problem.

Simple numerical bonuses or penalties will make certain attempts impossible or certain (I should point out that I'm trying to reduce these as I always like to leave the longshot; if you are comfortable with them, then I see less problems). The usual way around this is to reduce the scale for stats even further. Such that you'd want to have only, say 3 to 7 with these dice to allow for from -2 to +2 bonuses. But that takes the granularity far from what I'd call fine. Again, you might be comfortable with this, but I'd advocate a larger range.

But how to avoid the original problems with the range being too large. Well, the easy answer is to go to d20. This would allow you to, say, rate attributes, in the 6 to 14 range with -5 to +5 modifiers, and still have all results fall in the possible range. TN 21.

Then there's the possibility of bonus/penalty dice. This can be made to sorta work, but most systems I can think of are going to heavily skew things if you go that way. Before we get too deeply into that, however, I think we need to look at a potentially much deeper problem (which I should probably have mentioned first).

What if you want an open-ended range of abilities (again, this is not problematic if you don't mind certain or impossible results)? For example, if Strength is a stat, then you might want to allow it to go much higher depending on what you're rating. If 16 is max human, then an elephant is going to have about a 21 strength assuming an exponential sort of curve, or an 80 assuming a linear (all other reasonable possible curves will be somwhere in between). And if you need to include stats on a 30 ton dragon, it's going to be worse (elephants rarely exceed 5 tons).

The usual way to get past this problem is to allow for rolls to explode somehow. Making the die roll itself open-ended. This could be a use for those doubles and tripples. Doubles are pretty common (26.29%), actually, while tripples occur only on in 100 rolls. The option that I see that presents itself most readily is to allow for an additional roll on a roll of doubles that is read just like the first, and added on. This makes the unskilled able to do anything, potentially, but, geven the curve, makes most things still unlikely. This will work especially well with difficulty modifiers.

Save Tripples for something unrelated to normal resolution, but which is a focus of the game. So if the game is about "Saving the Earth From Certain Doom", have a roll of tripples mean the apocalypse is beginning. Or in a game about spies, a tripples means that somehow, something serendipitous happens to reveal the villain's plans. Or something like that.

Anyhow, there are other options, but the above should give yu something to chew on for awhile. Given all that, what sounds good?

Mike

What Mike said. Mostly.

I think that part of the concern here results from the tendency to be stuck using target numbers. I see quite a few problems with TNs. They largely stem from the problem that tasks simply don't have a set difficulty. The difficulty of a task is dependent upon a person't skill level.

Brain surgery should be impossible for someone with no medical training. It should be hard for a recent med-school graduate. It should be routine for a brain surgeon. Can a TN-based system model this? Not if it is open-ended (unless you tack on some override rules) and not if the range of possible values is the same for those with low skill as for those with high skill (as in your system).

I suppose you could make some tasks require minimum skill levels to perform, but that gets very cumbersome.

This is something I've been struggling with myself. If anyone has a good solution...

~szilard

I disagree slizard.

I think that you have some cognative problem with how this can work (especially given your design focus on it), but it can. In the end, any system is going to assign a final odds to succeed, and it will either come up success or failure. The criteria you desire can be produced with open ended systems, and target numbers (to be techical such a system can't create a zero odds of success, but it can make the odds so long that one might as well not bother rolling). I assure you. It may not be easy to produce, but it can be done.

Further, the proposed system is an excellent candidate for this. It focuses quite heavily on the skill portion of whatever is used to do the task at hand. It just needs to tone down the base chance of success for high stats. Which with the open-ended system can be partially accomplished simply by raising the TN for brain surgery and the like. Thus, if we make the TN for brain surgery 15, this means that a Dex 8 Highly skilled surgeon has better than 80% chance of success. While the same man, Unskilled (forget trying at all if you aren't at a high Dex; patient will die almost certainly) has only a bit more than 15% chance. We can jack up the TN further for a more dangerous surgery to say 18, which changes the odds to a bit more than 30% for the talented surgeon, and .2% or so for the Unskilled guy.

Again, I agree that this is still too close, but that can be fixed, as I said by reducing a bit further the maximal effect of stats. I think with a few tweaks we can get the "easy" brain surgery down to a small fraction of a percent, and the "dificult" surgery down to a fraction of that.

BTW, another way to make something like Brain Surgery hard is to have two rolls for it. One to know what to do, and another to do it right. That can make the outcome nil.

And in the end, there is always the simple rule that lots of games use that says that certain unskilled tasks are, very simply, impossible. I agree that it's a croc, but one that isn't too unpalatable. Personally, I like to allow rolls to do amazing things, and upon success, just narrate in some expanation:

"You've never been trained in brain surgery; how did you save her?"

"Well, it turns out that the problem was vascular, and as a plumber, the solution seemed obvious. A little duct tape, and she'll be up and about in no time."

Happens in the movies all the time. Heck, if you're MacGuyver, it happens several times an episode.

Mike

Thanks for the comments.

Sorry about the misplaced post, I guess I should have read the sticky first. OTOH, I do kind of have a game in mind for this. Several years ago I wrote an entire RPG based on the storyteller D10 system (I had delusions of sending it to WW). It's an "invasion of the body snatchers" type game, but the characters are the body snatchers instead of the humans trying not to be snatched. Anyway, this system would potentially be used for a re-write of that, so it is kind of for a game in development. :-)

Looking at the "medium" mechanic, I wonder why you don't use the median? IOW, if you roll doubles on a medium roll, use the double value. It's easier to process mentally, and will produce a more naturalistic result (the one y0ou have is skewed high).

Actually, I considered both ways of doing the middle value. The reason I didn't keep the median value is that it works out as almost exactly median, i.e. it's functionally the same as rolling one die and taking the result - average 5.5. But now that you mention it, I guess that's not necessarily a bad thing, makes determining average values easier (since the median roll on "middle skill level" is 5.5, work from there to determine TN's perhaps.)

For attributes, I had considered a range of 9 (-4 to +4, something like FUDGE), so that's functionally exactly what you were proposing with your 9 scale system with everything 4 points lower (so instead of 11, use a TN of 7 for something that should be very difficult but not impossible for someone with a poor skill but relatively easy with a high skill. If your attribute is human average, then it's zero and the roll is all you need, simply because that means less calculation.

But how to avoid the original problems with the range being too large. Well, the easy answer is to go to d20. This would allow you to, say, rate attributes, in the 6 to 14 range with -5 to +5 modifiers, and still have all results fall in the possible range. TN 21.

Actually, it's not so much that the range is so large, but more that the curve is VERY skewed.. with a low roll, you're just as likely to roll above 3 as below 3, making 4-10 very unlikely. With a high roll, you're just as likely to roll 9-10 as you are to roll 1-7. I can easily enough do the figures again with D20 dice, but by increasing the range surely I'm just increasing the reliance on attributes and thus making the roll less relevant? In other words (ignoring attributes and just looking at the roll itself) it's bloody difficult to get a good result when rolling low dice with D10's. With D20's it's going to be even more rediculous to get a decent roll, but for a highly skilled person, every roll will be a great one. That's kind of my concern for the entire system in a nutshell, actually :-)

I do like your idea of doubles re-rolling and adding. Quite apart from giving the unskilled a chance to do anything (which I think is a good thing) it gives the players something to get a bit excited about when it happens, and that can never be bad either.

Again, I agree that this is still too close, but that can be fixed, as I said by reducing a bit further the maximal effect of stats. I think with a few tweaks we can get the "easy" brain surgery down to a small fraction of a percent, and the "dificult" surgery down to a fraction of that.

OK, I'm just going to sound dense, but what do you mean by the maximal effect of stats? I assume you mean that a high stat makes too many things easy even with a medium or low skill? It's tricky, because lowering the effect of stats means lowering the stat range. Say 1-5 (-2 to +2) instead of 1-9; this makes skill very much more important in most tasks though, making unskilled people think twice before attempting most tasks (although with doubles exploding it does improve their chances).

Something else I have been mulling over is how to have more than three levels of skill in this system. Even the meanest systems usually have 5 (unskilled, novice, apprentice, journeyman, master), three doesn't seem enough. What do you think of the idea expending it to 5 like this:

unskilled - low die, doubled
novice - low + middle die
apprentice - middle die, doubled
journeyman - middle + high die
master - high die doubled

By doubling the typical result, you have to rethink attribute levels. I don't really want to have to double rolls or add dice together, but this is the only way I can think of short of saying "novice is the average of low and middle and journeyman is average of middle and high" because that means players have to think, and we all know how much they hate that. I also don't know what you would do with doubles or a triple in this system, and it would make exploding dice harder, wouldn't it? Or would you just say that in the case of doubles, you get whatever the double was, plus another standard roll. The math gets a bit complicated.

Any further thoughts?

Brian.

Hmm.. that's interesting.

I just modified my program to work out the figures for the 5-level system I mentioned above.

The averages are interesting.

Under 3-skill level system, the averages are 7.97 (call it 8) for high, 5.5 for medium and 3.02 (call it 3) for low.

Under the 5 level system I described above, the averages are 15.95 (call it 16) for doubled high, 13.47 (call it 13.5) for medium+high, 11 for doubled medium, 8.52 (call it 8) for medium+low and 6.05 (call it 6) for doubled low.

Maybe it's obvious that it should be this way for those who are very mathematically inclined, but I'm interested to note that there's a variance of ~2.5 between the averages in the 3-skill system 8/5.5/3 and ALSO a ~2.5 variance in the 5-skill system (16/13.5/11/8.5/6). Thinking about it, I guess it is obvious since all I have done is doubled the variable of the 3-skill system (from 2.5 to 5) and then put another variable in the middle of each one, breaking it into 2.5 again.

My point? Well, knowing those figures makes it easier to work out what attributes should be, right? If I go with the 5-skill system there's 10 difference between a highly skilled roll and a totally unskilled roll, and the split points between them are at even points. Because they're only 2.5 different in each case, there's not a lot of room for varying TN's. So I'm thinking this makes attributes more important, if I keep the range low it makes the skill level far more important (thus a 5 or 7 point range would be better than a 9 or 19 point range), or do I have it cockeyed?

Any thoughts?

Brian.

Mike Holmes wrote: I disagree slizard.

I think that you have some cognative problem with how this can work (especially given your design focus on it), but it can. In the end, any system is going to assign a final odds to succeed, and it will either come up success or failure.

Further, the proposed system is an excellent candidate for this. It focuses quite heavily on the skill portion of whatever is used to do the task at hand. It just needs to tone down the base chance of success for high stats. Which with the open-ended system can be partially accomplished simply by raising the TN for brain surgery and the like. Thus, if we make the TN for brain surgery 15, this means that a Dex 8 Highly skilled surgeon has better than 80% chance of success. While the same man, Unskilled (forget trying at all if you aren't at a high Dex; patient will die almost certainly) has only a bit more than 15% chance. We can jack up the TN further for a more dangerous surgery to say 18, which changes the odds to a bit more than 30% for the talented surgeon, and .2% or so for the Unskilled guy.

Again, I agree that this is still too close, but that can be fixed, as I said by reducing a bit further the maximal effect of stats. I think with a few tweaks we can get the "easy" brain surgery down to a small fraction of a percent, and the "dificult" surgery down to a fraction of that.

BTW, another way to make something like Brain Surgery hard is to have two rolls for it. One to know what to do, and another to do it right. That can make the outcome nil.

And in the end, there is always the simple rule that lots of games use that says that certain unskilled tasks are, very simply, impossible. I agree that it's a croc, but one that isn't too unpalatable.

"You've never been trained in brain surgery; how did you save her?"

"Well, it turns out that the problem was vascular, and as a plumber, the solution seemed obvious. A little duct tape, and she'll be up and about in no time."

Happens in the movies all the time. Heck, if you're MacGuyver, it happens several times an episode.

I like the basic mechanic a lot. If I were using it, I'd forget about adding a number to the result based on Attribute, and forget about Difficulty determining the target number. Instead, Skill would create the target number, Difficulty or Attribute would be handled only by small skill modifiers, and which die is used would determined by your character's Personality Traits and Motivations. (Maybe a trait wagering system, like in Story Engine.)

Doubles and Triples would affect narration rights; perhaps the player normally narrates only his characters action, but with a double he can narrate his action including results and the reactions of NPCs and such, and with a triple he basically grabs GM status for a bit.

You could also spend a certain number of Story Points (five, maybe) to change the result of a die other than the one you're using, after you roll. This way, you could spend five to create a double, and ten to make a triple.

Story points can also handle the brain-surgery problem; normally, you can only roll for things that the GM thinks you ought to be able to do, given your background, but a Story Point lets you roll for anything. Incidentally, Story Points are acquired by impressing your fellow players; each player can hand out a certain number per session to the other players, at their own discretion.

Needless to say, my approach is rather different from Mike's.