I know there are games that use a "blackjack" dice method where you try to roll between two numbers, but are there any that use multiple rolls in this fashion:

Say I want to roll as high as I can, but not over 20. I first roll 3d6. Then I roll 2d6. Then I roll 1d6. I can stop at any time if I'm satisfied with my roll. Like if I roll 17 on the first roll, I can just stop there. But if I roll 9, I'm going to take the second roll. If I roll a 14 on the first roll, I'm faced with a tough call: it might be easy for my opponent to beat me, but I'm likely to go over on my second roll. 

It would work on a Margin of Success, Margin of Failure basis. Going over 20 would mean a MoF of 5 + the amount by which you went over.

For opposed rolls, the player would go first, then the GM would roll and would have to keep rolling until he beat the player's roll or went over or took all three rolls. For unopposed actions, you'd just have a difficulty number to beat, which may or may not be known.

You could rank attributes 1-3. For a 1 you'd get just the first roll. For a 2 you'd get the first and second. For a three you'd get all three rolls if you need them. Skill levels could add to the "ceiling" number. So a skill of 2 would raise your ceiling to 22. At a skill of 5, you'd get an extra die on the first roll, or 4d6. At a skill of 10, you'd get two extra dice on the first roll, for 5d6. An attribute of 2 would have a maximum skill of 5. An attribute of 3 would have a maximum skill of 10. If you have a 1 attribute, you can't gain rank in any related skills.

I'm not sure what sort of setting to apply it to or anything. Right now I'm just looking for a way to put some extra risk-taking and decision-making into the act of rolling dice.

I've also considered a starting ceiling of 18.

My concern is that it's too much rolling. Also, fudging rolls would take away any added fun, so the game would either have to be especially kind or brutal to players' characters.

Because it's based on gambling a few things that come to mind are like an old west setting or something like gritty underground. I think you have a solid start for a mechanic starting with 3 dice and working down. I don't think you need to make it that much more complicated. If anything past the basic mechanic I'd say set your target, lets say 20 your original suggestion then offer something like Bust forgiveness based on the level. So if you had a skill of 3 and you rolled 22 you could still count it as a 20. That would mean the highest skill would be a 10 and allow them to roll over with impunity. To limit the power of Bust Forgiveness you might only allow a total forgiveness in a session up to the skill level or only allow it once in a session. For example a character with a skill of 5 rolls over by 3 he forgives the 3. Later in the session he rolls over by 3 again he does not have enough skill to cover the overage and now has to decide if he wants to hold at 23 or reduce his Margin of Failure.

Skill could also be used to determine winners in ties.

Neat stuff. I like the choice of staying or rolling in new dice. One thing you could do here to avoid players feeling like the rolling takes away from the role-playing is have them narrate a part of the action each time they make a roll. For example, if Harry is trying to win a fencing match, on the first roll he would narrate some actions (i.e. sidestepping, landing a point. or, if you want less player-narration; attempting to strike his opponent, trying to dodge out of the way). The GM then fills in with how effective the result is (based on how high their initial roll was). Repeat this process each time you roll in new dice, with the ultimate success or failure being determined by the dice but narrated by the GM.

Really like this guy

You could rank attributes 1-3. For a 1 you'd get just the first roll. For a 2 you'd get the first and second. For a three you'd get all three rolls if you need them.

Skill levels could add to the "ceiling" number. So a skill of 2 would raise your ceiling to 22. At a skill of 5, you'd get an extra die on the first roll, or 4d6. At a skill of 10, you'd get two extra dice on the first roll, for 5d6. An attribute of 2 would have a maximum skill of 5. An attribute of 3 would have a maximum skill of 10. If you have a 1 attribute, you can't gain rank in any related skills.

Good suggestions on keeping it simple and narrating after each roll. I was initially considering having skills allow the player to fudge the roll by the same number of points, but when I tested it myself, it felt like it took too much of the risk away. But I could definitely see having a limited number of chances to fudge instead of a higher ceiling on every roll. The end effect would be to give the player a number of rolls equal to skill that always score the top number.

Maybe I should make the ceiling number 21, to emphasize the blackjack connection. I do think it felt a little riskier at 20, though. 

The rolls are actually surprisingly consistent in their clusters, so I'm worried that, if the two sides in a contest have the same skill, the contest might take too long to win, because the margins of success and failure are so small. Of course it depends on how many points make a victory. But I'm thinking of adding one more gambling option:

After an outcome is determined, you can press your luck and roll a single die: on a 1-3 you cut the margin in half; on a 4-6 you double it. It would be the same whether you succeed or fail. Of course, if you succeed you're hoping to double it, and if you fail, you're hoping to cut it in half. But either could always happen. Hence the risk.

Another thought to consider is whether there would be a rule difference between an extended challenge and an immediate one. Would a chase across the country side use the same mechanics as a sneak attack from behind? Also, would a life-or-death battle be solved in one round of action? This is something I have been struggling with myself as far as game mechanics go. One of the ways I thought of solving it is distinguishing the number of conflicts needed to succeed based upon how difficult (or important) the act is. For example, a cross-country chase (lets say hunting down Dracula) might be a best-of-three sort of challenge. Spotting a decoy he throws up to mislead you in the chase would be resolved with a single challenge. Then, when finally encountering him, the fight is a big deal, so to best-of-three again, or maybe best-of-five. I actually feel like a single challenge answer for each of these scenarios would be cool, but my fear is that major conflicts would be settled almost immediately, and particularly with fights, there may be a drastically high mortality rate.

I know what you mean. I'm trying to think of something in between the single-roll pass/fail method with MoS or MoF determining how you narrate it. Not that there's anything bad about it. It's perfect for jumping over something or breaking something.  And with good enough narration -- and maybe a hidden die roll -- you could even make it suspenseful. But some challenges call for more drama and more elapsed time at the table. And of course the typical mechanic for that is the hit-point battle. Which is great, except when there's a lot of misses or when it starts to feel like a tree chopping contest.

One way to avoid the misses is to have head-to-head die rolls and the winner hits. So there's always a hit each roll. This makes battles shorter.

Another idea I'm toying with is what I think of as a "tug-of-war" model. Picture a line marked from -10 to 10 with 0 in the middle. Call it a "battle-status chart." You start with a marker at the middle, then move it a numer of "spaces" to the positive or negative side determined by the MoS or MoF. This might lend itself well to dramatic narration as it's easier to see which way the battle is going and when someone makes a comeback. The unusual thing about it is that you're not keeping track of hit points, so there's no carry-over "damage" from one conflict to another. To use it for a fight, there may need to be another aspect to it: like, say, you get hurt each time the marker passes -5. I think it would be cool for a chase mechanic, because it lends itself visually to the idea of falling behind or getting ahead.

A third concept is the "tennis model." You need a certain number of victories -- say, three --  but you can't win unless you have at least two more victories than your opponent. If close matches start to drag out, a rule where a very good die roll counts for 2 victories might help avoid endless back and forth.

Ah yes, the tree chopping contests... *Shudder*

I like the "tug-of-war" model and the "head-to-head" model. Also, I'm not familiar with some of the lingo you used. What is "MoS" and MoF?"

With the head-to-head model, this does allow for a much quicker means of combat. Someone's going to get hurt every round, which increases the benefit of cutting and running early, and could also make players think twice before getting into a lethal brawl. The downside of this method is that a battle-grid style play would be almost impossible - each round would represent a series of actions (including, I would assume, movement), ending with one landing a blow.

With the tug-of-war model, you have a lot of play as to what happens during the combat. Special abilities, tactics, spells, ect. could all be used as special modifiers as to who gets the edge on a given "tug."
   
    On that, I was just playing a board game I found at a pub based (quite satirically) on two Scottish families having a traditional tug-of-war contest. During a players turn, they could place down people on the rope which modify how much that side had. Hence, you could place big, high number people on your side of the rope, and weak, low numbered people on your opponents side. You could also place player cards over the player cards of others, as long as the color matched (green to green, red to red, blue to blue). Additionally, there were certain people cards that had the word "heave!" on them. When this card was placed, the two players would compare the total number of the people cards on their sides, and the person with the highest number gained a number of "points" up the rope equal to the amount his or her number was higher than his or her opponent. So, you could place a 1-point heave card on your opponent (especially over a powerful 4 point person card), or a high 4-point heave card on your side, ect.
  Following this model, there is a lot for you to work with. Lets use a physical combat as an example. Instead of colors, there could be sections for "tactics" (for fighters and rogues), "magic" (for casters), and "circumstance" (for other factors). Characters' abilities could be used to place down "cards" (or whatever) over a section (magic for casters, tactics for fighters, ect). Players would get cards (or whatever) based on how powerful they are (maybe a power pool?), and each side would take turns placing what they have. Then (maybe every 3rd round, or maybe with a specific ability), there could be a "Hit!" card, where whoever has the advantage moves their marker whatever points to their side. The only thing I thought of here is that you wouldn't want to have a huge deck of cards every time you went into combat, but you could probably remedy this by making different abilities (Cone of cold as a negative ability to place on the opponents side) usable every X number of rounds (or something).
    Haven't thought about how you would handle multiple combatants with this one, but I'm sure there is a way you could work it out.

With the "Tennis Model," I would caution against painfully long conflicts and encumbering needs for score keeping and the like. That said, I'm not a tennis fan, so there ya go.

now this is interesting. I like both the idea of skill increasing the Bust number and the idea that it could be used to swing the number either way, though I think it should be per roll not per day, skill is more important than luck to me.

I see this as like, maneuvering, attacking and correcting. The first roll is footwork, where you stand, were you put your weight, the basic roll that sets up the action, the second roll is the attack itself and the third is to correct in mid swing, well to put it in combat terms. Its also maybe the run, the jump and the willpower to claw your way over the edge. To better illustrate, the first roll is gathering speed, the run to the ledge, if you have enough speed it will carry you over the gap, but most people wont make that. The second is the jump itself, again if you have good legs you make the other side, with the third roll being a will power test. Can you pull yourself onto the next rooftop after making that jump and falling short. I would make difficulty based around how many rolls you get.

An interesting premise and something very fun, I like the Gambling aspect a lot.

The above, but not over dice mechanic is pretty cool. You should check out Pendragon: http://www.pendragon.mu/, I believe they use a system something like that.

whoops, I mean http://www.nocturnal-media.com/games/pendragon I grabbed the bands website by mistake.

I worked on a game a while ago that used a dice mechanic mostly like this and I changed it completely after few test sessions because, like you anticipated, it was painfully slow because of too much rolling.

Though, reading your ideas, it gave me another idea that could be much quicker and smooth. What about having to choose how many dice to roll and that's it. It would work somehow like a risk-reward mechanic where the player choose how many die he wants to roll. If he chooses to throw more die, he could getter better results but risk "busting" (could turn a check into a catastrophic failure or something). So, instead of throwing 3d6, then choose to throw 2d6, then 1d6, you'll have to choose between 1 to 5 die to throw before the check then throw all the die at once.

Here's another idea that could or could not work with this mechanic : The type of dice rolled could be tie to the skill level of the characters. Low skill level character would be forced into 1 type of die, let's say d6. A more skilled character could choose between d4, d6 and d8, for example and could mix and match the type of dice.

Hi guys,

The discussion's content is fine, but this thread can't continue until the author provides an external link to source material of some kind, as described in the sticky.

Everyone, please remember not to dive into discussion unless you see that the thread is conforming to that requirement.

Best, Ron