What are the advantages and, more importantly, disadvantages of such a system (as used in L5R, for example)?

A great question I wish I knew the answer to.
 
In L5R/7th Sea as far as I can tell, you roll Trait+skill dice (d10s), keep as many as your trait score, with 10s exploding but counting as single dice still (i.e. roll a 10, another 10 and a 7 and you get 27 for that one die).
It seems that Trait is much more important than skill (gives higher numbers), perhaps slightly offset by greater chance of an explosion the more Rolled dice you have.
i.e. Highest stat wins usually, but with skill increasing chance of a lucky fluke mega win. Seems odd.

The main disadvantages are the long handling time - roll dice, select those to keep, re-roll for explosions (in extreme cases you won't even select which to keep until after this stage) and then total the amount - and the rather opaque probabilities produced by the system (fixed with lookup tables sure, but it makes on-the-fly TN setting tricky).

As implemented in 7th Sea we also found that it had a tendency to favour attributes over skills (you keep traits/attributes in 7th Sea and roll trait+skill, IIRC), and also it produces a wide range of competence over a small range of Trait values, which means that rapidly you reach a situation where tasks are trivial for one PC, hard for the others, and impossible for the rest.

As an example - any PC on Xk1 will find a TN of 15 to be challenging. 6% chance at 1k1 rising to 29% at 5k1 (not that you'd ever be 5k1). If you're keeping two, then it's hard if you're unskilled, or a fifty/fifty shot with just a point of skill. If you're keeping three, mind, you've a 66% chance unskilled, and 85% with just one point of skill. Drop the TN to 10, and it's still a challenge for Mr Xk1, but is now practically an auto-sucess for the other two players.

So the average task (such as "hit the bad guy") would often be frustratingly hard for one PC and trivially easy for another. As the problem got worse through character development it eventually became a campaign ending issue for us (if we'd wanted to play on it could have been fixed over time by other players changing their XP spend, obviously, but having got ourselves in to a hole we decided to stop digging instead).

This granularity is however a specific flaw in 7th Sea rather than an inherent one in the system, I think (presumably R&K using d6s would produce a smaller range of results, for example).

I always toyed with the idea of flipping roll and keep, where you keep your skill and only roll your trait. There was always so many more skills than traits, especially with skill specialization, that it seemed like the obvious way to diffuse die keeping power.

The only catch is that every untrained skill would have to read as keep half die...

I'd have to have a statistician run the numbers to see what that would do to a campaign. The immediate concern is a flattening of skill purchasing. I don't know if that would be a big concern, though.

But trait keep 1/2 die on unskilled should still retain the wild results from explosive dice, I think.

I.e. Air 3, etiquette 2, untrained calligraphy (IIRC) should read: Etiquette 3k2, Calligraphy 3k0.5, leaving unskilled high trait a still high chance to exploit die explosion.

Thoughts?

Quote from: Grymbok;629409The main disadvantages are the long handling time - roll dice, select those to keep, re-roll for explosions (in extreme cases you won't even select which to keep until after this stage) and then total the amount - and the rather opaque probabilities produced by the system (fixed with lookup tables sure, but it makes on-the-fly TN setting tricky)...

I agree with this analysis. It can also lead to huge dice pool syndrome, particuylarly if you have a system that allows people to buy extra dice from stunts etc.

Anything with this kind of dice pool approach has a problem with small dice pools and high TN's. I generaly (depending on the game) like the players to think they at least have a chance. If a TN is 12 and you can only roll 2k1, then that's a bit disheartening.

Some variation of Roll and Keep is about all I use in designs these days, though not the L5R iteration, as it's cumbersome both to sum and remember.

Quote from: Anon Adderlan;629817Some variation of Roll and Keep is about all I use in designs these days, though not the L5R iteration, as it's cumbersome both to sum and remember.

What other variations are there?

One Roll Engine (ORE) is "roll and keep" if you take the words literally: You roll a number of dice, and keep some of them.

I like the idea behind ORE better than R&K because, while the hazzle you go through is similar, the method provides more information than a single number. The hazzle is even less, becuase calculation is not necessary.

I have my gripes with the ORE's other rules, though.

Overall, I still don't see any advantages to Roll n' Keep, in any form. Simplicity? Transparency? Ability to modify inputs variously to get different outputs? It doesn't do any of those that I can see. If I squint a bit, perhaps the point of it is to decrease the granularity of a dice-pool system since you can inflate the number of dice rolled without hugely increasing the average result obtained, but I think if you wanted that a wholly different mechanic would be the way to go.
 

Quote from: Opaopajr;629631Thoughts?

The 'keep half die' thing is a bit messy. Actually, looking at your system perhaps explains why they went with Trait as the basis for the 'keep' function, since traits are always at least 1.
Another option might be "keep second best" for Xk0 rolls, instead of halving the die? Or just having Xk1 as the default for untrained, and skill adding extra dice (the biggest jump in probability of success/ average result should probably be at the threshold between untrained/trained).

High Valor (and several forthcoming games) use a system of rolling and keeping--but its a "Top Die" system where you only keep the highest value rolled. If it is a 10 you may keep the next highest value once.

It allows for epic battles, but not regular super-explosions.

Quote from: Silverlion;629838It allows for epic battles, but not regular super-explosions.

How so?

Don't such systems have a limit to the number of dice that can be rolled before results start to become predictable (ie if you roll enough d10, youre gonna get a ten guaranteed).

Quote from: Ghost Whistler;629851How so?

Don't such systems have a limit to the number of dice that can be rolled before results start to become predictable (ie if you roll enough d10, youre gonna get a ten guaranteed).

Generally speaking mathematically yes. However, in actual play? Not so much. It comes up from time to time, but not often enough to blow the games curve so to speak. A few great epic moments is the point.

 Plus the game isn't just "get a 10" it has traits which add to that 10, a ten will usually get a success for the players against MOST foes, but once you start pushing towards the Legendary and Mythic Numbers (TN22 and 28) it becomes a LOT less likely to happen on a d10.

In L5R there was a point where taking risks with raises was supposed to come in. Taking an extra 5, 10, or 15 on a task was to be a more common feature. Unfortunately people rapidly figured out the key was the kept dice and soon 'untrained sure things v. iffy skill masters' threw a lot of that bidding out the window.

Quote from: Ghost Whistler;629824What other variations are there?

Well for one you can roll a number of dice equal to an Attribute, and keep any dice which roll a number less than or equal to an Ability. This started out as a fix to the huge dice pools I kept finding in White Wolf games. Another is to roll multiple pools and keep a number of dice equal to the smallest sub-pool. There's actually quite a lot of variations possible, and I like them because they provide inspiration as to WHAT happens in addition to IF.

I have a great idea for a 'roll and keep' game system where you roll a d20, and then you have to keep it