A proposal for tiered skill training [very long]

[Ainamacar]

I've got time for a quick clarification: with my tweaked proposal it was intended that you would have a smaller, broader set of skills, so the example would be as your Decipher Script example. The number of successes achieved would determine the outcome. I would be very wary to include chain skill checks under this, or your, system because of the imaginary bonus dice spell you came up with. If a task genuinely involves abilities that you can't put together in one skill, I would probably tell the player to 'aid another' themselves with their better skill and roll normal dice with their worse skill. In fact that's a great idea for chaining skills.

OK, that wasn't clear to me. I was under the apparently mistaken impression that the number of successes required was equal to the number of different skills used because breadth (as I understood you) related the number of skills to the number of successes.

In my mind a check is always a single use of a single skill, so I don't think of "chained" skill checks as anything but separate checks unless one adds completely new structure on top, like Skill Challenges did. Aiding oneself with a secondary skill would probably be a rare occurrence in any game I ran, but I think it's an intriguing idea.

The chance of failure can drop below 5% in this system. Either you remove the idea that a 1 is an automatic fail, or you throw in a skill trick that allows you to reroll dice, or increase their face value. Even with 1 as an automatic failure, 5% only comes up when rolling 1 dice looking for 1 success.

I think we talked past each other here. I was assuming that every check with a single skill required a single success (as per my above misunderstanding). I did not assume that a 1 is an automatic failure. Also, my objection wasn't that the probability of success can't drop below 5%, it is that it could be 0% or 5% but it could never be in between those. This is so because, for any check that succeeds on a single success, if p=0 then the probability of success is 0 regardless of how many rolls are attempted, and if p=.05 each additional roll increases the probability of getting a success. In other words, on checks where a single success is required it is impossible to get, say, a 2.5% chance of success or a .75% chance of success. My objection is removed because apparently you allow that some checks using a single skill may require more than a single success.

My fear with your newer proposal is that adding +5 to a user's skill every level makes the skill advancement even more exponential.

The new table is not cumulative, so each element in the table lists the total benefit for that level of training. Mechanically, it is identical to the first table. For example, whether one is Competent or a Master, the total skill bonus is just +5. A +15 just for training would be disastrous. (The idea for the new table was that a player looking up training bonuses in a table wants to see exactly what they get, not "sum" it up over all the lower levels of training as well. It was part of my attempt to give it a better presentation.)

It also conflates the breadth/depth situation again - as a DM I can require either more successes or a higher DC, but it's not clear to me which to use when (and as much as you or I can calculate the binomial probability of success, it's not fun at the table!).

I clearly have some issues with the breadth/depth situation, because of how badly I misinterpreted your tweaked system in the first place. Perhaps you can explain again how one determines results in your system. If I now understand you correctly, a character might roll 4 dice on a skill check using a master skill. If he gets 1 success then that has some result, and if he gets 4 successes that is some superior result (which naturally only a master could normally get by virtue of his many dice). I honestly don't see how that fixes anything. As DM I could always think the situation might warrant giving on 2 successes what I might otherwise have given on 3. Or on 3 what I might otherwise have given on two. Suddenly one is adjusting both the DC and the result, and I don't see any way to avoid that unless the DC is fixed for all checks or the number of results is exactly 1 for all checks. And if the result can't vary on the number of successes, what is the point of even contemplating a system which counts successes?

As I've said, for me the main use of multiple outcomes isn't to adjust the probabilities of success, it is to define outcomes that match the theme of untrained, competent, expert, or master training, i.e. what such a character could reasonably be expected to do, and then define a DC appropriate to that outcome.
*or*
In cases where there are a clear fixed set of outcomes (like in the Decipher Script example with minimal, partial, and complete results) to pick a DC which gives each result with frequency appropriate to the difficulty of the task.

In both the above cases, however, the results are defined first and then the DC is picked. Most additional modifiers should affect the DC. But really important ones, ones that are so big they should make an Expert as good as a Master in a particular task, should cause the DM to adjust the results of success itself. This isn't a mathematical calculation to get the right probabilities, it is a qualitative assessment on how that modifier changes the situation.

I'd also note that my proposal is specifically designed to *not* require using the multiple successes and the related binomial math for figuring things out. A master character obtains an appropriate master-level result by rolling a single die. The probability of that happening is just p, and lets the DM set appropriate probabilities for success in a natural way. Yes, there is an increased probability that the character gets an Expert result, but one should set the probabilities for Expert results assuming the person making the check is an Expert, and so on. The only time the math of multiple required successes rears its head is when the character is attempting to obtain a result above what their training makes "normal". That represents a special occasion, and in any case the DM should still set the DCs based on what an appropriately trained person would get. It relieves the DM of the burden of trying to calculate the exact probabilities, and it happens only in cases where the PCs know they are stretching beyond their natural ability.

Distinct levels of skill tricks is a great idea though. The way you have described the acrobat's skill check makes me wonder if rather than 'more dice, more successes', a simple granting of rerolls would simplify the system.

It might be possible to recast the multiple dice as rerolls. Certainly the probability of getting at least one success is identical either way. The difference would come in explaining when one can use skill tricks. Currently they represent ways to spend extra successes. If the additional dice represent additional rerolls, however, one would always stop after the first success and would never bother to roll the other rerolls. So how does one know how many, if any, skill tricks could be used?

I guess the problem I have is that I think allowing for different numbers of successes provides more scope in resolving skill checks. By allowing superior skill users to get the same effect from 1 success as those beneath them would get from multiple successes, they can't access the granularity of the system - admittedly your 'roll another dice to get the next best outcome' tweak resolves this but I really think it's too complicated and drawn out. Multiple dice are at least resolved quickly!

Allowing different numbers of successes does provide more scope for resolving skill checks. Isn't that the point? If that is an issue the only sure solution is to make every check binary again.

I also think you're underestimating how quickly the 'roll another dice to get the next best outcome' works. A master rolling a check that has a Competent, Expert, and Master outcome rolls exactly 1 die with probability p, exactly 2 dice with probability (1-p)p, and exactly 3 dice with probability (1-p)^2. That means the average number of a dice rolled is 1p + 2(1-p)p+3(1-p)^2=p^2-3p+3. Assuming all p's occur with equal probability, the average check requires just ~1.85 rolls. Not only that, if one fails the roll there is no deliberation, one immediately rolls again. So on average it requires fewer rolls and less math than rolling 3 dice on every check. Yes, there is some lag because the rolls are performed separately, but one or more rerolls only happens (on average) 50% of the time. Unless the player is a serious slow-poke when rolling both will take a comparable amount of time. And remember, this is only the case when all three outcomes have been defined. If the master rolls a skill with only an Untrained/Competent result (which should be most of them) he can simply roll his 3 dice.

Thanks for your clarifications!

 

[Chris_Nightwing]

I think we are confusing each other. You talk of Competent, Expert and Master outcomes - this is granularity in success, no?

I propose that if you roll multiple dice together, the number of successes determines the 'outcome level' of your example. If there is no granularity in the possible outcome (can I jump that gap) then I propose extra successes be used for skill tricks (such as you jump the gap and still get to move, or you needn't roll a check for that gap again).

I also don't understand how the newer version of your system handles the lower skill guy trying for higher skill outcome - he still needs multiple successes?

 

[Crazy Jerome]

I would be very wary to include chain skill checks under this, or your, system because of the imaginary bonus dice spell you came up with. If a task genuinely involves abilities that you can't put together in one skill, I would probably tell the player to 'aid another' themselves with their better skill and roll normal dice with their worse skill. In fact that's a great idea for chaining skills.

This is more or less how Burning Wheel handles this issues. BW is a dice pool system, so that its "linked tests" add or subtract one die to the next test in the chain, for success or failure, respectively.

In the proposed system, I'd use the current 3E/4E Aid Another mechanics as explicitly "Linked Tests" mechanics, then use the new Aid Another mechanics int his proposal for that purpose. That is, using a "linked" skill gives you a chance to modify the next DC by 2.

Note that another way that BW makes this work is that you don't automatically get to use your best skill as the main one, with the others chipping in. You make the linked tests in an order determined by the fiction--and sometimes may be forced to do so even when this isn't your best mechanical order. If you have separate checks for languages and Decipher Script, then translation may require the language check first, strictly as a modifier to your final Decipher Script roll.

Among other things, this addresses Ainamacar's concern about blurring the boundaries of what each check means. If you succeed in the the language check but fail in the Decipher Script roll, then you read the plain words well enough, but failed at some deeper level of meaning to translate--perhaps idiom or concept. For the reverse, you did so poorly with the language that you really didn't get every nuance, but your great Decipher Script check managed to puzzle out the main meaning anyway. If you just throw all the checks in a bucket and roll for a set number of successes, you lose these distinctions.

 

[Ainamacar]

I think we are confusing each other.

Probably. At least it's cordial confusion, as opposed to stereotypical internet RAGE!

You talk of Competent, Expert and Master outcomes - this is granularity in success, no?

Quite so.

I propose that if you roll multiple dice together, the number of successes determines the 'outcome level' of your example. If there is no granularity in the possible outcome (can I jump that gap) then I propose extra successes be used for skill tricks (such as you jump the gap and still get to move, or you needn't roll a check for that gap again).

Yes, this is what I would consider the standard way of handling granularity in success-based systems (although skill tricks are a bit of a twist on the classic formula). The system I have played in for the last several years operates in just this fashion, except the notion of "extra successes" does not exist because there is a unique result for every possible number of successes.

I, on the other hand, propose a system where outcomes corresponding to one's training or lower only require a single success, and the training relative to the outcome determines the number of dice rolled. On the other hand, outcomes corresponding to training above what one has use the classic success system just as you do, with untrained/competent requiring 1 success, expert requiring 2 successes, and master requiring 3 successes. If desired, the DM can continue increasing that number for ever more dramatic outcomes.

I also don't understand how the newer version of your system handles the lower skill guy trying for higher skill outcome - he still needs multiple successes?

Yes, the lower skill guy requires multiple successes, and as since he does not get to roll multiple dice by default he must acquire resources or be in very favorable circumstances which provide extra dice on his check. (Note that the only difference between the "old" and the "new" system should be in the presentation. I did not intend to change the mechanics at all, so if you can point out what looks different I'll either fix my mistake or try and state what I mean more clearly.)

All resources or circumstances can be represented by modifiers to the skill check, of which there are 3 major types I'll summarize. (For simplicity I assume bonuses, since I'm still thinking about the best way to represent penalties.)

1. Increasing the skill bonus (i.e. what one adds to each d20 roll). These represent resources which make a character more likely to succeed at what he can already do, but do not enable him obtain more powerful outcomes. The bonus should be limited to +3 at most so it doesn't move to a wildly different DC or outshine the effect of the other two modifiers. Most minor resources and circumstances fit here.
2. Increasing the number of dice rolled. These represent resources which enable a creature to obtain more powerful outcomes than normally possible, but the primary effect is to put these outcomes in reach, not make them likely. Notice that the extra dice also makes success more likely at the things the creature is already good at. Most moderate resources and circumstances fit here.
3. Increasing (for this check only) the creature's level of training. These represent resources which enable a creature to obtain more powerful outcomes than normally possible, but the resource is so powerful as to effectively make the creature as good as someone with better training. If the resource or circumstance seems like it would be more important than the creature, this modifier may be appropriate. Such resources are very rare and extremely potent.

So lower skill guy must obtain modifiers of type 2 or 3 to obtain higher skill outcomes, since these either add extra dice or simply make one die sufficient.

We're already familiar with the type 2 modifiers because taking extra time and aid another usually work this way. The DM is free to imagine more. For example, if the PCs are researching something they might go to a library and receive an extra die due to the resources of the library. All these type 2 resources are, in a sense, various forms of aid another. Aid another itself represents help from creatures. Taking extra time is rather like "aid self" by being thorough. Access to a relevant library is like having the library aid on the check, and so on.

These types of modifiers are flexible, and it is up to the DM to determine which is most appropriate. For example, the bonus from a library is not fixed. If the PCs are searching for lore about a specific dead god a noble's vast collection of literature and family records might provide a +1 bonus, the national library an extra die, and the library of a major church once allied with that dead god an effective increase in training. On a check to find specific knowledge about a certain noble's financial affairs the church library might give a +2 bonus (the noble was an occasional supporter of that church), the national library an extra die (it includes extensive political and financial records), and the noble's collection an effective increase in training (the noble who owns the library is a longtime business partner with the noble in question).

This is more or less how Burning Wheel handles this issues. BW is a dice pool system, so that its "linked tests" add or subtract one die to the next test in the chain, for success or failure, respectively.

That's really interesting. At least at first glance I like your suggestion of using aid another for extra dice within checks and "links" using +2 between checks. It feels natural to me that aid another normally grants a stronger bonus than the indirect aid from a linked test, except in those cases where success with one skill is a prerequisite to even making a check with another. And it certainly seems to maintain a strong boundary between checks, which means I can start to see how to self-consistently fit individual checks into a framework of many skills working toward larger related goals.

It could open up some new mechanical territory like "cross-check" skill tricks, for example. Unlike normal skill tricks these would grant benefits to linked checks, but otherwise they would have a similar flavor. I think I could see using these to define scenarios with special interactions between skills specific to the context. Trying to make a general list of "cross-check" skill tricks would be a fools errand, because the unpredictable interactions between skills are a primary driver of the emergent behavior we observe during the game. Good DMs often do this sort of thing naturally, and Skill Challenges sometimes show this kind of thinking, where a success with one skill yields some special result in the Skill Challenge without counting as a success per se. It's a tough balancing act, though, giving just enough structure to support good play without making things overly rigid.

 

[Chris_Nightwing]

Yes, the lower skill guy requires multiple successes, and as since he does not get to roll multiple dice by default he must acquire resources or be in very favorable circumstances which provide extra dice on his check.

I think the only confusing part of the newer presentation was in rolling one dice at a time. The unskilled user trying to do something skilled (but taking longer/getting help) rolls their first dice, and if it's a success he rolls the second looking for another success, but if it's a failure then he still rolls the second dice, but can only get the lower outcome, right?

I'm still in favour of rolling all the dice at once

 

[Ainamacar]

I think the only confusing part of the newer presentation was in rolling one dice at a time. The unskilled user trying to do something skilled (but taking longer/getting help) rolls their first dice, and if it's a success he rolls the second looking for another success, but if it's a failure then he still rolls the second dice, but can only get the lower outcome, right?

I'm still in favour of rolling all the dice at once

Almost. Extra dice from taking a long time apply to every level of outcome, so they can always be rolled together.

For example, suppose an expert takes extra time to gain 2 additional dice on a check that has results for competent, expert, and master. The expert can roll the first 3 dice at once. If all three are successes then he obtains the master result. If exactly 1 or 2 of the dice are successes he gains the expert result. If none of the first three dice are successes he can then roll the extra die from training to see if he gets the competent result. (So in the end this character rolled 3 dice and needed 3 successes to get master, 3 dice and needed 1 success to get expert, and (potentially) 4 dice and 1 success to get competent.)

And despite your smiley, please note that in all cases with multiple outcomes one can choose to roll all the dice at once if they are labeled beforehand. In the case above, for example, one could say that three white d20s apply to all outcomes, and a red d20 only helps to achieve competent results.

Likewise, a master could always roll an arbitrary check with all dice at once if the dice are labeled. For example,
White die - The die that applies to all results
Red die - The extra die that applies to both competent and expert results.
Blue die - The extra die that applies only to competent results.

Clearly this mapping is identical to rolling in sequence: the white die is like the first die rolled, the red die is like the second die rolled, and the blue die is like the third die rolled. Equivalently, the white die is like the single die rolled to get the master result, the white die and red die are like the two dice rolled to get the expert outcome, and all three dice apply to getting a competent result. This has the advantage of working for all checks, regardless of what outcomes they have (and resolves the metagame issue I raised in the very first post because the DM doesn't have to give any indication on how many dice to roll). And if there are extra dice from other sources that apply only to specific levels, simply give them the correct label. (For example, if a character used Aid Another on this check but only for the Competent and Expert outcomes that could be represented as another red die.)

I personally tend not to care for labeled dice, because in play I note that players have a tendency to roll first and then announce the labels they used. Usually it's an honest mistake which requires either a reroll or randomly determining the labeling after the fact, while at worst it is outright cheating. However, for people who make a convention and stick to it without exception this kind of thing can work extremely well.

Similar labeling situations already crop up in D&D. Imagine the times a player uses another die to fake rolling a "d2". Some people use the rule that odd results are a 1 and evens are a 2, while others make rolling low a 1 and rolling high a 2. Or vice versa. Or something else entirely with a 50% chance of being either result. Mathematically it just doesn't matter as long as the convention being used is known before rolling.