The D&D Experience (or, All Roads lead to Rome)

[Dannyalcatraz]

I had an error in my original chart. The corrected data set is below the explanation.

So Nagol, those are your numbers, but what is your evaluation of the SC from a purely mathematical standpoint?

(Side note- having never run a SC, I don't know: is 6 successes before 3 failures the standard SC structure?)

 

[BryonD]

Hang on a tick though. That's a gross oversimplification. For one, the DC's of actions are not necessarily defined solely by the level of the skill challenge. There's absolutely no reason that some parts of the skill challenge might not have different DC's based on the in game situation.

If nothing else, modifiers based on the situation will change the DC (or give bonuses to the die roll, which is effectively the same thing).

Additionally, right in the skill challenge write-up, and this one is explicit, is the idea that you can tie various skills into the challenge that don't directly resolve the issue, but rather give bonuses to other rolls. So, that intimidate check doesn't really calm the bear down but gives a +2 to Nature checks because it worked.

You've even admitted that you can double up on successes, so, the whole you must get 6 before 3 isn't true even by the explicit rules.

Yes, if you run a SC as a mindless check list, then sure, it's going to be boring as all get out and give wonky results. The solution though, isn't to chuck the framework, it's to not run it like a mindless check list.

In exactly the same way that a DM will ignore mechanics in any other edition that get in the way of a good narrative (Fonzie Bump anyone? Can't do it in 3e unless you ignore the mechanics), why does running 4e suddenly mean that I am locked into a single stream straightjacket?

None of these points address the issue.

The nature of "the bear" or whatever is defined as I described. The fact that you may get a circumstance or aid another bonus doesn't nothing to change that.

The merits, and lack thereof, of "double credit" has already been well covered.

And again, if you are going to offer ignoring mechanics as a defense of the mechanics, then you are better off without the mechanic in the first place.

There is no gross oversimplification of the skill challenge mechanic itself.

 

[Nagol]

So Nagol, those are your numbers, but what is your evaluation of the SC from a purely mathematical standpoint?

(Side note- having never run a SC, I don't know: is 6 successes before 3 failures the standard SC structure?)

From a math perspective, I think the SC system is dangerous to design/run for people without experience since the probability of success suffers such a quick drop-off as the minimum success number rises. People don't grok probability easily and most won't run the math to see what the expected success rate will be.

I haven't kept up on the errata to the skill challenge system, but I believe the 6:3 set is expected to be a "hard" challenge. Depending on the table, aiming for the typical skill check to be around 9+ leads to less than a 1 in 3chance of eventual success and is probably a decent rate for a "hard" challenge with liberal opportunity to get bonuses from secondary sources. If every check gets +2 the chance of success rises to just over 50%; a +4 on every check moves it up to 80%.

A consequence of this design is every point matters. Never select a secondary ability unless you expect a bonus to the roll sufficient to raise it above your best ability. If you don't have the highest effective ability, try to find some way to give a bonus to someone who does rather than taking a shot yourself. The system penalises lower than maximal abilities harshly.

 

[Patryn of Elvenshae]

That isn't bad faith, that's statistics.

What is being shown is that if Joe has a +8 modifier

I understand the math that is being shown; it, however, has more than just a slight tinge of "lies, damn lies and statistics" about it.

The comparison is so facile as to be meaningless.

But so far, the model may be incomplete. SCs typically follow a format of "X successes before Y failures," at least IME.

Specifically, this, and the fact that it is, at best, only sideways applicable to skill challenges.

Consider a 6-successes-before-3-failures skill check (which, BTW, isn't supposed to be an easy thing for the party to pull off).

The universe of successful results are, as Nagol states:

(6 successes in a row)
(6 successes + 1 failure where the failure occurs anywhere except the final roll)
(6 successes + 2 failures where the failures occur anywhere except the final roll)


0.7^6 + (0.7^6*0.3*6) + (0.7^6*0.3^2*21) = 55.18%

... or, more simply, the total number of unique, successful combinations is the sum of:

No Failures: 6 choose 6
1 Failure: 7 choose 6 - 6 choose 6
2 Failures: 8 choose 6 - 7 choose 6

And the specific odds for any given combination coming up are pretty easy to calculate (e.g., an 8 choose 6 combination has a [Chance of Success]^6 * [Chance of Failure]^2 probability of occuring, and there are 21 unique 8 choose 6 combinations).

All this goes by way of saying that Nagol's second, corrected chart, is actually correct.

[sblock]
Further explanation:

There is one 6 choose 6 success - 6 successes in a row.

SSSSSS

There are actually seven 7 choose 6 success:

FSSSSSS
SFSSSSS
SSFSSSS
SSSFSSS
SSSSFSS
SSSSSFS
SSSSSSF

However, the bolded one is the same as the 6 choose 6 success, and therefore is discarded. Thus, the total number of unique 7 choose 6 successes is 7 choose 6 - 6 choose 6.

You can also do this with counting failures, which are 3 choose 3, 4 choose 3, etc., up to 8 choose 3.[/sblock]

However, even that doesn't really tell the whole story, because, just like in combat encounter design, you don't just look at "Total Monster HP / [Fighter Damage per Attack * Fighter Attacks per Round]" and say, "Yep, this combat will last 7 rounds."

It ignores the possibility for players to choose mechanically optimal actions - e.g., doing things like making secondary skill rolls, or using powers to gain automatic successes, or even tailoring their approach to highlight their own strengths and the weaknesses of the particular skill challenge.

In other words, this is, at best, an entry level tool, and presenting it as some sort of revelation is, at best, misleading.

And, again, it ignores the capability of players to change their approach midstream, just like in a combat scenario, players can elect to reposition themselves, close a door, expend additional spells, fight defensively, etc.

If a negotiation is resolved as a single die roll and some aid another attempts, then the players all make their rolls simultaneously (probably at the end), and the main roller rolls, and ... the end! Either they succeed or they fail. There is no opportunity to say, "Hmm - we started arguing with the Noble using tactic A, and he isn't buying it; maybe we should mention B, too?" At that point in the one-die-roll method, it's too late.

 

[Patryn of Elvenshae]

Double post.

 

[Raven Crowking]

I was wondering if there is a pbp or two here on EnWorld where we might be able to read how some good skill checks went in actual play?

 

[Nagol]

I understand the math that is being shown; it, however, has more than just a slight tinge of "lies, damn lies and statistics" about it.

The comparison is so facile as to be meaningless.


<snip>

However, even that doesn't really tell the whole story, because, just like in combat encounter design, you don't just look at "Total Monster HP / [Fighter Damage per Attack * Fighter Attacks per Round]" and say, "Yep, this combat will last 7 rounds."

It ignores the possibility for players to choose mechanically optimal actions - e.g., doing things like making secondary skill rolls, or using powers to gain automatic successes, or even tailoring their approach to highlight their own strengths and the weaknesses of the particular skill challenge.

In other words, this is, at best, an entry level tool, and presenting it as some sort of revelation is, at best, misleading.

And, again, it ignores the capability of players to change their approach midstream, just like in a combat scenario, players can elect to reposition themselves, close a door, expend additional spells, fight defensively, etc.

If a negotiation is resolved as a single die roll and some aid another attempts, then the players all make their rolls simultaneously (probably at the end), and the main roller rolls, and ... the end! Either they succeed or they fail. There is no opportunity to say, "Hmm - we started arguing with the Noble using tactic A, and he isn't buying it; maybe we should mention B, too?" At that point in the one-die-roll method, it's too late.

The assumption is the players always choose mechonically optimal choices, but that those choices involve a consistent expected minimum roll to succeed. If multiple avenues exist with differing minimum rolls to succeed, the group is best served by concentrating only on those with the best return.

Changing your approach mid-stream only serves a purpose if the new approach has a better expected outcome (i.e. is more optimal). If the effective chance per roll does not improve, changing your approach offers nothing to the challenge.

I expect the biggest revelation for most people is how non-linear the decay in ultimate success is compared to the linear decay in the the single-roll comparison.

As I wrote in the last message, the implication is to discover who has the best native chance and work to grant those people bonuses since attempting a sub-optimal gambit definitely reduces the overall success probability considerably.

 

[Patryn of Elvenshae]

The assumption is the players always choose mechonically optimal choices, but that those choices involve a consistent expected minimum roll to succeed.

I don't agree that this is the case - given the ability to vary the DCs on a skill-by-skill and attempt-by-attempt basis (e.g., "The first Stealth attempt is Easy; the 2nd is Moderate; and the 3rd is Hard" or "Only one Stealth check can provide a success," or "Stealth is Easy; Insight is Moderate; Intimidate auto-fails and makes NPC Y hostile in encounter Z later on").

Changing your approach mid-stream only serves a purpose if the new approach has a better expected outcome (i.e. is more optimal). If the effective chance per roll does not improve, changing your approach offers nothing to the challenge.

Yes? Is that not obvious?

The difference is that a skill-challenge (or, at least, a skill-challenge-like) framework is that it provides an opportunity for that to occur, which "Roll once, all in" does not provide.

 

[LostSoul]

So, in 1e we have a combat system which is firmly rules-first, but which has tried to mitigate itself to be as fiction-first as possible.

Since I am calling my 4E hack "Fiction First" and I use skill challenges, I wonder what you think of how I use them:

First of all, I only use skill challenges in a few places. So far those have been social conflicts and evasion/pursuit.

In a social conflict, it begins by encountering an NPC (of course). Typically this happens when a wandering monster shows up. When the DM does not know how the NPC is disposed to the PCs, he rolls on a table (called Reaction Roll, stolen from B/X; at the moment it's only modified by Charisma but a list of other suggested modifiers would make sense).

That roll determines the NPC's disposition and thus the number of successes that any type of social conflict will require. Once again, this is only when the DM doesn't already know the NPC's disposition; he can set it as he chooses, based on the NPC's personality and such things. Metagame concerns are already worked into the system, as of course they must be!

Players only make checks in social conflict when the actions they have described for their PC trigger an internal conflict within the NPC, as determined by the DM by his understanding and play of said NPC. Checks resolve those actions, as normal checks do; the modifiers to the roll (Stat + ability modifiers + any other modifiers) are based on what the PC is doing. DCs for checks are set by either a Defence of the NPC (AC, Fort, Ref, Will) or a passive skill modifier (10 + the NPC's relevant modifiers) based on the PC's action; usually this ends up being the NPC's Will Defence.

Checks are not required to acquire successes or failures. Actions are what's important. Sound argument (against rational opponents!) and "deals you can't refuse" can give you "successes", while the wrong words or actions can add up failures. Once again, this depends on how the DM plays the NPC.

When the determined number of successes or failures have been met, this indicates that the NPC is done with the conflict, and whatever end state it's in at the moment is how things stand. The idea of success or failure as a whole on the skill challenge has nothing to do with the game world; it's possible to get what you want and still "fail" a skill challenge, and not get what you want and "succeed".

Since the system only deals with conflict, it's possible that both parties can come to an agreement/resolve the conflict without making any checks, or decide that they can't reach an agreement/resolve the conflict without making any checks. Whenever the conflict ends, it ends.

The set number of checks needed is a tool to help the DM play the NPC impartially, and by doing that they help generate unexpected results.


Evasion/pursuit works in a similar fashion. The number of successes needed to evade pr catch NPCs is set by their persistence, determined by a morale check (if needed). DCs are based on speed and terrain and other such factors. I've only run three of these and the first one didn't work so well, so I have less to say on this as it hasn't been playtested that much.

 

[Nagol]

I don't agree that this is the case - given the ability to vary the DCs on a skill-by-skill and attempt-by-attempt basis (e.g., "The first Stealth attempt is Easy; the 2nd is Moderate; and the 3rd is Hard" or "Only one Stealth check can provide a success").



Yes? Is that not obvious?

The difference is that a skill-challenge (or, at least, a skill-challenge-like) framework is that it provides an opportunity for that to occur, which "Roll once, all in" does not provide.

One of my points is the group should be trying to use the optimal choice every round if the group has complete insight into the challenge (i.e. it is presented to them to read). If the challenge is run 'blind', options that exist that aren't optimal and can lead to a failure are effectively a trap; they reduce the overall chance of success and the comparative difficulty may not be easily discernable by the group.

Glancing at The Jester's "finding the white lake skill challenge" there are DCs ranging from 19 to 29 on actions the characters can attempt. Assuming similar levels in the appropriate skills, the PCs should restrict themselves to the DC 19 checks. If the PCs has a skill bonuses of +17, i.e. they fail on a 2 and they never gain a failure from the sodden ghouls, but don't take actions that cost time of resources (i.e. optimal play), then they have less than 2 percent failure chance. If they gain one failure from the ghouls and blindly choose some other interesting sounding options, such as disbelief (DC 27) and use Perception to navigate instead of Nature (DC 23), their chance of success falls below 40%.