Game Design: Good average chances of success for all situations?

[Kariotis]

When designing new systems or homebrewing existing ones, one of the areas where I still often question both my experience and my gut feelings are the target values for average chances of success on a given roll. One thing I've been doing for a long time now is use a simple matrix that always remained the same, whether I was designing dice pool systems, roll-under-mechanics, used D100s, D20s or even playing cards. I only controlled for skill level and circumstances of the roll and left everything else out in order to not make it too complicated.

I didn't use the exact values I came up with, I used them as rules of thumb (depending on the dice used, it wouldn't even have been possible to hit the precise values). In some cases I deviated a lot from individual values, but rarely much from the general distribution as a whole.

It has always seemed pretty important to me to consider the average success rates at different skill levels from the absolute lowest to the absolute peak. If the success rate is too high, the game may become too easy, and players may become bored or disengaged even if they're happy their character is decked out. If the success rate is too low, the game may become frustrating and discouraging instead of challenging.

So whenever I designed a new game, or started homebrewing an existing one, this has been one of the first things I did and then I always keep my eye on it during the design process. Trying to make sure that the development over the skill levels is smooth with a couple of bumps here and there to signal skill thresholds, and that even difficult tasks are sometimes accomplished while peak performers rarely fail provides players with a satisfying level of challenge and a sense of accomplishment. My hope is always that by balancing the difficulty of tasks and the modifiers that affect success rates, the system becomes more engaging and immersive.

So if you've been using something similar, this thread is for any GM and designer who wants to share their target values (or wants to take it as an opportunity to write them out in detail). This is essentially a poll thread - I don't mind a bit of discussion, but our focus here should be on getting as many replies with concrete numbers as possible to get an overview. I removed my own values for now in order to not poison the well.

 

[pemerton]

I don't have examples, but will cheerfully bump your thread!

But also have a question: mightn't it depend on other features of the system? Eg Burning Wheel deliberately has quite low chances of success (and hence quite high failure rates) compared to 4e or 5e D&D, because of the different way that failures unfold and the different significance low chances of success have for PC development.

 

[Fenris-77]

My first instinct here would be to start pulling the system and character design apart to look for tiers and core bits. Set next to that would be whatever the genre expectations are for the game in question. So a zero-to-hero OSR game has very different expectations for starting character competence than does a game like, say, Delta Green. Once you have those general expectations in hand you can start talking about how many skills at what level best represent that. A related addendum is a hard look at character advancement - the faster characters gain significant 'power' the lower they can start on the power scale, almost regardless of the above concerns.

 

[GMMichael]

When designing new systems or homebrewing existing ones, one of the areas where I still often question both my experience and my gut feelings are the target values for average chances of success on a given roll. . .

Why not take another step back and ask, "what is success?" " What if the game uses fuzzy outcomes instead? "

It has always seemed pretty important to me to consider the average success rates at different skill levels from the absolute lowest to the absolute peak. If the success rate is too high, the game may become too easy, and players may become bored or disengaged even if they're happy their character is decked out. If the success rate is too low, the game may become frustrating and discouraging instead of challenging.

This sounds like a problem for GM-less games, but I don't see how rolls can be too easy or too hard with a GM (or in my case, a GM armed with dice) who determines what the roll results should be/exceed.

Is this another way of saying "I don't want my monster manual to have statistics that are too low?"

 

[Leatherhead]

Given the principles of loss aversion (where a loss is twice as bad, as a win is good): A character that is supposed to be competent at a task should produce a desirable result ~67% of the time. Desirable results include;' success, exceptional/critical success, and falling forward with minor or negligible setbacks. Undesirable results include; null results (whiffing), falling forward with significant setbacks, failure with resource drain, and exceptional/critical failures.

 

[Krachek]

Repeating task like hitting in melee should aim 2:3 or 3:4.
note that these odds are cultural. 40 yers ago hitting 1:2 or even less was much more expected.

 

[kenada]

My homebrew system uses degrees of success and (mostly) fixed target numbers, so I calibrated the results around the spread I wanted and how that interacts with progression. I use AnyDice and write software to look at the results (depending on the complexity of what I want to examine). The latter is mostly necessary when exploring abilities that do something a little weird that doesn’t model nicely on AnyDice.

My system uses fixed difficulties (for the most part). It’s 2d10 against the following spread:

• Critical Success: 23+
• Complete Success: 17–22
• Mixed Success: 11–16
• Failure: 10−

Characters have attributes ranging from −3 to +3 and skills with ranks from +1 to +5. Making a Skill Check untrained imposes a −4 penalty. Working Together can add an additional modifier, and you can sacrifice things to increase the degree of success (e.g., bribing a guard, adding extra materials to a craft). Below are the percentage chances of success broken down over the range of modifiers (AnyDice link).

Modifier	Failure	Mixed Success	Complete Success	Critical Success
−5	85%	14%	0%	1%
−4	79%	20%	0%	1%
−3	72%	27%	0%	1%
−2	64%	33%	2%	1%
−1	55%	39%	5%	1%
+0	45%	45%	9%	1%
+1	36%	49%	14%	1%
+2	28%	51%	20%	1%
+3	21%	51%	27%	1%
+4	15%	49%	33%	3%
+5	10%	45%	39%	6%
+6	6%	39%	45%	10%
+7	3%	33%	49%	15%
+8	1%	27%	51%	21%
+9	1%	20%	51%	28%
+10	1%	14%	49%	36%
+11	1%	9%	45%	45%

Note: Percentages are not cumulative.

Characters have starting attributes of +0, so one with at least a +1 skill rank is looking at a ~64% chance of some kind of success. As you go up, the distribution shifts to better quality success. This spread was inspired by PbtA games. I feel like it gives a nice sense of progression while still working within the constraints of static difficulties, which I want as part of maintaining the neutrality of the referee. There are ways to shift the difficulty up and down, but it’s meant to be based on the framing of the scene and/or consensus. If a challenge is a HQ one, it seems like it should be harder. Still a WIP there though.

Combat is a little different in that the target Armor is not fixed (= 11 + armor proficiency + one of Dodge, Block, or Parry), but it’s intended to follow a similar spread (more or less). I could use a fixed target, and the Armor provides a modifier to your roll, but then it feels like I’d have just reinvented THAC0 as THAC17 or something wacky like that.

 

[soviet]

There is no correct answer. You must first ask questions like:

• Who (thematically) are the characters?
• What does an individual roll represent?
• What other ways exist to manipulate the outcome of a roll?
• What is the purpose of play?
• What is the desired rhythm of play?

 

[Darth Solo]

The correct answer is 50% base chance of success doing anything modified by character ability.

 

[Lanefan]

I'd say there is no good blanket answer that can apply to everything in a game. Each individual case is different, and has a different expected success rate.

I mean, look at baseball. From the hitter's perspective, consistently succeeding at getting on base 1/3 of the time is pretty damn good. But that also means a pitcher's expected success rate in not allowing the batter to get on base is, ideally, higher than 2/3.

 

[GMMichael]

I mean, look at baseball. From the hitter's perspective, consistently succeeding at getting on base 1/3 of the time is pretty damn good. But that also means a pitcher's expected success rate in not allowing the batter to get on base is, ideally, higher than 2/3.

That's why I don't play baseball. You spend most of your time in the field watching other players failing to succeed.

Wait.

I think I just described D&D...

 

[aramis erak]

For me, the most intuitive is the DPG-CT, 2300 AD, and MegaTraveller task system It's called the DGP-CT as it was introduced in Digest Group Publications' supplements for Classic Traveller; the MegaTraveller system has some additional nuances, but is the same fundamental math. GDW (makers of CT, MT, and 2300) changed the math a touch by the die swap; the odds are actually better for harder tasks under 2300.

I'll note that I intuited the baseline pretty early in... like the first session I used it.
Note that all three games define skill of 1 as a level employable in trades or as ship crewmen; Doctors and Lawyers are defined as level 3... and certain CT programming tasks need Computers 8... which I've only seen in MT...

The labels are built for the average NPC... CT/MT Stat 7 or 8 (providing +1), skill 1; 2300 stat 10-11 (providing +2), skill 1.
CT and MT, it's 2d6, 2300 is 1d10.
This gives Min/Median/Max of CT/MT 4/9/14, 2300 of 3/8.5/13
Rushing shifts up a difficulty and reduces time; caution shifts down difficulty but doubles time.
Simple 3+ Given the minimum roll, the PC cannot fail.
Routine 7+Usually doable (CT/MT Nat 5+ 5/6 for 83.333% chance; 2300 nat 4+ for 70%); extra time makes it automatic.
Difficult 11+ doable most of the time if you can do it cautiously, risky otherwise; rushing is impossible. (CT/MT nat 9+, `27%; 2300 nat 8+, for 30%)
Formidable 15+ Requires taking extra time or having better than the base 1 rank employable skill and/or attribute... (CT/MT max 14 is below the needed; 2300 max 13 is below the requirement) Note that higher attribute (10+) can hit this rarely (3%); in 2300, this requires professional (as in, Graduate School level) skills or experienced employable skill and high attribute.
Impossible 19+ - the average guy working in field cannot do this, not even by extra time. The MT max allowed die roll gives this a slim chance in both.
I like this kind of analysis; I don't care for the CT/MT using att/5, as it's a 2d6 att; 2300 using att/5 is much better for me, as it's 4d6-4 for atts..., so I up the CT/MT TN's by 1, but use att/3 instead of att/4

 

[Voadam]

For me the sweet spot is generally an at the table baseline of 3/4 success. Generally as a DM and a player I prefer succeeding more often with a chance of failure over big chances of failure where either nothing happens in the story (a missed attack) or there is an actual incompetency.

With too big a chance of failure the incentive is to avoid the mechanics and shift to narratively dealing with stuff where you can, but for something like D&D combat I generally enjoy doing out the mechanical stuff as part of the game.

The tone I enjoy is a bit more action movie or superhero/Conan comic book level of action and competency.

I find the players generally provide enough bad choices and bad luck on rolls that I do not want the system to grind them down as well as a baseline.

 

[payn]

I'm liking Travellers flat progression and 2D6 base more and more. Though, I will say I am a fan of bounded accuracy in D&D. Seems not only the deigning is easier this way, but also easier to understand and feel as a player/GM.

 

[Pedantic]

I think it's as or more important to consider the player build/resource side. What is the set of spreads a player can access, and what is the resource cost of being good at something?

The incentives to specialize hinges on a pretty small window, where the marginal improvement at a given check need to be significantly higher, or the resource cost needs to be quite cheap. Pushing either of those too far and player conservatism will favor generalists.

 

[LegendaryAdventureStudios]

Hey, thought I could contribute for how Array of Champions has it.

First, some explanation about the system itself. Characters have what I would define as 'Strong, Normal and Weak' statistics. Level 1 Strong is 100 in the statistic, 85 is normal and 70 is weak. It's a D100 system where you subtract an enemies defence from your offence to find the DC (so if your Weapon Attack of 100 is targeting a Physical Defence of 35 you need a 65 to hit). It's worth noting the player also rolls for defensive rolls, not the GM - taking the enemy offensive statistic and reducing it from their defence (so a fireball might involve subtracting an enemies Supernatural Attack from your AOE Defence to find the target DC)

In practice when facing a same level enemy the following math occurs

Strong vs. Strong - 65% chance of success
Strong vs Normal - 80% chance of success
Strong vs Weak - 95% chance of success

Normal vs. Strong 50% chance of success
Normal vs Normal - 65% chance of success
Normal vs. Weak - 80% chance of success

Weak vs. Strong - 35% chance of success
Weak vs. normal 50% chance of success
Weak vs. Weak 65% chance of success

Now this math does get skewed - your level up progression is going to increase your statistics above the prescribed math curve (a "min maxed" character could alter the math to get +20% in ALL strong statistic chances at the level cap [20], (meaning they, on paper at least auto hit when targeting normal or weak)

Where things get more interesting math-wise is how the tactical application of buffs and debuffs from both sides can skew this math to create guaranteed successes or ruin solid chances into becoming near impossible. Over-committing to creating either situation can leave you vulnerable in a different area too!

Now do keep in mind the tone is Heroic Fantasy with Void Knights slashing enemies 300ft away, Heralds running at high enough speeds to exit combat by running to the future and suddenly having two of them present at once and other nonsense. Hence the skew towards players succeeding - besides in a system where players make every roll, they're bound to fail some.

 

[aramis erak]

I'm liking Travellers flat progression and 2D6 base more and more. Though, I will say I am a fan of bounded accuracy in D&D. Seems not only the deigning is easier this way, but also easier to understand and feel as a player/GM.

MegaTraveller had bounded accuracy in 1987... (die roll mod limit +8)