Quote from: jhkim;857339Some other options:

1) Roll 3d20 and keep the middle result - variance 20.1
2) Roll 3d6 - variance 8.75

Note that 1d20 has a variance of 33.25The first can also work for percentile systems - roll 3d100 and keep the middle result.

Good options jhkim.
The "middle of three" seems like too much rolling.
Switching from flat (1d20) to bell curve (3d6) has other implications, which may or may not be appropriate.

Quote from: JoeNuttall;857348Good options jhkim.
The "middle of three" seems like too much rolling.
Switching from flat (1d20) to bell curve (3d6) has other implications, which may or may not be appropriate.

If rolling three dice and adding them together is OK, why is rolling three dice and taking the middle too much rolling?

Quote from: jhkim;857362If rolling three dice and adding them together is OK, why is rolling three dice and taking the middle too much rolling?

Actually you're right - the middle of three d20 is probably fine. I think it was your talk of three d% that sounded too much!

Jhkim you’re a genius :-)

Three rolls isn’t too many – in fact it’s just the right number – if all the rolls mean something.

I got to wondering if all three were over the target then that would be absolute success, and if all three were under that would be abject failure.
So zero successes = abject failure, one success = partial succes, two successes (i.e. middle of three) = success, three successes = absolute success.

This:
1) gives different levels of success.
2) builds up tension for an abject failure or absolute success.
3) doesn’t require you to change the target number or do any maths.

If you use it with my open 2d10 system (replace any zeros with another 2d10 – below 50% chance of success it drops away at x80% per +1 difficulty), then in addition:

4) increasing difficulty has more impact on chance of partial success than absolute success. That is, an inept can fluke a partial success, but not an absolute success. (+1 makes partial success 80% as likely, but success only 64% as likely, and absolute success only 50% as likely. Similarly there is a reduced standard deviation from partial success (4.7) through to absolute success (2.9)).

Here are some numbers:

If you need to roll 8+ then you have only 99% chance of partial success, 88% chance of success, 48% chance of absolute success.

If you need to roll 11+ then it's 89% partial, 54% success, 14% absolute.
 
If you need to roll 15+ then it’s 48% partial, 12% success, 1% absolute.

If you need to roll 20+ then it’s 28% partial, 1% success, 0.01% absolute.

Applying this to climbing, absolute success means you got up it in double quick time, a failure would mean you didn’t climb it, but partial success could mean you get half way up and can either choose to abandon it, or try again. A second partial success would get you to the top, but a failure means you fall…

Applying it to arm wrestling, the multiple rolls idea makes it more of a contest, so it should be adapted to be a generic grappling solution. Each round you are grappling you make three opposed rolls (each side rolls only a single d10, not 2d10, so it’s the same odds). The first person to get three successes in a row wins. This means you *can* win in a single round. For evenly matched opponents average duration = 2.3 rnds, 99% won in 8 rounds or less. If you have +3 then you’ll win 95% of the time, +5 and it’s 99% of the time. So if you’re good you’ll beat all the village.

Quote from: Larsdangly;855933I just meant that, if you line up 10 people chosen at random on the street and have the strongest one arm wrestle the weakest, the outcome will be so one sided you would hesitate to even let them proceed. This is basically equivalent to having a ST 6 person arm wrestle a ST 15 person in D&D. In the versions of the game I can easily think of, this would be close enough you would have to roll, with a decent chance the weaker person would win.

Taking a base of 50-50 for equal ST, and reflexively shifting a point each way on the d20 (so sum is always 100%), a 9-point difference gives 19:1 odds.

That's how Chaosium has done it ever since 1st ed. RuneQuest (1977 or 78).

Going back to Metamorphosis Alpha (1976) the usual factor comparison is similar but on 3d6, so a mere 5-point difference gives about 21:1.

In Tunnels & Trolls (1975), if you merely had each try to make a Level 1 Saving Roll, the odds would be only about 9:1; but more probably you'd compare the actual totals with the higher winning, which I'm guessing would do much more than roughly double the ratio.

These are all using 3d6 for normal "man on the street" spread.

Quote from: Phillip;857462Taking a base of 50-50 for equal ST, and reflexively shifting a point each way on the d20 (so sum is always 100%), a 9-point difference gives 19:1 odds.

That's how Chaosium has done it ever since 1st ed. RuneQuest (1977 or 78).

Going back to Metamorphosis Alpha (1976) the usual factor comparison is similar but on 3d6, so a mere 5-point difference gives about 21:1.

Yes, there are games that do this better than others; I was referring to the 'base line' of various D&D editions. Though you could argue that even the Chaosium approach is unrealistic in this particular case. It is like imagining that a withered, 100 pound old man could beat a 200 pound dock worker 1 time out of 20 at arm wrestling. In reality, he would get his arm broken a million times out of a million.

In reality magicians fail to work magic a million times out of a million. Assuming an affirmative answer to whether a Giant could even exist, ankle-biters might seem to be in rather more of a fix to fight it than typically depicted.

Quote from: Larsdangly;857464Yes, there are games that do this better than others; I was referring to the 'base line' of various D&D editions. Though you could argue that even the Chaosium approach is unrealistic in this particular case. It is like imagining that a withered, 100 pound old man could beat a 200 pound dock worker 1 time out of 20 at arm wrestling. In reality, he would get his arm broken a million times out of a million.

Depends on what the roll includes. I'd say there's a greater than 0/1,000,000 chance that the old man does something sneaky to win the contest, or that the dock worker does something stupid or clumsy or has bad luck or an old injury messes him up or something. But if almost all of those things are out of scope for the roll, then yeah the Arm Wrestling Rules or GM Discretion should include that a certain amount of difference in actual arm strength results in a win. And lazily using a game's standard mechanic, especially when it includes a minimum 5% chance of a surprising result, is flawed. Especially if the consequences are serious (e.g. chance not to fall off a cliff to your death, regardless of skill or precautions). That's when lazy rules lazily applied mess things up.

Well, really the line where things are messed up is determined by what players are sensitive to. You can give some young kids an immersive experience by leading them to an arcade video game with an "attract" mode, letting them touch the controls and not even putting a quarter in. Older players may complain when you do or don't use certain types of rules, according to their personal tastes and ideas.

I'm really not able to be around much, so this slipped by.

QuoteOne of my pet peeves in RPG design has to do with the effect of skill. In the real world - and in many fictional worlds - someone with expert skill can reliably do tasks that a beginner has no chance at. Examples I gave from the earlier thread include,


Originally Posted by jhkim  View Post
1) An expert professional acrobat can do a back flip hundreds of times on stage without failing. A beginner can try a back flip over and over and never succeed.

2) An expert computer programmer can write a quick program to do something that someone new to computers has no chance at. Say, find the frequency that each of a given list of names appears in a given ebook.

3) An expert locksmith can reliably pop open a door that a beginner can't get through at all.

4) A grandmaster at chess can reliably beat someone who is middle-ranked. In turn, a middle-ranked chess expert can reliably beat someone who is a beginner to chess.

5) An expert sniper can reliably make a shot that a beginner can't hit even after dozens of tries.

6) An expert mountain climber can reliably make it up a cliff face that a beginner can't get up after dozens of tries.

This is not reflected by many systems. For example, in BRP, rifle skill starts out at 25% for someone with no training, and 90% is considered high expert skill.
My preferred way to deal with this is to have this built into the skill system. For example, suppose my resolution mechanic is to take stat total + 1d10 and compare to a difficulty number (like Eden Studio's Unisystem).

I can say that a backflip is difficulty 15, an expert professional acrobat has stat total 14 or more, and a beginner might have a stat total more like 3 or 4. Then I can easily scale this, so that a legendary acrobat might have a skill of 22, and reliably perform feats that even experts regularly fail at.

Your system must take into account the difficulty level of the skill as well as rate of gain in that skill.  Cumulative skill trees/pyramids make this easier.  SO that you can have Stat+ basic skills+ advanced skills.
This makes the most sense in your computer example, where computer programming is an advanced skill, and all the familiarity in the world won't allow you to do something without that advanced knowledge.

 So maybe 0-5% attribute bonus for your rifle example, basic gun skill at +1-4% per level of the skill, rifles at 2-7 per lvel of the sub skill, and maybe sniper skill as a sub of rifle at 2-12 per level.  If your difficult shot is a -5% difficulty, and you have a beginner with no skill, or even someone with 3 or four levels of gun skill, at an average of 2.5% ability per level, that's going to need a critical hit basically.
But your expert, with level 7 gun, level 4 rifle, and level 3 in sniper (sub skills have to be lower level than parent, more general skills)?  (7*2.5)+(4*4.5)+(3*7)=52% skill    

Also a % skill system will have greater gradation and allow for smaller incremental growth.  Note that the more specific and advanced the skill, the slower the level gain.

Quote from: LordVreeg;857491So maybe 0-5% attribute bonus for your rifle example, basic gun skill at +1-4% per level of the skill, rifles at 2-7 per lvel of the sub skill, and maybe sniper skill as a sub of rifle at 2-12 per level.  If your difficult shot is a -5% difficulty, and you have a beginner with no skill, or even someone with 3 or four levels of gun skill, at an average of 2.5% ability per level, that's going to need a critical hit basically.
But your expert, with level 7 gun, level 4 rifle, and level 3 in sniper (sub skills have to be lower level than parent, more general skills)?  (7*2.5)+(4*4.5)+(3*7)=52% skill

That still means that expert skill is making fairly little difference. To take another example, let's say there's a slightly easier shot, +20% difficulty.

Now the beginner has 7+20=27%, while the expert sniper has 56+20=76%. That means that the beginner is going to match or outperform the sniper pretty regularly.

Quote from: JoeNuttall;857446Jhkim you're a genius :-)

Three rolls isn't too many – in fact it's just the right number – if all the rolls mean something.

I got to wondering if all three were over the target then that would be absolute success, and if all three were under that would be abject failure.
So zero successes = abject failure, one success = partial succes, two successes (i.e. middle of three) = success, three successes = absolute success.

This:
1) gives different levels of success.
2) builds up tension for an abject failure or absolute success.
3) doesn't require you to change the target number or do any maths.

Dear Lord! I think you've reinvented Mutant Chronicles' "2d20 system", but with 3 dice. Points for finding a use for it, though.

Quote from: Bloody Stupid Johnson;857513Dear Lord! I think you've reinvented Mutant Chronicles' "2d20 system", but with 3 dice. Points for finding a use for it, though.

As far as I can tell the Mutant Chronicles system is a universal resolution system, that's not what's needed nor what I'm proposing. I'm sure the mechanic of "count number of successes" has been around a long time.

I'm suggesting that we combine

Quote from: Eric Diaz;854577One solution is simple: just DO NOT use the same system for throwing a punch and for climbing.

with

Quote from: AsenRG;854819Then again, Unknown Armies 2e manages to have three different kinds of rolls for different situations. And they all use the same d100 system and the same skill number;).

i.e. two variants of a single system, using one variant in one sort of situation, another variant in another. As in

Quote from: Eric Diaz;854618Things that take a moment should be "swingy" (throwing a punch, shooting an arrow), things that take more than that shouldn't (playing a song, climbing a wall, etc).

The argument on the issues is still going on in the thread, some arguing you can have the same system for combat and all other contests

Quote from: Phillip;857462That's how Chaosium has done it ever since 1st ed. RuneQuest (1977 or 78).

And some insisting it should be different

Quote from: Larsdangly;857464It is like imagining that a withered, 100 pound old man could beat a 200 pound dock worker 1 time out of 20 at arm wrestling. In reality, he would get his arm broken a million times out of a million.

And some blaming the system or the players

Quote from: Skarg;857476That's when lazy rules lazily applied mess things up.

I wrote down a set of requirements for such a system, in particular (as I posted here) I wanted something where +1 skill meant twice as much or three times as much as it did in combat.
My system has an exponential tail off (+1 means x80%) so I wanted +1 to mean x80%, x80% squared, x80% cubed, i.e. x80%, x64%, x51%. Via a complete fluke not only did jhkim's suggestion solve all my other requirements, it matches the probabilities I specified *precisely*.

That is I decided to follow Phillips advice:

Quote from: Phillip;853978Whatever your real assessment of the situation, put that first -- and anyone who can write a fraction can make a randomizer if need be.

This isn't rocket science unless you've got the whole thing ass-backwards.

Make an assessment of the situation, worked out what I wanted the randomiser to be, and came up with one.

It might not be rocket science, but I'm very pleased with the resultant system.

Quote from: jhkim;857506That still means that expert skill is making fairly little difference. To take another example, let's say there's a slightly easier shot, +20% difficulty.

Now the beginner has 7+20=27%, while the expert sniper has 56+20=76%. That means that the beginner is going to match or outperform the sniper pretty regularly.

Again, easy to adjust the numbers to get the exact feel you want.  In my games, I prefer having chance have more of an effect.  Do rifles at 2-12 per level and sniper at 2-20 per level (since sniper is a rarer skill and requires time and preparation) and you end with 7% skill vs 81.5% skill.

On rifle (or other firearm) skill: To the extent that statistics have been collected, a 90% hit rate in actual combat (which is what most game rules are concerned with) would be not just 'expert' but amazing.

It's different enough from being on a range shooting at targets that don't shoot back, that one is not terribly predictive of performance in the other.

A) If you've already made up your mind, WTF ask a book for an opinion you don't want? Never mind that it might not be intended to give that opinion in the first place, so you're the rules' bitch by playing your fool self twice.

B) In most cases, you're supposed to be playing with others. In a group of 4 or 5, odds of 3:1 or 4:1 for the best vs. the rest mean the rest can make a worthwhile contribution. Again, WTF with arguing over numbers pulled out of your ass when what you want is simply ALL YOU BASE IS BELONG TO ME?

"Only Brad Hammer could handle this, and he's dead, so game over."
"Not so fast! Mack Steele can't disarm a bomb, but he can make an anti-nerve-gas bomb with common household ingredients and a lottery ticket."