I have not seen any disdain for roll under in actual play, except where it is mixed with roll high mechanics. In actual play, BRP is very well received.

D20 was smart to have "roll high" for everything. Back in AD&D, we had D20 roll high for combat and then D20 roll low for ability rolls and that was schitzo. Palladium is schitzo with the % skills and D20 for combat and then nothing for ability rolls.

Our old Traveller games were schitzo too when 2D6 roll high was used with skills and combat, but lots of GMs used roll under for ability scores and the inevitable complaint of why not roll under vs. high INT vs. my only +1 skill roll.

Quote from: Herr Arnulfe;731878Right, earlier in the thread we talked about "diminishing returns" for XP spent at higher skill levels, and it was pointed out that systems with linear randomizers achieve this through escalating XP costs, instead of diminishing skill improvements.

Point is that skills with nonlinear randomizers can also achieve this--or close enough--if the costs are chosen carefully. And once you take this into account by looking at the "skill level" as the cost rather than the numeric value, you can see that an increment in skill level is very different from a situational modifier.

I think this discussion was presaged on rec.games.frp.advocacy, which is where I originally posted my argument starting with this: https://groups.google.com/d/msg/rec.games.frp.advocacy/ODHOm3sVqbo/NcEP97jfSqYJ

(Note that I corrected some details as the thread moved on.)

Quote from: Herr Arnulfe;731865OK, I didn't actually need an explanation of what a bell curve does, but thanks anyway for the elaborate demonstration.

Then why the fuck did you ask?

Quote from: Adric;731714I'm just not convinced that increasing chances of success by such small increments is necessary. On a d20, each number has a flat 5% chance of occurring. That means that the smallest amount you can improve boosts your chance by 5%.

For D%, the smallest amount you can improve is 1%. If skills regularly increase by more than 1% in a given system, why track it at such fine detail? Just round it off to the nearest 5% and use D20 or the nearest 10% and use D10.

Another problem with a pass/fail system that uses d% is that 9 times out of 10, the second die won't matter. If the target is say, 55%, the 10's die needs to be a 5 for there to be any tension on the 1's die. If the target number is a flat multiple of 10, and there are no modifiers, the second die never matters at all.

Because percentile dice allow you to parse it out straight up without the needed calculations. And because sometimes rolling a d20 just to be rolling a d20 can feel a bit... lame.

As for the one roll making the second roll not needed. How is this even a valid argumant? Its not. You are usually rolling percentile dice both at once so if you need to read it you read it and of you dont you dont. Rarely are percentiles rolls in increments of 10. Also mostly invalidating that argument.

Some games do not follow a 5 progression. Palladium games comes to mind. Star Frontiers is another. Universe I believe as well.

Other games use a 5 progression deliberately so you CAN swap in a d20 if you dont have a pair of d10s handy.

Quote from: ZWEIHÄNDER;731863Critical Success and Critical Failure
If you roll matching dice on percentiles and succeed your chances, you generate a Critical Success. If you roll matching dice on percentiles and fail your chances, you generate a Critical Failure.

What are the odds on this? How often does it happen?

Quote from: Shipyard Locked;731909What are the odds on this? How often does it happen?

This is reasonably easy to work out:

0-10%: No chance of critical success, 10% chance of critical failure.

11-21%: 1% chance of critical success, 9% chance of critical failure.

22-32%: 2% chance of critical success, 8% chance of critical failure.

33-43%: 3% chance of critical success, 7% chance of critical failure.

44-54%: 4% chance of critical success, 6% chance of critical failure.

55-65%: 5% chance of critical success, 5% chance of critical failure.

66-76%: 6% chance of critical success, 4% chance of critical failure.

77-87%: 7% chance of critical success, 3% chance of critical failure.

88-98%: 8% chance of critical success, 2% chance of critical failure.

99%: 9% chance of critical success, 1% chance of critical failure.

100%: 10% chance of critical success, 0% chance of critical failure.

Note that it takes 22% in a skill before you raise the chance of critical success to 2% or more, whereas the chance of critical failure doesn't go below 2% until you're at 99%. Grim and perilous indeed!

Quote from: Justin Alexander;731892Then why the fuck did you ask?

I never asked for an explanation of bell curves. I asked why it's important for the results to be clustered towards average, and how that impacts the feel of a game.

Quote from: Arminius;731891Point is that skills with nonlinear randomizers can also achieve this--or close enough--if the costs are chosen carefully. And once you take this into account by looking at the "skill level" as the cost rather than the numeric value, you can see that an increment in skill level is very different from a situational modifier.

OK, so we're back to the topic of situational modifier effects being diminished for extreme values on a bell curve, as discussed earlier. From a gameplay perspective I can understand why some players (e.g. Brander) might prefer their ultra-skilled sniper PC not have to worry about stuff like wind, moving targets etc., although from a realism perspective there are many examples to the contrary (e.g. an average hockey player won't notice the difference between a wood and a composite stick, whereas an NHL players does).

For what it's worth, Dungeon World and its clones are one of the bell-curve-based systems in which the clustering of results around the mean is tangible IMO, because the curve is used primarily to regulate degrees of success and failure. However, I think DW would work just as well with a linear randomizer. The bell curve creates some "warping" of DoS/DoF bands at different proficiency levels, which doesn't necessarily map to reality IMO.

Quote from: Herr Arnulfe;731917For what it's worth, Dungeon World and its clones are one of the bell-curve-based systems in which the clustering of results around the mean is tangible IMO,

It is very tangible as the most likely results are targeted at the "yes, but" range of 7+ (you need high stat/skill bonuses to get a 10+ full success).
So even when a character is moderately proficient the game system aims to hit the "hard choices" of "choose two out of three results" to ensure "drama".

Quote from: Dirk Remmecke;731921So even when a character is moderately proficient the game system aims to hit the "hard choices" of "choose two out of three results" to ensure "drama".

Agreed, although most bell curve systems don't use degrees of success/failure AFAIK (with the exception of diepool systems where you count successes towards a target number, but that's a whole other discussion). You could still cluster the majority of rolls in the "yes, but" category in DW using a linear randomizer though.

Quote from: Herr Arnulfe;731914OK, so we're back to the topic of situational modifier effects being diminished for extreme values on a bell curve, as discussed earlier. From a gameplay perspective I can understand why some players (e.g. Brander) might prefer their ultra-skilled sniper PC not have to worry about stuff like wind, moving targets etc., although from a realism perspective there are many examples to the contrary (e.g. an average hockey player won't notice the difference between a wood and a composite stick, whereas an NHL players does).

I don't know much about hockey. But you're mischaracterizing the effect. An ultra-skilled sniper won't have to worry about wind IF the shot is one that (absent wind) a moderately-skilled marksman has a fair chance of making. If it's a challenging shot for the ultra-skilled, then wind will make a big difference to him, but it won't matter much to the moderately-skilled who was facing an impossible shot anyway.

Quote from: Arminius;731931I don't know much about hockey. But you're mischaracterizing the effect. An ultra-skilled sniper won't have to worry about wind IF the shot is one that (absent wind) a moderately-skilled marksman has a fair chance of making. If it's a challenging shot for the ultra-skilled, then wind will make a big difference to him, but it won't matter much to the moderately-skilled who was facing an impossible shot anyway.

Is that more realistic than a flat modifier? I don't know much about sniping, but I can think of several examples where a normal person would be penalized more heavily by circumstantial factors while attempting a difficult task than a skilled person would be. (e.g. normal climber vs. skilled climber trying to scale a sheer cliff without a top-rope).

It seems there's no single "rosetta stone" die-system for reality simulation, because reality follows a wide variety of differently-shaped curves depending on the task, some of which are actually the opposite of a bell. Hence my preference for flat probabilities - at least they fall somewhere in the middle-ground of the various possible curves.

And to be clear, I don't have strong feelings about die-systems, as long as the GM understands them and employs them well. Usually they fade into the background once the game gets underway anyway. I do have strong feelings about people with strong feelings about die systems.

Quote from: Herr Arnulfe;731927Agreed, although most bell curve systems don't use degrees of success/failure AFAIK (...). You could still cluster the majority of rolls in the "yes, but" category in DW using a linear randomizer though.

Yes, but you would have to make the "yes, but" range broader to achieve that.
For someone not versed in bell curves DW's "yes, but" range is "just three numbers", not "those three numbers".

I wondered that this feature of DW didn't come up in the many DW threads we had on this forum. This feature that still gives away the Apocalypse World story game heritage of DW. ("Put the player between a rock and a hard place, push the player towards hard choices!")

It's really not an issue of flat modifiers as much as how results tend to cluster near the mid-range.

Here's a simple system for the purpose of a thought experiment.

Let's say we determine Jump by rolling 1d20+STR, and let's say we divide that result by 2 to determine distance jumped (ie, ever point on our scale works out to 6 inches).  Assume a STR modifier of +4.

We can roll a 1 (total result of 5) which works out to 2.5 feet.  We can roll a 20 (total result of 24) which works out to 12 feet.  

Jump three or four times.  How far did you jump each time?

Regardless of the specific distance, the jumps probably clustered in a narrow range.  A d20 makes the minimum result just as likely as the maximum result (5% chance for each) - every result is equally likely - even an 11.  Take a 2d10 and while the chance for a minimum result is equal to that of a maxium result (1%), you're 10x more likely to roll an average result (10% chance of an 11).  

Using the same formula, the odds of rolling an 8-14 are ~60%, giving our hypothetical jumper a range of 6-9 feet for the majority of their jumps.  There's only a 20% chance that they jump less than that or more than that per jump.  

By forcing the dice to create the bell curve, you can use a simple formula rather than something like a table lookup: ie, 1-4=4 feet, 5-8=5 feet, 9-10=6 feet, 11-12=7 feet, 13-16=8 feet, 17-20=9 feet etc..  

I personally like consistently applied formulas (I think they're easy to remember), and I think that most skilled people should usually do 'about as well as average' on most things they do.