An Argument against Stat Normalization

[AlexanderHowl]

I do not believe in state normalization. In part, it is because of my training in research methods, which I taught for several years. Anything below two standard deviations from the average is meaningless, and the larger the sample, the larger the divergence must be to have real meaning. When it comes to populations as large as humanity's, six or more standard deviations are necessary to be remarkable.

In games, I believe that standard deviations are represented by attribute and skill levels. 12-13 is two above, 14-15 is three above, 16-17 is four above, 18-19 is five above, etc. Since attributes are harder to develop than skills, skills will usually be higher than attributes.

So, in my games, antagonists and protagonists are exceptional people. They have attributes much higher than the average, but even average people can have exceptional skills. In any case, most NPCs and PCs will have one or attributes at 14+ and one or more skills at 20+.

The end result? Heroic characters that can deal with legendary challenges. It is nothing to inflict modifiers of -10 in a campaign and modifiers up to -30 are not unknown. I guess that is my main argument, state normalization is boring, and it probably does not reflect reality any better than the standard system.

[Anthony]

Originally Posted by AlexanderHowl

I do not believe in state normalization. In part, it is because of my training in research methods, which I taught for several years. Anything below two standard deviations from the average is meaningless

No, anything below two standard deviations is not good evidence of being non-random. In any case, stat normalization usually just assumes the standard distribution for stats is low -- if IQ has an average of 10.5 with a standard deviation of 0.5, max human is probably about 14.

[RyanW]

Originally Posted by AlexanderHowl

In games, I believe that standard deviations are represented by attribute and skill levels. 12-13 is two above, 14-15 is three above, 16-17 is four above, 18-19 is five above, etc.

The conclusion follows logically from your premise, but you haven't made any effort to justify the premise. Stat normalization isn't saying that human intelligence (for example) is more limited than believed, but that IQ 18 represents far higher intelligence than is assumed. Or perhaps in some normalizers' opinion that IQ 18 leads to less interesting and/or balanced gameplay.

[coronatiger]

I feel that it makes no sense that (for instance) mouthpieces should be naturally good artists and scientists. Beyond a certain limit (which can change between settings/styles), players should take talents for their fields of expertise (and be allowed to buy extra levels during play) instead of improving DX/IQ.

That isn't as relevant for HT, but someone pointed out somewhere that a high HT opponent would need to take 6*HP damage before they'd go down. And that also stretches my suspension of disbelief.

[AlexanderHowl]

Multiple Talents are unrealistic and inefficient, as well as being unlikely, as someone who received reaction bonuses from multiple endeavors would likely be quite famous. In the case of someone like my fiancee, it makes more sense to give her IQ 16 than to give her IQ 12, Per 16, Will 16, Academic 4, Mathematical Ability 4, Musical Ability 4, Craftiness 4, etc. Yes, that means that she is likely the smartest person in the room except in very exclusive settings, but she pays for it in other ways. Disadvantages are a better way to represent realism than stat normalization.

[SilvercatMoonpaw]

I thought keeping stats low was about not having them overshadow skills?

[Kromm]

Originally Posted by Anthony

In any case, stat normalization usually just assumes the standard distribution for stats is low -- if IQ has an average of 10.5 with a standard deviation of 0.5, max human is probably about 14.

Pretty much exactly this. Something like μ = 10.5, σ = 0.5 (which I like) or μ = 10, σ = 1 is the usual assumption. There's no special reason to use a huge σ.

Originally Posted by AlexanderHowl

Multiple Talents are unrealistic and inefficient

Regarding realism, "slightly above-average attributes plus multiple Talents" is probably more realistic than generic high IQ (or DX) for polymaths, because except for the most extreme and noteworthy examples – and fictional ones – believable multidisciplinarians still have only a modest number of broadly related areas where they're strong, and far too many areas where they're weak to say, "Well, that's just an Incompetence quirk." There are exceptions, but they shouldn't be the rule.

As for efficiency, as I wrote elsewhere:

Originally Posted by Kromm

Point-cost efficiency is just an artifact of a game system that's designed to be a "close enough is good enough" model for fictional characters in adventure genres, and that isn't intended to be any kind of reality simulator. In particular, point-cost efficiency isn't equivalent to conservation of effort, evolutionary efficiency, or anything else in the real world. Minimizing the GURPS point cost of the attributes, Talents, and skills needed to get certain target scores on a GURPS character sheet isn't something sought by real people bent on self-improvement or something favored by nature; it's something desired by GURPS players.

It's okay to seek efficiency for a character as long as you realize that has nothing to do with realism. You're doing so to get a more capable character for your points. You're gaming . . . which is 100% acceptable when playing a game!

—

I say all this as someone who's on the record as having run very high-powered campaigns where every PC had high attributes. I'm not against high scores being prevalent among PCs! I'm just against arguments to justify that as "realistic." It isn't . . . it's a game construct.

[Tomsdad]

Originally Posted by AlexanderHowl

I do not believe in state normalization. In part, it is because of my training in research methods, which I taught for several years. Anything below two standard deviations from the average is meaningless, and the larger the sample, the larger the divergence must be to have real meaning. When it comes to populations as large as humanity's, six or more standard deviations are necessary to be remarkable.

The problem is what ever your statistical* definition of meaningless is, and the system's definition of meaningless is not necessarily the same.

for instance:

Originally Posted by AlexanderHowl

In games, I believe that standard deviations are represented by attribute and skill levels. 12-13 is two above, 14-15 is three above, 16-17 is four above, 18-19 is five above, etc. Since attributes are harder to develop than skills, skills will usually be higher than attributes.

So by that definition and the preceding assertion, 12-13 is the first meaningful difference from 10. (basically 11 is below meaningful difference, since 12 & 13 are the first meaningful ones)

so let's see:

10 or less is 50% chance of success on the 3d6 bell curve, 11 is 62.5% (a 25% proportional increase in likelihood of success)

Sw damage is 1d (avg 3.5) with ST10, but 1d+1 (avg 4.5) with ST11, or a 30% proportional increase on average

of course the significance of a difference of 1 means different proportional changes at different points in the bell curve (or other ranges used), but well game systems ≠ population distributions

Originally Posted by AlexanderHowl

So, in my games, antagonists and protagonists are exceptional people. They have attributes much higher than the average, but even average people can have exceptional skills. In any case, most NPCs and PCs will have one or attributes at 14+ and one or more skills at 20+.

The end result? Heroic characters that can deal with legendary challenges. It is nothing to inflict modifiers of -10 in a campaign and modifiers up to -30 are not unknown. I guess that is my main argument, state normalization is boring, and it probably does not reflect reality any better than the standard system.

Ultimately the bit in bold is the most important thing, you want exceptional people who can do legendary things? Then that's great and yep that's a great way to get that. And if that gives you what you want that's great too. But that doesn't necessarily extend to making broader statements like "stat normalisation in boring" in abstract, or that it doesn't reflect reality, or support your definition and extrapolation of what 2SD is in GURPS

Especially as your using the legendary and exceptional as proof, but well by definition most things, actions and people are neither.





*which is rather dependent on what your looking at rather than some blanket statement like anything below 2SD is meaningless, For example in the US a chap who is 6ft 3" is 1.9 SD and in the 97.4 percentile, but by your assertion you consider this a meaningless difference from the average 5' 9"?

I can't comment on your Statistical background (mine's in biology), but in some normal distributions 2 SD can be quite meaningful.

Also I'm not sure quite what you mean by:

and the larger the sample, the larger the divergence must be to have real meaning. When it comes to populations as large as humanity's, six or more standard deviations are necessary to be remarkable.

Assuming the distribution remains roughly the same the size of the population doesn't really matter? But I guess it depends on how your defining remarkable, do you mean it in terms of rarity or result/ability

Put it this way in the US of 1 in 39 chaps will be 6' 3" chaps in the US. Now that's not massively rare but does that also mean 6' 3" isn't meaningfully tall in comparison to the 5' 9" average? Or put it this way in a larger population there will be more 6' 3" chaps about, but they'll all be as good at reaching high shelves and better than their 5' 9" mates

Basically a lot of this comes down to how you define remarkable.

[Tomsdad]

Originally Posted by Anthony

No, anything below two standard deviations is not good evidence of being non-random. In any case, stat normalization usually just assumes the standard distribution for stats is low -- if IQ has an average of 10.5 with a standard deviation of 0.5, max human is probably about 14.

Exactly right

Cheers

TD

[Kromm]

Originally Posted by Tomsdad

I can't comment on your Statistical background (mines in biology), but in some normal distributions 2 SD can be quite meaningful.

Not to mention that normal distributions tend to be associated with things found in the real world, which have thermodynamic, sociodynamic, or similar reasons to take that shape. Given that GURPS attributes are artificial constructs intended to facilitate dramatic storytelling, and dramatic storytelling in even the most familiar settings and least over-the-top genres includes a good dose of imagination and fantasy, I'd be very, very hesitant to assume that those attributes are normally distributed. I can state as one of the designers that we certainly never imagined them that way. "Attributes are normally distributed" is one of those persistent myths like "GURPS IQ = (real-world IQ test score)/10" or "1 point always equals 200 hours of study" that isn't actually among the designers' assumptions for the game.