| 2 people marked this as a favorite. |
"Each ability score starts at 10, representing human average"
From Ability Score Overview.
| 3 people marked this as a favorite. |
Remember that half the people you meet are above average. That is literally what average means.
| 5 people marked this as a favorite. |
It's kind of pedantic, but technically no NPC has any ability scores, they just have modifiers. This obscures the fact that the modifiers are just set to provide an appropriate challenge or utility to the PCs, as the story unfolds.
The commoner statblock is also for a physical laborer. A librarian has similarly lopsided modifiers, just toward mental. So if you do want to interpret the modifiers as indicative of an underlying score, and you want to use that to calibrate your in-world expectations for ability scores, maybe "human average" could be more accurately described as "human baseline".
Additionally... yeah, that's kind of the conceptual niche of level 1 characters, a cut above the 'average person' without being superheroes... yet.
As something of an informed attribute, 10 is declared the average. Now that human adventurers get modifiers the ancestral average is at least 10.7 whereas the aforementioned turbo-commoner is rocking a 12.3 before rounding. Given the standard rules for rounding, the ancestral average still clocks in at 10 but our turbo-commoner is plainly a 12. This sounds like a clear cut case of power creep. The ancestry, in a vacuum, maintains a thin pretense of 10 is the average but as soon as they step out from behind this curtain of ambiguity and get a job in your campaign they lay bare the lie. Please note I am not complaining. It is only reasonable to assume that working for a living would shape one's attributes. Honestly, going from 10 to 12 after several decades, at least four revisions and a change of ownership or two is very mild power creep. Come to think of it, they might not be keeping pace with inflation.
| 2 people marked this as a favorite. |
Every person you meet is exceptional in some way. Talking about average or median ability scores is not ultimately very useful.
Like if you encounter a village where everybody is unusually literate that's more of a "fun worldbuilding quirk" than "something is up here, common people aren't supposed to be this well-read."
| 4 people marked this as a favorite. |
Remember that half the people you meet are above average. That is literally what average means.
That's what below average people believe.
Above average people know the difference between average and median. You're talking about the latter.
| 2 people marked this as a favorite. |
There was a discussion on a discord chat about it and someone pointed out that the basic commoner stat block has a +3 mod to str, +2 to con, and +1 to dex and wis, and I find it amusing that apparently your 18 str level 1 barbarian is only slightly stronger than a commoner
is that statblock representative of average joe schmoes around the world or are these jsut turbo commoners?
Worth noting that a level 1 character isn't that far above just having a background, and farmer is a reasonable background choice for a first level barbarian. The basic idea that people who do physical labor all day are stronger than average isn't particularly wild.
Strength +3 seems a bit high. +2 strength with the hefty hauler skill feat seems like a better fit. I also don’t see farmers as that dexterous, so a +0 is fine. +1 wisdom might be appropriate to represent the old farmer archetype, but maybe not the young farmer type.
So +1 str/+1 con from ancestry, +1 str/+1 con and hefty hauler from background with maybe a bonus to wisdom.
| 5 people marked this as a favorite. |
Remember that half the people you meet are above average. That is literally what average means.That's what below average people believe.
Above average people know the difference between average and median. You're talking about the latter.
Assuming a normal distribution the mean and median should be very very close. And for most people when they think of average they really are talking about the arithmetic mean. It's just that if we had an actual data set for everyone on Golarion it's unlikely the arithmetic mean would be exactly 10.0 (it's probably going to be between 9.5 and 10.5). But when casually talking about things people will probably just say 10.
You're technically correct, but your post comes off as condescending.
It's also worth noting that in a perfectly symmetrical distribution the mean and median should be equal.
Remember that half the people you meet are above average. That is literally what average means.That's what below average people believe.
Above average people know the difference between average and median. You're talking about the latter.
Assuming a normal distribution the mean and median should be very very close. And for most people when they think of average they really are talking about the arithmetic mean. It's just that if we had an actual data set for everyone on Golarion it's unlikely the arithmetic mean would be exactly 10.0 (it's probably going to be between 9.5 and 10.5). But when casually talking about things people will probably just say 10.
You're technically correct, but your post comes off as condescending.
That's Xeno's whole deal, don't take it personally. It is fun to watch when you aren't on the other side of it.
| 4 people marked this as a favorite. |
That's Xeno's whole deal, don't take it personally. It is fun to watch when you aren't on the other side of it.Remember that half the people you meet are above average. That is literally what average means.That's what below average people believe.
Above average people know the difference between average and median. You're talking about the latter.
Assuming a normal distribution the mean and median should be very very close. And for most people when they think of average they really are talking about the arithmetic mean. It's just that if we had an actual data set for everyone on Golarion it's unlikely the arithmetic mean would be exactly 10.0 (it's probably going to be between 9.5 and 10.5). But when casually talking about things people will probably just say 10.
You're technically correct, but your post comes off as condescending.
I'm not taking it personally, but I don't really like to see the board be a hostile place to (new or infrequent) posters when they don't come in posting in a hostile presumptive manner themselves.
|
| 1 person marked this as a favorite. |
Remember that half the people you meet are above average. That is literally what average means.That's what below average people believe.
Above average people know the difference between average and median. You're talking about the latter.
Some people wonder about "Your friends have a greater number of friends than you do." I wonder if you are more likely to meet people with above average charisma. Low charisma people probably avoid meeting new people.
Low Con people are probably more likely to die young.
Low Wisdom or Dex might make you more likely to die in a driving accident.
So I think most people you "Meet" are above the median.
|
| 1 person marked this as a favorite. |
Remember that half the people you meet are above average. That is literally what average means.That's what below average people believe.
Above average people know the difference between average and median. You're talking about the latter.
Some people wonder about "Your friends have a greater number of friends than you do." I wonder if you are more likely to meet people with above average charisma. Low charisma people probably avoid meeting new people.
Low Con people are probably more likely to die young.
Low Wisdom or Dex might make you more likely to die in a driving accident.
So I think most people you "Meet" are above the median.
Not me. I'm average at everything. I'm just lucky to be here. I probably should have died quite a few times already.
| 1 person marked this as a favorite. |
Not me. I'm average at everything. I'm just lucky to be here. I probably should have died quite a few times already.
Same, Trixleby. Same.
| 1 person marked this as a favorite. |
So I think most people you "Meet" are above the median.
It depends how you calculate the median, what is the sample chosen to determine what is the "average human".
Because depending on the society they live in, humans don't have the same stats. Rural societies will certainly increase Strength, Constitution and Wisdom over the others when urban societies will favor Intelligence and Charisma much more. So the stats of the "average human" can be very different depending on where this "average human" lives.I think we should not overthink stats. 10 is the average, so if you're at 12 you are above average and at 8 you are under average. Now, you choose how to play your character depending on what you consider to be "average".
| 5 people marked this as a favorite. |
Speaking of children, the CRB doesn't specify that 10 is the adult human average, so all those babies and toddlers could be dragging it down... ;-)
| 3 people marked this as a favorite. |
Wow, that's gotta be rough. Me? I'm amazing at everything I do, always and forever.Not me. I'm average at everything. I'm just lucky to be here. I probably should have died quite a few times already.Same, Trixleby. Same.
We all know about your perfect, golden, sexy soul Perpdog. No need to rub it in!
| 1 person marked this as a favorite. |
Speaking of children, the CRB doesn't specify that 10 is the adult human average, so all those babies and toddlers could be dragging it down... ;-)
Beat me to the punch =P
To say nothing of the sick, elderly, and disabled... Even if none of those traits strictly affects ability scores for PCs.|
|
| 1 person marked this as a favorite. |
Well at least before being very old gave +3 to all mental scores and -3 to all the physical scores. I forget if PF2 has those rules.Speaking of children, the CRB doesn't specify that 10 is the adult human average, so all those babies and toddlers could be dragging it down... ;-)Beat me to the punch =P
To say nothing of the sick, elderly, and disabled... Even if none of those traits strictly affects ability scores for PCs.
PF2 does not. The only thing age does is qualify for certain elf feats, as far as I remember.
| 6 people marked this as a favorite. |
You're technically correct, but your post comes off as condescending.
I thought it was funny.
| 1 person marked this as a favorite. |
You would.
| 2 people marked this as a favorite. |
Just look at the NPC stats:
Commoner:
Str +3, Dex +1, Con +2, Int +0, Wis +1, Cha +0
Beggar:
Str +1, Dex +3, Con +2, Int +0, Wis +1, Cha +1
Farmer:
Str +3, Dex +1, Con +3, Int +0, Wis +2, Cha +0
Those are level -1 or 0 NPCs and the range is from seven to nine ability boosts. Higher level NPCs get more.
PCs start with 2+2+1+4 ability boosts.
PCs thus are at best two boosts above the average man.^^
NPCs rarely have a -1 and less seems to not really appear on a humanoid creature.
So, 10 may perhaps be the human average in theory, but in practice, a 10 is putting you on the level of the lowest score a typical person on the street may have, with maybe a rare few having less.
It was already said early in the thread, but I think it bears repeating, NPCs don't calculate their ability scores the same way that PCs do, nor do they use them for the same things.
A +3 to strength doesn't mean that from an outsider's perspective that NPC should be seen as having a Strength score of 16 and thus being able to be compared with your Barbarian with a Str of 18, or the human average of 10. The numbers are there because the Commoner is a level -1 NPC and those are what -1 creature stats look like.
Personally I've always felt that an ability score of 10 represents human average, while 20 is the natural human maximum, and beyond that is only achievable with supernatural means.
| 4 people marked this as a favorite. |
If you think of it as a Gaussian distribution ...But that would be silly, wouldn't it? The Gaussian distribution is a continuous distribution defined on the whole real number line. The "amount of children" distribution is clearly discrete, and also defined on a half-open interval (no such thing as negative children) - basically on all natural numbers including zero. There's a bunch of distribution that you could fit the data to here, with Poisson being the most well-known. Gaussian isn't one of them.
We mathematicians routinely convert the continuous Gaussian distribution into a discrete distribution by dividing it into intervals. Given a Gaussian distribution such as f(x) = (1/sqrt(2pi))e^{x^2/2), we could turn it into a discrete distribution over integers by F(n) = the area under f(x) between n-0.5 and n+0.5. Thus, Jacob Jett made no error in calling a discrete distribution Gaussian.
Of course, classic Dungeons & Dragons instead used the distribution 3d6 to approximate the discrete Gaussian distribution. 3d6 has the advantage of having integer values. It also has mean 10.5, variance 35/4, and standard deviation 2.96. So I will talk about the 3d6 probability distribution instead.
Its values range from 3 to 18 with 3 occurring at a rate of 1/216, 4 at 3/216. 5 at 6/216, 6 at 10/216, 7 at 15/216, 8 at 21/216, and then we lose the triangular numbers and continue at 9 at 25/216, 10 at 27/216, 11 at 27/216. 12 at 25/216, and then back to the triangular numbers in reverse order, 13 at 21/216, 14 at 15/216, 15 at 10/216, 16 at 6/216, 17 at 3/216, and 18 at 1/216.
The title question, "Is 10 in a stat still the basic human average?" is answered "No" in that distribution. The average score is 10.5 rather than 10. A lackluster commoner in the 3d6 distribution would typically have three 10s and three 11s for ability scores rather than six 10s.
Then we factor in racial adjustments. Those values reflect not the individual character but the entire population that shares the character's ancestry. If an ancestry had a fixed +2 in Constitution and another fixed +2 in Intelligence, then we would have the probability distribution 3d6+2 for Constitution and Intelligence among the people of that ancestry. Those scores would average 12.5 rather than 10.5. The other four scores would still average 10.5. We can imagine a typical person of that ancestry to have one 12 and one 13 in Constitution and Intelligence and two 10s and two 11s in the other ability scores.
Humans, on the other hand, place their two +2 bonuses into two ability scores of the player's choice. For a non=player character, we treat that choice as random. Each ability score as a 1/3 chance of getting a +2, so the distribution for each ability score would be (2/3)(3d6) + (1/3)(3d6+2). The average score of that distribution is 11.166666..., which is 11 and 1/6. Thus, we can imagine an average human with five 11s and one 12. However, really it ends up as taking the three 10s and three 11s and randomly bumping up two of them, for either one 10, three 11s, and two 12s, or two 10s, two 11s, one 12, and one 13, or three 10s, one 11, and two 13s.
However, Pathfinder 2nd Edition made one big chance to the 3d6 probability distribution: It got rid of the odd numbers. Due to the symmetry of the 3d6 distribution where every odd number had an associated even number with the same probability, this set all odd-number probabilities to zero and doubled all the even-number probabilities. 4 occurs at a rate of 3/108, 6 at 10/108, 8 at 21/108, 10 at 27/108, 12 at 25/108, 14 at 15/108, 16 at 6/108, and 18 at 1/108. Even though the lowest value of 3d6 is an odd number and the highest value is an even number, the even-only 3d6 has the same mean at 1134/108 = 21/2 = 10.5.
Thus, a lackluster distribution of scores for a human commoner would be three 10s and three 12s, or two 10s and four 12s. Three 10s, two 12s, and a 14 is pretty common, too. The bonuses of a typical human commoner should sum to around +3 or +4. This is before I factor in background, class, and the four free ability score boosts at 1st level.
I don't know whether I ought to factor in in background, class, and the four free ability score boosts at 1st level. The bonuses from background could represent people who already have relevant high ability scores flocking to that background. Non-adventurers don't have classes. And I presume that the four free ability score boosts come from practice and training, since they repeat at 5th, 10th, 15th, and 20th levels, but do commoners have that practice?
It is mostly the old Nature versus Nurture question.
To answer those questions, let's look at some data, as Ryuhi did above. Archives of Nethys provides some negative 1st- and 0th-level humans for me to examine. Level measures threat rather than the life experience that would improve ability scores. The two are closely linked in adventurers but for non-adventurers the threat would be well below their life experience. A master pastry chef who studied in many countries and has great acclaim would still be only a 1st-level threat. But high threat means higher life experience, so our data should pull from low threat. I also split my data by source, since characters encountered during adventure paths often have some significance, which is correlated to life experience.
Level -1 from Gamemastery Guide
Adept Str +0, Dex +2, Con +0, Int +3, Wis +2, Cha +1 sum = +8
Apothecary Str +0, Dex +1, Con +1, Int +3, Wis +3, Cha +1 sum = +9
Apprentice Str +1, Dex +2, Con +1, Int +3, Wis +0, Cha +0 sum = +7
Barrister Str +0, Dex +1, Con +1, Int +3, Wis +2, Cha +4 sum = +11*
Beggar Str +1, Dex +3, Con +2, Int +0, Wis +1, Cha +1 sum = +8
Commoner Str +3, Dex +1, Con +2, Int +0, Wis +1, Cha +0 sum = +7
Harrow Reader Str +1, Dex +2, Con +1, Int +2, Wis +3, Cha +3 sum = +12*
Judge Str +0, Dex -1, Con +1, Int +4, Wis +4, Cha +2 sum = +10*
Librarian Str +0, Dex +1, Con +0, Int +4, Wis +3, Cha +1 sum = +9
Merchant Str +2, Dex +0, Con -1, Int +2, Wis +2, Cha +4 sum = +9
Physician Str -1, Dex +1, Con +1, Int +4, Wis +2, Cha +2 sum = +9
Servant Str +1, Dex +3, Con +1, Int +0, Wis +1, Cha +2 sum = +8
Server Str +1, Dex +4, Con +0, Int +0, Wis +1, Cha +2 sum = +8
Tax Collector Str +0, Dex +1, Con +0, Int +4, Wis +2, Cha +3 sum = +10*
Teacher Str +0, Dex +0, Con -1, Int +4, Wis +2, Cha +3 sum = +8
Urchin Str -1, Dex +3, Con +0, Int +1, Wis +1, Cha +2 sum = +6*
Level -1 from Adventure Paths
Abberton Ruffian Str +3, Dex +1, Con +2, Int -1, Wis +2, Cha +1 sum = +8
Antaro Boldblade (cleric) Str +2, Dex +0, Con +1, Int +0, Wis +2, Cha +2 sum = +7
Black Tear Cutthroat Str +2, Dex +3, Con +1, Int +0, Wis +0, Cha +0 sum = +6
Burnbearer Str +2, Dex +1, Con +2, Int -1, Wis +2, Cha +1 sum = +7
Foolish Hunters Str +1, Dex +3, Con +1, Int +0, Wis +1, Cha +2 sum = +8
Gold Tank Investor Str +2, Dex +3, Con +2, Int +0, Wis +0, Cha +1 sum = +8
Level 0 from Gamemastery Guide
Dockhand Str +3, Dex +1, Con +3, Int +0, Wis +1, Cha +1 sum = +9
Farmer Str +3, Dex +1, Con +3, Int +0, Wis +2, Cha +0 sum = +9
Miner Str +2, Dex +1, Con +3, Int +0, Wis +2, Cha +0 sum = +8
Torchbearer Str +2, Dex +3, Con +1, Int +0, Wis +1, Cha +1 sum = +8
Level 0 from Adventure Paths
Bone Shard Toughs Str +1, Dex +3, Con +2, Int +0, Wis +2, Cha +0 sum = +8
Burning Mammoth Hunters Str +2, Dex +3, Con +2, Int -1, Wis +3, Cha +0 sum = +9
Eunice (apprentice wizard) Str +0, Dex +4, Con +0, Int +4, Wis -1, Cha -1 sum = +6
Happs Bydon (bandit) Str +2, Dex +3, Con +2, Int +0, Wis +0, Cha +1 sum = +8
Logger Str +3, Dex +1, Con +2, Int +0, Wis +2, Cha +0 sum = +8
Barrister and Judge are prestigious positions, so they are not base commoners. Their +4 scores and sum of 10 or higher also marks them are not normal, so I will reject them from my data. The Tax Collector is similar enough to them to also be rejected. The Urchin is clearly a child, so he is out of my data set, too. The Harrow Reader has a weirdly high sum, so I remove her as an outlier. I wondered why the Black Tear Cutthroats from Kingmaker are so pathetic, but I left them in. That leaves 26 people with the following distribution of sums.
+6: 2
+7: 4
+8: 13
+9: 7
And the following distribution of 6*26 = 156 scores.
-1: 8 (5%)
+0: 39 (25%)
+1: 41 (26%)
+2: 37 (24%)
+3: 24 (15%)
+4: 7 (5%)
That does not approximate a Gaussian distribution. The distribution is too uniform at +0, +1, and +2 and it has a long tail with the 15% chance of +3. The mean is 1.33, the variance is 1.60, and the standard deviation is 1.27.
Since the odds are so close to multiples of 5% we could duplicate this with a d20 roll. Rolling 1 means a -1 ability penalty, rolling 2-6 means a +0 ability bonus, rolling 7-11 means a +1 bonus, 12-16 means a +2 bonus, 17-19 means a +3 bonus, and rolling 20 means a +4 bonus.
If we wanted a Gaussian distribution, then the binomial distribution (8/15+7/15)^5 gives a good approximation, with mean 2.33, so I shift it by -1 to get mean 1.33.
-1: 4%
+0: 19%
+1: 33%
+2: 29%
+3: 13%
+4: 2%
On the other hand, imagine using the system of building a player character, but choosing random ability scores at each level. However, to get a sum of +8 rather than +9, we use only three free ability boots. That gives a ability bonus a 1/2 chance of a free ability boost, a 1/3 chance of a boost from human ancestry, a 1/3 chance of a boost from background, and a 1/6 chance of a boost from class. That, too, has a mean of 1.33 but its shape is more like a Poisson distribution:
-1: 0%
+0: 20/108 = 18%
+1: 44/108 = 41%
+2: 33/108 = 31%
+3: 10/108 = 9%
+4: 1/108 = 1%
Of course, if we deliberately stacked them to favor high scores, we would end up with +0,+0,+0,+1,+3,+4.
I am accustomed to thinking of 1st-level player characters having better ability scores than the non-player characters in young-adult roles, but a sum of +9 is only a little better than a sum of +8, and some NPCs have that sum of +9, too.
| 1 person marked this as a favorite. |
This was in the context of the "amount of children" distribution, which has a finite lower bound (of exactly zero). Using a discrete quantitation of the Gaussian distribution won't get rid of the problem. You'd need to use a log-normal distribution if you insist on using Gaussian "underneath".
Yes, in theory the Gaussian distribution has an infinite domain of all real numbers. In practice, when I was a cryptanalyst trying to break codes by trying billions of partial keys and checking for a statistical bias in output that indicated a correct partial key, a finite population of billions falls short of the infinite domain.
The even-only 3d6 distribution generates ability modifiers from -3 to +4 with mean 0.25. And its standard deviation is slightly less than (1/2)(2.96). For simplicity, let me treat the standard deviation as 1.5. The bounds for a +0 bonus would be (-0.5,0.5). If we move the mean to 0, that moves the bounds to (-0.75,0.25). If we shrink the standard deviation to 1, that moves the bounds to (-0.5, 0.17). By the Z-table for the Gaussian distribution with mean 0 and standard deviation 1, the chance of landing in that interval is 0.57-0.31 = 26%. The next Z-table interval for +1 is (0.17,0.83). The chance of landing in that interval is 0.80-0.57 = 23%. The Z-table interval for +2 is (0.83,1.5). That interval give probability 0.93-0.80 = 13%. The Z-table interval for +3 is (1.5,2.17). That interval gives a probability 0.985-0.933 = 5.2%. Let me switch to a table:
Modifier, Gaussian, Even-only 3d6
-5, 0.07%, 0%
-4, 0.5%, 0%
-3, 2.7%, 2.8%
-2, 8.7%, 9.3%
-1, 19%, 19%
+0, 26%, 25%
+1, 23%, 23%
+2, 13%, 14%
+3, 5.2%, 5.6%
+4, 1.3%, 0.9%
+5, 0.2%, 0%
+6, 0.02%, 0%
The 0.02% probability at +6 is 1 out of 5000. The 0.07% probability at -5 is 1 out of 1428. Generating 238 NPCs would require 1428 ability scores, but very few GMs want to generate over 200 NPC stat blocks. Therefore, those low probabilities might as well be zero. The domain of the Gaussian distribution is infinite, but the part we care about is finite.
The Z-tables I found on the internet went up to only 4 standard deviations. The ones I used as a cryptanalyst when up to 12 standard deviations, because a 1 out of a trillion (American trillion 10^12) chance mattered in that field.
P.S. I am writing this answer mostly for the readers who will one day learn about Gaussian distributions and vaguely remember that the stats in Pathfinder 1st Edition Dungeons & Dragons provide practical examples. I used to be a mathematics professor, too, and don't want to pass up a teaching opportunity.
Ultimately, Mathmuse provides an elegant answer to the OPs question. Is 10 in a stat still the basic human average---No. But it never really was. The OPs question is really the wrong question (which Mathmuse is nice enough not to vividly point out).
I think that Blissey1's view that 10 is the base stat for commoners comes from the point-buy system. In Pathfinder 1st Edition, the Purchase option for ability scores said that 10 cost 0 points, 11 cost 1 point, 12 cost 2 points, 13 cost 3 points, 14 cost 5 points, 15 cost 7 points, 16 cost 10 points, 17 cost 13 points, and 18 cost 17 points. I gave my players 20 points to spend, the High Fantasy option. Anyone accustomed to point-buy would view 10 as the baseline.
Usually, they would buy their key ability score up to 16 by spending 10 points, and then convert that 16 to 18 with a +2 racial modifier.
I once crunched the point-buy numbers to see how well they fit the 3d6 curve. The costs fit the probabilities well, except that the 18 should cost more that 17 points.
I think a case could be made that a common human's dump stat averages to 10 in PF2.
PF2 is different from PF1 in that the latter has NPC classed individuals using an ability score array of 13, 12, 11, 10, 9, 8 before any racial adjustments. Even this array has ability scores averaging just under 11 after racial adjustments.
| 1 person marked this as a favorite. |
When you look at creatures ability bonuses, you'll see that they increase with level more than they increase with actual power. Seeing a small Silent Stalker having more strength than a Gargantuan Megaprimatus doesn't make much sense.
They are an abstraction.
I think possibly looking at distributions of stats are not the right way to think about the baseline.
An alternative interpretation is that "difficulty classes" for skill rolls are calibrated so that the average person gets a +0 from their stats (and whatever other bonus from their level of training.)
It was already said early in the thread, but I think it bears repeating, NPCs don't calculate their ability scores the same way that PCs do, nor do they use them for the same things.
A +3 to strength doesn't mean that from an outsider's perspective that NPC should be seen as having a Strength score of 16 and thus being able to be compared with your Barbarian with a Str of 18, or the human average of 10. The numbers are there because the Commoner is a level -1 NPC and those are what -1 creature stats look like.
Personally I've always felt that an ability score of 10 represents human average, while 20 is the natural human maximum, and beyond that is only achievable with supernatural means.
Them not being created the same does not suddenly mean that +3 is not really a +3 is not really a 16. The only reason they have the simplified stat is to make it easier on GMs, not because their attributes are any different from players.
As for the natural maximum (ignoring rage) was 24 (10 base +2 race +8 point +4 levels). This has been lowered to 22 (10 base +8 points +4 levels). Of note the magical maximum (ignoring rage) was lowered from 30 to 24.
A totally average human should theoretically, if you add up all their ability score modifiers, should have a net 0. Or at least in the range of -1 to +1. But all of them have a total a lot closer to +9, as Mathmuse showed. It would fundamentally break the assumptions of the world if we assumed those numbers are representative of how we should look at human commoner stats.
But if they didn't have stats like that, they'd break the mold of the math given for level -1 to level 1 monsters. That wouldn't be the worst thing in the world, since the ability scores aren't actually how Paizo calculates things like the to hit bonus for monsters/npcs, but it would be somewhat head-scratching why a level -1 creature with a Str score of 0 has a +5 to hit. It would feel wrong, especially when the monster-making guide does recommend that there should be a correlation between comparative stats and their associated features.
So they give them unusually high stats because it doesn't actually matter. Either the ability scores should be taken literally and their features have weird numbers, or the features correlate with the ability scores well but those scores are vague generalizations, not meant to be taken as seriously as those of the PCs.
Admittedly ALL of this is very much my own conjecture and I have no quotes from the designers to say that this is what they do, but it doesn't really matter. Scores for monsters have ALWAYS been weird. A grizzly bear no more strength than a level 1 Fighter who maximizes it?
As a side note, when I say that to me 20 represents the natural human maximum, I am assuming that PCs who level up through their adventures are exceeding the human maximum. Without any Apex items or anything, I am assuming that a level 20 Barbarian with a Str score of 22 is superhuman.
I think possibly looking at distributions of stats are not the right way to think about the baseline.
An alternative interpretation is that "difficulty classes" for skill rolls are calibrated so that the average person gets a +0 from their stats (and whatever other bonus from their level of training.)
I don't see that in the actual NPCs. For example, the Librarian at level -1 with ability modifiers Str +0, Dex +1, Con +0, Int +4, Wis +3, Cha +1 has Library Lore +13. In order for a PC to have a lore skill at +13, we would need at least a 5th-level expert with Int +4. Thus, I find it easier to believe that the librarian's Int +4 matters to boost her Academic Lore, Library Lore, and Arcana skills.
The librarian is essentially a 5th-level expert in a narrow field. They are creature -1 because as a combat challenge, the librarian merely hits people with a book. A few NPCs in the Gamemastery Guide have two levels. For example, the Barrister, creature -1, has a sentence, "In a court case or other legal proceeding, the barrister is a 4th-level challenge." So the barrister is a 4th-level NPC that fights as poorly as a street beggar, but is a worthy adversary in the courtroom. The barrister uses Cha +4 to achieve Diplomacy +12 for Sway the Judge and Jury ability that can win cases with a Diplomacy check.
The NPCs are designed to look like they use their high ability modifiers.