Point Value of Stats Beyond 8-15?

[Tosamu]

Edit: Since it's come up a few times, I don't plan to actually use a point-buy system with these values. I want to model the results of rolling stats, but I need to have a way to handle scores beyond the real point-buy system.

I'm curious to calculate the average point-buy values of some of the more common dice-based stat generation methods - 3d6, 4d6b3, etc. The problem is that I'm not sure how to value rolls outside the allowable point-buy range. For example, if a player rolls 17/14/12/10/9/6, is that better or worse than the standard array of 15/14/13/12/10/8?

Following the PHB's pattern would yield something like the table below, but I'm not sure I'm taking the right approach. I left the official values from the book in bold for reference.



Score
Modifier
Point Value


3
-4
-7


4
-3
-6


5
-3
-4


6
-2
-2


7
-2
-1


8
-1
0


9
-1
1


10
0
2


11
0
3


12
+1
4


13
+1
5


14
+2
7


15
+2
9


16
+3
12


17
+3
15


18
+4
19



Does this seem like a reasonable guess?

[Naanomi]

I'd work it to dis incentivize dumping instead of encourage it... -1/-1/-(1/2)/-(1/2)/-(1/3)?

[Tosamu]

I'd work it to dis incentivize dumping instead of encourage it... -1/-1/-(1/2)/-(1/2)/-(1/3)?

I don't need to incentivize or disincentiveize anything, since I'm only using these numbers to estimate the point value of stats generated through rolling.

For example, if I rolled 4d6 drop lowest, and got the array 17/14/12/10/9/6, I can't meaningfully compare it to the results of point buy (or other stat generation methods) without knowing the value of a 17 and 6.

[Naanomi]

Ah fair enough... Still the penalty of very low stats (outside of Constitution) is marginal at best, I'd be tempted to just say -1 for each point below 8 even just as an evaluatory system

[Witch of Whimsy]

The way I tend to look at it, the current system is set up so each point in a stat costs an amount equal to the modifier provided, with a minimum of 1. up to 13 costs 1 point each, as each point is still working with a modifier of 1 or lower, but taking it to 14 for a +2 costs that same 2, and the 15 is working inside a +2, so also costs 2.

By this logic, it assumes 16 would cost 3, 17 another 3, and 18 costs 4.

As for reducing beyond 8, this is similar, but a bit more nuanced. With the way things are currently, you start at 8/8/8/8/8/8, requiring 12 points to get to an average of all 10s. Essentially, the 27 point buy assumes a 15 point buy after you put 12 in all your stats to get all 10s. This is so you are never "taking away" stats, but instead "neglecting to increase" them, which is a more positive/optimistic style. In this way, you can treat 9-8 as the system already taking the -2 to reduce each stat to 8, providing that extra 12.

You then follow the reverse logic from increasing stats, assuming you decrease the modifier when you hit last threshold for the current modifier. Doing that, you get another -1 for 7, -2 for 6, -2 for 5, -3 for 4, and another -3 for 3.

You then feed those 12 points back into the current logic of the 27 point buy system, and turn the -1s for 8 and 9 into what they are now, and end up with...



Score
Modifier
Point Value


3
-4
-11


4
-3
-8


5
-3
-5


6
-2
-3


7
-2
-1


8
-1
0


9
-1
1


10
0
2


11
0
3


12
+1
4


13
+1
5


14
+2
7


15
+2
9


16
+3
12


17
+3
15


18
+4
19



...Does that suffice?

[BurgerBeast]

The logic of the previous post seems pretty spot on, but I don't think it wise to count stats below 8 for bigger negatives. Having a dump stat of 8 compared to 3 makes less difference to character than a 2-point difference in your primary or secondary stat. Maybe it's not that extreme but I think you get my logic.

[Naanomi]

The logic of the previous post seems pretty spot on, but I don't think it wise to count stats below 8 for bigger negatives. Having a dump stat of 8 compared to 3 makes less difference to character than a 2-point difference in your primary or secondary stat. Maybe it's not that extreme but I think you get my logic.
I agree, lets take it to an extreme and see what it looks like...

Dropping three stats to 3 gives us -33 point equivalent. Add that to a basic 27 and we have 50 points to have as an equivalent.

So lets say a... half-elf Paladin can rock 18/3/18/3/3/20 and have equivalent value to a 'normal' character? Or a Mountain Dwarf Barbarian with 20/16/20/3/3/3?

[Witch of Whimsy]

It's less about the situational condition of whether or not the stat is as valuable as others (since the system doesn't care about any of that; every stat is worth just as much as the others, regardless of class), but rather how the system evaluates the "worth" of certain values.

Also, I strongly disagree that a 3 is only marginally worse than an 8 in a dump stat.


If you have a 3 in Strength, you are physically incapable of jumping (you actually have a -1 modifier to high jumps on a running start). You are mechanically still able to jump 3 feet forward on a running start, but if you are incapable of getting off the ground, you aren't jumping. It also imposes a maximum carry capacity of 45, and if you use variant encumbrance, your light load comes in at 15, making you borderline incapable of wearing cold-weather clothing while carrying a single weapon, let alone having a pouch of gold, a kit, or anything else. DMs who care less about RAW and more about realism will also question how a medium creature weighting over 100 lbs is able to support their own weight with a strength stat that low.

If you have a 3 in Dex, you better have heavy armor, or that's a -4 to AC, while also being a -4 to initiative, dex saves (common), and stealth, making you a perpetual escort mission for your party. Don't be an escort mission.

If you have a 3 in Con, that's -4 HP per level. If you assume a minimum 1, this means a 20th level wizard averages at 23hp. If you assume you can get 0 hp by rolling 0 or below after modifiers, that wizard only has a 1 in 3 chance of gaining hp at all. Furthermore, if you use the "take 4" method for every level beyond the first, that wizard will never have more than 2 max hp.

If you have 3 Int, you are not sentient; you are classified as of animal intelligence, and unless you're someone's pet, you are a feral creature.

If you have 3 Wis, you have a passive perception of 6, and can be expected to become an active liability every time a control caster shows up. You will be charmed nearly every time. You will be feared nearly every time. If it's save or suck or save or die, one way or the other, you will almost always fail, and force the other players to deal with it.

If you have 3 Cha, you are basically incapable of roleplaying socially. Charisma as is is the least terrible stat to have this low, but it still makes it so your character should be left behind any time the party plans to do anything diplomatic, and begs the question of why anyone would put up with you in a group.




That said, those are the reasons you aren't allowed to drop below 8. Characters in 5e are expected to, as a baseline, be capable of participating in checks in a meaningful fashion. A -1 is thematically interesting, and objectively poor, but it's functionally not much different from a +0. No matter where you put it a -1 (not taking class into account), it isn't going to make you useless. To illustrate, the worst case scenarios are A.) Dex: -1 AC, negated by heavy armor. and B.) Con, for a grand total of -20 HP at lvl 20, which at that point just means it might take 1 less hit to drop you.

Regardless, this was never about letting people fudge the system to min-max their characters. This was about reverse-engineering the system and determining logical values for the unrepresented numbers. Low stats are supposed to suck, and a 3/3/3/12/18/18 should not be considered equivalent to a standard array. Perhaps allowing players to use this table to get 11 more points by turning themselves into cripples would be a terrible idea (for a multitude of reasons, not all mechanical), but that's precisely why it's not an option.

[DeAnno]

I agree with the top half of your table, but would only give points for going down in actual modifier below 8 in a stat. A low stat should be worth less, but it usually isn't a huge deal to have stats of your choice be super low.



Score
Modifier
Point Value


3
-4
-3


4
-3
-2


5
-3
-2


6
-2
-1


7
-2
-1


8
-1
0


9
-1
1


10
0
2


11
0
3


12
+1
4


13
+1
5


14
+2
7


15
+2
9


16
+3
12


17
+3
15


18
+4
19

[Lollerabe]

More or less identical to the table you made, but me and my group use this: http://1d8.blogspot.dk/2014/04/point-buy-calculator-for-d-next-5e.html

And we do allow 16s in creation as well. Hope that helps :)

[Anonymouswizard]

The logic of the previous post seems pretty spot on, but I don't think it wise to count stats below 8 for bigger negatives. Having a dump stat of 8 compared to 3 makes less difference to character than a 2-point difference in your primary or secondary stat. Maybe it's not that extreme but I think you get my logic.

I can agree with the bolded part from experience. I once played in a homebrew system where stats cost more points the higher they got, but scores below 10 gave back points in identical proportions. Now, stats above 13 ate into your downtime (so there was only one in the party), and stats below 7 gave you a minor disability (so there was only one 6 in the party), but all that meant was that pretty much everyone dropped a stat to 7 in order to afford an additional two 12s. Ever thought you'd see someone specify their legs were stronger so they didn't have to roll to walk? To have minmaxed to the point of being unable to harm someone with a knife? Both came up in that game, as well as the investigator with almost disabled Faith (but he was an atheist so that didn't matter). Once you get to 'bad' there's really no disadvantage to going worse, it's not like you'll be succeeding at most of the rolls anyway.

[Cybren]

If you have 3 Int, you are not sentient; you are classified as of animal intelligence, and unless you're someone's pet, you are a feral creature.

If you have 3 Cha, you are basically incapable of roleplaying socially. Charisma as is is the least terrible stat to have this low, but it still makes it so your character should be left behind any time the party plans to do anything diplomatic, and begs the question of why anyone would put up with you in a group.



Two insufferably pedantic points and a regular one:
1) animals ARE sentient in the vast majority of cases. Most of them however are not sapient.
2) int doesn't directly correlate to sapience in 5e. Previous editions had "under 3 int is nonsapient" but that is absent from 5e, and given that it is a possible result in the default rolling method (4d6b3), with no particular rule or blurb pointing anything out about not putting 3s in int, we can deduce that in 5E int does not directly correlate to sapience
3) begging the question is a logical fallacy of invoking your argument as its own proof, not simply the raising of a question.
(Also, charisma i assume is the easiest to rationalize as a still-functioning adventurer with a severe weakness. As a DM if someone wanted to say they had 3 cha because they were covered in pox scars from a disease that left them with a really unpleasant voice then i'd be mostly satisfied with that, even though it's a bit of a kludge)

[Witch of Whimsy]

Alright, I'll bite.

First of all, consider that the majority of responses in this thread seem to be missing the entire point; this is to determine how to compare rolls to point buy, and is in absolutely no way trying to recreate point buy to allow players to utilize the numbers that were intentionally left out.

You are allowed stats below 8 and above 15 while rolling because you have no control over what numbers you get, and this lack of control prevents abuse. It can create more powerful characters, and often does, but the player has to actually roll those stats, rather than cherry-pick them.

You are only allowed to operate between 8 and 15 within point-buy and arrays because you have absolute control over the values. Control invites abuse, so it was made far more tame in order to minimize that abuse.

Consider that rolls are just as likely to roll odd as they are even, and that these odd numbers are effectively wasted unless you invest points in them later. With point buy, if you have odd stats at all, you're probably making sure it's somewhere you already plan to boost through a half feat, or in 2 different scores so you can +1/+1 them at lvl 4. The former is a problem you either have to work around via strategic placement&eventual boosts, or simply let remain despite being odd. The latter is a calculated "problem" that you created for yourself while already having a tailor-made solution, allowing you to get that other 12 or 14 for essentially no cost.

In this way, it becomes a choice between the potential of RNG, and the certainty of a manual approach, where neither is objectively superior. I believe that is how it should be.



Secondly, there are systems within the rules that imply PCs were never meant to have scores below 8.

There is no (minimum 1) clause to rolling HP, nor a (minimum 0), so by RAW, if you have 6 CON and roll a 1 for hp while leveling, you lose 1 HP. Taking this to an extreme, if you have a -4 Con mod on a wizard, and you roll a 1 or 2 when hitting level 2, your Max HP becomes 0 and you die because you leveled.

Similarly, there is no (minimum 1/0) clause for jumping, meaning a -4 Str mod provides -1ft on a running high jump, which, because it goes beyond the written extreme of the system, also means you round down -1/2 to provide that same -1 on standing jumps after rounding (0 would be rounding up), while standing. Does this mean you can't jump? No, not jumping would be 0, and -1 is not 0. RAW says you don't just not jump with a 0, but you break physics whenever you attempt to jump with a -1. Even a -2 STR means you round to a 0ft without a running start, and cannot jump up at all without running. That said, you can still leap forward in either case by breaking physics again to jump without jumping (You could argue that just because you don't jump a full foot doesn't mean you can't jump at all, but RAW says to round down any fraction, and since jumping only works in feet, inches would count as a fraction of feet, and thus be rounded to 0)

I won't delve into Int because it is only implied through various spells and the Monster Manual that 1-3 is animal intelligence, and never outright stated in RAW.



Thirdly, 5e uses a different system from previous editions, in which every point below 8 was just another -1. 5e employs Bonded Accuracy rather than Scaling. Now, things still scale, but at a radically reduced rate (with the exception of health and damage, which the system is balanced around), and floating modifiers have almost been removed entirely. What this ends up doing is making every +-1 more meaningful than before. For the sake of creation, I think it bias to assume negative modifiers are somehow not as bad as positive modifiers are good. Maybe the way the game plays might make it so, or your class might minimize the detriment of that exceptionally poor stat, but should the system care about whether or not you might be able to cope? Should it factor in that maybe this stat means less to you because of your build? Should a system that objectively measures the value of a number regardless of class automatically assume stats below a certain threshold is useless to you?

I for one strongly believe it should not; if it doesn't consider your main stats to be worth more points, it shouldn't consider your off/dump stats to be worth less.

Also, just to reiterate, This is purely for comparative purposes, and is in no way meant to be used as an actual point-buy method. I can't stop you from trying to use it as one, but while going above 15 might be mostly harmless, you should avoid letting players drop below 8 while buying stats; WotC disallows it for a reason.

[Cybren]

You are allowed stats below 8 and above 15 while rolling because you have no control over what numbers you get, and this lack of control prevents abuse. It can create more powerful characters, and often does, but the player has to actually roll those stats, rather than cherry-pick them.

You are only allowed to operate between 8 and 15 within point-buy and arrays because you have absolute control over the values. Control invites abuse, so it was made far more tame in order to minimize that abuse.


The average roll for 4d6b3 is slightly higher than the standard array. You give up power in exchange for using the array or point-buy yes. You have control, yes, but whether or not that control is abusive is purely a matter of how it works. You wouldn't say being offered your choice of between $1 and $5 is not a better deal than being given a 50/50 shot at $20, because it's not. The EV for the first option is $5 (assuming you take the maximum) and the EV for the other one is $10. In all cases the second option is a better choice.



Consider that rolls are just as likely to roll odd as they are even, and that these odd numbers are effectively wasted unless you invest points in them later. With point buy, if you have odd stats at all, you're probably making sure it's somewhere you already plan to boost through a half feat, or in 2 different scores so you can +1/+1 them at lvl 4. The former is a problem you either have to work around via strategic placement&eventual boosts, or simply let remain despite being odd. The latter is a calculated "problem" that you created for yourself while already having a tailor-made solution, allowing you to get that other 12 or 14 for essentially no cost.
How is this a problem? Odd scores are no more or less wasted when rolling as they are with point buy. You're calling the same thing a sin in one case and a virtue in another


Secondly, there are systems within the rules that imply PCs were never meant to have scores below 8.

There is no (minimum 1) clause to rolling HP, nor a (minimum 0), so by RAW, if you have 6 CON and roll a 1 for hp while leveling, you lose 1 HP. Taking this to an extreme, if you have a -4 Con mod on a wizard, and you roll a 1 or 2 when hitting level 2, your Max HP becomes 0 and you die because you leveled.

This is more a hole in the rules. Rolling 4d6b3 is the default option, so some rules that in extreme circumstances not handling it is a problem with those rules, not the ability score range. They may solve that in an updated version of 5E by making point buy or the array the first presented option, but they haven't. Odds are they will return the line from the playtest/old editions that you gain at least 1HP when leveling up.



Thirdly, 5e uses a different system from previous editions, in which every point below 8 was just another -1. 5e employs Bonded Accuracy rather than Scaling. Now, things still scale, but at a radically reduced rate (with the exception of health and damage, which the system is balanced around), and floating modifiers have almost been removed entirely. What this ends up doing is making every +-1 more meaningful than before. For the sake of creation, I think it bias to assume negative modifiers are somehow not as bad as positive modifiers are good. Maybe the way the game plays might make it so, or your class might minimize the detriment of that exceptionally poor stat, but should the system care about whether or not you might be able to cope? Should it factor in that maybe this stat means less to you because of your build? Should a system that objectively measures the value of a number regardless of class automatically assume stats below a certain threshold is useless to you? Certainly confining starting stats to a narrower range helps 5E achieve bounded accuracy by letting the designers have a better prediction of where stats fall, but given that 5E works fine with rolled stats that can range from 3-18 (with most DMs probably not forcing you to use a 3 anyway because they are Weak Warriors), having high starting stats doesn't really break bounded accuracy. Or even bend it

[BurgerBeast]

Alright, I'll bite.

First of all, consider that the majority of responses in this thread seem to be missing the entire point; this is to determine how to compare rolls to point buy, and is in absolutely no way trying to recreate point buy to allow players to utilize the numbers that were intentionally left out.

This point wasn't lost on me. I'm saying a 15 14 13 12 10 4 is not as effective a character as a 17 14 13 12 10 8, even from the point of view of an NPC. It's not about the ability to trade up or down. It's that 4s aren't as debilitating as they might appear to be. The game is designed so that characters can maximize the benefits of their highest scores while minimizing the deficits in their lowest score. Granted, not always, but most of the time. Low strength? Use a finesse weapon. Low dexterity? Wear heavy armour.

I might say to keep the system, but only apply it to an NPC's top four scores, since most characters in the world have fond ways to compensate for their two lowest scores, either through technology (e.g. armor) or behaviour (e.g. playing to your strengths).

[Witch of Whimsy]

This point wasn't lost on me. I'm saying a 15 14 13 12 10 4 is not as effective a character as a 17 14 13 12 10 8, even from the point of view of an NPC. It's not about the ability to trade up or down. It's that 4s aren't as debilitating as they might appear to be. The game is designed so that characters can maximize the benefits of their highest scores while minimizing the deficits in their lowest score. Granted, not always, but most of the time. Low strength? Use a finesse weapon. Low dexterity? Wear heavy armour.

I might say to keep the system, but only apply it to an NPC's top four scores, since most characters in the world have fond ways to compensate for their two lowest scores, either through technology (e.g. armor) or behaviour (e.g. playing to your strengths).

Hmm... I see what you're trying to say, and understand your point. Personally, I still disagree that the system should bias itself into undervaluing lower scores, but you raise a fair point.

To use your example, I agree that a character at 17/14/13/12/10/4 might be more effective than one at 15/14/13/12/10/8 (which I believe is what you meant) at the things they seek to excel in. Objectively, a +3 beats out a +2, and makes you better at something you are likely to utilize very often. It is also true that a -3 in a stat you aren't planning to be making use of much anyway might not be considered that extreme.

However, I prefer to look at a character in terms of strengths and weaknesses. What are they best at, and where can an enemy make them hurt? I like to think a monster, especially an intelligent boss monster, will make sure to find and exploit the kinks in a party's literal and metaphorical armor, since it's the rational thing to do. A -1 to a dump stat makes you vulnerable in situations that require that stat, but you're really only 5% less likely to succeed than a 10, so you shouldn't feel completely powerless when it's tested. That said, each addition -1 is another 5% lower chance to succeed, and makes you less and less able to participate in those challenges.

Another angle is to look at the actual DC. At 1st level, a DC 13 is about what you can expect a PC to have. -1 means you have a 35% chance to succeed. -3 means you have a 25% chance to succeed, and that chance only becomes smaller as you progress in levels. That -3 means you only have a 15% to succeed against DC 15, and require a 20 to succeed against DC 17. Eventually, you hit the point where, unless your DM uses "20 always succeeds" for skill/ability checks and saves, you automatically fail, and with a -3, that happens quite a bit sooner.


Now, of course, it would be perfectly fair to mention that, in regards to saves, not all ability scores are created equal. If you decide to tank the two most popular stats to tank, Intelligence or Charisma, you could quite easily expect to simply get away with it due to the scarcity of things that target those stats. Still, I think it's the DM's job, in a manner of speaking, to make sure those weaknesses come into the light in some fashion; that low Int character might enjoy their time in the limelight for quite a many levels, which is fine, because a player should feel powerful, but later on, some of your enemies should feature spells like Phantasmal Force and Feeblemind, or throw a ghost with Possession at them for Charisma, just to illustrate that there is a consequence for having stats that low.


So I suppose I look at particularly high stats as "What I'm very good at and will be my greatest assets in this adventure" and particularly low stats not "What I don't plan on using anyway" so much as "What I expect will eventually kill me."



Also, I don't think there's anything wrong with keeping a total that might end up being considered "lower than standard point buy" by my chart. That aforementioned 17/14/13/12/10/4 would be considered 25 points, but because someone personally cares more about having that 17 than said person is worried about that 4, it doesn't strike them as at all "bad." It might give them a bit of pause and think "yeah, a -3 is pretty terrible, but honestly I think it's worth it to get a +4 in my primary after racials." Which is fine, and that hypothetical person in my example is free to continue being awesome, using that higher main stat to feel really cool. Still, conversely, the hypothetical in my example should be reminded every now and again that that power came at a cost, and in those moments, that cost wasn't cheap.

In the end though, it's really more about using what you believe to be correct. My chart makes perfect sense to me, and is what I personally use to compare, but it isn't the official chart by any mean. Such a thing does not exist. If mine seems incorrect in some fashion to you(plural, addressing anyone reading this), you might adjust it so it doesn't start scaling downwards at the same rate, and instead totals at -7 at 3(-1,-1,-1,-2,-2). I do not personally believe that properly illustrates how truly terrible it is to have a stat so low, but if it is more in line with your views, it is just as valid/invalid as my own. If it serves your purposes, and you're satisfied with the results, then I'd consider that problem solved.


PS: I apologize for the veritable fortress I've been erecting out of text walls.

[Knaight]

Regardless, this was never about letting people fudge the system to min-max their characters. This was about reverse-engineering the system and determining logical values for the unrepresented numbers. Low stats are supposed to suck, and a 3/3/3/12/18/18 should not be considered equivalent to a standard array. Perhaps allowing players to use this table to get 11 more points by turning themselves into cripples would be a terrible idea (for a multitude of reasons, not all mechanical), but that's precisely why it's not an option.

The system is intentionally designed so that for each class, there are some stats that are much more important than others. This is the reason the price per point goes up after a while, as any stat getting pumped that high is likely to be one of the higher value stats, and as such worth more. Meanwhile, low stats hurt less, and while a 3 is a lot worse than an 8, the 3-8 gap in a bad stat matters less than the 17-12 gap in a good stat. For a rolled character, while the numbers are going to vary, the high numbers are still likely to show up in the good stats, so for a point buy equivalency the reasoning behind the price increase still holds. At the low end though, the per point gap shouldn't increase. It probably shouldn't diminish hugely, as that won't be relevant if it can't actually be used to optimize. An actual 2 point or higher gap for 1 point in an attribute seems like a bit much to me.

As an example test case, we'll take Jane Generic (or Gaon Generic, if we want to go with a more faux medieval spelling). She's an average person who then goes adventuring, and starts with a stat block of 11, 11, 11, 10, 10, 10. She's a 15 point character, based on point buy. Next up, we'll take David the Divergent. David rolled up some pretty extreme rolls, but has the same total, at 16, 16, 16, 3, 3, 3. This is also a 15 point character with the skew introduced (and this is before getting into how big of a jump there should be past 15, there's a case to be made that by the standards of the positive side ). In play though, David the Divergent is going to be a much stronger character mechanically, despite sporting three glaring weaknesses. 18, 15, 14, 3, 3, 3 is probably even stronger, and it's slightly cheaper.

There's also the matter of odd rolls. In high stats they are likely to matter, as high rolls are going to go to the attributes most likely to see increases, and a +1 stat getting a +1 mod down the line will matter. In low stats? Not really. So, lets look at DeAnno's table here, as it looks pretty solid. David the Divergent is a 27 point character on it, while Jane Generic is still as 15 point character. So, let's bump up Jane Generic a bit. 13, 13, 13, 12, 12, 12 is also 27 points. Now Jane Generic is an adventurer, not at all specialized but a cut above the common folk, in this case in every respect. Power wise, that stat block is likely to actually be roughly equivalent; maybe a touch better, but that just brings us into the matter of the jump from 15 to 16. If that is reduced by 1, Jane Generic gets straight 12s while remaining at equivalent points, which compensates.

I'd be inclined to go with DeAnno's table, if I had to only have one table, but with the 16-18 stats costing 11, 13, 16 respectively. Better yet would be to go with more tables. How many stats are important to a character does vary a bit, but it's usually pretty close to 3. So we can take the rolled scores, and divide them up into high 3 and low 3. The high 3 table would have heavy scaling on both ends, reflecting how in either case it's going to hurt. The low 3 scaling would be pretty flat; yes an 18 in a non essential stat is nice, but it doesn't add that much. What adds a lot is how all three of the high three scores have to be at 18 for a low three 18 to exist at all. The forum table software and I are not friends, so here's a "table" for this. You'll notice that this does diverge a bit from the standard table, but it takes a certain point value (28, to be precise) to be able to actually generate a stat block which can make this divergence appear, so it should be fine.

Score, High Three Value, Low Three Value
3, -8, -3
4, -6, -2
5, -4, -2
6, -2, -1
7, -1, -1
8, 0, 0
9, 1, 1
10, 2, 2
11, 3, 3
12, 4, 4
13, 5, 5
14, 7, 6
15, 9, 7
16, 11, 8
17, 13, 9
18, 16, 10


o use your example, I agree that a character at 17/14/13/12/10/4 might be more effective than one at 15/14/13/12/10/8 (which I believe is what you meant) at the things they seek to excel in. Objectively, a +3 beats out a +2, and makes you better at something you are likely to utilize very often. It is also true that a -3 in a stat you aren't planning to be making use of much anyway might not be considered that extreme.

However, I prefer to look at a character in terms of strengths and weaknesses. What are they best at, and where can an enemy make them hurt? I like to think a monster, especially an intelligent boss monster, will make sure to find and exploit the kinks in a party's literal and metaphorical armor, since it's the rational thing to do. A -1 to a dump stat makes you vulnerable in situations that require that stat, but you're really only 5% less likely to succeed than a 10, so you shouldn't feel completely powerless when it's tested. That said, each addition -1 is another 5% lower chance to succeed, and makes you less and less able to participate in those challenges.
The strengths and weaknesses look is fine, and I'd agree that enemies will to some extent intentionally exploit weaknesses when possible. With that said, it's a smaller effect than the increase in what they are actually good at. This makes a lot of sense too - the amount skills actually matter varies a lot for different people, and people tend to put some effort into putting themselves in places where the skills they are better at matter more. To use a real world example, I'm generally comfortable with math, the concepts involved in various sciences, etc. I also am an awful artist. Not coincidentally, I ended up studying for engineering, in which those math and science skills will be relevant, and where my artistic skills are good enough to hit the sort of bare minimum required there, where the absence of anything that actually resembles skill isn't a big deal at all. Meanwhile other people with the reversed skill set can try to end up in something more on the artistic end, where the absence of math and science skills that would be horribly painful for me will go mostly unnoticed. Someone with both is probably not going to get a huge amount of use out of one of them (although they will get some out of both, and there are areas where having both in mass matters), and someone with neither is closed off from a lot of good options, although there's so many other skills that they could easily still be fine.

Now, in the real world this frequently doesn't work perfectly, and there are plenty of cases where people's talents are completely wasted. There are some cases where the core skill set is utterly lacking in something people still have to do with some frequency. 5e PCs though are generally all adventurers in an adventuring class, so they've mostly dodged that bullet. They have successfully put themselves in a position where their good abilities matter more than their bad, and the point chart should reflect this. Hence my high three and low three model, where absolutely sucking at the low three (scores of 3) from average (score of 10), is equivalent to being a meaningful cut above average (14) but not hugely exceptional in the high 3.