On the significance of stats

[Mico Selva]

I have decided to do a little experiment to check the actual impact of stats.

 

This is my character, Sucky. I deliberately did not spend any stat points on character creation.

 

 

I went to the tavern and hired another character, Tagalong, with equally distributed stats:

 

 

Apart from stat distibution (gender, name, portrait and voice set), these characters are identical. So what do these additional stats give to Tagalong, and how much of a difference do they make, compared to levelling up?

 

Sucky at Level 1:

 

 

Tagalong at level 1:

 

 

Sucky at level 5:

 

 

SIGNIFICANCE OF INDIVIDUAL STATS

 

CONSTITUTION

 

Health and stamina bonus gives +9 both at level 1, while levelling up gave sucky +68 H/S (+17  on average), landing him a total of 119. As these values are percentage-scaled, according to my crude calculations, Tagalong would have around 141 H/S at level 5 (+20 per level up, 3 more than Sucky).

 

Note that characters in PoE basically 'start' at third level, which means they more or less double in power at level 4, and then at level 10.

 

So basically you have to put 10 points into CON to get around 3 H/S more per level.

In comparison, in D&D 3.5 you get 3 HP more per level if you up a stat by 6 points, but D&D damage scale is much lower overall.

For example, in D&D a light weapon deals 1-6 damage and a medium weapon deals 1-8. In PoE, a light weapon (hatchet) deals 10-16, while a medium weapon (battle axe) 13-21.

 

At low levels, higher CON will not even give you one level advantage, around level 5 it gives more than one level advantage, and this will only get higher further on.

 

Level 1: Sucky 51 Tagalong: ~60

Level 2: Sucky 68 Tagalong: ~80

Level 3: Sucky 85 Tagalong: ~100

Level 4: Sucky 102 Tagalong: ~121

Level 5: Sucky 119 Tagalong: ~141

Level 6: Sucky 136 Tagalong: ~161

Level 7: Sucky 153 Tagalong: ~182

Level 8: Sucky 170 Tagalong: ~202

Level 9: Sucky 187 Tagalong: ~222

Level 10: Sucky 204 Tagalong: ~243

 

In D&D 3.5 scaling of CON depended on your class. It could as much as double your initial HP (for mages) and since HP per level were partially randomized, it also had a larger impact during character development.

 

DEXTERITY

 

Dexterity to Accuracy is a flat bonus, so Tagalong has +9 ACC, because has has 9 more DEX.

Sucky got +3 ACC per level on average, so this is a flat advantage of three levels.

Accuracy significance is not very intuitive to grasp, but basically it operates on 1-100 scale, If I recall correctly, there is a random roll which like this:

1-10 miss (no damage)

11-50 graze (50% damage)

51-90 hit (100% damage)

91-100 hit (150% damage)

 

Your Accuracy is added to the roll, while enemy's defense (e.g. Deflection) is subtracted from it. In the broadest possible stroke +9 Accuracy is a +9% chance to increase your damage by 50% of your base damage (so +4.5% damage on average)

 

In comparison, +9 Might gives you a flat +18% damage. So yeah, like [uSER=15843]Sensuki[/uSER] said, Dexterity kind of sucks overall. Even though increasing it by 3 gives you a whole one level up advantage, that still does not mean a lot.

 

In D&D 3.5 increasing Dexterity by 2 gave you a +1 to hit (with ranged weapons), which meant at least one whole level up advantage - or more for non-combat classes, but to-hit checks worked very differently there and it is hard to make a direct comparison.

 

INTELLECT

 

Increasing INT by 10 gives you a +50% effect duration and AoE. The usability depends on abilities used. I can easily see the effect duration bonus being useful - extending mind control (Puppet Master ability) from 34.5 seconds to 46.5 seconds has a large impact.

 

AoE range increase is useful in some cases. Many abilities have good enough range with minimal INT, but some will benefit from increased range (like Priest's Restore Stamina spells).

 

Increasing INT is the only way to increase AoE and range of abilities. They do not scale with character level.

 

MIGHT

 

Might increases ALL damage and healing numbers. I can see it as very important for priests and damage dealers and as a possible dump stat for support/crowd control focused characters.

 

Might damage bonus is more visible in higher-damage weapons and abilities, but you want as much bonus as possible for low damage weapons too, because of DT mechanic.

 

Base damage 10 vs DT 5 = 5 damage.

 

With 30% bonus from Might

 

Base damage 13 vs DT 5 = 8 damage (60% increase).

 

Weapon damage comparison:

 

 

 

Ability damage comparison:

 

 

 

As with Intellect, increasing Might is the only way to increase weapon/ability damage (and healing), as these do not scale with level.

 

PERCEPTION

 

Hard to say how useful Interrupt is, but I would boost Perception for fast-attacking characters, who land a lot of hits, because it is almost useless for hard-but-slow-hitting strikers.

 

Interrupt does not scale with level AFAIK, so you only have as much as you get from Perception.

 

RESOLVE

 

I have no idea if Concentration is any useful. It seems logical that it is more useful for characters using abilities with longer casting time (higher-level abilities?) but I do not se how these can be identified outside trial-and-error.

 

Concentration also does not scale with level.

 

IMPACT OF MAIN ATTRIBUTES ON DEFENSE

 

This is a separate topic, because various forms of defense are based on all character's stats, grouped into three pairs.

 

Fortitude (FOR): Represents a character's endurance to "body system attacks" such as poison or disease. It is based on Might and Constitution.

 

Reflex (REF): Represents a character's ability to dodge area of effect attacks. It is based on Dexterity and Perception.

 

Will : Represents a character's strength of mind and resistance to mental attack. It is based on Intellect and Resolve.

 

There is also Deflection (AC equivalent), which is probably the most important of all defensive stats, but attributes do not affect it in any way.

 

The other three are basically D&D saving throws copied to PoE and rescaled (so they use the same scale as Deflection). They all oppose enemy ACCURACY.

 

Starting Defense stats are based on your class (only).

 

Every point you put into an attribute gives you a 1.5 points defence bonus in the defense stat.

 

Additionally comparison, all Defense stats get a +3 bonus on each level up.

 

As you can see in the screenshots, Lv. 1 Tagalong has significantly better Defense atributes than both Lv. 1 Sucky and Lv. 5 Sucky.

 

Sucky gained 4 levels, so his Defense stats were increased by 36 points overall. You get as much Defense from assigning 24 attibute points. Tagalong has 57 attribute points advantage over Sucky, meaning he will only catch up to her LV. 1 Defense around 10th level.

 

2 attribute points in relevant stats are worth one level up when it comes to Defense. This is a lot of value, which means Defense is the most important secondary stat affected by the main attibutes.

 

CONCLUSION

 

Based how often you Defense stats other than Deflection (I will call them saving throws below) are used, there are two conclusions.

 

1. If saving throws are often used, you can only safely dump one of the paired stats, or your corresponding saving throw will suffer. Basically, do not dump both CON and MIG, DEX and PER or INT and RES.

2. If saving throws are seldom used, I can only see MIG and INT as significant stats (with both being dumpable under certain circumstances), with the rest having very little impact on gameplay. Although with the number of points you get, you can easily max out these two and still have a decent number of points to divide among the rest.

 

--------------------------------------------------

 

tl;dr version:

Pump up Might, Intellect and Dexterity (if slow attack rate) or Perception (if fast attack rate), put the remaining points into Constitution and you are all set.

Edited August 23, 2014 by Mico Selva

[Infinitron]

Over at the RPG Codex, MicoSelva wrote about how gameplay went down with a character with minimum values in all attributes:
 
 

It went pretty well. Low statline does not impair your character very much. As I wrote in my crude analysis of main attributes (not sure if they can actually be called 'main' at this point), they mostly impact Defense against abilities and spells.
 
Offense-wise, the heaviest loss was Sucky's mind control (Puppet Master) only working for 34.5 seconds, when it would work for 46.5 seconds with INT maxed out. 35 seconds was enough to win the encounters I engaged in (that ability is a no-brainer first-strike spell for every combat - BALANCE), but I can see these 12 seconds matter in harder fights and/or higher difficulty.
 
Might's impact was rather minor. Weapon type and enemy armor is much more important and additional Might is useful to overcome DT, especially for lower-damage weapons. Sucky still dealt reasonable damage, especially with his abilities (which ignore DT altogether AFAIK).
 
BTW: axes currenty suck ass compared to other weapons. 10% additional damage on criticals (which happen 10% of the time on average) - big ****ing whoop. Other weapon classes get stuff flexible damage type (swords), which allows them to hit through lower DT, staright-up ignoring DT, etc. All of these are more powerful than 1% more damage on average.
 
Constitution's impact is negligible. You get around 2 levels of advantage in H/S at level 10, but having 240 H/S instead of 200 does not mean much. Could still be important in tougher fights/higher difficulty. I certainly did not feel like Sucky's 119 HP was not enough and that he would do much better with more.
 
Dexterity does not matter much, because missing rarely happens. Spellcasters can straight-up ignore it, because spells get an inherent +10 accuracy bonus, so it is as if you had 10 more DEX. For Sucky, fighting was mostly a means to get Focus, so it did not matter that 10% of his hits were turned into grazes and only dealt 50% damage.
 
I have not found any use for Perception. Maybe if combat gets more polished, less chaotic and you actually tell what is going on, you will be able to focus on interrupting enemies using powerful abilities. At the current state - nope.
 
Resolve can actually be useful for a melee fighter who also uses abilities, because you will get interrupted a lot. Someone mentioned a Rogue's escape ability being interrupted which resulted in a dead knocked out rogue. But this is speaking theoretically. Again, it was really hard to tell what exactly was going in combat. Sucky might have been interrupted a few times when using abilities, because they did not fire up. (Or it could have been a bug with game not accepting input correctly.)
 
---
 
As for Defense impact, I have not noticed Sucky suffering any more than his colleagues, even though his Defense stats were around 25 points lower than theirs. He got hit with webs and such as often as everyone, but did he get slowed down more than others or something like that - I could not tell.
I have to note that he defended against physical attacks just as well as everyone, because main attributes have no bearing on Deflection.

Edited August 25, 2014 by Infinitron

[Labadal]

Bump. This is a very interesting read and something the devs should take a look at.

[Mayama]

Its acutally quite hard to discuss attributes right now because the stats of everything are not balanced. Like Josh said in that balance interview.

 

Edit: so for example INT duration impact would be a lot more significant if spells had reasonable base duration.

Edited August 25, 2014 by Mayama

[Sabotin]

From what I can see the defining part of play customization is more or less entirely on class and gear. Attributes provide small bonuses to whatever and talents seem to have the same purpose (which I'm really disappointed about).

 

I don't even think it's so much about the quantitative aspects of the attribute bonuses, but that they don't really seem to have an impact on play style. Looking at this testimony, the character plays exactly the same way, just being a bit better/worse at stuff. Like a high might wizard and a high int wizard, where the int one would focus on debuffs and the might one on damage. Apparently both just pick the inherently better spells?

[Helm]

 

Over at the RPG Codex, MicoSelva wrote about how gameplay went down with a character with minimum values in all attributes:

 

 

It went pretty well. Low statline does not impair your character very much. As I wrote in my crude analysis of main attributes (not sure if they can actually be called 'main' at this point), they mostly impact Defense against abilities and spells.

 

Offense-wise, the heaviest loss was Sucky's mind control (Puppet Master) only working for 34.5 seconds, when it would work for 46.5 seconds with INT maxed out. 35 seconds was enough to win the encounters I engaged in (that ability is a no-brainer first-strike spell for every combat - BALANCE), but I can see these 12 seconds matter in harder fights and/or higher difficulty.

 

Might's impact was rather minor. Weapon type and enemy armor is much more important and additional Might is useful to overcome DT, especially for lower-damage weapons. Sucky still dealt reasonable damage, especially with his abilities (which ignore DT altogether AFAIK).

 

BTW: axes currenty suck ass compared to other weapons. 10% additional damage on criticals (which happen 10% of the time on average) - big ****ing whoop. Other weapon classes get stuff flexible damage type (swords), which allows them to hit through lower DT, staright-up ignoring DT, etc. All of these are more powerful than 1% more damage on average.

 

Constitution's impact is negligible. You get around 2 levels of advantage in H/S at level 10, but having 240 H/S instead of 200 does not mean much. Could still be important in tougher fights/higher difficulty. I certainly did not feel like Sucky's 119 HP was not enough and that he would do much better with more.

 

Dexterity does not matter much, because missing rarely happens. Spellcasters can straight-up ignore it, because spells get an inherent +10 accuracy bonus, so it is as if you had 10 more DEX. For Sucky, fighting was mostly a means to get Focus, so it did not matter that 10% of his hits were turned into grazes and only dealt 50% damage.

 

I have not found any use for Perception. Maybe if combat gets more polished, less chaotic and you actually tell what is going on, you will be able to focus on interrupting enemies using powerful abilities. At the current state - nope.

 

Resolve can actually be useful for a melee fighter who also uses abilities, because you will get interrupted a lot. Someone mentioned a Rogue's escape ability being interrupted which resulted in a dead knocked out rogue. But this is speaking theoretically. Again, it was really hard to tell what exactly was going in combat. Sucky might have been interrupted a few times when using abilities, because they did not fire up. (Or it could have been a bug with game not accepting input correctly.)

 

---

 

As for Defense impact, I have not noticed Sucky suffering any more than his colleagues, even though his Defense stats were around 25 points lower than theirs. He got hit with webs and such as often as everyone, but did he get slowed down more than others or something like that - I could not tell.

I have to note that he defended against physical attacks just as well as everyone, because main attributes have no bearing on Deflection.

 

Pillars of Eternity = Baby's first RPG. Even Dragon Age has more depth.

 

 

This is just so sad, they might as well deprecate these placebo attributes.

[Infinitron]

See that word that starts with "B" in this board's title? The one that isn't "Backer". That means something.

 

Edited August 25, 2014 by Infinitron

[IndiraLightfoot]

From what I can see the defining part of play customization is more or less entirely on class and gear. Attributes provide small bonuses to whatever and talents seem to have the same purpose (which I'm really disappointed about).

 

I don't even think it's so much about the quantitative aspects of the attribute bonuses, but that they don't really seem to have an impact on play style. Looking at this testimony, the character plays exactly the same way, just being a bit better/worse at stuff. Like a high might wizard and a high int wizard, where the int one would focus on debuffs and the might one on damage. Apparently both just pick the inherently better spells?

This is really worrisome. I share your concerns. For some reason, I want a single attribute point to be important. Why place them on a plain linear scale? How about a logarithmic scale or something? 

[Infinitron]

I don't understand how you could think them not having an impact is unrelated to the "quantitative aspect".

Edited August 25, 2014 by Infinitron

[Helm]

See that word that starts with "B" in this board's title? The one that isn't "Backer". That means something.

 

Are you implying that I am not a backer because I don't have a badge? Just because I haven't linked my backer account to my forum account doesn't mean I am not a backer. I have also played Pillars of Eternity Baby's first RPG.

[Infinitron]

Wat

[Mico Selva]

 

See that word that starts with "B" in this board's title? The one that isn't "Backer". That means something.

 

Are you implying that I am not a backer because I don't have a badge? Just because I haven't linked my backer account to my forum account doesn't mean I am not a backer. I have also played Pillars of Eternity Baby's first RPG.

 

I think Infinitron meant that the game is in beta, so it is still subject to change.

[Sabotin]

He means it's a beta .

 

I guess it would be more clear if I say that I think tweaking the amounts the attributes give would not solve the issues. It's my belief they need to change what some of them do or how they do it.

[hairyscotsman2]

 

From what I can see the defining part of play customization is more or less entirely on class and gear. Attributes provide small bonuses to whatever and talents seem to have the same purpose (which I'm really disappointed about).

 

I don't even think it's so much about the quantitative aspects of the attribute bonuses, but that they don't really seem to have an impact on play style. Looking at this testimony, the character plays exactly the same way, just being a bit better/worse at stuff. Like a high might wizard and a high int wizard, where the int one would focus on debuffs and the might one on damage. Apparently both just pick the inherently better spells?

This is really worrisome. I share your concerns. For some reason, I want a single attribute point to be important. Why place them on a plain linear scale? How about a logarithmic scale or something? 

 

Because then you'd need to weight the point buy system.

 

I get the impression they're a little bit more interested in crash bugs than balance just now.

Edited August 25, 2014 by hairyscotsman2

[Matt516]

Good post! Some of your numbers on the benefits of Accuracy and Might are a bit off though. I did the detailed math in this thread (and it was subsequently corrected and improved by the community) if you're interested:

 

http://forums.obsidian.net/topic/67761-dps-vs-accuracy-deflection-heres-the-maths-enjoy/

[Mico Selva]

I have seen that thread and I know some of the math in my posts is pretty crude. Thanks for the attention to detail.

[nikolokolus]

Really thoughtful breakdown and it echoes a lot of what I encountered in my twelve or so hours I've put into the game.

[Mr. Magniloquent]

I'm a bit surprised by all of the attention the attribute system is receiving with regards to whether it is significant or not. It's entirely besides the point.

1. The current design objective of "no bad builds" is implicit that builds will have little meaningful variation.
2. This is first and foremost, a low-level campaign. Attribute bonuses are percentage based. This will cause their influences to be less distinctive.
3. Character classes, abilities, talents, etc. are not yet fully implemented. We do not know how they will compound against attribute bonuses.

Just by those three points alone, we already know that attributes within the current state of Beta will not have much influence. The question is not about whether attributes are meaningful. The question is about whether players should be "allowed" to fail, or at the very least--perform poorly. If the answer is that they shouldn't, then classes should have been designed on rails like Diablo 3. If the answer is that they should, then the magnitudes of difference within build values will become self-evident in pursuit of "allowing" players to fail.

 

Beyond my belief that players should be allowed to fail, I truly do not think the answer to "fixing" the attribute system has much to do with changing magnitudes as per reasons #2 and #3 above. The problem is that some attributes, like Perception and Resolve, have very narrowly defined circumstantial purposes that lack the intrinsic and more broadly applicable benefits of other attributes. Then again, this very well may change. When talents are fully implemented, they may rely significantly on attributes values in ways that we currently cannot know.

[Yonjuro]

..

1. The current design objective of "no bad builds" is implicit that builds will have little meaningful variation.

 

 Not at all. Here's a made up one attribute system for fighters. You get a knob that trades offense with defense.

 

 You can have a dynamic range that makes it essentially do nothing. If the default is good then there are no bad builds, true.

 

 You can also give it a dynamic range where all the way to the left means hits almost everything but gets hit by almost everything (rogue++) or all the way to right means hits almost nothing but gets hit by almost nothing (tank++). Both of those extremes and everything in between can be good builds especially in a party based system. It requires that you play the more extreme characters according to their abilities.

Edited August 25, 2014 by Yonjuro

[Fluffle]

The question is about whether players should be "allowed" to fail, or at the very least--perform poorly.

 

I think the Gamescom videos show that players, in particular Adam, are allowed to fail, very poorly even

[Matt516]

 

 

..

• The current design objective of "no bad builds" is implicit that builds will have little meaningful variation.

Not at all. Here's a made up one attribute system for fighters. You get a knob that trades offense with defense.

 

You can have a dynamic range that makes it essentially do nothing. If the default is good then there are no bad builds, true.

 

You can also give it a dynamic range where all the way to the left means hits almost everything but gets hit by almost everything (rogue++) or all the way to right means hits almost nothing but gets hit by almost nothing (tank++). Both of those extremes and everything in between can be good builds especially in a party based system. It requires that you play the more extreme characters according to their abilities.

Interestingly, the Dark Eye game Blackguards does something similar to this. All the way to offense ended up being almost unilaterally the favored choice though, partly due to the small party size that pretty much required everyone to be capable of doing damage.

[Mr. Magniloquent]

 

 

..

• The current design objective of "no bad builds" is implicit that builds will have little meaningful variation.

Not at all. Here's a made up one attribute system for fighters. You get a knob that trades offense with defense.

 

You can have a dynamic range that makes it essentially do nothing. If the default is good then there are no bad builds, true.

 

You can also give it a dynamic range where all the way to the left means hits almost everything but gets hit by almost everything (rogue++) or all the way to right means hits almost nothing but gets hit by almost nothing (tank++). Both of those extremes and everything in between can be good builds especially in a party based system. It requires that you play the more extreme characters according to their abilities.

Interestingly, the Dark Eye game Blackguards does something similar to this. All the way to offense ended up being almost unilaterally the favored choice though, partly due to the small party size that pretty much required everyone to be capable of doing damage.

 

 

I immediately thought of Blackguards as well with his comment. At least with Blackguards, there is opportunity cost. Even if other circumstance (small party) favors offense, the choice is meaningful. With the current implementation within Beta, attributes only have two values--adequate, and more than adequate (at least within combat). This distinction is mediocre. This mediocrity is only magnified by the absence of talents (which may more heavily rely on and distinguish roles), and that some attributes have narrowly defined and marginal use.

 

The only places where these distinction are felt are where they are such a given that they don't matter. The wizard for example. Their class is designed as primarily focusing on AoE damage and crowd-control effects. More than half of their abilities have a duration, and more than half of those have an area of effect. Being that Intellect increases both AoE and effect duration, there is almost no reason (mechanically) to avoid maximizing Intellect. The choice presented to increase it or not is almost false. This is not satisfying.

 

The class structure plays against the attribute design. They need to increase the significance of each attribute to how each class is deliberately focused, while limiting the attribute distribution in such a way as to where opportunity cost demonstrates meaningful distinctions in play.

Edited August 25, 2014 by Mr. Magniloquent

[Yonjuro]

 

....

Not at all. Here's a made up one attribute system for fighters. You get a knob that trades offense with defense.

...

Interestingly, the Dark Eye game Blackguards does something similar to this. All the way to offense ended up being almost unilaterally the favored choice though, partly due to the small party size that pretty much required everyone to be capable of doing damage.

 

 

 

 Sure, that's an interesting example. In PoE I think both are viable. They would be more extreme versions of rogue and fighter.

[Danathion]

I'm a bit surprised by all of the attention the attribute system is receiving with regards to whether it is significant or not. It's entirely besides the point.

1. The current design objective of "no bad builds" is implicit that builds will have little meaningful variation.
2. This is first and foremost, a low-level campaign. Attribute bonuses are percentage based. This will cause their influences to be less distinctive.
3. Character classes, abilities, talents, etc. are not yet fully implemented. We do not know how they will compound against attribute bonuses.

Just by those three points alone, we already know that attributes within the current state of Beta will not have much influence. The question is not about whether attributes are meaningful. The question is about whether players should be "allowed" to fail, or at the very least--perform poorly. If the answer is that they shouldn't, then classes should have been designed on rails like Diablo 3. If the answer is that they should, then the magnitudes of difference within build values will become self-evident in pursuit of "allowing" players to fail.

 

Beyond my belief that players should be allowed to fail, I truly do not think the answer to "fixing" the attribute system has much to do with changing magnitudes as per reasons #2 and #3 above. The problem is that some attributes, like Perception and Resolve, have very narrowly defined circumstantial purposes that lack the intrinsic and more broadly applicable benefits of other attributes. Then again, this very well may change. When talents are fully implemented, they may rely significantly on attributes values in ways that we currently cannot know.

 

As I understand, a fundamental design goal has been to not allow players to 'fail' - or even to 'perform poorly'. To achieve that, classes 'have been designed on rails'.

 

We cannot think, as we did playing the IE games, that attribute choice is defining our character's success in a given class - the classes have already been built for success. All the player does using PoE's attribute 'modifiers', is tweak the classes efficiency - to improve performance in a desired role. Adding efficiency  in an area, however, does not change the fact that what we started with would be successful. So even if trying to play against our attribute choices the issue is not one of success or failure, but how efficiently we succeed, because success is guaranteed.   

 

Admittedly, it took me a while to get my head around that - I expected to have the same level of control here as I do in BG, but I've come to terms with it.

 

I'm now far more anxious to see how the skill and talent systems evolve, as they are where we will have control over customising our character 'build'. The stats, in PoE, are not important - it seems, to me, they were not designed to be. Attribute requirements for talents then, would be an undesirable contradiction. 

[sophismata]

As I understand, a fundamental design goal has been to not allow players to 'fail' - or even to 'perform poorly'. To achieve that, classes 'have been designed on rails'.

 

We cannot think, as we did playing the IE games, that attribute choice is defining our character's success in a given class - the classes have already been built for success. All the player does using PoE's attribute 'modifiers', is tweak the classes efficiency - to improve performance in a desired role. Adding efficiency  in an area, however, does not change the fact that what we started with would be successful. So even if trying to play against our attribute choices the issue is not one of success or failure, but how efficiently we succeed, because success is guaranteed.

 

There're a couple of notes to make here.

 

Firstly, choice should mean something. If you choose to focus on might and intelligence, it should affect your game in some visible, significant way. Otherwise the choice is pretty meaningless.

 

Secondly, there should be a real choice. The IE games had false choice: your stats were predicated by your class. A fighter never benefited from a high intelligence. Wizards and thieves could not gain any benefit from constitution scores above 16. And so on.

 

In a class based system, your fighting style will generally be determined by your class and class options. Attributes are there to customise that style. Therefore, the attribute system does not have to accommodate mechanical archetypes - melee wizards, holy warriors, etc. That's the role of the class system. But the choices made to attributes should be felt and they should be genuine choices.