The purpose of this post is to rebalance Resolve and Perception, which both have lost important effects (Concentration and Interrupt) since PoE1, with unmodified attributes such as Might and Dexterity in a very simple way.

 

There are many ideas to improve Resolve, which was considered almost useless without Concentration modifier. Options include damage split to Str/Res, kind of shield against new Interrupt, negative duration modifier to afflictions, moving duration from Int to Res, etc. None has been thoroughly tested so far, while some of these seem to be overpowered or ambiguous. The new proposal is based on a different perspective, on the almost equally nerfed Perception.

 

Without Interrupt modifier, Perception affects only Accuracy and the related hit chance in combat. +1 Perception grants 2% additional DPS in a typical case (Accuracy = Deflection) plus maybe some extra Penetration, clearly inferior to Strength and Dexterity, which give 3%. The 2% DPS modifier comes from the 1.25% miss-to-crit conversion and 62.5% expected damage (50% hit and 25% graze), and slighly varies with different Accuracy to Deflection scenarios. This weakness of Perception has been hardly recognized and discussed so far.

 

The simplest solution would be equalizing the DPS effect of Perception/Accuracy with that of Might and Dexterity. Simple math results that Accuracy should be increased by +1.5 per Perception point instead of only +1. Since the combat effect of Resolve is the inverse of Perception, Deflection should be increased also by +1.5 per Resolve point. Res/Per/Dex/Might are mathematically equalized in a such simple way!

 

I can imagine that many players would still not be satisfied with Resolve giving just 50% more Deflection, but this feeling might be related to the fact that defensive playstyle is generally unattractive. The real question to my opinion is whether Resolve/Deflection is balanced to Constitution's +5% Health increment. However, this question has been already there in PoE1...

So new Resolve gets spell damage and +1.5 deflection?

I think an increase to the accuracy/deflection bonus of perception/resolve would be a simple solution to try out during the beta. One thing I'd say is that, because of rounding, non-integer values aren't a good idea. Instead of suggest making them give +2 and residing enemy deflection/accuracy by a proportionately smaller amount.

So new Resolve gets spell damage and +1.5 deflection?

I assume this proposed change is instead of the might>strength change. Fix perception and resolve independently rather than changing Might.

To misquote:

 

For Every Complex Problem, There Is an Answer That Is Clear, Simple, and Wrong.

 

So new Resolve gets spell damage and +1.5 deflection?

I assume this proposed change is instead of the might>strength change. Fix perception and resolve independently rather than changing Might.

 

You are right, damage bonus should be kept at original Might. Resolve would be overpowered with damage there.

I think an increase to the accuracy/deflection bonus of perception/resolve would be a simple solution to try out during the beta. One thing I'd say is that, because of rounding, non-integer values aren't a good idea. Instead of suggest making them give +2 and residing enemy deflection/accuracy by a proportionately smaller amount.

If you don't like non-integer bonuses, let's tone down all other attributes by 33%. So +2% damage per Might, +2% action speed per Dexterity. One can find proportional (possibly rounded) values to Int and Con, as well.

There are two problems with this:

 

First is giving a non-integer value to the bonus from each additional point of Perception. Depending on whether the calculator rounds up or down or rounds at all, you're prioritizing either even or odd values of Perception, and it's best to avoid those kinds of break points in the ability scores.

 

 

Second is that Accuracy is vs. all defenses, while Deflection is just one defense. So mathematically you would have to give Resolve +(all defenses) to even out the accuracy bonus from Perception.

 

What you *could* do is give Perception and Resolve something like outgoing hit to crit bonus / incoming crit to hit bonus. You'd have to crunch the numbers to see exactly what percentages of each you'd have to give each point of Per/Res to even it all out. (1 point each of miss to graze, graze to hit, and hit to crit being roughly equal to one point of accuracy, and conversely, 1 point each of crit to hit, hit to graze, and graze to miss being roughly equivalent to one point in all defenses).

 

Or maybe that doesn't work either because if you give 1 of each conversion you're still dealing with the integer-value problem and if you give only some conversions and not others the math gets really overcomplicated.

 

Resolve would need a much bonus than Perception in order to bring it up to equivalency, since there's no direct defensive counterpart to Accuracy; i.e., if Perception gets +1 accuracy and +1 hit to crit, Resolve would need +1 (crit to hit, hit to graze, graze to miss) AND +1 crit to hit, to mathematically balance out.

 

There's also an effective "ceiling" to how much you can improve Perception; get more than +25 accuracy from any source, you're no longer missing (vs. equivalent defense targets), you're only shifting Grazes to Crits and that's a much lower effective damage bonus

 

 

  On the other hand crossing that accuracy threshold is very useful with crowd control spells! Perception and Dex do also benefit from buffing non-damage powers and abilities and spells, whereas Might doesn't. In Dex's case this is balanced out by the irreducible frames and a few other things, in Per's case it's probably overcompensated for.

Edited December 7, 20178 yr by Dr. Hieronymous Alloy

 

 

 

I think an increase to the accuracy/deflection bonus of perception/resolve would be a simple solution to try out during the beta. One thing I'd say is that, because of rounding, non-integer values aren't a good idea. Instead of suggest making them give +2 and residing enemy deflection/accuracy by a proportionately smaller amount.

If you don't like non-integer bonuses, let's tone down all other attributes by 33%. So +2% damage per Might, +2% action speed per Dexterity. One can find proportional (possibly rounded) values to Int and Con, as well.

 

 

 

That's an option too yeah but I'm  hesitant to downgrade Dex because it worked fine at 3% for the whole first game and I don't want to muck with something that works; so I'd rather bring everything else up to the 3% point if possible.

 

 

edit: hrm

 

What about if you changed critical hits to +50% damage, instead of +25%? Then, either (miss to graze + Graze to hit) OR (hit to crit) would be mathematically, roughly, against equivalent defense, equal to +(one half an accuracy point). So you could give Per an additional +1% hit to Crit, and give Resolve (+1 downshift everything)+(+1 crit to hit) and they'd be equivalent.

Edited December 7, 20178 yr by Dr. Hieronymous Alloy

There are two problems with this:

 

First is giving a non-integer value to the bonus from each additional point of Perception. Depending on whether the calculator rounds up or down or rounds at all, you're prioritizing either even or odd values of Perception, and it's best to avoid those kinds of break points in the ability scores.

 

Second is that Accuracy is vs. all defenses, while Deflection is just one defense. So mathematically you would have to give Resolve +(all defenses) to even out the accuracy bonus from Perception.

 

In order to systematically exploit the rounding issue, you'd have to know and adjust to the Deflection bonus of all your enemies. It's the actual Accuracy-Deflection what matters, not pure Accuracy. So it's impossible to exploit rounding except single occasions. It doesn't matter in the long run.

 

I can see the rationale in the second argument. Accuracy and Deflection bonuses were equal in PoE1 for few years, and it was widely accepted. But I'm not sure how Accuracy bonus was, and should be related to other defense bonuses.

 

 

If you don't like non-integer bonuses, let's tone down all other attributes by 33%. So +2% damage per Might, +2% action speed per Dexterity. One can find proportional (possibly rounded) values to Int and Con, as well.

 

 

 

That's an option too yeah but I'm  hesitant to downgrade Dex because it worked fine at 3% for the whole first game and I don't want to muck with something that works; so I'd rather bring everything else up to the 3% point if possible.

 

edit: hrm

 

What about if you changed critical hits to +50% damage, instead of +25%? Then, either (miss to graze + Graze to hit) OR (hit to crit) would be mathematically, roughly, against equivalent defense, equal to +(one half an accuracy point). So you could give Per an additional +1% hit to Crit, and give Resolve (+1 downshift everything)+(+1 crit to hit) and they'd be equivalent.

 

 

Perception and Resolve have been nerfed compared to PoE1 with the removal of Interrupt and Concentration bonuses. First simple rebalance option is compensation by increasing the remaining bonuses by 50%. Second simple option is to nerf all other attributes by 33%. If you don't like either options, how about boosting everyting, although with different amounts? For example, Resolve gives 2% Deflection, Perception gives 2% accuracy, Might gives 4% damage, Dexterity gives 4% action speed, etc.

Edited December 7, 20178 yr by Raenvan

 

Perception and Resolve have been nerfed compared to PoE1 with the removal of Interrupt and Concentration bonuses. First simple rebalance option is compensation by increasing the remaining bonuses by 50%. Second simple option is to nerf all other attributes by 33%. If you don't like either options, how about boosting everyting, although with different amounts? For example, Resolve gives 2% Deflection, Perception gives 2% accuracy, Might gives 4% damage, Dexterity gives 4% action speed, etc.

 

 

I'm hesitant to support changes any bigger than the minimum necessary to get things in rough balance . . . the more that gets changed, the more secondary effects there are going to be (for example, increase Might's bonus too much, Barbarians start getting way out of hand due to readily available Might bonuses, etc.).

 

One thing that could be done is giving Perception and Resolve non-numeric bonuses. Perception already has a bit of this in that it helps find traps (although that's relatively minor and should probably be combined with the mechanics skill *and* per both contributing to trapfinding). If they figure out a way to make Concentration & Interrupt benefit from Res & Per again, that could be a major help.

Resolve is going to handle all spell damage in the next iteration, that just leaves Perception out in the cold.

 

If you drop Strength and Dex to 2% they pretty much become meaningless. The game seems balanced around the 3% so best to leave that as is.

 

Leaving the deflection bonus to the new Resolve makes little to no sense, while adding it back to Perception has a precedence in earlier versions of the game, makes sense from a game logic standpoint and gives Perception that little something more it needs to be roughly comparable to Strength, Dexterity and Resolve.

Resolve is going to handle all spell damage in the next iteration, that just leaves Perception out in the cold.

 

If you drop Strength and Dex to 2% they pretty much become meaningless. The game seems balanced around the 3% so best to leave that as is.

 

Leaving the deflection bonus to the new Resolve makes little to no sense, while adding it back to Perception has a precedence in earlier versions of the game, makes sense from a game logic standpoint and gives Perception that little something more it needs to be roughly comparable to Strength, Dexterity and Resolve.

 

 

Eh, my preferred solution would be to reverse Str/Res move and go back to Might. I really don't like Str/Res for a lot of reasons, partly that Might was thematic with the setting, partly that the change makes hybrids much less viable.

 

I've also never really liked the "heap extra defensive bonuses onto a stat" method of balancing them. Every stat has some defensive bonus already. Heap deflection and (is it Reflex?) both on Perception and you're just making Perception a kind of flavorless empty stat for tanks (like Resolve was previously). 

 

That said I admit I don't have an answer to this puzzle that I'm fully happy with, either.  What I'd prefer to see is something that gives both Per and Resolve distinct, opposing "roles", like they had previously with interrupt/concentration  (i.e., "bonus critical hits" vs "incoming critical hit reduction" etc).

Edited December 8, 20178 yr by Dr. Hieronymous Alloy

What I'd prefer to see is something that gives both Per and Resolve distinct, opposing "roles", like they had previously with interrupt/concentration  (i.e., "bonus critical hits" vs "incoming critical hit reduction" etc).

 

I don’t think symmetry is required, personally. A different effect for Resolve will give it more identity than anti-Perception.

 

What I'd prefer to see is something that gives both Per and Resolve distinct, opposing "roles", like they had previously with interrupt/concentration (i.e., "bonus critical hits" vs "incoming critical hit reduction" etc).

I don’t think symmetry is required, personally. A different effect for Resolve will give it more identity than anti-Perception.

Yeah, that's fair, I'm just trying to make my suggestions minimal so they have as few unintended effects as possible, and if pet and res balance each other out, roughly, it should help prevent unintended secondary balance problems. Or that was my thought pattern anyway.

Let me summarize what has been already found. The change should be as minimal as possible, so Might and Dex bonuses remain +3% per attribute point. Perception and Resolve bonuses could be boosted various alternative ways, but preferably symmetrically (as they have been after early PoE1 versions):

1. Restore original Interrupt and Concentration mechanisms and modifiers.
2. Increase Accuracy and Deflection bonuses to +1.5% per point. Rounding is probably not a real issue in long term. Secondary defenses (Will, Reflex and Fortitude) might be scaled, although they are already larger than the necessary 1.5% to my opinion.
3. Adjust other combat mechanisms that amplify the gains from unmodified +1% Accuracy and Deflection increments. Like Graze interval and Crit bonus. This idea involves more variables, and could be more difficult to balance than option 2.
4. Invent new unique and/or separate values to Perception and Resolve. It will be tough to balance them with other attributes, and to get widespread approval (see the case with Might to Str/Res split). I've read many suggestions for Res, fewer for Per.

good summary.

 

Just to make things more complicated now that you've already typed the nice summary:

 

1) working out the math on a critical hits / miss proposal. 

 

What would it do to the math if Perception got an additional +3%  hit to crit bonus?

 

As before, the standard equation when Accuracy and [Defense] are equivalent is 

 

0 + (25/2) == 62.5% expected damage on a normal attack vs. equivalent defense, adjusted for accuracy / miss / graze rate

 

With the additional +1 Accuracy that a point of Per currently gives, that shifts up to 

 

50+ (25/2) + (1*1.25) == 63.75% with an additional point of accuracy (as one miss shifts up to a graze, graze shifts up to a hit, hit shifts up to a crit). 

 

That means that the additional point of Per/Accuracy has shifted your expected damage value per swing up by 1.25, or 2%.

 

If you add +3% hit-to-crit (so a point of perception is giving +1 Accuracy and also +3% hit-to-crit conversion, that becomes

 

47 + (25/2) + (4*1.25) == 47 + 12.5 + 5 = 64.5 % expected value per swing, so a difference of 2% flat, which is .032 or 3.2% of the previous expected value. close to the values you'd expect from Might or Dex but a little higher. (Once you have accuracy bonuses over +25 though that advantage will drop off). 

 

 

As to Resolve:

 

If we gave Resolve Graze-to-Miss, that would be a reduction of . . 

 

Again, if we figure each incoming attack is made by equivalent perception .  .. 

 

Normal:

 

50 + (25/2) == 62.5% expected damage on a normal attack vs. equivalent defense, adjusted for accuracy / miss / graze rate

 

If we adjust that by, say, 4% Graze-to-Miss, that becomes

 

50 + ( 21/2) == 61.5 = 2 point drop == .032 or a 3.2% drop in expected incoming damage per attack

 

This is because Graze-to-Miss is a more powerful effect (counterintuitively) than hit-to-crit is, because a graze reduces from 50% damage to zero, while a crit increases from 100% to 125% damage. 

 

point being, giving perception +1 accuracy and +3% hit to crit is (roughly) equivalent to giving Resolve +4% graze to miss. 

 

So that's my recommendation: add +3% hit to crit for each point of Per, and +4% graze to miss for each point of Resolve.

 

EDIT: that's all assuming that a "%" of hit to crit or graze to hit or whatever is a percentage off the attack roll, not a percentage of the crits, etc. 

 

 

2) One downside to current Perception

 

one problem with current perception: it now has the same problem mechanics had in the last game -- You have to have one party member with high perception, and more than one party member with high perception is useless.

 

Suggestions: allow traps to be found with either/or perception and mechanics, allow pooling of Per for trapfinding, allow pooling of Per and Mechanics for trapfinding, etc.

Edited December 10, 20178 yr by Dr. Hieronymous Alloy

Hey you bunch of elevated gamers,

 

First, I am loving these breakdowns. I always like seeing 'smarter' people than myself dive into our systems. I wanted to ask you if you have taken into account that Perception helps all attacks, while Might only helps damage attacks. It feels correct that Might should boost DPS more than Perception just because Perception also effects CC spells that do not do damage. Where as Might does nothing to help CC attacks. You seem to have a better grasp on the number than myself. I only ask because I didn't see you point to it in any of your breakdowns.

 

I got your backs

-SKing

Hey you bunch of elevated gamers,

 

First, I am loving these breakdowns. I always like seeing 'smarter' people than myself dive into our systems. I wanted to ask you if you have taken into account that Perception helps all attacks, while Might only helps damage attacks. It feels correct that Might should boost DPS more than Perception just because Perception also effects CC spells that do not do damage. Where as Might does nothing to help CC attacks. You seem to have a better grasp on the number than myself. I only ask because I didn't see you point to it in any of your breakdowns.

 

I got your backs

-SKing

 

 

Good point!

 

Yeah, one problem with "breaking out the math" is that it doesn't cover things that aren't math -- like, we can break the math out for Might vs. Per, but that will only show the damage numbers, not the non-damage stuff.

 

Right now / next patch, Might / Resolve / Strength will give the biggest bonus -- three percent  per point -- to all damage, multiplicatively. When you curve this out, each additional point of Might gives a slightly smaller relative damage boost (because 3% of base is a smaller fraction of the total with each additonal point), but roughly speaking it's a straight 3 % damage boost. Importantly, now that Might is multiplicative, though, the absolute damage boost from Might remains constant -- each additional point of Might gives you 3% additional of your base pre-might damage total.

 

Dexterity and Perception, though, both boost all attacks / spells / actions, including crowd control, etc. Dexterity reduces action speed by 3%, which is roughly equivalent to Might's 3% bonus, except . .  well, it gets really complicated, partly because you can only reduce action speed so much, and there is a minimum number of frames any action can take (and Dex doesn't reduce those five or six frames). So in practice, Dexterity gives somewhere between a 2.7% and 1.7% bonus per point, while also comparatively boosting CC etc. (see chart here: https://forums.obsidian.net/topic/86684-mechanics-the-big-attack-speed-conundrum/?p=1808225)

 

Perception, like Dex, boosts everything -- but because of the way the attack roll is calculated and the statistical effect of an additional point of accuracy, gives at most a 2% or so damage bonus; also, it drops off faster than Might or Dex -- once you get above +25 accuracy, you're only adding .8% damage with each additional point of per / Accuracy (because you're only converting grazes to crits instead of converting misses to crits, and because crits only give a 25% damage boost, while going from a miss to a graze gives an additional 50% damage).

 

So yeah, you're absolutely correct, the real comparison isn't Per with Might, it's Per with Dex, because those are the two stats that boost everything, not just damage. Dex gives a slightly smaller damage boost than Might, but gives a damage boost ranging from about 2.7 % to about 1.7 % per additional point .. while Per gives between 2.0% and 0.8% over similar ranges. That doesn't seem much, but it adds up, so that someone who stacks Per at character creation ends up doing roughly 10% or so less damage overall than someone who stacks Dex.

 

If I were *really* smart, I'd figure out how to generate a graph that showed the relative bonus-per-additional-point of Might, Dex, Per, and Str/Res. 

 

 

 

[  I'm *not* that smart though: I'm pretty sure my recommended percentage adjustments above are wrong, because I calculated "graze to hit" and "hit to crit" percentages off the overall attack roll, not as percentages of the grazes on the attack roll or percentages of the hits on the attack roll. Correcting for that, for my rec's above to work, you'd need something like 6% graze-to-miss per point of resolve, and 6% hit-to-crit per point of Per, and those would quickly get silly I fear -- at 20 Per, you'd have sixty percent hit to crit! which is just getting silly. So I retract my above recs -- I've got no idea how to solve this problem! :sheepish: ]

 

Thanks for reading all our armchair rambling! 

Edited December 9, 20178 yr by Dr. Hieronymous Alloy

Hey you bunch of elevated gamers,

 

First, I am loving these breakdowns. I always like seeing 'smarter' people than myself dive into our systems. I wanted to ask you if you have taken into account that Perception helps all attacks, while Might only helps damage attacks. It feels correct that Might should boost DPS more than Perception just because Perception also effects CC spells that do not do damage. Where as Might does nothing to help CC attacks. You seem to have a better grasp on the number than myself. I only ask because I didn't see you point to it in any of your breakdowns.

 

I got your backs

-SKing

 

Mmmm, but Perception doesn't help heal spells, and old Might boosts heal spells. so both affect attacks, Perception affect CC, Might affect heals, I think its fair to give them same number, if Perception's gain to damage is 2%, Might should be 2% too. (One also affect CC, one also affect healing, fair enough)

 

Dex, however, is the stats that almost affect everything tho, not like Perception and Might. So it should have a little less gain than Might and Perception imo.

 

The new Strength can be more powerful because it only affect phys damage and no heal bonus tho.

Edited December 9, 20178 yr by dunehunter

I really hope they find a solution that doesn't involve fractional values of Accuracy or whatever. I think the goal of making each point of an attribute matter is a good one, and if one/two of them suddenly have bonuses that only really go up every 2 points, that will be a big step back.

 

I don't actually think they would do it though, I'm pretty sure making each point matter was a stated goal of the system back in the day.

I really hope they find a solution that doesn't involve fractional values of Accuracy or whatever. I think the goal of making each point of an attribute matter is a good one, and if one/two of them suddenly have bonuses that only really go up every 2 points, that will be a big step back.

 

I don't actually think they would do it though, I'm pretty sure making each point matter was a stated goal of the system back in the day.

 

Fractional values, say +1.5 Accuracy, do matter. Imagine an encounter where your Accuracy equals your opponent's Deflection. A roll of 50 means you just grazed. Had you +1.5 Accuracy, it would be a hit similar to the effect of +1 or +2. In another fight, your opponent's Deflection is one higher, and you again grazed with a roll of 50. Had you +1.5 Accuracy, you would still graze (51 is the lower limit of Hits) similarly to +1, but different to +2. In the long run, +1.5 is equivalent with +1 in half of the fights, and with +2 in the other half.

 

Having attributes integer modifiers is only for beauty, not for game mechanics.

 

Edit: The above explanation is incomplete, it only shows why +1.5 is not equivalent to +2. To prove the other part, that it can differ also from +1, one assumes that enemies can have integer+half Deflection. Like -1 lower base, and +1.5 due to Resolve. In an encounter where your opponent's Deflection is higher by 0.5 than your Accuracy, you graze with a roll of 50. You would still graze with +1 Accuracy, but hit with +1.5 or with +2.

Edited December 9, 20178 yr by Raenvan

 

I really hope they find a solution that doesn't involve fractional values of Accuracy or whatever. I think the goal of making each point of an attribute matter is a good one, and if one/two of them suddenly have bonuses that only really go up every 2 points, that will be a big step back.

 

I don't actually think they would do it though, I'm pretty sure making each point matter was a stated goal of the system back in the day.

 

Fractional values, say +1.5 Accuracy, do matter. Imagine an encounter where your Accuracy equals your opponent's Deflection. A roll of 50 means you just grazed. Had you +1.5 Accuracy, it would be a hit similar to the effect of +1 or +2. In another fight, your opponent's Deflection is one higher, and you again grazed with a roll of 50. Had you +1.5 Accuracy, you would still graze (51 is the lower limit of Hits) similarly to +1, but different to +2. In the long run, +1.5 is equivalent with +1 in half of the fights, and with +2 in the other half.

 

Having attributes integer modifiers is only for beauty, not for game mechanics.

 

Edit: The above explanation is incomplete, it only shows why +1.5 is not equivalent to +2. To prove the other part, that it can differ also from +1, one assumes that enemies can have integer+half Deflection. Like -1 lower base, and +1.5 due to Resolve. In an encounter where your opponent's Deflection is higher by 0.5 than your Accuracy, you graze with a roll of 50. You would still graze with +1 Accuracy, but hit with +1.5 or with +2.

 

 

Yeah I suppose if enemies could have fractional Deflection from Resolve that would work. I know values don't have to be integers, but it does tend to feel cleaner and probably less scary to new players.

 

Generally if a system requires 0.5 values I'd just double the numbers involved, but I know number inflation has a lot of issues of its own (and in this case it wouldn't really work for Dex since Recovery has a lower bound it can go to... zero), and I also know that a lot of people absolutely hate having large numbers in CRPGs.

Edited December 10, 20178 yr by Answermancer

One thing that could help somewhat:

 

If we returned to a 50% bonus for critical hits, as it was in the first game, rather than a 25% bonus, that would increase the per-point value of perception up to 2.4% rather than 2.0% flat. Which doesn't solve the whole problem but does bring things in closer parity -- it's only 0.3% below the value of a point of Dex at that point. 

 

It might even actually completely solve the issue depending on how the falloff values for each change as you add additional points of per/dex.

 

 

In a way, Per has been nerfed twice: once by the removal of interrupt, once by the shift to a 25% critical bonus.

 

So I think that's my recommendation for this problem: return to a 50% bonus for critical hits, rather than the current 25%. 

Edited December 10, 20178 yr by Dr. Hieronymous Alloy

One thing that could help somewhat:

 

If we returned to a 50% bonus for critical hits, as it was in the first game, rather than a 25% bonus, that would increase the per-point value of perception up to 2.4% rather than 2.0% flat. Which doesn't solve the whole problem but does bring things in closer parity -- it's only 0.3% below the value of a point of Dex at that point. 

 

It might even actually completely solve the issue depending on how the falloff values for each change as you add additional points of per/dex.

 

 

In a way, Per has been nerfed twice: once by the removal of interrupt, once by the shift to a 25% critical bonus.

 

So I think that's my recommendation for this problem: return to a 50% bonus for critical hits, rather than the current 25%. 

Exactly! And I'd consider adding an inherent Interrupt to critical hits, as well. If I'm right that's equivalent with 1 of the old 3 points from PoE1, so not overpowered. 50% critical bonus + auto-interrupt + penetration bonus (whose expected extra damage hasn't been calculated to my knowledge) should be attractive enough to raise Perception. Or to raise Resolve to avoid received crits.