What with current posts on whether PTG is balanced and the 'Gunkulator'...

[Lemon Knuckles]

I was riffing on this...

7 hours ago, gjnoronh said:

If I had to guess Jervis looked at this history after 30 years of being involved in rules design and thought - ****** it no one is ever happy with the balance when we include points - let players decide what's right for their tabletop.   I think that "open play is the best way" model depends on people being grown ups in how they approach the game, and underestimates the various difficulties we all had when AoS launched in setting up tournaments, pick up games etc.    But I think it's where his heart is.    

[gjnoronh]

Yeah but as Joe pointed out neither I or Jervis as saying go with half assed points.  They went with NO points on launch of AoS which we'd all agree was a mistake.

The points system we have is minimally granular.  It's really almost entirely 20 point increments with a few 10 point increments.   You could divide by 20 or so and get the actual increments rather then the window dressed because people are used to thinking about 2000 point armies increments (liberators are worth 5 points an 80 point unit is worth 4 and the rare 70 is worth 3.5).   Unit sizes are usually 1-5-10 per cost increment with a few exceptions so again we have less granularity then we could.  

The choice to go with a minimally granular system indicates they are intentionally not trying to be super specific about the math. Compare this to 7th/8th ed where you had 0.5 point increments for some units/upgrades in a 2000-2500 point typical army size.  That was very very granular - and still not very accurate.  

[Lemon Knuckles]

The current level of granularity is fine imo (although I don't see the harm in allowing themselves to break down to increments of 10 a little more often than they currently do).

I'm interested to know what you mean exactly when you talk about "accuracy."  I'm not super clear on what you are saying there.

[gjnoronh]

Accuracy = how well do the points actually represent table top value.    My assertion is always getting that right is very very hard (or arguably impossible) as the game has too many variables beyond what's on the warscrolls that determine table top value.  

[Fulkes]

1 hour ago, gjnoronh said:

Accuracy = how well do the points actually represent table top value.    My assertion is always getting that right is very very hard (or arguably impossible) as the game has too many variables beyond what's on the warscrolls that determine table top value.  

Table top value can be all over the place. They could have started with models worth 1 point each and our games could be based on a 200 points limit, the dead horse I keep beating is that no matter what you decide to do to determine how much something is "worth" the basic metrics used to value one model should be the same ones used for every model or else we just end up with wildly varying points values caused by shifting design metrics. Either the game needs to use the same guidelines on every model or it might as well not be using points at all and we could be determining game size by wound count or something else instead.

[Lemon Knuckles]

3 hours ago, gjnoronh said:

Accuracy = how well do the points actually represent table top value.    My assertion is always getting that right is very very hard (or arguably impossible) as the game has too many variables beyond what's on the warscrolls that determine table top value.  

I agree with you are explicitly saying, but I somewhat disagree with what I think your implied assertion is.

Explaining why (and doing it well) is perhaps complicated.  It shouldn't have to be complicated, but there is a surprising dearth of theory regarding aos.  Theory is not the same thing as theory-crafting, of which there is some.   Theory-crafting is a mathematical exercise, but theory is a conceptual one. 

Points are a mathematical problem.  Value is a theoretical problem.  I agree with you that it is impossible to solve the problem of value through points.  But I would probably disagree with someone who said that it would be impossible to solve given a definite game-state.

This leads me to a few (maybe provocative) points:

1. It is possible to calculate value within a definite game-state 
2. It is therefore possible to calculate a discrete series of values across a discrete series of definite game states
3. It is therefore possible to construct a generalized theory of value such that the problem presented by any specific game state is solvable
4. Some such theories will be more accurate than others; the validity of any such theory can be tested in game through the application of derivative strategies, and judged based on results

I started messing around with gunk originally because I was focused, not on the idea of "accuracy," but rather on the idea of "internal consistency."  But in doing the actual work, and interacting with people around these forums about the work, I am starting to pay more attention to the question of accuracy.

The original points system based on Wounds was absolutely consistent, but it definitely wasn't accurate in the sense you mean.  Our current points system is definitely more accurate than Wounds, but much less consistent.  I am starting to think hard about why it can't be both.

To talk about that kind of thing well requires a theoretical framework so that concepts hang together in a consistent and inter-related way.  It requires some agreed-upon levels of abstraction.  A common set of concepts and key principles.  Or else we just talk past each other, or in circles.  

I'm working through such a theory now, and will post somewhere in the forum or in the blogs, not because I think it is "right", but because I think it is fertile and it can start the kind of conversation I'd really like to have with many people here.

 

 

[gjnoronh]

22 hours ago, Fulkes said:

Table top value can be all over the place. They could have started with models worth 1 point each and our games could be based on a 200 points limit, the dead horse I keep beating is that no matter what you decide to do to determine how much something is "worth" the basic metrics used to value one model should be the same ones used for every model or else we just end up with wildly varying points values caused by shifting design metrics. Either the game needs to use the same guidelines on every model or it might as well not be using points at all and we could be determining game size by wound count or something else instead.

That's my point - the basic metrics to define value aren't consistent because as Lemon Knuckles points out in the following post - value would depend on a definitive game state.

Definitive game state = for example what scenarios are used, what the terrain looks like, what the likely opponents are, what the skill level of the players is.  See also my previous posts in this thread.  How do you value two models with the same stat line, one flies the other doesn't? How do you value two models with the same stat line one  with a command ability that has a very specific situtation in which it's very  useful (but that scenario takes a great deal of skill to set up) one with a more generally useful but less powerful command ability.  If you don't know how much vertical terrain is on the board or how skilled the player using it will be it's hard to say how valuable that model is.  There are plenty of tables  with almost no vertical terrain some with lots.

 

@Lemon Knuckles I get your point on definitive game state but you would have to be very careful not to generalize your math to the non definitive game state, or specifically to a game state that differs from your carefully defined game state.  Easiest way to do it is - no terrain, no scenarios, models just smash in the middle of the board starting at say 24 inches apart so no skill involved.  That's not how AoS is played but would give you a defined game state to derive math from.  

[Lemon Knuckles]

2 hours ago, gjnoronh said:

That's my point - the basic metrics to define value aren't consistent because as Lemon Knuckles points out in the following post - value would depend on a definitive game state.

Definitive game state = what scenarios are used, what the terrain looks like, what the likely opponents are, what the skill level of the players is.   How do you value two models with the same stat line, one flies the other doesn't? How do you value two models with the same stat line one  with a command ability that has a very specific situtation in which it's very  useful (but that scenario takes a great deal of skill to set up) one with a more generally useful but less powerful command ability.  If you don't know how much vertical terrain is on the board or how skilled the player using it will be it's hard to say how valuable that model is.  There are plenty of tables   with almost no vertical terrain some with lots.

 

@Lemon Knuckles I get your point on definitive game state but you would have to be very careful not to generalize your math to the non definitive game state, or specifically to a game state that differs from your carefully defined game state.  Easiest way to do it is - no terrain, no scenarios, models just smash in the middle of the board starting at say 24 inches apart so no skill involved.  That's not how AoS is played but would give you a defined game state to derive math from.  

@gjnoronh, let me just start by making it clear that I value the conversation, as I do all of the various conversations that have occurred here and elsewhere around this topic.  It is all useful.

That said, I feel like are reaching a point where we are just talking past each other and recycling the same things over and over.

I could add a lot to your list of what to include in a definite game state:  material state of the board, composition of my army, composition of my opponent's army, current round, current score, realm rules in effect, etc., etc.  As I have said before, I agree with you.  I'll make it clear again now:  it is impossible to compute a value ahead of time that will hold constant across all possible iterations of definite game states.  But where we don't seem to agree is regarding the implication of this fact.  I think you are suggesting that this is some sort of show-stopper.  To me it's both obvious and irrelevant.

The problem you are describing is an irreducible problem in almost everything around us.  I've talked briefly about economics earlier in this thread, but we could just as easily talk about something as arcane as wave function collapse, or as common as figuring out what you want to do when you get out of bed in the morning.  Or... literally any game ever.

To me it comes across as tautological.  Imagine for a minute that you could compute a single constant (theoretical) value that would always maintain 100% fidelity across a near-infinite number of definite game-states.  What would that mean?  To me, it would mean that there is no game there at all.  All there is is an automated script.   

EDIT:  I forgot to mention, but do think it's important to call out.  I disagree that opponent skill qualifies as a variable for a definite game state.  I think that is a category mistake.

[Fulkes]

2 hours ago, gjnoronh said:

That's my point - the basic metrics to define value aren't consistent because as Lemon Knuckles points out in the following post - value would depend on a definitive game state.

Definitive game state = for example what scenarios are used, what the terrain looks like, what the likely opponents are, what the skill level of the players is.  See also my previous posts in this thread.  How do you value two models with the same stat line, one flies the other doesn't? How do you value two models with the same stat line one  with a command ability that has a very specific situtation in which it's very  useful (but that scenario takes a great deal of skill to set up) one with a more generally useful but less powerful command ability.  If you don't know how much vertical terrain is on the board or how skilled the player using it will be it's hard to say how valuable that model is.  There are plenty of tables  with almost no vertical terrain some with lots.

 

@Lemon Knuckles I get your point on definitive game state but you would have to be very careful not to generalize your math to the non definitive game state, or specifically to a game state that differs from your carefully defined game state.  Easiest way to do it is - no terrain, no scenarios, models just smash in the middle of the board starting at say 24 inches apart so no skill involved.  That's not how AoS is played but would give you a defined game state to derive math from.  

I feel like you're conflating points values with effective value. One is a means of determining the cost of a model or unit that should be based on math, the other is how much a model or unit is worth at any point in the game. The second determines things like target priority, the first determines how much of something you can put into an army list.

The fact that a game state can effect how much "value" a unit has on the table should not dissuade the designers of a game from coming up with a metric to assign those models a points value that is applied equally across the entire game. Ideally the points cost should take into account all of a unit's abilities, and in the case of large models how much board space they take up, and only adjust it up or down if fine-tuning is found to be needed from there. 

Basically, you can't dissuade me from beating my drum over and over that in a game that involves numbers and math that math should be a core part of the design instead of rapidly shifting design metrics that make the way the game is designed drastically shift that leads to the newer stuff being better and cheaper than it would have been if it'd been created earlier in the edition. Not applying a standard metric is what has traditionally caused problems like codex creep (or whatever the WHFB version was called), and has lead to some newer Stormcast being the same cost as older unit choices despite being far better.

A well designed game can't leave these sorts of things up to personal feelings and shifting doctrines if it's going to have an semblance of being fair to the players playing it.

 

[gjnoronh]

I think we're talking past each other possibly but I hear you both.  

I am not arguing that the game shouldn't be well designed!

I am arguing that in a chaotic data set it's essentially impossible to get highly precise math to guide you in design.   You can get  rough math at best but you can't get better then that. 

@Fulkes I don't understand why we wouldn't want points value to reflect in game value.   I also see I can dissuade you.

 

Here's the background. I've been playing Warhammer in various incarnations for 28 years.  I've participated in competitive composition rules/ points value/unit value discussion for essentially that entire time.  In the 1990's I helped run one of the early WFB tactics websites at the 'dawn' of the warhammer internet.   I've watched a lot of math hammer discussion in almost 30 years  and suggestions that with tighter math we can do better design.     There are at least two different highly detailed and highly iteratively desiged add on math systems I've seen WPS and Swedish comp that were additions to the listed GW point values that tried to better capture unit value (because we all knew the GW points weren't quite right.) I've played in large scale tournaments run with those systems and debated them (and other approaches to balance the game) on the internet over the years.   These highly detailed math systems have  got their pros and reflect a huge investment of work on their designers- but they never reached a high degree of accuracy IMO.   

Here's an old swedish comp rules set. The details may vary in the AOS world but take a look at the careful conditional thought processes they used to help reflect incremental value in different situations (i.e. one thirster is X value on the table top 3 blood thirsters may not be 3 X value) This team put a lot of work into this system to think through those types of  issues.   I never liked the system but I respect the work involved - it had a world wide following in 8th ed.  

http://buckeyebattles.com/wp-content/uploads/2015/06/Swedish-Comp-System-v16.0.pdf

AoS IMO is much more complex in terms of  unit value and game play then WFB was.   As a result the math to get it 'right' is much more complex then it ever was in the past.   At the same time the point granularity on unit pricing is  intentionally far less precise then it ever was before.    I'm coming from the perspective I've seen it (intense effort to develop the perfect math of game value) before and seen the pitfalls before.    I am simply based on that history arguing do not   overestimate our ability to 'get the math right.'  I do think "getting the design right" is absolutely important but high precision math  to guide design may functionally not be possible given that complexity.  

[Keldaur]

Points are something we accept as fair when in reality, even when backed up by a mathematical model, they are terribly inaccurate. 

[Keldaur]

On 11/28/2018 at 5:05 PM, Fulkes said:

I feel like you're conflating points values with effective value. One is a means of determining the cost of a model or unit that should be based on math, the other is how much a model or unit is worth at any point in the game. The second determines things like target priority, the first determines how much of something you can put into an army list.

The fact that a game state can effect how much "value" a unit has on the table should not dissuade the designers of a game from coming up with a metric to assign those models a points value that is applied equally across the entire game. Ideally the points cost should take into account all of a unit's abilities, and in the case of large models how much board space they take up, and only adjust it up or down if fine-tuning is found to be needed from there. 

Basically, you can't dissuade me from beating my drum over and over that in a game that involves numbers and math that math should be a core part of the design instead of rapidly shifting design metrics that make the way the game is designed drastically shift that leads to the newer stuff being better and cheaper than it would have been if it'd been created earlier in the edition. Not applying a standard metric is what has traditionally caused problems like codex creep (or whatever the WHFB version was called), and has lead to some newer Stormcast being the same cost as older unit choices despite being far better.

A well designed game can't leave these sorts of things up to personal feelings and shifting doctrines if it's going to have an semblance of being fair to the players playing it.

 

He can correct me if I am wrong. But he is not saying that you should leave it up in the air. He is stating that point values are not precise enough and never will be. 

Just as a little example, how do you assign an exact point cost to a bloodsecrator?  

Its effective value relies on:

- Number of models it provides a +1 attack buff. It's not the same to buff 10 Bloodreavers than 80 Bloodreavers.

- Quality of the attacks those units have. It's not the same to buff 30 bloodletters than 30 bloodreavers.

- Possibility to fizzle spells. 

- Ignore battle shock tests.

- Battallion that increases its range.

You can't. Its point cost is only an approximation or a mean to certain scenarios, it does not represent its effective value in the tabletop, and therefore it's not accurate.

The variance of the possible scenarios alone does not allow you to solve its point cost with absolute precision. And there is no mathematical model with a fixed point cost that can fix that.

[Fulkes]

Irreducible complexity is a bad argument against trying to have a standard metric based in math instead of feeling. This is a game that can be largely predicted using probability. Sure there are statistical outliers, but for every game state that exists we can predict with a strong degree of certainty the probably outcome, and thus we should be using a metric that can account for those probable outcomes. Hence my argument for math over whatever they're currently using (which based on some claims I've read seem to drastically differ from one design team to the next responsible for each book).

The only argument about complexity that I'd agree with is the determination of weighting the values used in order to get a final outcome. Once everyone is on the same page though by using the same guidelines then the points changes from unit to unit can be based on changing of the weighting of those values (say shooting is becoming too strong in the game so you weight it more heavily and apply the change evenly across the entire game instead of the rather haphazard approach we've seen from time to time were two nearly identical units end up being balanced differently).

Points will never be perfect, but that should never stop us from chasing perfection in the hopes of having the best game we can. Honestly I don't get this arguement AGAINST math or standardized balance being used. A more balanced game means better games for everyone. And no, it doesn't mean we all have the same things, it just means the approach used to determined how much any given thing should cost us is approached the same way for every model in the game giving us a more level game state where 2k of one army is balanced against 2k of another army and we're not finding that it's more like 2.5k of that other army in ability.

[Keldaur]

 

13 minutes ago, Fulkes said:

 

Points will never be perfect, but that should never stop us from chasing perfection in the hopes of having the best game we can. Honestly I don't get this arguement AGAINST math or standardized balance being used. A more balanced game means better games for everyone. And no, it doesn't mean we all have the same things, it just means the approach used to determined how much any given thing should cost us is approached the same way for every model in the game giving us a more level game state where 2k of one army is balanced against 2k of another army and we're not finding that it's more like 2.5k of that other army in ability.

3

The first part I bolded is what I am stating. Points will never be perfect. And I agree with @gjnoronh in that point values have gotten less precise (how many units have weapons that are obviously better than the other options in every single scenario, yet they are priced at the same cost). It also doesn't help that we are talking point costs in increments of 10s, rather than 1s. That's precision lost.

The second part is something I had never stated, and from reading his post, neither did he. 

But I will give it another try, do you really think you can even get a reasonably precise (i will drop the word "exact" since we are on the same page) value for a unit like the blood secrator no matter what else do you bring with him? If so, how?

[Fulkes]

So you want me to write an entire system of how we could points thing and then present how much a single model could be priced just to prove a point?

Because while it's possible you're being pretty daft to demand that. Forget current points in that situation because with a system like that everything in the game would have to be re-pointed. A single model in a vacuum compared to what we have now is meaningless in any context of this discussion.

[Kramer]

1 minute ago, Fulkes said:

So you want me to write an entire system of how we could points thing and then present how much a single model could be priced just to prove a point?

Because while it's possible you're being pretty daft to demand that. Forget current points in that situation because with a system like that everything in the game would have to be re-pointed. A single model in a vacuum compared to what we have now is meaningless in any context of this discussion.

The question is not the points you would give him. (You’re right, vacuum). But how. What metrics would you take, what to compare. What influences what. 

So method not outcome. 

[Fulkes]

2 minutes ago, Kramer said:

The question is not the points you would give him. (You’re right, vacuum). But how. What metrics would you take, what to compare. What influences what. 

So method not outcome. 

I've posted something that gave a rough outline of a method which was largely based on statline metrics (+1 point per inch of movement, +1 point per attack with a range of 1" (+1 per additional inch of range) ect, ect). Abilities and battalions would likely start from there (+1 point per inch of range (double it if the unit doesn't have to be "fully within", ect) but that'd take a fair amount of time to sit down and design and then test things over and over again to determine if the points costs need to be adjusted for how much something costs. Once a standard metric is created though then it's just adjusting the weight of individual things and applying the adjustments across the entire game.

With that I'd argue that spells and command traits should cost points as well or an extra charge leveled on top of the model based on the strongest ability they can take for each. Because let's be honest, most people would be taking that strongest thing anyways.

[Keldaur]

 

6 minutes ago, Fulkes said:

So you want me to write an entire system of how we could points thing and then present how much a single model could be priced just to prove a point?

Because while it's possible you're being pretty daft to demand that. Forget current points in that situation because with a system like that everything in the game would have to be re-pointed. A single model in a vacuum compared to what we have now is meaningless in any context of this discussion.

1

No, I want you to state how would you create a system that gives a reasonable precise point cost to units that have a very wide range of effective value (since their effective value depends on the army roster) such as the bloodsecrator.  

9 minutes ago, Kramer said:

The question is not the points you would give him. (You’re right, vacuum). But how. What metrics would you take, what to compare. What influences what. 

So method not outcome. 

Correct. Thank you.

[Keldaur]

15 minutes ago, Fulkes said:

I've posted something that gave a rough outline of a method which was largely based on statline metrics (+1 point per inch of movement, +1 point per attack with a range of 1" (+1 per additional inch of range) ect, ect). Abilities and battalions would likely start from there (+1 point per inch of range (double it if the unit doesn't have to be "fully within", ect) but that'd take a fair amount of time to sit down and design and then test things over and over again to determine if the points costs need to be adjusted for how much something costs. Once a standard metric is created though then it's just adjusting the weight of individual things and applying the adjustments across the entire game.

With that I'd argue that spells and command traits should cost points as well or an extra charge leveled on top of the model based on the strongest ability they can take for each. Because let's be honest, most people would be taking that strongest thing anyways.

5

So finally, do we agree that that system wouldn't give an accurate point cost for "value multipliers" such as the blood secrator then?

The problem is that you have to change how the bloodsecrator and units like him work, because their value is always reliant on external factors (the units it buffs). To be able to include them in your system, you need to strip away every single ability that interacts with other units (yours or the opponent) in such ways. I could had used any other unit, like for example the frost phoenix (his ability of -1, is it the same against a 4+ hit roll, or a 3+ hit roll ? How do we define its efficiency? A mean of every single unit? Oh, and if we get into account rerolls? ).

[Fulkes]

Do I agree it can't be "accurate"? Not really. I think you're relying too much on the belief that the points values can't be based on the ability and not on the things it could effect. You charge based on how it can effect things, not on what it's targets are. You do that for everything and suddenly the entire game approachs pointing things the same way and Bob's your uncle.

[Keldaur]

54 minutes ago, Fulkes said:

Do I agree it can't be "accurate"? Not really. I think you're relying too much on the belief that the points values can't be based on the ability and not on the things it could effect. You charge based on how it can effect things, not on what it's targets are. You do that for everything and suddenly the entire game approachs pointing things the same way and Bob's your uncle.

3

I asked you why and how it is possible to get an accurate value for those units. The example you gave doesn't fix the issue I talked about (not even remotely touched it) when it comes down to pricing. I am relying on facts, the fact is that a unit whose value relies on external factors which get exponentially better or worse in different ways cannot be measured accurately. It's like trying to measure how much does a heater increase the temperature in a room where there are several cooler systems that can be changed at any given time and provide different amounts of heat dispersal.

It's impossible to have any kind of accuracy when it comes down to pricing a unit such as the bloodsecrator, so I will do the work for you.

First, you strip the point cost for the ability from the bloodsecrator. (That doesn't make sense, but you want accuracy, right? the bloodsecrator won't be the one benefiting from its buff).

What you do is increase the other unit cost based on how they benefit from such buffs. What you do is assume such units are being given the buff based on the likeness for such units to benefit from it. You will have to use some trigonometry (the khorne memes about math aren't a coincidence) to calculate the maximum number of units benefiting from the buff, and then you will have to gauge the for how long they will benefit from it (congolining can only get you thus far, tabletop scenery, likeness of being shot off the table), you will have to use probabilities (which aren't accurate at a given time, but overall). And you have to do this individually with every single unit of the book because every single unit benefits from the bloodsecrator buff asymmetrically and at different lengths (ie: some units can't congo-line properly).  And afterwards you will have to take into account the existence of a battalion that increase the likeness of those units to benefit from it. And then you will have to take into account the possibility of more buffs stacking up (+1s to hit, +1 to wound, rerolls, extra attacks, etc...) that compound themselves with the bloodsecrator buff.

And this wouldn't even be remotely accurate, because even if you got the numbers right, first it shoehorns your army into using that battalion always to be more optimal, it also forces you to use the bloodsecrator, and it also forces you to take buffers. Luckily you at least could take any given unit and they accurately cost. Or not, because now we will need to calculate them against each other unit of the game if you want to properly cost them, or you leave assymetrical balance between units (which goes in detriment of accuracy of the point system) 

That the game developers should try to get more accurate with the point system is a given, but not always desirable, otherwise we could just use our models for chess and call it a day, not perfectly balanced either tho. The argument you made about "taking the higher cost because people will use that one anyways¨ should highlight when something is inaccurate, or at least, I don't agree with it.

Or maybe our definition of what is accurate is very different. And what I would call arbitrary eye gouging to you seems good enough. That comes down to my original post, that point costs in these games are there mostly to provide us with a system that we believe on its fairness. A belief.

Also, this isn't how this kind of discussions should go. I shouldn't have to be the one proving to you that it is not possible, but you should be the one proving its possible. I am sure you have read Russel's teapot analogy.

[Lemon Knuckles]

6 hours ago, gjnoronh said:

I think we're talking past each other possibly but I hear you both.  

I'm glad you are.  Sometimes it's hard to tell.  Perhaps because this is a rather polyphonic conversation and it's easy to lose track of who said what and what we are responding to.

6 hours ago, gjnoronh said:

I am arguing that in a chaotic data set it's essentially impossible to get highly precise math to guide you in design.   You can get  rough math at best but you can't get better then that. 

Going back and re-reading the conversation so far, I'm certain that we agree much more than we disagree, but we are spending a lot of time disagreeing on things that are really just a confusion of language.  Part of that is us being sloppy with the words we are using; and part of that is because, as I've mentioned before, the theory that would normally help us be more precise with our words is under-developed in this game.

I think three of the words that are causing the most problems here are:  value, balance and accuracy.  Let me try to reframe:

1. Value has been used in both the context of "points value" and "situational value."  The same word is pointing to two very different things, and that is creating confusion.  It would be better if we instead used "points efficiency" and "value."
2. Both Balance and Accuracy are combining together in really confusing ways as well.  I was struggling a bit with this because as I said in the earlier post, I agree with what you are saying but find it irrelevant.  By sharing the Swiss Comp PDF you might have actually helped me to understand why we seem to be talking past each other.

Points, to me, are all about resource allocation.  It is a strategic element that forces the player to make decisions about how they want to invest their resources.  Different units can be seen as differently bundled resource allocation curves.   As long as the points system is capable of aggregating stats into conceptual resources in a meaningful way, and prices those resources consistently, it is, to my mind, balanced.  The question of whether or not each individual unit cost is accurate (in your definition of the word) to any of its infinite situational values is irrelevant, because the overall economy is generally accurate and internally consistent. 

Again, I'll go back to the simple comparison:

1. There was a time when AoS was comped based only on Unit Wounds.  This system was not good because even though it was very precise and internally consistent, the aggregation of resource was woefully lacking.  Wounds is not a good proxy for the different kinds of things that can give a unit value.  If we played a game comped with wounds capped at 14 wounds, and you brought a 14 wound monster with movement 6, and I brought a 14 wound monster with movement 14 that could fly, I have essentially received an unbalanced amount of resource at no cost.  In effect, my 14 wounds are more points efficient that yours.
2. The current iteration of AoS points is much better in terms of aggregating resource.  There is a base method there that I feel is pretty good.  But it is not always internally consistent, and so we have a (admittedly smaller) version of the same problem, but for a different reason.

Neither 1 nor 2 has anything to do with accuracy really, at least not in the way you've been using it, and not in the way that Swiss Comp seems to want to use it.  There goal statements are revealing:

"Goals The main goal of the system is to evaluate the strength of an army. There are however also a few secondary goals: - Promote creativity in list writing - Award tactical play by: 1. Minimize rock/paper/scissorslist types 2. Direct the meta game away from play styles that is perceived (by many) as boring (such as very defencive castle play, deathstars, avoidance-point denial, a few enormous units etc)"

Notice that they want the point system to evaluate the strength of the entire army, not as a means of establishing a balanced economy.  This is completely different to what I'm talking about, and it is confirmed when you read the secondary goals.  The Swiss Comp also aimed to minimize strategy in favor of promoting tactics, and eliminate min-max type list-building.

I don;t know if you have any experience with MTG, but to me, the Swiss Comp idea is like trying to create a system where the mana cost of any individual card in my deck would change based on what other cards I put in my deck.   In effect, it would want to punish me for being a good deck-builder, or take pity on me if I just threw an unsophisticated pile of 60 together.  That's not balance.  That's some sort of competitive communism.

[Lemon Knuckles]

1 hour ago, Kramer said:

The question is not the points you would give him. (You’re right, vacuum). But how. What metrics would you take, what to compare. What influences what. 

So method not outcome. 

Wow, lots happened while I was responding to gjnoronh.  Guess I have some reading to catch up on.

But this caught my eye and I can't resist butting in.

This question (method) is exactly what gunk is trying to explore.  Working on gunk has led me to start thinking about theory of warhammer in general, which has sparked a bunch of s--- that is starting to really rattle around this old noggin and is now dying to come out.  I'm anxious to share with all of you.  I imagine reactions will be mixed, but hey, looking forward to some interesting conversation nonetheless.

[Fulkes]

I find it interesting that it all keeps circling back to me needing to provide proof of something being possible meaning that I would need to sit down and spend all my unpaid free time just to win an argument on the internet with people who refuse to believe the points I've raised.

Basically it's not worth it and I'm tired of arguing that there is nothing in this game that can't be quantified numerically, it's just a misguided belief that these things need to be made so complex that it can't be expressed in such a fashion. Even when I've presented how I'd do it all I see is the same argument paraded back out: "prove it". I point out that it's daft to demand someone to actually do a lot of work for free just to prove the possibility of something working as presented and I'm told that I just need to explain the "how". I explain how I'd approach the "how" and then I get told to prove it all over again.

[Lemon Knuckles]

38 minutes ago, Fulkes said:

I find it interesting that it all keeps circling back to me needing to provide proof of something being possible meaning that I would need to sit down and spend all my unpaid free time just to win an argument on the internet with people who refuse to believe the points I've raised.

You are being asked to provide proof for something that is, I think, irrelevant to the point you are trying to make--that's why you are finding it frustrating.  I'll go back to MTG again for analogy because maybe talking about another game makes it less volatile.  Wrath of God can have immense value if I am sitting across from an Aggro-deck.  It can have no value at all if I am sitting across from a creatureless Combo deck.  That variability in the value of a card like Wrath of God is irreducibly true, but that fact really has nothing to do with how the game should cost cards like Wrath of God.

I also don't think this is an argument (at least I hope not... I don't take it that way).  Not everything written on the internet is motivated by malice or feelings of self-entitlement.  

Lastly, you are right that you would probably need to sit down and do some work to figure it out.  I've spent a lot of time working with gunk, for example.  Because of that time spent, I can say things now that I wouldn't have been able to say before.  I'm definitely not saying I'm right, or that gunk is right, or anything like that.  But I am saying something that I said very early on in this post in response to the titular question...  the point of gunk, right now, is really to do gunk, and see where it leads.