Forum adverts like this one are shown to any user who is not logged in. Join us by filling out a tiny 3 field form and you will get your own, free, dakka user account which gives a good range of benefits to you: No adverts like this in the forums anymore. Times and dates in your local timezone. Full tracking of what you have read so you can skip to your first unread post, easily see what has changed since you last logged in, and easily see what is new at a glance. Email notifications for threads you want to watch closely. Being a part of the oldest wargaming community on the net. If you are already a member then feel free to login now.

Was thinking about balance and thinking one of the ongoing problems is POINTS seem to be random and meaningless. The impact of points costs is HUGE.

So when seeking balance I think points would be the best starting place. By that we need to assign a cost to stat lines. Start with a clean slate as well. No "prior" point values. Just pick a point for the median.

Why median? Points adjustment are not a "point of stat" basis but on a "statistical impact" basis. So the cost of a 2+ vs a 3+ to hit or as a save, should be reflected based on it's basic statistical improvement over the previous step. So a 4+ to hit = median an increase of points based on the Statistical increase of going from 4+ to 3+ is applied. Going from a 4+ to a 5+ again you get a cost "reduction" based on the statistical shift. A model cost is the sum total of it's statistically adjusted statline costs added up.

Weapons costs should also adjust to the "effectiveness" of the weapon. Not just some pat singular value for a Heavy Bolter or Lascannon. Again.. a Bolter in the hands of a 4+ should cost less than the same weapon in the hands of a 3+ and it should cost even more in the hands of a 2+ to hit.

The idea here is that, excepting "special ability abuse and bad writing" the armies will be statistically balanced, but not rock-paper-scissors balanced, nor "Hard Balanced". Since the game involves chances basing things on statistical average shifts makes for a kind of "soft balance" where you have a fair shake at any given game or really, any given combination there of because the points costs are shifted per model per upgrade based on the statistical effectiveness so every unit becomes a statistical wash.. except dice are never statistical so the game relies on the skill of the player and the luck of the dice rather than arbitrary points costs that make some units insanely cheap for their effectiveness leading to a lack of variety in armies being fielded because there is only ONE effective build rather than something where everything costs in balance to possible effectiveness so in theory.. any build should at least have a chance.

That is very rationale. Hence GW will ignore.

 XeonDragon wrote:
That is very rationale. Hence GW will ignore.

Then to Hell with GW's points. We can publish them and have one of the Tournaments pick it up and from there GW's points values will diminish.

Imagine being able to bring just about any combination and while not being a "sure thing win" at least have a solid statistical crack at it. Mind you it's only the beginning. From there other adjustments are needed. Like assigning points to "abilities" and to "global or army wide" Specials as they are extremely unbalancing. Having to "pay" for them in some way allows armies without UBER army wide automatic specials to be balanced fairly against those with.

Imagine having to pay for Space Marine global abilities. Or really any army's global. Like Disgustingly Resilient or Resurrection. But starting with the stat based costs is like the tip of the iceberg but it's a good place to plant the flag and work downward.

It will never be "perfect" balance or "hard" balance but that's not really a bad thing. Perfect balance is bad and can make things dry and boring. The idea would be to keep flavor and uniqueness, have that be "effective" on the board, and yet balanced.

Space Marines should play like SMs. Sisters like sisters. Orks like Orks. All armies should have the needed abilities, specials, options, wargear, etc to give that proper flavor and yet be "as fair as possible" on the board. But I'd have to bow to someone with more "mathammer" skills to work out an algorithm that allows compiling the effect of statistical shifts and I know our community has said people because they are doing just that all the time. They've got it down pretty pat about the statistical impact of +1BS vs re-rolling to hit vs re-rolling 1s and the statistical variance between each.

Honestly, the more I think about it, the more monumental a task it seems and one of exponential complexity scaling. But there should be a way to break it down into workable segments and then work from that segment to the next, etc.

If you were creating a game from scratch the idea would work and have merit. It would take a lot of careful testing to ensure that the stat you value as y per point always functions that way even at points x and z on the graph, but it's a start. 40k can't do that. It's far more an emotional investment at this point than a logical one.

 Canadian 5th wrote:
If you were creating a game from scratch the idea would work and have merit. It would take a lot of careful testing to ensure that the stat you value as y per point always functions that way even at points x and z on the graph, but it's a start. 40k can't do that. It's far more an emotional investment at this point than a logical one.

That's kind of the point. You are talking about hard balance were the statistics of any given thing is universally balanced. Which is impossible. New game or not.

The goal is soft balance through rational points values that reflect that units actual statistical performance. For that you have to set base values and work your way outwards trying to balance at each point along the way by resolving things on a less granular level the higher you go.

As I know the mathammer guys already know the specific performance of any given stat and how better or worse each value is by percentage.
But yeah it will be a lot of work but not as much as it might seem, well, except for the ridiculous model range breadth at this point in time.

But you got to start somewhere.

So let's start with BS, a 4+is with how many points. A wound is worth how many points. And step by step develop the larger structure. If you do it right got can plug in about any stat line and get a points value for a model with those stats. So once you get the right algorithm you can auto generate values as things go along.

It won't be perfect. But it will be better than the, profit motive points value and drunken and utterly smashed on hard liquor points system GW employs currently.

I know a guy with a background in data analytics, writing algorithms. A couple of years back we spent a couple of weeks trying to figure out if GW used an algorithm for point values, by playing around with the factors you suggested.

TBH, most of what he said I didn't understand, but the TL;DR version was "I think they have an algorithm, but it is either not very good, or they treat the results as guidance only and vary the points to sell more models".

I love the idea by the way, I think it would be an interesting project. If it were to start, I think starting with basic troops (e.g. Imperial Guard) as the basis for "costing" BS4+ etc. rather than more "elite" troops might make sense. The reason I say that is the more "elite" a unit, the more the points are (INMHO) likely to reflect a more complex "mix" of factors. So to my mind, take the simplest units first to figure out base costs of key skills, then work up for more complex attributes from there.

 meatybtz wrote:
So when seeking balance I think points would be the best starting place. By that we need to assign a cost to stat lines. Start with a clean slate as well. No "prior" point values. Just pick a point for the median.

Why median? Points adjustment are not a "point of stat" basis but on a "statistical impact" basis. So the cost of a 2+ vs a 3+ to hit or as a save, should be reflected based on it's basic statistical improvement over the previous step. So a 4+ to hit = median an increase of points based on the Statistical increase of going from 4+ to 3+ is applied. Going from a 4+ to a 5+ again you get a cost "reduction" based on the statistical shift. A model cost is the sum total of it's statistically adjusted statline costs added up.

Weapons costs should also adjust to the "effectiveness" of the weapon. Not just some pat singular value for a Heavy Bolter or Lascannon. Again.. a Bolter in the hands of a 4+ should cost less than the same weapon in the hands of a 3+ and it should cost even more in the hands of a 2+ to hit.

The idea here is that, excepting "special ability abuse and bad writing" the armies will be statistically balanced, but not rock-paper-scissors balanced, nor "Hard Balanced". Since the game involves chances basing things on statistical average shifts makes for a kind of "soft balance" where you have a fair shake at any given game or really, any given combination there of because the points costs are shifted per model per upgrade based on the statistical effectiveness so every unit becomes a statistical wash.. except dice are never statistical so the game relies on the skill of the player and the luck of the dice rather than arbitrary points costs that make some units insanely cheap for their effectiveness leading to a lack of variety in armies being fielded because there is only ONE effective build rather than something where everything costs in balance to possible effectiveness so in theory.. any build should at least have a chance.

You cannot calculate the statistical impact of a stat change, changing the Damage characteristic of heavy bolters from Damage 1 to Damage 2 increases damage against 1/2/3-Wound models 0%/100%/50%. WS and Attacks are useless if you never get into melee, BS is useless if you don't have any ranged weapons.

I think the best way to do balance is to determine balance goals, find a unit that you want to anchor points in, you never change the base cost of your anchor unit's cost. When units are a lot better than your anchor they get nerfed, a lot worse they get buffed. For each individual unit you set up a range of scenarios you want that unit to be more or less pts-efficient. You could say you want Intercessors to be pts-effective against Tau Empire Strike Teams in a scenario where they don't move, Fire Warriors don't get to Rapid-fire and both units are in cover because Intercessors will have a massive 2+ Sv while Fire Warriors only get a 4+ and half their maximum number of shots. In a scenario where both units get to RF and neither is in cover it makes sense that Intercessors should be inefficient because they are not abusing their 3+ Sv by going into cover for a 2+ Sv, double-tap at long-range ability and their AP is not as meaningful against 4+ Sv as it is against 3+ Sv. You set up these scenarios and find a points range that makes the unit effective and ineffective in the scenarios as you desire. This ensures that every unit has a role in the game of countering some number of units in some scenarios and ensures that it in return gets countered in certain scenarios against certain units.

But all of that is meaningless without playtesting. Starting from scratch ensures you will have a bazillion issues in your pts and you will most likely have 0 playtesters and therefore no way to improve. Starting with one of GW's sets of points (CA19 or CA21) makes a lot more sense because you can use tournament data and professional opinion to shape your opinions on what needs changing. Still, it will fail to be perfect unless you get a dedicated playtest team that want to test out the majority of points costs. Then it'll take a couple of hundred hours to a couple of thousand hours to lay down the statistical groundwork for a balanced edition and you still need a dozen playtesters playing a dozen games each to factor in how missions alter the value of different units, much more if your groundwork was poor. I think you are underestimating the task of balancing every 40k unit.

I stopped half-way through my last project trying to start with CA19 and making common-sense changes to account for over and under-powered units and the rules changes of 9th because I like the 9th edition format of pts better and because I thought it would be more useful to have a set of points that would be usable against an official points cost. I'll probably end up spending 30 hours on my CA21,5 without doing any groundwork and without playtesting anything. My version of CA19 altered for 9th required both players to make their list specifically for that format and nobody was going to care or help me test it even if I finished it https://www.dakkadakka.com/dakkaforum/posts/list/790475.page. Just copy-pasting the costs for every unit in the game takes 20-ish hours and you have to account for style changes you might want to end up making that can add tens of hours to a project, like when I rewrote every datasheet in my Necron fandex. Twice.

Everybody is also dead-set on helping GW make the official game the best it can be, ITC has said that they're not going to make custom missions anymore, fat chance that they'll accept a fan-edit of the Munitorum Field Manual. If you want to spend your time effectively forget about all the statistical bull, talk with your playgroup about units you want to use but think are crap and have them agree that you don't have to pay full price for them and then ask them what units you use they think are OP and agree to a pts increase for them. Even if the only thing you can agree on is increasing the price of one supremely overpowered model that can help out the balance in your local community a lot. Hitting the Wraithknight with a 100-pt nerf bat was the easiest and best house rule my group ever made back in 7th.

Pointing units in a wargame is more of an art than a science - there's just too many factors which are all highly variable.

I don't entirely disagree with anyone here except that it's impossible or that it is all art.

I agree the task is impossible for a human to complete. Too many iterations are required.

However. You can both run the statistics and compile results as well as play test using AI.
In fact someone recently taught an AI to play kill team and it beat him.. once that AI learned to play.

So fully play-testing is not too much of an issue at least you can iterate rapidly through a trained AI.

Which means time to get my Google AI code back up and running and dive into methodology to generate lots of statistical data quickly, iterative changes, and compile results.

Oh yeah, what a time to get back into AI. My local processing power should be enough, if not I can always beg some time on giant AMD super computer banks.

Oh and for art vs science, the ART comes in the form of weighting the individual characteristics. For example you would "weight" BS to be worth more than WS. Weapons with more shots are weighted more than weapons with fewer dice to roll per turn. LD is almost a dump stat. Brackets can be more important than actual wound values.

 meatybtz wrote:
...In fact someone recently taught an AI to play kill team and it beat him.. once that AI learned to play.

So fully play-testing is not too much of an issue at least you can iterate rapidly through a trained AI....

That's not statistics, that's machine learning. You need a breaker and a builder to balance the game, the builder assigns points values and the breaker finds the most broken stuff, it doesn't matter whether that's two AI's or a group of game designers and a group of playtesters. If you have access to a dozen talented software engineers they could solve the problem, but GW's software engineers cannot even compete with a single Russian fan in terms of making easily navigatable rules, I doubt they could write AI that could solve any meta and make pts that generate a meta that is hard to solve. Art is a lot of things, I think comparing assigning points to painting a Monet or playing the piano makes sense, but comparing it to abstract art or banging pots and pans together does not make sense. I guess you could say GW is making abstract balance with their Munitorum Field Manuals but I think they're just playing the piano badly.

I definitely agree that AI has potential for balancing games.
That there would definitely be an art in balancing the numbers you feed it.
If you get it wrong; garbage in, garbage out.

 meatybtz wrote:
I don't entirely disagree with anyone here except that it's impossible or that it is all art.

I agree the task is impossible for a human to complete. Too many iterations are required.

However. You can both run the statistics and compile results as well as play test using AI.
In fact someone recently taught an AI to play kill team and it beat him.. once that AI learned to play.

So fully play-testing is not too much of an issue at least you can iterate rapidly through a trained AI.

Which means time to get my Google AI code back up and running and dive into methodology to generate lots of statistical data quickly, iterative changes, and compile results.

Oh yeah, what a time to get back into AI. My local processing power should be enough, if not I can always beg some time on giant AMD super computer banks.

Oh and for art vs science, the ART comes in the form of weighting the individual characteristics. For example you would "weight" BS to be worth more than WS. Weapons with more shots are weighted more than weapons with fewer dice to roll per turn. LD is almost a dump stat. Brackets can be more important than actual wound values.

How would your AI account for balance in a world where we can assume that for any game played each player may only have access to a random, to you, selection of models and nothing else. You'd need to solve for random selections of unsynergistic junk playing against hyperoptomized lists and that's where this falls apart. If you balance a unit around synergies it will be weak without them if you balance without them it will be too strong. Unless you want to assign each unit a value that changes as you add other units to your list your balancing task literally can't work.

 Canadian 5th wrote:

 meatybtz wrote:
I don't entirely disagree with anyone here except that it's impossible or that it is all art.

I agree the task is impossible for a human to complete. Too many iterations are required.

However. You can both run the statistics and compile results as well as play test using AI.
In fact someone recently taught an AI to play kill team and it beat him.. once that AI learned to play.

So fully play-testing is not too much of an issue at least you can iterate rapidly through a trained AI.

Which means time to get my Google AI code back up and running and dive into methodology to generate lots of statistical data quickly, iterative changes, and compile results.

Oh yeah, what a time to get back into AI. My local processing power should be enough, if not I can always beg some time on giant AMD super computer banks.

Oh and for art vs science, the ART comes in the form of weighting the individual characteristics. For example you would "weight" BS to be worth more than WS. Weapons with more shots are weighted more than weapons with fewer dice to roll per turn. LD is almost a dump stat. Brackets can be more important than actual wound values.

How would your AI account for balance in a world where we can assume that for any game played each player may only have access to a random, to you, selection of models and nothing else. You'd need to solve for random selections of unsynergistic junk playing against hyperoptomized lists and that's where this falls apart. If you balance a unit around synergies it will be weak without them if you balance without them it will be too strong. Unless you want to assign each unit a value that changes as you add other units to your list your balancing task literally can't work.

Not really, remember we are a D6 system which vastly simplifies things. As for "synergy". Part of the current synergy is an artifact of bad points balancing where you have JUNK UNITS. The idea here is that since the "stastical" effectiveness of a unit based on D6 dice statistical changes per point you can approach a level of balance where things are not JUNK.

In the real world you will get deviation from that, which is to be expected and even encouraged. There MUST be something more to it than flat balance. That is the "art" others have talked about. The statistical averages are a good place to start. But beyond GIGO you also have to account for "A BAD FOUNDATION LEADS TO A BUILDING FALLING DOWN". That is, if the points are crap, you have crap units, if you have no foundation then any change you make causes the building (game) to sway wildly, if not fall over and sink into a swamp. That's GW's problem. There is no "foundation" from which they derive balance from. So any change they make, be it Chapter Approved, Munitorium, FAQ, things sway dangerously and unpredictably.

I know, it's really hard to conceptualize a rock solid foundation and rational basis for points in any given game, as.. well you know.. money is more important.

So lets look at it this way.

What is the value of a 4+ BS vs a 3+ BS? what is the specific deviation in % success between them? That's a "base value" it is a hard value. The AI complexity appears when you have to take multiple things and value them in respect to each other.

When is a 4+ BS worth say, 10 points. But when is a 4+ BS maybe worth 7 points, or twelve?
What is a T4 worth, how much "better" is it than T3? What about T5? How does that interact statistically with the value of the 4+ BS?

What are armor saves worth? An Armor save is not a flat value. It can't be, just as T4 can't be given a flat value the interaction between the three listed stats produces the "real value" of a given unit.. or in our case a LOT of other Stat Line values. Initially I thought I might be able to make each stat line value have a "hard value" in isolation. But, as you have pointed out to me, that's not possible or reasonable As a T4 unit with a 3+BS but a 5+ armor save does not perform the same as a T3 unit with a 3+BS but a 3+ armor save.. or whatever.

It will be one hell of a deep dive.

As with all things AI learning and AI based statistical analysis you have to get the Question right before you go coding for it. Ask the wrong question you will get the wrong answer. Computers are pretty dumb things. It's up to the human to "phrase" the question correctly.

There's also relative value as well - relative to the weapon/stat of another unit in the same army, as well as against the stats of the opposing army.


For a truly 'balanced' ruleset, you would need to generate realtime points values.

Basically, you'd have to have more units than you need, and run an algorithm for both forces to determine how much each is 'worth' fighting the other - heavy bolters being more valuable against orks than lascannons for example.

And then you'd have to whittle down your list to an agreed value.

I think if a mathematician built this into an app it would work - you have your army lists and put them in and it does stat comparisons (and factors in terrain layout, relative effect of unit speeds, cover bonuses etc) and then spits out a balanced matchup from the forces you have.

Have fun with your stats, but this is a fools errand. There are too many variables in 40K to allow for a single system to value units. Just all the special rules throw everything out of balance, yet alone the synergistic effects of a well designed unit versus a badly designed one.

Whenever a system attempts to give players rules for point values of their own units, they either overvalue everything or allow for the design of optimal units by buying what you need while leaving everything else at minimal values.

 meatybtz wrote:
Not really, remember we are a D6 system which vastly simplifies things. As for "synergy". Part of the current synergy is an artifact of bad points balancing where you have JUNK UNITS. The idea here is that since the "stastical" effectiveness of a unit based on D6 dice statistical changes per point you can approach a level of balance where things are not JUNK.

Nonsense. Examples of synergy could be as simple as saturating a single type of target profile reducing the efficiency of the weapons common to a meta by offering some of them significantly suboptimal targets. That doesn't require some units to be bad and others to be good it just requires a good meta read.

Speaking of meta, how do you account for people skewing lists to face Marines because they're such a large part of the meta? How do you then stop an 'anti-meta' list such as Goff hordes from exploiting that skew? There's a lot more to balance than simply looking at unit effectiveness and this is before you get into special rules, auras, doctrines, stratagems, etc.

What is the value of a 4+ BS vs a 3+ BS? what is the specific deviation in % success between them? That's a "base value" it is a hard value. The AI complexity appears when you have to take multiple things and value them in respect to each other.

When is a 4+ BS worth say, 10 points. But when is a 4+ BS maybe worth 7 points, or twelve?
What is a T4 worth, how much "better" is it than T3? What about T5? How does that interact statistically with the value of the 4+ BS?

What are armor saves worth? An Armor save is not a flat value. It can't be, just as T4 can't be given a flat value the interaction between the three listed stats produces the "real value" of a given unit.. or in our case a LOT of other Stat Line values. Initially I thought I might be able to make each stat line value have a "hard value" in isolation. But, as you have pointed out to me, that's not possible or reasonable As a T4 unit with a 3+BS but a 5+ armor save does not perform the same as a T3 unit with a 3+BS but a 3+ armor save.. or whatever.

It will be one hell of a deep dive.

If this is an example of your 'deep dive' I think you may want to put away the diving board lest you wind up in a neck brace. Basic unit stats and points might be the foundation of 40k's issues but they aren't all of, some units will always be over or underpriced due to meta concerns, or due to a special rule. Take eradicators as an example, people are afraid of them and they have a ton of targets they can get good returns against, so they got nerfed. Yet even before that nerf good players weren't often maxing out on them in favor of other options so are they too good, do Marines just have better units, or is the meta (that they may have helped create) such that they simply aren't required? How do you propose to answer this with your focus on stats and AI learning?

Eradicators always suffered from low flexibility and low range.

That being said, this smacks of trying to micromanage everything. For casual play, who cares? For tournament play, it would be real easy to give a total points handicap or bonus based on the average winlos percentage rankings army by army against all armies.
Since tourney play is usually already internally more or less optimized (or thinks it is) and more or less competent at the rules (or thinks they are) this gives you a decent measure stick of "is this army overpowered".

Then hike the ones that are down 20 points from 2000, or from (next month) 1980, etc until they seem to perform as well as the sucky ones that you have been giving 20 extra points each month to to try to improve them.

Call it "a levelized 2000 point army" instead of just 2000 points. It might be 2200 for tau, 2100 for orks, and 1600 for the very toughest marine group.
The meta will get pulled by the armies of not death gaurd and not ravenwing that have extra to spend, so you will in fact have to make real take all comers lists and not JUST optimize to kill 10 ravenwing terminators.

Is an idea, anyway, and more easy to implement (and less hard to justify) than some algorithm that spits out "eradicators should cost 98 now."

Dukeofstuff wrote:
Eradicators always suffered from low flexibility and low range.
They actually kind of suck at those two important variables, which are much harder to quantify than their S8 or ap-4.

Exactly but they also really can't cost any less in terms of points or they become good in spite of all that. This is the point I was aiming to make.

Dukeofstuff wrote:

Call it "a levelized 2000 point army" instead of just 2000 points. It might be 2200 for tau, 2100 for orks, and 1600 for the very toughest marine group.
The meta will get pulled by the armies of not death gaurd and not ravenwing that have extra to spend, so you will in fact have to make real take all comers lists and not JUST optimize to kill 10 ravenwing terminators.

That could be interesting. Depending on how you set the points adjustments, you're liable to make life a little harder for the relatively casual players that show up at tournaments though. (And yes, those exist.) But the concept is interesting. Ideally, you'd identify how "strong" an army is by a combination of factors including how many popular units and combos it includes, but that's a very difficult balancing act even before you factor in how often the meta shifts these days.

I've been thinking that it might be interesting, in a tournament setting, to adjust a player's total tournament score with a modifier based on their army. So playing tau or orks might give you +20% modifier to your total score for a game while playing marines might give you +0%. So the high score at the tournament might be the guy who lost one of his games, but did so with a less powerful army than the guy who won all of his games. Alternatively, you could just have that apply to someone's ITC points or what have you.

Mathhammer intro https://1d4chan.org/wiki/MathHammer. Increasing BS from 4+ to 3+ increases hit rate from 3/6 to 4/6 1-4/3*100%=33% increase. Increasing a unit's damage output and durability by 100% makes the unit worth 100% more in a fight. Increasing a unit's damage output or durability by 100% makes the unit worth 41% more in a fight (100%*1-sqrt(2)). Durability is useless unless the unit gets attacked, offence is useless unless the unit attacks something. Holding uncontested objectives has value.

You can make an algorithm for pts based on stats, GW has used this method in the past, if you want to let people engineer their own unbalanced rules this is fine but ultimately you need playtesting to get good balance whether you generate pts initially via algorithm or my method of deciding winning and losing scenarios. You could also start with everything costing 5 pts and playtest and iterate on pts via AI until the balance is good.

Wyldhunt wrote:

Dukeofstuff wrote:

Call it "a levelized 2000 point army" instead of just 2000 points. It might be 2200 for tau, 2100 for orks, and 1600 for the very toughest marine group.
The meta will get pulled by the armies of not death gaurd and not ravenwing that have extra to spend, so you will in fact have to make real take all comers lists and not JUST optimize to kill 10 ravenwing terminators.

That could be interesting. Depending on how you set the points adjustments, you're liable to make life a little harder for the relatively casual players that show up at tournaments though. (And yes, those exist.) But the concept is interesting. Ideally, you'd identify how "strong" an army is by a combination of factors including how many popular units and combos it includes, but that's a very difficult balancing act even before you factor in how often the meta shifts these days.

I've been thinking that it might be interesting, in a tournament setting, to adjust a player's total tournament score with a modifier based on their army. So playing tau or orks might give you +20% modifier to your total score for a game while playing marines might give you +0%. So the high score at the tournament might be the guy who lost one of his games, but did so with a less powerful army than the guy who won all of his games. Alternatively, you could just have that apply to someone's ITC points or what have you.

Swedish Comp did this for WHFB, it might be possible to dust-up the creators.

The idea in general will work, but applied by design is particularly difficult to use because modelling complex relationships isn't easy.

Another approach is to do exactly what you propose, but from the end rather than the start, basing it on game results. You need to collect a huge amount of data and unfortunately the systematic errors due to either known or unknown causes will tend to pile up (and, also, it will probably generate some amplifying effects where armies goes up and down in efficiency when their value reach the breaking point).

Both are definitely a better system than the current one, but GW is an obsolete company in its own essence, strict to the depressing idea of not change anything that isn't broken, so that's it.

I personally think that a with a collection of all the tournament data in Python se should be able to train an AI to forecast the result of any given game: considering the huge amount of information that a single game implies it would became probably very effective.
At that point, you use the prediction to make assessment on the meaningful pricing points.

Hi folks.
Using points values to improve balance in random pick up games has been used to good effect in lots of other wargames.
Perfect balance is not the goal, but to improve the game play experience as much as possible, without being overly restrictive should be !

Anyhow to this end , I would like to make the following suggestions.

1)Assign points at the level of interaction.
As all the interaction in 40k is at the unit level.
Any attempt to cost individual elements of the unit ,soldiers guns armour etc. Means all the very complex interactions between all these different elements have to be calculated. And then variable costs would have to be assigned based on the synergistic enhancements of each possible configuration.

If however, we just cost the units as compete entities, (reduce the variants to some standard options everyone takes.)
It reduces the variance in synergies to the army composition.

2)Use ratio based army composition.
This is the best way to get themed armies without the excesses of min -max composition.
Eg For every 4 Core units you may take 1 Special unit. For every 2 Special units you may take a Rare Unit.

What a unit is classified as depends on the theme.
Eg Defencive Infantry, armoured assault ,fast moving recon , strike force etc

3) Have the same resolution methods for all units.
Unfortunately, the current 40k rules are bogged down in far to many special exceptions to sell models .
It would be possible to re write the rules in a much simpler way , allowing far more intuitive play and more detailed tactical interaction.
But it would mean loosing most of the 'fluffy special rules '' as most 40k players often have sunk cost mentality about rules re writes, it is not really a practical option.

I happy to discuss in more detail these ideas if anyone wants to?

Lanrak wrote:
1)Assign points at the level of interaction.
As all the interaction in 40k is at the unit level.
Any attempt to cost individual elements of the unit ,soldiers guns armour etc. Means all the very complex interactions between all these different elements have to be calculated. And then variable costs would have to be assigned based on the synergistic enhancements of each possible configuration.

If however, we just cost the units as compete entities, (reduce the variants to some standard options everyone takes.)
It reduces the variance in synergies to the army composition.

So you mean instead of a unit of 10 Termagaunts buying individual models and individual special weapons you can either have 10 basic Termagaunts, 30 basic Termagaunts, 10 devourer Termagaunts, 30 devourer Termagaunts or a unit with 15 of each? I don't think tiny things matter that much, there just needs to be an upside and a downside to every option in terms of pts-efficiency and role on the board. Simplifying options does not automatically balance the game, it becomes easier but there are still lots of units with one option and that option is either terrible or overpowered and lots of units with zero options that are terrible or overpowered. I think the cost of moving to a simplified system is too great, people should be able to take 19 Termagants and 8 devourer Termagants if they want.

2)Use ratio based army composition.
This is the best way to get themed armies without the excesses of min -max composition.
Eg For every 4 Core units you may take 1 Special unit. For every 2 Special units you may take a Rare Unit.

What a unit is classified as depends on the theme.
Eg Defencive Infantry, armoured assault ,fast moving recon , strike force etc

I don't understand what your classifications change. How do you avoid min-max compositions with your system? I'll take as many Core and Special units to maximize how many overpowered Rare units I can get. How does your system prevent or punish spamming and force or reward me for building themed armies?

3) Have the same resolution methods for all units.
Unfortunately, the current 40k rules are bogged down in far to many special exceptions to sell models .
It would be possible to re write the rules in a much simpler way , allowing far more intuitive play and more detailed tactical interaction.
But it would mean loosing most of the 'fluffy special rules '' as most 40k players often have sunk cost mentality about rules re writes, it is not really a practical option.

That's what 8th did by removing the special resolution rules for vehicles. I don't know how you can simplify the system any more. I don't see what special rules change the way you roll to hit and to wound. GW is already getting rid of FNP, a rule that slowed down the game because of how the game system handled FNP rolls.

It is more to do with the overall effect of a unit in game.
It was why common units with single specialist weapon, have a much lower cost per specialist weapon. Than an elite squad with multiple specialist weapons .The specialist unit is much more effective in game as unit with a specialist role, rather than a slight boost to a bog standard unit.

Some units have a wide variety of number of models in them. But there is often an optimal number in terms of number of models to unit effectiveness in game.

The synergy of ''in unit composition'' makes accurate costing of units based on adding up the composite parts needlessly complicated.

If we establish a base cost for a 'standard unit'. Then simply cost ''complete units'' in comparison to the ''standard unit,'' it gives us a much better chance to balance the game play at the level of the interaction.

40k codex books like to give the illusion of lots of choice. But as it takes ''more competitive players'' next to no time to find sub optimal options , and just use the most in game effective options. Why not just list the most used options by more competitive players for random pick up games?


If there are units that are just not effective in game, or are ''instant win''. Then that is terrible game development. These units should be removed.

However if a units points cost does not reflect it in game effectiveness, then simple re evaluating it in comparison to other units is the simplest way to balance it.
(Narrative gamers who want more wild armies and match ups do not use point values anyway!)

Mainly the idea is Core units are the units that follow the main theme of the army .
Specialist units are the 'on theme' support units.
And Rare units are counter theme units that could still fit the narrative theme of the army at a push.
Some units that are so opposite to the theme of the army may not be available .

I am happy to bring back the 0-1 or 0-2 restrictions on units that are actually game breaking when taken in multiples in specific lists. This can be applied to Core , Specialist or Rare units.)

I am assuming that people wanting better game play balance from point values are interested mainly in random pick up games. And if they want
better game play then limiting list options to useful on theme choices could be seen as worth while?

Current 40k rules limit the interaction to mainly roll a 3+,4+or 5+ on a D6.This artificially restricts the possible game play.
I would much rather use comparison of opposed values to determine the dice roll required.(With modifiers,)To allow a much wider range of interaction but that is wandering off topic so I will stop there.

40k should give you rules to represent your miniatures, miniatures should not represent the rules of 40k.

I think it's impossible to accurately cost everything in points, because a units effectiveness is determined by the units around it.

For example: you can easily say a space marine is worth more than a guardsman - they have better stats and are both basic troopers.

You cannot, for example, put an accurate points cost on Kaptin Badrukk. He has an aura which buffs local flashgitz, but this is worth nothing if there are no flashgitz nearby, and is worth a lot if there are loads of flashgitz nearby. Another example's the deffkilla wartrike - it's great in vehicle heavy lists, and rubbish in infantry heavy lists.

Synergy between units is what makes points become arbitrary.

This is the point I was trying to make.
The synergistic interaction of the components within a single unit is very complex far more difficult to evaluate. If you assign points to individual parts of the unit.(Each option would need multiple costs depending on other choices.)

Where as comparing complete unit to complete unit ,(The game is played at the unit vs unit level, eg the game level of interaction.)
Gives is far more accurate comparative values between units.

The next level is the synergy between units. This is why Army composition methods in other games control the amount of ''synergistic bonuses'' to result in more balanced match ups for random pick up games.
GW just want to sell models that look and sound cool .They assume the majority of customers do not actually play games.(Collectors/Painters.)

So to make headway into making a 40k PV system good enough to result in balanced random pick up games.
It is important to focus on the methods that are used to develop good game play , not model sales. I.M.O.

 XeonDragon wrote:
I know a guy with a background in data analytics, writing algorithms. A couple of years back we spent a couple of weeks trying to figure out if GW used an algorithm for point values, by playing around with the factors you suggested.

TBH, most of what he said I didn't understand, but the TL;DR version was "I think they have an algorithm, but it is either not very good, or they treat the results as guidance only and vary the points to sell more models".

I love the idea by the way, I think it would be an interesting project. If it were to start, I think starting with basic troops (e.g. Imperial Guard) as the basis for "costing" BS4+ etc. rather than more "elite" troops might make sense. The reason I say that is the more "elite" a unit, the more the points are (INMHO) likely to reflect a more complex "mix" of factors. So to my mind, take the simplest units first to figure out base costs of key skills, then work up for more complex attributes from there.

I would go with this. They’d probably use a formula as a guideline, then go up or down by 10%. Then anything 50 points and over they keep at 5 point intervals.

The problem with trying to do this is that the actual value of a stat is dependent on all of your stats and the stats of the thing you're aiming at; the upgrade from -0 AP to -1 AP is bigger than the upgrade from -1 to -2, more AP means more if you've got more Strength, and going from S4 to S5 matters against a T5 target but does nothing against a T3 target.

You could make a formula for costing things purely based on their stats if you really wanted to (Rogue Trader had one) but in practice it'd be about as arbitrary and based on guesswork as the points GW makes up.