Crowdsource Points Project: AoS Gunkulator

[SuperHappyTime]

Why is the weighted total added at the end? Shouldn’t it be averaged (or kept separate depending on model/weapons)?

[Lemon Knuckles]

It's interesting because units have a single fixed points cost, however many warscrolls inherently have a range of possible values based on things like weapon customization options or conditional buffs.  The weighting is an attempt to cost each of these possibilities individually and then collapse them all back to a single unit cost. 

The simplest example is a unit like Hexwraiths:

Hexwraiths have two weapon lines, neither is customizable, and neither of them has any conditional states that modify their expected output.  So to calculate the expected output for Hexwraiths, we just take 100% of the value of Scythes and 100% of the value of the Hooves & Teeth and add them together to get the value of the unit.

Slightly more complicated examples would be units like Grave Guard or Grimghast Reapers:

Grave Guard can only carry one weapon, but they have a choice to make between two available options.  So we evaluate each option, but weight each at 50% then add the results together to cost the unit as a whole.

Grimghasts don't have have weapon options, but the unit leader has a different weapon, so we value those two separately and make sure to assign the correct number of models to each stat-line.  Also, the effectiveness of the Reapers' Slasher Scythe attacks can be influenced by an ability like Reaped Like Corn, so we indicate this alternate state in the Dimension field and evaluate and weight each separately.  We then sum both the weighted values for the two different Scythe attacks and the Knell attack in order to value the unit as a whole.

For some units, this can get really complicated.  Liberators have 12 different weighted stat lines that have to be added together, for example.  Arkanaut Company has 17.

There's a post somewhere earlier on in this thread which lists the current weightings being assigned to the different Dimension types (aura, target, on the charge, etc.).

There's also a separate topic started by @Black Blade on The Lingering Shadow of AoS 1 where he talks about older units tending to get overshadowed by newer units because the range of abilities that newer units tend to have.  I find his thoughts there really interesting in their implications for this project.  For example, Hexwraiths are worth 160 points, but can only ever be worth 160 points.  Their value is completely linear.  Conversely, with other units whose values are being driven by multiple different weighted dimensions, it is possible to make them effectively worth more than their points through smart play.  (It is also possible to make them worth less!).  

Maybe the best way to say it is that for many units, their actual, explicit point cost is a weighted value occupying the middle of a range of virtual values it is capable of.  

 

[Lemon Knuckles]

14 hours ago, Superninja said:

I would like to see ironjawz units put through the gunkalator.  We feel behind the curve.   Celestar ballista is spot on?!?  It has always seemed overpowered/underpriced to me.  oh well, I'll chalk that up to me needing to play more/learn more/play better.   Part of that feeling was the ability to drop it anywhere(allegiance abilities)/the other part was the d6 hits on each hit and the -2 rend.  I wonder how it compares to gitmob rock lobber?  The rock lobber has great range, no need for line of sight, -2 rend as well, and 3 damage, but still seems lesser to the ballista. (at least in my mind)

 

Unit/GunkPoints/GHB18Points/Difference/GHB18vsGunk Variance%

• Grot Rock Lobber/99/100/-1/1%
• Cellestar Ballista/101/100/1/-1%

They work out practically identical.  But there's something to your intuitions about the comparison, because they get to these points a little bit differently:

Unit/ExpectedDamagePoints/EffectiveWoundsPoints/BaseUnmodifiedPoints

• Grot Rock Lobber/123/97/110
• Cellestar Ballista/148/84/116

So, yeah, the Ballista has a 15 point edge on the Rock Lobber if we only consider expected damage output.

As for the Ironjawz, the results are... surprising?

Unit/GunkPoints/GHB18Points/Difference/GHB18vsGunk Variance%

• Orruk Ardboyz/176/160/16/-9%
• Orruk Brutes/213/180/33/-15%

The closest analogue I can find in our current list of sample units is Putrid Blight Kings for the Ardboyz, and Kurnoth Hunters for the Brutes:

Unit/ExpectedDamagePoints/EffectiveWoundsPoints/BaseUnmodifiedPoints

• Orruk Ardboyz/140/252/196
• Putrid Blightkings/108/252/180 
• Orruk Brutes/293/180/236
• Kurnoth Hunters/237/218/228

Interested in thoughts/reactions.

[JPjr]

@Lemon Knuckles Random thought, I was just looking at the spreadsheet and it occurred to me that whilst it wouldn't input any values it might be worth adding columns for GA & Faction. Might be useful for later stages (and easier to add now than later) but also just interesting to be able to quickly compare and have a running tally of how much each is +/-. 

 

 

[Lemon Knuckles]

On 11/19/2018 at 3:13 PM, VBS said:

Further explanation: one increases point cost for a unit based on its synergies alliance abilties, etc... as one might reason that the unit can/will benefit from it. But someone that uses the unit without making use of these factors that have an impact on the units point cost --> from that individual's point of view/unit usage, it will not reflect an accurate cost.

This same principle holds even at Level One for many units.

Simple example:

Dragon Ogres, Storm Rage ability:  re-roll hit rolls of 1 for this unit while it is wholly within 12" of a friendly Dragon Ogre Shaggoth.

This is a Level One calculation since it is wholly war scroll contained, and it directly impacts the stat-line in a mechanical way.  It can and should have value in the system.

If Storm Rage was always active, it would be worth 28 points on a unit of 3.  When trying to cost the unit, then, there is a 28 point range of points values possible. We're currently solving for this by weighting the different states according to defined values by condition (in the case of auras like this, it's 50%).  What that effectively means, is that someone running a unit of 3 Dragon Ogres in a list with no Shaggoths is down 14 points right off the bat.  Someone who is tactically great at managing the battlefield and who manages to always keep the buff up on his unit of 3, is effectively gaining 14 points.

All of this suggests that units that A) appear under-costed, and B) contain a high potential for unlocking virtual points, merit special consideration during list building.  (Example, Chaos Marauders look kinda juicy right now as they seem to check both boxes pretty hard).

[Lemon Knuckles]

12 minutes ago, JPjr said:

@Lemon Knuckles Random thought, I was just looking at the spreadsheet and it occurred to me that whilst it wouldn't input any values it might be worth adding columns for GA & Faction. Might be useful for later stages (and easier to add now than later) but also just interesting to be able to quickly compare and have a running tally of how much each is +/-. 

I definitely think you are right.  Thanks.

[Superninja]

oh wow so the ardboys and brutes are coming out better than expected (for me)?!?!?!   Interesting indeed! TY  for your work.

 

[Black Blade]

In my local scene, everyone seems to really think Evocators are super OP. Have they been run yet?

[Lemon Knuckles]

Okay, massive update.  33 new units added (@Black Blade, including Evocators).  9 new diagnostic fields added to enable analytics.  Classification system added to group point divergences in a functional way.

What this all means is that we are getting to a place where we can drill into the data better and start to identify interesting issues and formulate preliminary hypotheses.

Technical details of the changes will be captured in the updated Changelog in the first post of the thread later today.  

A series of analytic takes follows.

High Level Thoughts

Spoiler

Currently at 55% accuracy overall (63% if we exclude battle-line units).  Interim goal is 66% accuracy.  It is becoming increasing likely, however, that a significant portion of inaccuracy is due to “fudge” in the system.

• Death is currently the most calibrated alliance with 71% accuracy.  12% of Death units are considered “Very” over- or under-costed.
• Chaos is the next most calibrated alliance with 56% accuracy, however 31% of Chaos units are considered “Very” inaccurate, all on the under costed side.
• Order is calibrated at 50%, with 23% of units considered either “Very” over- or “Very” under-costed.
• Destruction is only calibrated at 40%, but also only 10% of units are considered “Very” inaccurate.
• Heroes are calibrated to 57% accuracy.
• Wizards are calibrated to 61% accuracy (Monstrous Wizards are calibrated at 89%, Non-Monstrous Wizards are only calibrated at 33%)
• Non-Wizards are calibrated to 53% accuracy (63% excluding battle-line)
• Monsters are calibrated to 62% accuracy.
• Battle-line is only calibrated to 35% accuracy.
• Artillery is calibrated to 100% accuracy, but only across a sample of 3.

Since there is material variation in accuracy across different unit types (see specifically Battle-line units, and Non-Monstrous Wizard units), it will be important to structure a consistent methodology for unit selection by alliance in order to better make alliance comparisons.

The Significant of Fudge

Spoiler

Right now, I am seeing 5 possible drivers of inaccuracy:

• Mistakes.  This is fudge, but potentially the kind that is most easily correctable.  See “On Dragons and the Dragonlord” analysis below.
• Unreasonable inconsistency.  This is fudge.  See “The Curious Case of the Fungoid Cave-Shaman” analysis below.
• Problems with the Points Update Process between Annual GHBs.  This is fudge.  See “On the Move from GHB2017 to GHB2018” analysis below.
• Battle-line Units.  This is potentially a calibration opportunity with the math specific to Battle-line.
• Non-Monstrous Wizards.  This is potentially a calibration opportunity with the math specific to Non-Monstrous Wizard heroes.

Fudge is important because it increases the risk of compounding errors in the points system.  If the points cost of Unit X is being incorrectly skewed by fudge, and if the points cost of Unit Y is predicated in some measure by comparisons to Unit X, then the risk of Unit Y being pointed incorrectly increases.     To put it another way, Fudge has a tendency to trickle and multiply throughout the system until it is all a goopey mess.

On Dragons and the Dragonlord

Spoiler

Let’s look at Dragonlord, and a set of close comparibles.

UnitName/ExpectedDamagePoint/EffectiveWoundsPoint/UnmodifiedBasePoint/GunkPoints/GHB18Points/Variance/Rating

• Dragonlord/421/239/330/419/340/79/Undercosted
• VampireLordonZombieDragon/394/189/292/433/440/-7/SpotOn
• SorceressonBlackDragon/254/142/198/293/280/13/SpotOn
• ChaosSorcerLordonManticore/75/156/116/185/200/-15/SlightlyOvercosted
• ChaosLordonManticore/133/192/163/210/250/-40/Overcosted

The absolute closest comparable in the list is Vampire Lord on Zombie Dragon.  The Dragonlord is superior in expected offense (421 v 394) and defense (239 v 189), but we would still expect VLoZD to cost out more due to the spell casting ability.  Gunk costs VLoZD at 433 and Dragonlord at 419.  However, GHB18 values Dragonlord at just 340 points!  That’s 79 points under the gunk value.

It’s not a question of the gunk calculation.  It is consistent across the broader compatibles of Sorceress on Black Dragon and Chaos Sorcerer Lord on Manticore.  The non Wizard Chaos Lord on Manticore is a bit off, but in the other direction.

To put it simply, Dragonlord stands out as a definite problem.  It would be pointed correctly if you cut its movement in half and took away its ability to fly.  But I think it’s possible that there’s a simple but embarrassing answer here:

VLoZd’s breath attack is listed out as a missile weapon on its war scroll.  Dragonlord’s breath attack is one of those older types that is listed as an ability.  If we take Dragonlord’s Dragonfire ability out of the gunk calculation, it’s calculated points value drops to 360, only a 6% variance to GHB18.  Is it possible that the problem with Dragonlord’s points value is that GW simply forgot or ignored this ability when pointing it?  It’s a very feasible answer to the obvious disconnect between Dragonlord and VLoZD.

The Curious Case of the Fungoid Cave-Shaman

Spoiler

Let’s look at Fungoid Cave-Shaman, and a set of close comparibles.

UnitName/ExpectedDamagePoint/EffectiveWoundsPoint/UnmodifiedBasePoint/GunkPoints/GHB18Points/Variance/Rating

• FungoidCave-Shaman/24/56/40/106/80/26/VeryUndercosted
• Necormancer/11/62/37/80/110/-30/VeryOvercosted
• Battlemage/11/47/29/73/120/-47/VeryOvercosted
• AbhorrentGhoulKing/57/72/65/133/140/-7/SlightlyOvercosted
• Butcher/59/71/65/109/120/-11/SlightlyOvercosted
• Arch-Warlock/82/92/87/178/140/38/VeryUndercosted
• VampireLordonNightmare/65/60/63/145/140/5/SpotOn

We can start by saying that this whole sub-category of Non-Monstrous Wizards is a mess.  I feel fairly confident in saying that this group is plagued with material inconsistency.

Probably the best comparison for FCS in the list is the Ghoul-King or the VLoN.  They are all single model wizards with a command ability.  Ghoul King’s point cost is about 5% inefficient, VLoN is 3% efficient while FCS is a whopping 33% efficient.  But both AGK and VLoN are more combat capable mages than the FCS, so let’s try some different comparables.

Necromancer and Battlemage are the closest comparables if we ignore the presence of the FCS’s Command Ability, and they are both very points inefficient (Necromancer by 27% and Battlemage by a 39%).  

There’s really no way to reconcile any of this.  FCS is pointed correctly if we ignore its Command Ability.  But methodologically, if we choose to ignore the Command Ability on other compatibles like AGK and VLoN, we break the calibration on those units.  And although ignoring the value of the CA on FCS would fix it’s points, we would have to assume the presence of a CA for Necromancer and for Battlemage in order to bring them closer in line to their actual GHB18 costs.

The Non-Monstrous Wizards sub-group seems to be plagued with inconsistency in how points values are calculated, and the Fungoid Cave-Shaman is probably the most problematic of the bunch.
 

On Moving from GHB17 to GHB18

Spoiler

There are currently 18 units in the sample list that experienced points updates in GHB18.  44% of the time these updates were improvements that brought the unit more in line with calculated points values.  33% it made the situation worse.  And 22% of the time it accomplished nothing.  Let’s look at the latter case first.

There were 4 units whose values changed without really accomplishing anything:  Beastlord, Abhorrent Ghoul King, Sorceress on Black Dragon and Keeper of Secrets.  In the case of the last three, their points values all changed by 20.  However in all three cases, while the direction of the change was correct, the actual adjustment needed was 10.  Changing by 20 simply resulted in moving a small points inefficiency from one side to the other.  In the case of Beastlord, there is no correct answer as long as points continue to be held to multiples of 10.  The calculated value of Beastlord is 85 points.  It went from 80 points in GHB17 to 90 points in GHB18.  We are basically treading water with all 4 of these units.

8 of our sample units underwent changes in GHB18 that actually brought them more in line with gunk values:

• Savage Boar Boys went from Very Overcosted to Overcosted
• Irondrakes went from Very Overcosted to Overcosted
• Blood Knights went from Overcosted to Slightly Overcosted
• Huskard on Thundertusk went from Overcosted to Overcosted, but by less margin
• Ironblaster went from Overcosted to Spot On
• Hammerers went from Slightly Overcosted to Spot On
• Lord of Change went from Undercosted to Slightly Overcosted

7 of the 8 changes above were adjustments of 20 points.  It seems like 20 is a default go-to value for most GHB adjustments.  This is probably negatively impacting their effectiveness.

5 units underwent changes that actually made them less calibrated with gunk:

• Kurnoth Hunters went from Spot On to Slightly Undercosted
• Bestigors went from Undercosted to Very Undercosted
• Tzaangor Skyfires went from Spot On to Slightly Overcosted
• Orruk Ardboyz went from Spot On to Undercosted
• Dragon Ogres went from Undercosted to Very Undercosted

It’s interesting to focus on the two most extreme cases above where unit valuations went from bad to worse.  Both Bestigors and Dragon Ogres were Undercosted in GHB17.  The move to GHB18 made them Very Undercosted.  Even if we take the stance that there was a legitimate method for GW thinking that Bestigors and Dragon Ogres needed to be made better, we have to consider that the Beasts of Chaos battle-tome was set to come out, and in every way that context alone would be making Bestigors and Dragon Ogres better units.  The points change seems unjusitifiable, unless perhaps it was also meant to increase enthusiasm for the new battle-tome and help a new army make a splash.

All in all, the current GHB method for periodic points changes does seem non-methodical and somewhat arbitrary.  It is not completely arbitrary, as many of the changes make some sense.  But it is definitely not calculated, and definitely not consistent.  

Just a Bunch O' Graphs

Spoiler

 

As always, would love to hear thoughts, ideas, questions, suggestions or to see what you can puzzle out from the data.  I know the spreadsheet is not the easiest thing to work with, but I believe in the value of having this as accessible as possible for crowdsourcing any insights.  If you can think of ways to enable better collaboration, please let me know!

[Lemon Knuckles]

Easy Reference Current Gunk Output

Spoiler

 

[Superninja]

The fungoid cave shaman is an excellent figure.  However, as an Ironjawz player, if I take the Fungoid as my general to use the command ability, I lose out on the Command trait and Artefact that would normally go on my general.  This may factor in the cost somewhere.  I have used the fungoid many times but only had him as my general a couple.

[Lemon Knuckles]

Hi Superninja,

I have a field in the gunkulator to track whether a unit is unique as a means of testing whether things like this have any bearing on current GW costing.  

My take is that it does not figure into the cost currently, and that it should not be considered in the gunk valuation.  I also think this is correct.  But I do agree that ithese kinds of things influence value in some way, and should figure into each player’s case-by-case evaluation.  This relates directly to the difference between points cost, points value and what we can call meta value.

A good example to explore this is Nagash.

Nagash has a GHB18 value of 800.  This is his Points Cost.  
Nagash has a gunk score of 789.  In the context of the gunkulator, this is his Points Value.

Before the LoN battle tome rolled out, the GHB had Nagash costed at 800, and his gunk score would still have been 789.  So no change there.  But here’s the interesting part.  There is definitely a before and after in terms of how players value Nagash pre/post LoN.  His points cost didn’t change, his points value didn’t change, but the context in which players value what he brings to the table did.

Half of Nagash’s points value/cost comes from his spell-casting ability.  Pre-LoN, the game was such that it was difficult for players to cash-in on that value.  Often, the full point value of his spell-casting was not realized as the Nagash player would simply run out of spells to cast.  Post LoN, Nagash benefited from access to spell lore, and later endless spells and then realm spells.  In effect, the game changed such that it was much easier for Nagash to derive full value from his points investment in spell-casting.      
  
Simply:  his point cost didn’t change, his point value didn’t change, but his meta value certainly did.

I think lots of things fit best under meta value.  Allegiance abilities, artifacts, command traits, army build and strategy, the meta game, etc.

Calculating meta value is much more complex than calculating points value, for reasons like the ones @VBS explained so well earlier in this thread.  It is not impossible, though, and there is no theoretical reason why in-roads couldn’t be made into that problem.  There are some practical reasons, however, that complicate things, including a lack of time and resources applied to the problem.

[Superninja]

All true.  I am really enjoying your method and your madness.  Keep on rockin.

[scrubyandwells]

Hi @Lemon Knuckles, thanks very much for this effort. Looking at your latest list of units, I didn't find much to disagree with. Your outputs largely matched up with my personal evaluations.

The only area that seemed problematic were some of the shooting units. Your results often seemed to overvalue shooting units relative to my personal sense. The nature and overall landscape of AoS2 has lowered the aggregate value of shooting. I think we've seen this reflected in tournament results over the last six months.

To look at a few examples:

* Arkanauts - Gunk: 153 | GH18: 120
* Judicators - Gunk: 159 | GH18: 160
* Kurnoth Hunters - Gunk: 211 | GH18: 200
* Savage Orruk Arrowboys - Gunk: 165 | GH18: 140

In each of the above, Gunk doesn't quite match up with my personal valuation, given the collective set of variables in AoS2 that lowers the value, on average, of shooting in the game. (FWIW, personally, I prefer a weighting in AoS that favors, moderately, combat over shooting/magic.)

E.g., while I agree 210pts is a reasonable reflection of the value of Kurnoth Scythes and Swords, I think 210pts, or 200pts, is an overestimation of the value of Kurnoth Bows. The same applies with Judicators: In GH17, 160pts seemed spot-on for them, but in GH18, it seems 20pts too high, whereas Gunk thought it was still spot-on. 

For reference, here's some of the currently-problematic units in the game that I thought your system largely nailed:

* Grimghast Reapers - Gunk: 170 | GH18: 140
* Morrsarr Guard - Gunk: 189 | GH18: 160 (note: your spreadsheet had 140, the pts for Ishlaen Guard in GH18)
* Witch Aelves - Gunk: 136 | GH18: 100 (I could see some argue 136 is too high; a couple of DoK rules make Witch Aelves especially terrifying currently)
* Dire Wolves - Gunk: 79 | GH18: 60
* Sequitors - Gunk: 143 | GH18: 120
* Evocators - Gunk: 232 | GH18: 200

cc: @Thomas Lyons

[Lemon Knuckles]

@Superninja, thanks man, I appreciate it.  

@scrubyandwells, hi, and thanks for sharing your thoughts in such a detailed and well-written post.  Super helpful, and very appreciated.

Going to take a bit of time to digest it, along with some interesting things that have been mentioned by others here and in some other conversations around this forum.

[Thomas Lyons]

On 11/24/2018 at 12:24 AM, scrubyandwells said:

Hi @Lemon Knuckles, thanks very much for this effort. Looking at your latest list of units, I didn't find much to disagree with. Your outputs largely matched up with my personal evaluations.

The only area that seemed problematic were some of the shooting units. Your results often seemed to overvalue shooting units relative to my personal sense. The nature and overall landscape of AoS2 has lowered the aggregate value of shooting. I think we've seen this reflected in tournament results over the last six months.

To look at a few examples:

* Arkanauts - Gunk: 153 | GH18: 120
* Judicators - Gunk: 159 | GH18: 160
* Kurnoth Hunters - Gunk: 211 | GH18: 200
* Savage Orruk Arrowboys - Gunk: 165 | GH18: 140

In each of the above, Gunk doesn't quite match up with my personal valuation, given the collective set of variables in AoS2 that lowers the value, on average, of shooting in the game. (FWIW, personally, I prefer a weighting in AoS that favors, moderately, combat over shooting/magic.)

E.g., while I agree 210pts is a reasonable reflection of the value of Kurnoth Scythes and Swords, I think 210pts, or 200pts, is an overestimation of the value of Kurnoth Bows. The same applies with Judicators: In GH17, 160pts seemed spot-on for them, but in GH18, it seems 20pts too high, whereas Gunk thought it was still spot-on. 

For reference, here's some of the currently-problematic units in the game that I thought your system largely nailed:

* Grimghast Reapers - Gunk: 170 | GH18: 140
* Morrsarr Guard - Gunk: 189 | GH18: 160 (note: your spreadsheet had 140, the pts for Ishlaen Guard in GH18)
* Witch Aelves - Gunk: 136 | GH18: 100 (I could see some argue 136 is too high; a couple of DoK rules make Witch Aelves especially terrifying currently)
* Dire Wolves - Gunk: 79 | GH18: 60
* Sequitors - Gunk: 143 | GH18: 120
* Evocators - Gunk: 232 | GH18: 200

cc: @Thomas Lyons

Love this project - thanks for the CC @scrubyandwells.

@Lemon Knuckles, I tend to agree with Scruby on your potential over valuation of range attacks in 2.0.  I wonder what a blanket -1 to hit would do to the valuations on all of the shooting units (simulating a consistent Look Out Sir! effect on the targets you are most likely to be targeting).  Most of these valuations that Scruby has pointed out on ranged units seem a touch too high (i.e. I would price most of those at 20 points lower than most of the Gunk values).  Just a thought.  I'll be continuing to follow this.

[Black Blade]

The one I am most surprised by in my faction are Fulminators, but maybe thats just because of how I roll. I've had the discussion with folks at my local that most cav units are overcosted disproportionately with other units UNLESS the cav unit can also fly. So the eel cav results ring true to me. Then again maybe it's just the meta narrative like you were saying about ranged hunters being costed right here but not being equal to its other versions in the current meta.

[Lemon Knuckles]

On 11/26/2018 at 11:29 PM, Thomas Lyons said:

@Lemon Knuckles, I tend to agree with Scruby on your potential over valuation of range attacks in 2.0.  I wonder what a blanket -1 to hit would do to the valuations on all of the shooting units (simulating a consistent Look Out Sir! effect on the targets you are most likely to be targeting).  Most of these valuations that Scruby has pointed out on ranged units seem a touch too high (i.e. I would price most of those at 20 points lower than most of the Gunk values).  Just a thought.  I'll be continuing to follow this.

One of the challenges with troubleshooting under the methodology established is that the impact of any change cascades across the entire sample in a consistent way.  It's like a water-balloon where if you press in on one side, the other side bulges out in response.  I think this is procedurally correct, but it does make tweaking things a bit challenging at times.  I could easily change things to drop the units you mention by roughly 20 points... but it would break a bunch of other units downstream.

I agree with you and @scrubyandwells that there's something with range.  The Arkanaught valuation, for example, is one that really bugs me.  But there are other issues that bother me as well, including a tendency to over-value battle-line units, an unusually high accuracy across Death units coupled with an unusually low accuracy across Destruction units, and a hot mess when looking at non-monsterous wizards.

I've taken a step back and adjusted the actual framework of the model a bit (not just tweaking key variables) in a way that seems potentially promising.  This provisional new framework accounts for Bravery (which helps the battle-line and Death/Destruction variances), and it also changes some of the methodology around range and movement.  I'm not posting an updated spreadsheet yet as I want to clean it up and try a few different iterations of different tweaked variables out first, but I'd be interested in any thoughts on this early updated output.

There are still definite problems, but many of them I believe are fixable.

Spoiler

 

[Thomas Lyons]

On 11/28/2018 at 1:23 PM, Lemon Knuckles said:

I agree with you and @scrubyandwells that there's something with range.  The Arkanaught valuation, for example, is one that really bugs me.  But there are other issues that bother me as well, including a tendency to over-value battle-line units, an unusually high accuracy across Death units coupled with an unusually low accuracy across Destruction units, and a hot mess when looking at non-monsterous wizards.

Regarding weighting, does the model take into account that ranged units only get 5 possible "turns" of ranged combat contribution while melee combat has 10 possible "turns" of contribution (since you can fight on both your turn and the opponent's).  I wander if the addition of this 2-1 contribution possibility could fix their contributions.

[Bellfree]

This is nice and all, but I have to ask why? GW have proven, with ABSOLUTE SURETY that they do not give a single F**K about balance by jamming the Realmscape rules into matched play.

Before, when you didn't have to worry about your opponent rolling a 6 at the end of his combat phase and just winning the game for free, maybe something like this would matter.

[Ravinsild]

If you figure this out for this game do you think it would also apply and/or work for 40k? 

[Lemon Knuckles]

On 12/1/2018 at 11:35 AM, Thomas Lyons said:

Regarding weighting, does the model take into account that ranged units only get 5 possible "turns" of ranged combat contribution while melee combat has 10 possible "turns" of contribution (since you can fight on both your turn and the opponent's).  I wander if the addition of this 2-1 contribution possibility could fix their contributions.

Short answer, no it doesn't currently, but it needs to.  @Retro brought up that point early on in the process, and I punted on the issue a bit as I was struggling with other things at the time.  And for the record, this is the kind of thought I find especially valuable as it's not just an arbitrary tweak, it's a coherent and fundamental principle.  Thanks for posting it.

[Lemon Knuckles]

21 hours ago, Bellfree said:

This is nice and all, but I have to ask why? GW have proven, with ABSOLUTE SURETY that they do not give a single F**K about balance by jamming the Realmscape rules into matched play.

Before, when you didn't have to worry about your opponent rolling a 6 at the end of his combat phase and just winning the game for free, maybe something like this would matter.

Haha.  That's a can of worms.  There's a whole thread about why that has produced all kinds of interesting conversations from all kinds of thoughtful people.  That the conversation is happening is enough of a why for me I think.

[Retro]

Hey @Lemon Knuckles, sorry I haven't contributed much here for a while.

Yes I have ghb2017, do you want me to add the values alongside your latest version and flick it back to you?

 

[Lemon Knuckles]

Thanks!