Jump to content
MajorJuggler

New record for squad jousting value (Fair Ship Rebels 2.0), and a peek at MathWing 3.0

Recommended Posts

Not sure this can be done with a succinct answer, but why zero value on R4-D6? It really is, as you say, tremendous value to assign 0 value to it (if I follow, and I am trying to as the formulae are really interesting).

 

Mostly just curious, if you don't want to divulge, I understand. Some nice math here. 

Edited by Scopes

Share this post


Link to post
Share on other sites
On 7/29/2017 at 1:29 PM, MajorJuggler said:

This new Biggs squad is so ruthlessly efficient that it's quantifiable power creep, and extraordinarily difficult for most lists to fight.

You said it. While I am enjoying the challenge of trying to compete with it, man, "extraordinarily difficult" describes it to a 'T'. Yeesh. I'd almost rather face Paratanni. 

Edited by Scopes

Share this post


Link to post
Share on other sites
10 minutes ago, Kaptin Krunch said:

Thanks for showing us this! It's really in-depth. I like your idea for a Biggs change- mine was going to be "other friendly ships at range 1 of you count as being obstructed when defending".

 

It's bugging me that you said 'peak' instead of "peek" in the title though.

 

 

 

Oops. Fixed. I'm an engineer, we can only do math, not spell!

Share this post


Link to post
Share on other sites
37 minutes ago, VanderLegion said:

Out of curiosity, what's lowhhricks jousting value without the free evade when being attacked (since it's also being assumed that he's using his token to give Biggs an evade every round)

 

Same as quoted above, because his action economy for attack doesn't change (he always reinforces), and then you use his actual durability (reinforce applies to an otherwise 0 action x/1/6/3 statline). If Biggs gets attacked first in the round and then Lowhhrick gets attacked, then his reinforce is gone, so his numbers would be lower. [Edit: he drops to abs JV = 19 and PS JV = 21.2] You can file that under 'overestimate' with a couple of other items.

 

36 minutes ago, Scopes said:

Not sure this can be done with a succinct answer, but why zero value on R4-D6? It really is, as you say, tremendous value to assign 0 value to it (if I follow, and I am trying to as the formulae are really interesting).

 

Mostly just curious, if you don't want to divulge, I understand. Some nice math here. 

 

It's a first-order approximation, and sets the "floor" for what we can expect the squad to do. In this case the lowball estimate probably more or less offsets some of the other highball estimates. There's some variance in the certainty of the final answer, but I'm comfortable stating its at least 115 - 120.

Edited by MajorJuggler

Share this post


Link to post
Share on other sites

What if you took the zero value things,  like Primed Thrusters, and instead changed Commando crew to Chewbacca crew?  Would the overall value be even higher?  Negating an additional damage card *and* recovering another shield seems pretty hilarious in this already annoying damage mitigation build.  

EDIT: I suppose to get the other 2 points you'd have to change AT on Jess to IA as well, so might be a wash

Edited by piznit

Share this post


Link to post
Share on other sites
3 minutes ago, piznit said:

What if you took the zero value things,  like Primed Thrusters, and instead changed Commando crew to Chewbacca crew?  Would the overall value be even higher?  Negating an additional damage card *and* recovering another shield seems pretty hilarious in this already annoying damage mitigation build.  

EDIT: I suppose to get the other 2 points you'd have to change AT on Jess to IA as well, so might be a wash

 

Lowhhrick really wants to keep his damage output as high as possible. Other variations include Rey crew instead of Commandos. So long as you have the Rey focus token stack, you have an even better attack economy. The problem is that after a few rounds it runs out and you're better with Commandos.

Share this post


Link to post
Share on other sites

Glad you're finally coming around to looking at squad values, and loss exchange ratios, rather than specific ship values -- though I do think that trying to norm to single comparative values (JV) leads to very rough approximations where better explicit results are obtained by comparing two specific opposing forces. Then you can judge relative LER changes based on tactics (focused fire or not, for example) or squad element adjustments. 

We use force coefficients (a more general term for your jousting values) quite a bit in high level conflict modeling, but the problem with them is that so much has to be aggregated in order to normalize forces to values that can be compared, that you lose granularity of detail.  So you end up with either a nice model whose numbers look good but can't reliably account for the effects of small changes in unit composition, or models which address small changes well but don't scale or generalize. You end up in a position where you need several scales of conflict model for insight -- and even then it's at best informative, not conclusive -- and in the end well-calibrated PMJ is almost as reliable as the models in terms of qualitative conclusions; the quantitative conclusions have meaning in terms of direction and order of magnitude but differences below a level of magnitude are essentially meaningless.

Will be encouraged to hear what else you learn.

Share this post


Link to post
Share on other sites
26 minutes ago, Hawkstrike said:

Glad you're finally coming around to looking at squad values, and loss exchange ratios, rather than specific ship values -- though I do think that trying to norm to single comparative values (JV) leads to very rough approximations where better explicit results are obtained by comparing two specific opposing forces. Then you can judge relative LER changes based on tactics (focused fire or not, for example) or squad element adjustments. 

We use force coefficients (a more general term for your jousting values) quite a bit in high level conflict modeling, but the problem with them is that so much has to be aggregated in order to normalize forces to values that can be compared, that you lose granularity of detail.  So you end up with either a nice model whose numbers look good but can't reliably account for the effects of small changes in unit composition, or models which address small changes well but don't scale or generalize. You end up in a position where you need several scales of conflict model for insight -- and even then it's at best informative, not conclusive -- and in the end well-calibrated PMJ is almost as reliable as the models in terms of qualitative conclusions; the quantitative conclusions have meaning in terms of direction and order of magnitude but differences below a level of magnitude are essentially meaningless.

Will be encouraged to hear what else you learn.

 

One of the nice things about X-wing is that it is a relatively straightforward combat model compared to real life engagements, which you are far more familiar with than I am! The upshot is that the idealized equations hold up pretty well for X-wing, and reality can't stray very far away from the Square Law's result of X = Y*(ky / kx)0.5, or in this case XN = ( (ky / kx) * (YN2 + YN) )0.5.

 

The trouble arises when modeling the action economy, because it is highly nonlinear. I wasn't modeling a real action economy at all in v2.0, so it didn't differentiate action economies between various stat lines, let alone pilot abilities (hello Soontir Fel!). v3.0 is better in that it models it, but still has to use a linearized approximation with a blender-style approach. You're right that more accurate results can certainly be obtained with another model that truly looks at squad A vs squad B in all its nonlinear glory. I don't know if I'll ever get to that point though. (knock on wood). Last I knew, Vorpal Sword switched his PhD dissertation over to analyzing X-wing, presumably using a more direct squad approach as you're describing. Don't know anything about the details.

 

There's more work that I could do on mixed-unit squads with discrete units vs each other, at which point I need to check the literature again so I don't reinvent the wheel. Obviously it always makes sense to kill the glass cannons first and leave the tanks for last, assuming a simple linear system (non-linear value of regen can flip this on its head).

 

But, again, even a first-order approximation of the action economy in 3.0 is going to get you pretty close to reality, the Square Law makes it impossible to be radically wrong unless you seriously messed up your assumptions somewhere. With almost zero playtesting and just pure theory it's still sufficient to demonstrate that the list itself is extraordinarily powerful.

Edited by MajorJuggler

Share this post


Link to post
Share on other sites

The tendency of Square Law and similar approximations in deterministic solutions to wash out details is why with our major modeling tools we've gone almost entirely stochastic. Plus that allows some input to variations that purely attrition-based models can't account for (effects of maneuver, for example). Feeder elements in tools like CASTFOREM and COMBAT XXI have deterministic elements, but sometimes there's just no substitute for brute force and thousands of model runs when you have to aggregate.

(As an aside -- just the engagement priority problem in an interesting one. Using the real world scenarios I work with -- most of our models are "tanks first, infantry later" models.  But when you change how the infantry are equipped, or how AFVs protect themselves, some of the assumptions invert and you end up with very interesting, and counter-intuitive results. That assumes you model the defensive systems adequately, which is a challenge we're dealing with now. Attempting to translate some of thse problems to X-wing -- how does a change in apparent capability affect the decision of a player to choose which target to prioritize? The math can lead to a more optimal solution, but can't always account for the game state under which the decision is made -- so how do you use the math to better inform the intuition of the player so he or she can make a better decision under the more uncertain real-world conditions?

Edit: Put another way -- for example, how do you account for the choice of whether or not to shoot Dengar in arc, or when to shoot and risk a return shot from Quickdraw? Done well, the model should be able to suggest the conditions under which shoot is preferrable to don't shoot, but I think the player needs to understand the assumptions beyond just a heuristic that says "if X, do Y".)

Even X-wing is a bit too complex, I think, for deterministic precision -- as I recall just adding indirect fire into the conventional vs. insurgent differential models was a huge breakthrough back in the '60s, and at a tactical level X-wing is far more complex than the very simple assumptions those models were reduced to.

I think one thing you need to do is caution your audience -- these models, no matter how advanced, still hinge so much on underlying limitations and assumptions that while they provide insight, insight isn't an exact solution.

Edited by Hawkstrike

Share this post


Link to post
Share on other sites
12 minutes ago, ficklegreendice said:

I'm not going to bother with the math as it might as well be the gibbering from monolith from 2001 A Space Odessy to me 

 

Instead, I just wanna know why in the blue **** it's called "fair ship rebels"

http://meta-wing.com/ship_combos/2508

The 1.0 version of the list is using an ARC, a TIE, a T-65 and a T-70. Sounds like a lot of fair rebel ships. And the 2.0 version is just an update of that archetype bringing it to unfair levels based on the new stuff. Though even the 1.0 version was a lot of fun to fly, I loved thane in that list. 

Share this post


Link to post
Share on other sites

I faced a Miranda Nym bomber list with my variant today.

Biggs with R2-D2 and IA

Rex

Wookie Defender with Jyn and Jan

Lowhhrick with Selflessness and Chopper.

He got Rex early with a bomb, 2 hits + Sabine love this game. I eventually chased down Nym and killed it.

 

I realized that I may have been better off just running all game and going to final salvo. Maybe not, I'd probably die to too much plink damage from TLT and I can't run away and get proc Jyn.

IDK, just accept the 50~ MoV loss in that matchup and hope you don't face it too often? With your 180 degree arcs, you do have some good anti-Miranda tech, idk.

Share this post


Link to post
Share on other sites

So I just did a quick test run of the heaver version against my squad of choice at the moment.  I played loose and quick, and obviously playing against yourself kinda skews results.

None the less it was a very close game, though I did lose.  I killed rex and lowhrick, and had had jess at 2 hull, and biggs at 2* hull (*because of IA).

I might have even won if my 5 dice cruise missile attack at the beginning didn't get 2 hits/crits total after mods....

 

Share this post


Link to post
Share on other sites
51 minutes ago, Hawkstrike said:

The tendency of Square Law and similar approximations in deterministic solutions to wash out details is why with our major modeling tools we've gone almost entirely stochastic. Plus that allows some input to variations that purely attrition-based models can't account for (effects of maneuver, for example). Feeder elements in tools like CASTFOREM and COMBAT XXI have deterministic elements, but sometimes there's just no substitute for brute force and thousands of model runs when you have to aggregate.

....

I think one thing you need to do is caution your audience -- these models, no matter how advanced, still hinge so much on underlying limitations and assumptions that while they provide insight, insight isn't an exact solution.

 

Incidentally my first foray into modelling any sort of combat was to write a script to calculate the exact stochastic solution of a pair of Axis and Allies armies fighting each other, which I quickly found out yields a classic bimodal distribution.  So a whole other way of looking at the problem is to determine when one side has a 50% chance of winning vs the other side, looking at the entire distribution of all possibilities. The results should not be too far off from the deterministic method, but it absolutely would be better to look at both. 

If you're trying to fit ship cost to a power curve as a game designer, then the curve itself should not be stochastic, but the ship values certainly are. (Soontir rolled 5 blanks through a rock! Ha! sucker!). The game designers obviously can't use a stochastic variable for cost -- they have to choose one value, which hopefully turns out to be in the Goldilocks zone in real-world play. 

Certainly every model has its limitations. If time and resources were unlimited, I could direct a team of programmers to go implement a bunch of ideas, heck, we could work together on it for kicks. But its just me in my spare time, so I have to come up with something that's good enough to be useful, but not so good that it's my entire life's work. Arguably it already takes up far too much of my spare time as it is already!

 

 

To use a golf analogy, it's like we're discussing how to best land the ball directly on the green from the tee on a Par 3, and what we need to do to go from sticking it +/-30 feet from the hole, to +/- 10 feet from the hole. Then meanwhile FFG takes a swing and not only misses the green and the fairway, but they shank it wide left and end up 50 yards into the woods.

 

 

 

36 minutes ago, ficklegreendice said:

I'm not going to bother with the math as it might as well be the gibbering from monolith from 2001 A Space Odessy to me 

 

Instead, I just wanna know why in the blue **** it's called "fair ship rebels"

 

I think "Fair" sounds like "Four", and the list is, well, actually not very fair, so it's a bit of an intentionally ironic name.

Edited by MajorJuggler

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×