9 minutes ago, f0rbiddenc00kie said:

MajorJuggler has a much different expertise than I do. His is math, mine is pattern recognition.

His MathWing thread can tell you approximately how much pilot skill is worth for different price ranges and how pilot abilities like Scum Boba Fett affect jousting value. I can't do that.

What I can tell you is what you're paying for and why. What you can learn from MajorJuggler and I are very different things.

I would argue that MJ can do that very well too. ^^

And how we the public use that data might be very similar.

4 minutes ago, Blail Blerg said:

I would argue that MJ can do that very well too. ^^

And how we the public use that data might be very similar.

Most likely. However, our approach and perspectives differ widely. As Marinealver pointed out, I took a very "trait-based" view of the matter. Different is a good thing because it gives the public a much more holistic view of things, and it's wise to respect peoples' differences and learn what you can from them.

Just to share a bit of my perspective, the reason why I have such confidence in the formula and my "trait-based" view of the matter comes from the overwhelming amount of perfect results I achieved (once I understood the reasoning, that is). It's like having a graph of 56 points (the # of ships I did an analysis on so far, including variants), drawing a straight line, and having over 30 of them line up not even a fraction of a percent off where I expect the official value to be (and a lot of the ships I couldn't perfectly predict had unique qualities). Even if I wasn't able to pinpoint with perfect accuracy a ship's official value, I could explain why the variance exists with logic and reasoning consistent with all other ships.

I'm curious how the public will use this data. Now that we're talking about this, however, I do think it's a good idea for me to be a bit more clear about the limitations of the formula, what the guide can do for the public, and what my confidence level is when necessary since different areas of the guide I see as fact and others as strong speculation.

I believe that, whether or not this is FFG's formula (and I think it is), this is a useful tool for team planning.

From this, I personally will be creating a guideline for which ships I can run from the base model and which I should always modify.

Ex.- I wouldn't 'fill space' with a basic Y-Wing, but may seriously consider if I need an elite pilot and EPT on my B-Wing, or if a base model would do.

Quote

Before I answer this, you should understand a few things about me as a person: I am very skeptical, analytical, I enjoy asking questions and learning, I always want to understand the "why," and I don't like making claims that I can't prove. So to answer your question... yes. Yes, I am that confident. If I wasn't, it would bother me to say it.

So were I to email FFG right now and they were to say no I can have the keys to your bank account?

"I don't like making claims that I can't prove"

Yet you've made one you can't prove. You don't have an inside source at FFG which is the way to prove it.

"I am very skeptical"

But you're jumping to the conclusion that your formula is FFG's just because it, with a fair bit of fudging where needed, usually works. That's treating corellation as causation. Pirates prevent global warming logic.

There are a few more holes to consider in the claim that this is FFG's own formula.

• If this is FFG's own formula why do they constantly defy it in order to get their pricing wrong?
• The wave that should fit the formula just fine (before stuff started to be revised) is Wave 1. The TIE fighter fits it but not one of the other Wave 1 ships does. They all come out higher despite having inferior dials.
• Wave 2? Also sketchy. The A-wing and Firespray come out wrong (very wrong for the A-wing). The TIE interceptor comes out right despite having the same dial as the A-wing. The Falcon works despite having a ton of slots (which on other ships you use to justify price increases) and a PWT on it.

I think this is a bit of a case of looking through the enamoured eyes of its creator. It's a decent model there's no evidence to suggest you took if from the FFG offices. It's confirmation bias. You went looking for a pattern, you found one and made the fallacious leap that a pattern that works must be there by design.

Quote

What I can tell you is what you're paying for and why.

I disagree.

You're applying a formula that loosely works and then rationalising the errors. That usually results in sensible conclusions but they're still hypotheses.

Quote

the overwhelming amount of perfect results I achieved

From the small ships alone it says four of the naked ships are perfect fits and twenty don't. Similar story with the large ships. Not exactly overwhelming. You appear to be latching onto the successes and ignoring the failures. That approach is how the "vaccines cause autism" mess started.

I don't understand how, if you're truly the skeptical analyst you claim to be, why you'd present this as FFG's own formula with the more confidence than someone who swiped it from their offices? We don't even know if they use one any more. How can you claim with absolute certainty that this is the formula and then say you don't make claims you can't prove?

^--- This. +1,000,000

EDIT: As I have stated before, I can come up with any number of formulas that "match" the data as well, if not better than yours. It makes it no more, nor no less valid of a model than yours.

What it doesn't make them is: "DISCOVERED: FFG's OFFICIAL FORMULA".

5 hours ago, Blue Five said:

From the small ships alone it says four of the naked ships are perfect fits and twenty don't. Similar story with the large ships. Not exactly overwhelming.

One of the first sentences in my guide is " This guide will NOT tell you how FFG determines the final value of their ships ." A lot of your misunderstanding comes from you reading the data incorrectly. I don't expect the formula value to align with an official value unless I believe the ship is very low on the utility side. However, now that I see consistent patterns in the pricing of dials, upgrade bar slots, pilot skill for PS2-PS3, and glass cannon bonuses, I can use that information to very accurately predict prices.

Like the explanation I gave for the E-Wing, some combinations are just really powerful making using a straight formula for pricing everything just plain stupid. It's a combination of formula for a guideline and judgment calls for final values.

You are right about me not being able to ask someone at FFG to confirm it, therefor not being able to "prove." With that being said, I'm still confident that the formula is correct.

Quote

A lot of your misunderstanding comes from you reading the data incorrectly.

It's quite difficult to misunderstand "DISCOVERED! Official formula for ship stat pricing" and " I am 100% confident that this formula is what FFG uses as the basis for determining how much a ship's stats are worth". The first statement is a claim you can't back up in any way and the second is a statement of evidence-free belief in the first.

That would be irritating in of itself were it not for the subsequent statement that you're a skeptic who doesn't make claims they can't back up.

"Like the explanation I gave for the E-Wing, some combinations are just really powerful making using a straight formula for pricing everything just plain stupid. It's a combination of formula for a guideline and judgment calls for final values."

A completely fair statement. But then why do you hold to be self evident that FFG uses your base formula? Everything you say is sensible apart from the official part.

16 hours ago, Blail Blerg said:

This and @MajorJuggler 's jousting value give you a very large account of what is good in the meta. MJ, can we have fun information again about the game? Or did you "solve" the game? Cuz Xwing seems to be in a finer position than it ever was. (with some crazy scum attanii and the punisher being bad bad problems. )

Solving X-wing is really a 2-step process:

• identifying all the new pilots and combos in a wave (not done automatically by MathWing.... yet)
• placing a value on all those new pilots and combos

X-wing is essentially "solved" to a very good degree of certainty in principle, but every time new ships and combos get released it has to get re-analyzed. So the good news is that I can pretty accurately say how good something like Parattanni is, because I am now modeling a complete action economy that is unique per pilot. That's basically what's new in "v3.0" that I have not published. There's also some updates to the underlying mathematical model that is pretty boring to discuss here but could make for a good academic paper somewhere. The bad news is that my current scripting infrastructure is labor intensive when I want to add new pilots and abilities in each wave. I am spot-checking a few things now, but otherwise am procrastinating trying to stay up-to-date on everything. I might, someday, completely revamp my scripts to make it easier to add new stuff, and also run each pilot / squad through snapshots of all the different metas, complete with knobs for changing what a given pilot could get for action economy and range bins. It would be a kind of "ultimate X-wing design toolbox". Problem is I can't exactly give that away for free even if I get around to making it -- nobody in the industry has anything like this!

16 hours ago, Marinealver said:

Again much like Mathwing Jousting values it only factors mostly stats and a little bit of the dials but not by much. However in that case every maneuver is given a value to calculate it's expected cost. Here you have more of a "trait" system like Glass Cannon Bonus or Large Ship bonus which is abstract at its best. Which it still works since this game is nothing more than an abstraction of a "fictional" setting.

Incidentally I have gotten away from trying to directly make final point cost predictions in my most recent MathWing version. I take a more purist approach of simply calculating how much more (or less) damage a ship needs to do, relative to its jousting numbers, to make it equivalently powerful to a reference ship. I.e. Ship "B" has a lower jousting efficiency than ship "A", so ship "B" needs to get free shots on "A" X percentage of the time. You can then use data analytics from in-game playtesting tape to figure out what "X" actually is. I.e. if you know that a TIE Defender's white-K gives it a free shot 30% of the time compared to an academy TIE, then you can work it backwards to figure out what it's jousting numbers should be. Plus you'll need to tweak the TIE Defender's action economy knobs, as discussed above in the designer toolbox, since it has a white K-turn -- thankfully you can get this data from analytics if you know what to look for, and have a large enough data set. So the real way to get an accurate ship cost is to use a combination of mathematical models, with some coefficients that you directly measure from playtesting.

2 hours ago, Blue Five said:

It's quite difficult to misunderstand "DISCOVERED! Official formula for ship stat pricing" and " I am 100% confident that this formula is what FFG uses as the basis for determining how much a ship's stats are worth". The first statement is a claim you can't back up in any way and the second is a statement of evidence-free belief in the first.

I'm sorry, but you're misunderstanding something again. Prior to this you mentioned that only about 4 ships aligned with official values in my guide, correct? I was simply saying that the formula was never intended to give you "official values." Why? Because the formula by itself fails to account for utility, special combinations, and unique attributes. However, now that I understand the trends and prices for most forms of utility, well over half of the ships I analyzed are exactly at the price point I would imagine them to be.

2 hours ago, Blue Five said:

A completely fair statement. But then why do you hold to be self evident that FFG uses your base formula? Everything you say is sensible apart from the official part.

Thank you. I try to be a sensible person.

To be 100% clear, I know I can't get a thumbs up from FFG that this is what they use, and so I can't "prove" it that way. My approach would make more sense if you looked at it from the approach of an investigator at a crime scene; All the evidence, fingerprints, DNA, eye-witness accounts, weapons and wound marks, credit history, everything points overwhelmingly to a particular suspect. Now, I can't get the suspect (FFG) to admit it, but I can prove beyond reasonable doubt that I know who did it.

I think what bothers you is that what I believe constitutes as "proof" is different from your own. Semantics.

Quote

To be 100% clear, I know I can't get a thumbs up from FFG that this is what they use, and so I can't "prove" it that way. My approach would make more sense if you looked at it from the approach of an investigator at a crime scene; All the evidence, fingerprints, DNA, eye-witness accounts, weapons and wound marks, credit history, everything points overwhelmingly to a particular suspect. Now, I can't get the suspect (FFG) to admit it, but I can prove beyond reasonable doubt that I know who did it.

But you can't prove beyond all reasonable doubt. You're right in that you don't need FFG to admit it's their formula to prove beyond all reasonable doubt. The correlation proves very little: if FFG prices ships proportionally to their value then regardless of their method a correlation will exist. What you do need to do is eliminate the alternate possibilities: a different formula and no formula at all.

If you ignore your anomalous assignment of 50% as the offense value for a 1 attack primary then every step of your current process can be linked with simple mathematical relationships and the entire thing cancels down to one simple function.

p = 2 a ( ( (1 h + 1 s ) / (5 - 1 d ) ) + 1.5)

a / d / h / s is the statline and p is price.

Try it. Naked HWK aside it gives the same results for everything.

In this form it also tells us a few things we already know from observation: a higher value of agility increases the value of the hull and shield beneath it for example. You may also have noticed it breaks at Agility 5, but follow your trend for assigning hull values to your hypothetical base ships and you'll notice the exact same thing happen: your hypothetical Agility 5 ship has a hull value of zero.

I find it highly plausible that FFG's initial Wave 1 statline formula is based off a similar relationship between the four statline values. I don't think it's likely that this is the exact formula.

I'm not certain on anything. I can't eliminate the possibility that nowadays they pluck values out of the air based on intuition and tweak them through playtesting as they do for upgrade cards. It'd explain the Punisher.

47 minutes ago, Blue Five said:

p = 2 a ( ( 1 h + 1 s / 5 - 1 d ) + 1.5)

a / d / h / s is the statline and p is price.

Try it. Naked HWK aside it gives the same results for everything.

Okay.

Let's start with the most basic of the basic, the TIE-Fighter.
p = 2 [2] ( ( 1 [3] + 1 [0] / 5 - 1[ 3] ) + 1.5)
p = 4 ( ( 3 + 0 - 3 ) + 1.5)
p = 4 ( 0 + 1.5)
p = 4 (1.5)
p = 6 =/= 12

Sup?

6 hours ago, f0rbiddenc00kie said:

Okay.

Let's start with the most basic of the basic, the TIE-Fighter.
p = 2 [2] ( ( 1 [3] + 1 [0] / 5 - 1[ 3] ) + 1.5)
p = 4 ( ( 3 + 0 - 3 ) + 1.5)
p = 4 ( 0 + 1.5)
p = 4 (1.5)
p = 6 =/= 12

Sup?

Pretty sure he missed some parenthesis.

If you do:

p = 2[2] ((1[3] + 1[0])/(5-1[3]) + 1.5)

p = 4 (3 / 2 + 1.5)

p = 4 (3)

p = 12

7 minutes ago, VanderLegion said:

Pretty sure he missed some parenthesis.

If you do:

p = 2[2] ((1[3] + 1[0])/(5-1[3]) + 1.5)

p = 4 (3 / 2 + 1.5)

p = 4 (3)

p = 12

Oh, thank you very much. His post makes a lot more sense now.

Wow, he's right. This is the exact same formula I use but in a different form. The results match perfectly. This version also supports my theory that the 4-AGI equivalent of the TIE-Fighter would have exactly 1.5 hit points. This is very useful information. My version of the formula makes it easier to understand trends, but this version is superior for calculating extremities such as 1-ATK (which I was never 100% sure on, and I wanted to go back and edit that) and 4-AGI (which I was still working on). Fantastic work! @Blue Five Did you do this?

Quote

Pretty sure he missed some parenthesis.

That's what I get for copying it over from a version in normal mathematical notation late at night.

Attack values map to your offense values by 100a - 100 and the agility of your theoretical ships maps to their hit points by h + s = 7.5 - 1.5d if I remember correctly. This allows offense value to be expressed as 100a - 100 and defence value as 100((h + s) / (7.5 - 1.5d)). Substitute those into your main formula and it cancels down to what you have above.

Those are the functions your offense and defence values follow at the moment at least. I suspect the optimal functions aren't linear.

Quote

superior for calculating extremities such as 1-ATK

Superior in the sense that it's easier to do, yes. Superior in the sense that it gives meaningful results, not as much. Extremes lean into extrapolation which is where linear interpolation (the trend your offense and defence values currently follow) breaks down.

While I'd say your model strongly supports the claim that 3 attack is roughly twice as valuable as 2 attack those are the only two points in which we can have reasonable confidence. There are only three ships with four native attack, none of which are straightforward, and only one with one attack.

Extrapolation without very high confidence that the function is linear can give odd results. For example, what if the actual function was exponential?

The green line is your offensevalue = 100attack - 100 line. The exponential line gives almost identical results in the 2 to 4 attack range. However, it diverges significantly for 1, 5 and 6 attack which are outside the range of interpolation.

If you want to calculate the true value of extreme attack and defence you'd be better served by an approach such as MathWing's that goes into the actual die mechanics.

That being said, this has given me an idea for a HWK "fix". A HWK only turret slot upgrade that gives it +1 attack and +1 agility would put it bang on 16 and on intuition I don't think it'd be particularly strong or weak in practice.

@Blue Five

does your formula work with the large ships? (I'll check myself later today, need to keep a white screen up right now, in class).

Love the unified equation!

@f0rbiddenc00kie

awesome insight into pricing. Thanks for the post!

1 hour ago, Blue Five said:

That's what I get for copying it over from a version in normal mathematical notation late at night.

Attack values map to your offense values by 100a - 100 and the agility of your theoretical ships maps to their hit points by h + s = 7.5 - 1.5d if I remember correctly. This allows offense value to be expressed as 100a - 100 and defence value as 100((h + s) / (7.5 - 1.5d)). Substitute those into your main formula and it cancels down to what you have above.

Those are the functions your offense and defence values follow at the moment at least. I suspect the optimal functions aren't linear.

Superior in the sense that it's easier to do, yes. Superior in the sense that it gives meaningful results, not as much. Extremes lean into extrapolation which is where linear interpolation (the trend your offense and defence values currently follow) breaks down.

While I'd say your model strongly supports the claim that 3 attack is roughly twice as valuable as 2 attack those are the only two points in which we can have reasonable confidence. There are only three ships with four native attack, none of which are straightforward, and only one with one attack.

Extrapolation without very high confidence that the function is linear can give odd results. For example, what if the actual function was exponential?

The green line is your offensevalue = 100attack - 100 line. The exponential line gives almost identical results in the 2 to 4 attack range. However, it diverges significantly for 1, 5 and 6 attack which are outside the range of interpolation.

If you want to calculate the true value of extreme attack and defence you'd be better served by an approach such as MathWing's that goes into the actual die mechanics.

That being said, this has given me an idea for a HWK "fix". A HWK only turret slot upgrade that gives it +1 attack and +1 agility would put it bang on 16 and on intuition I don't think it'd be particularly strong or weak in practice.

I'm not sure it's exponential but it's definitely not a flat line. Each red dice you roll is worth more than the one before it because it's less likely that the result it rolls are going to be cancelled by defense dice.

I covered this last year when I looked at the maths of combat on my blog.

12 hours ago, Blue Five said:

I'm not certain on anything. I can't eliminate the possibility that nowadays they pluck values out of the air based on intuition and tweak them through playtesting as they do for upgrade cards. It'd explain the Punisher.

I'm 95% certain this is how it's done. This isn't me being snarky and cynical, this is me having been on FFG playtest groups for other games in the past and seeing how they work.

IMHO it's more likely to be a group discussion about appropriate cost than any formula output. They'll be looking at how much it costs to equip X with upgrades Y & Z and then if that resulting ship is too powerful alongside ships A & B, then making decisions that maybe they need to make X cost 2pts more so you can't build that squad?

I don't think ships are costed in a vacuum, they're costed in the context of what you can build for 100pts.

Quote

does your formula work with the large ships? (I'll check myself later today, need to keep a white screen up right now, in class).

It's forbiddencookie's base formula condensed into a single equation. It doesn't include any of his contextual multipliers.

I 100% agree about the extrapolation part. The results for 4-ATK/3-PWT are extremely consistent, although I never had high confidence in 1-ATK or 5-ATK/4-PWT, and the only 4-PWT that currently exists in game (Outrider w/HLC) has extremely unique qualities that make it difficult to isolate the information needed.

4 hours ago, Stay On The Leader said:

I don't think ships are costed in a vacuum, they're costed in the context of what you can build for 100pts.

Yeah, this is the reasoning I followed for why the T-65 X-Wing was priced at 21 instead of 20. The formula I use is for stats only, but final values need to account for everything that you mentioned.

I do notice you don't convert to PS1 equivalent. That's fairly straightforward to do for the early ships as almost every ship followed +1PS +1 point and those that don't usually can be worked out from the other pilots.

At a brief glance PS1 equiv usually improves the match of your formula to the actual values.

On 12/28/2016 at 5:26 PM, f0rbiddenc00kie said:

aaa-3] THE FORMULA :
Always start with the TIE-Fighter as the base.
12 * [( offense value + 100) / 2]% * [( defense value + 100) / 2]% = base stat line cost

i think you mean *100 not +100 as then the TIE fighter is approx 3 pts

nvm forgot value is actual value compared to the TIE Fighter

10 hours ago, Blue Five said:

I do notice you don't convert to PS1 equivalent. That's fairly straightforward to do for the early ships as almost every ship followed +1PS +1 point and those that don't usually can be worked out from the other pilots.

At a brief glance PS1 equiv usually improves the match of your formula to the actual values.

Na, it's just the method I decided to take. The formula result is supposed to be for a PS1 version of the ship in question, anyways, so I felt doing that would be a bit redundant.

Also if I did that for a few, then I'd feel inclined to do that for all ships that don't have a PS1 pilot. You know... for consistency. That's a lot of work, and I think the MathWing results would give a lot better insight with regards to pilot skill and ship prices. The only thing I feel I can contribute with regards to pilot skill is that EPT slots often seem to cost around 1 point.

I'm not sure I understand why a few people above are so hung up on him calling it "official" and not paying attention the fact that this works in 98% of cases, explains really well a lot of why certain ships are perceived as really good or bad, and offers amazing predictability for list building and for future ships. Especially with the latest updates. This is an excellent post/thread.

Great work, a very interesting read.

On 2/20/2017 at 6:06 PM, Blue Five said:

But you can't prove beyond all reasonable doubt. You're right in that you don't need FFG to admit it's their formula to prove beyond all reasonable doubt. The correlation proves very little: if FFG prices ships proportionally to their value then regardless of their method a correlation will exist. What you do need to do is eliminate the alternate possibilities: a different formula and no formula at all.

If you ignore your anomalous assignment of 50% as the offense value for a 1 attack primary then every step of your current process can be linked with simple mathematical relationships and the entire thing cancels down to one simple function.

p = 2 a ( ( (1 h + 1 s ) / (5 - 1 d ) ) + 1.5)

a / d / h / s is the statline and p is price.

Try it. Naked HWK aside it gives the same results for everything.

In this form it also tells us a few things we already know from observation: a higher value of agility increases the value of the hull and shield beneath it for example. You may also have noticed it breaks at Agility 5, but follow your trend for assigning hull values to your hypothetical base ships and you'll notice the exact same thing happen: your hypothetical Agility 5 ship has a hull value of zero.

I find it highly plausible that FFG's initial Wave 1 statline formula is based off a similar relationship between the four statline values. I don't think it's likely that this is the exact formula.

I'm not certain on anything. I can't eliminate the possibility that nowadays they pluck values out of the air based on intuition and tweak them through playtesting as they do for upgrade cards. It'd explain the Punisher.

Again all stats, no dials, no actions nothing else. As I said before it is not a complete picture.