Every game has a formula. Yes, they playtest, but I guarantee they have something to start from that isn't guesswork and that in most cases a formula can be applied to all ships. Playtesting ahould be to avoid creating things that break the formulas efficacy.

I guarantee they have something to start from that isn't guesswork and that in most cases a formula can be applied to all ships.

Do you have insider information, or are you just speculating?

Edit: It would be really interesting if someone did know something about the secret sauce FFG uses, but I think when people claim that FFG "has a formula" it's just wishful thinking...

Yes, they playtest, but I guarantee they have something to start from that isn't guesswork and that in most cases a formula can be applied to all ships. Playtesting ahould be to avoid creating things that break the formulas efficacy.

Even granting your point for the sake of argument, it would still be inaccurate to say "this formula predicts a cost of X, but it's actually Y, which means the ship is undercosted by Z". In your frame, there's a formulaic cost which is then adjusted up or down during playtesting to correspond more closely to the ship's actual value in play--but that would mean the "right" value isn't what's predicted by the formula but what's actually printed on the card.

And even then, metagame shifts alter cost balance later. It doesn't change the fact finding a fitting regression value is good for the community and an understanding of the game as a whole. I will whole heartedly approve every bit of research and analysis of this game.

And it isn't wish fulfillment. It's basic game design strategy.

Clearly there is a formula. I'm surprised some of our math-wing posters haven't cracked it yet. lol

Clearly there is a formula. I'm surprised some of our math-wing posters haven't cracked it yet. lol

"Traveling through hyperspace ain't like dusting crops, boy! Without precise calculations…"

If there is a formula, but the point values it spits out are typically modified by playtesting, that's going to look a lot like no formula at all. As MajorJuggler said upthread, it's trivially easy to calculate a linear regression model--any piece of statistical software all the way down to Excel will do it. But that doesn't mean the model that gets spit out will be meaningful or helpful, which is why modeling is at least as much an art as a science.

For instance, I can plug the following values for all ships in Waves 1-3 into Excel, and out pop the following coefficients:

Attack: 3.8

Agility: 5.7

Hull: 3.0

Shields: 2.0

Pilot Skill: 1.1

Pilot Ability (coded 1/0): -0.2

Elite Pilot Talent (coded 1/0): 1.4

Large Ship (coded 1/0): 1.8

Rebel Ship (coded 1/0): 5.1

Intercept (flat value): -22.3

In this formula, each point of Attack is worth 3.8 points on the final cost, each point of Agility is worth 5.7, and so on. And it does an okay job of matching the costs assigned to each ship--the largest residual (difference between predicted and actual cost) is less than 4 points.

But the model doesn't make any sense . Is FFG really increasing the costs of Large ships by almost 2 points? Are they charging us extra points to belong to the Rebel faction? Are they offering a discount for ships with pilot abilities? Are shields really worth less than hull? Probably not, in all cases. This model is "right", in the narrow sense that it was computed correctly according to a well-defined statistical process, but it's not meaningful at all.

There are reasons to suspect that this isn't a good list of values to use to predict the cost (it just happens to be a list of stats I already had put together in a spreadsheet). But I suspect no list of predictors is going to work well in a linear-regression context, because there's too much about this game that just isn't linear.

(Incidentally, in case you were wondering: this quickly hacked-together regression which is certainly and completely meaningless says the "right" value for the TIE Defender is just 22.3 points.)

Clearly there is a formula. I'm surprised some of our math-wing posters haven't cracked it yet. lol

After my "Predicting Costs by Lanchester's Law" post, FFG reduced the cost of the A-wing of the A-wing by 2 points. This puts the jousting value of the A-wing at almost 100% according to my formulas. So clearly they're just using my formulas now anyway!

(I kid)

If there is a formula, but the point values it spits out are typically modified by playtesting, that's going to look a lot like no formula at all. As MajorJuggler said upthread, it's trivially easy to calculate a linear regression model--any piece of statistical software all the way down to Excel will do it.

You can also do it using linear algebra to get a perfect fit, as long as you have at least as many coefficients as you have ships, and the resulting matrix is invertable. I would contend that anyone who can't derive the result using linear algebra for at least wave 1 (4x4 matrix) has no business blindly using regression formulas. This would actually be a pretty simple linear algebra homework problem, it would be fair game for the first test. (OK, probably reduced to a 3x3 matrix for a test so you could compute it by hand, but you get the idea...)

If you can get this far, then try extending it to waves 2, 3, and 4. It should become clear why linear regression is largely useless in this case.

And it still says nothing about what a ship's value is. Even if a formula WAS used, it was therefore used to generate the TIE Advanced's point cost... which implies how useful/consistent said formula is.

Unless of course the formula was changed from wave to wave or the new data points have to be worked out. I mean the Wave 1 regression only covers 4 ships. It needs to be reworked.

No, it needs to be scrapped. It's trivial to find a best-fit curve that predicts point costs, but meaningless; there are just too many wrenches thrown in the works by FFG getting imaginative with dials, new actions, etc.

Unless you think FFG makes up the point values they have some kind of system. The more data points there are to work with the better the formula can be determined. 4 and even 12 Data points are nowhere near enough, but anybody who wants to put in the effort and who has the knowledge can go for it. The last formula didn't look at dials, actions, or mod slots very well.

If the formula changes at all, the number of data points become meaningless also, but as people have pointed out, while it's an interesting exercise it's not very useful overall anyway other than for trying to determine what formula FFG uses and, as pointed out, doesn't determine actual tabletop value.

The Defender is slightly undercosted if anything.

Base Cost = 2

Attack ((Atk-2)x8)= 8

Agility ((Agi-2)x8)= 8

Hull ((Hull-3)x4.25)= 0

Shields(Shields x 4.5)= 13.5

Pilot Skill (PS x 1)

Final Cost for PS1 = 32.5

Check out this thread: Reverse engineering Squad Point Formula. http://community.fantasyflightgames.com/index.php?/topic/72935-reverse-engineered-squad-point-formula/?hl=%20reverse%20%20engineer

It works out very accurately within a point or two, for every wave and PS, and it shows that upgrade slots and actions are not a factor. Dials only seem to factor if there's more red than green, giving a 4 point reduction to cost (B-wing, Lambda). There are a few other factors that modify cost, a maximum of eight generic pilots (TIE Fighter & Z-95), -10 points for a large base ship, etc. unique pilot skills either add to the cost or the generics recieve a 1 - 1.75 point discount.

The regression formula worked really well in Wave 1. It worked only sort of well for Wave 2: it says A-wings are overpriced by 1-3 points, and that one turned out to be right. But it also says the named Falcons are underpriced by 8.75 points (or overpriced by 1.25 if you use a 10-point discount for a large base), and the ORS is overpriced by 1.25 points (or underpriced by 8.75 if you use a 10-point discount).

And it worked downright badly for Wave 3: B-wings are underpriced by 4.5, HWKs are overpriced by 11.25, and Lambdas are underpriced by 14 (or 4 if you use the 10-point discount).

Bottom line: the regression formula does an increasingly bad job of predicting point costs after Wave 1, and there's no reason to expect that it will do a good job at valuing ships in Wave 4.

If you allow for a couple of modifing factors and playtesting, then it still works up through wave 4. Try this out:

1) if the ratio between White and Red maneuvers is 1:1, reduce the cost by 4 (Lambda, B-Wing)

2) lowest PS ship can't cost less than 12 points. 8 ship maximum per 100pt squad (Tie Fighter, Z-95)

3) lowest PS ship with a turret slot can't cost less than 16 points. 4 ships w/ 360 R2 arcs maximum (HWK-290, Y-Wing)

4) lowest PS ship with a turret primary can't cost less than 26 points. 3 ship w/ 360 R3 arcs maximum (YT-1300)

If you allow for a couple of modifing factors and playtesting, then it still works up through wave 4. Try this out:

1) if the ratio between White and Red maneuvers is 1:1, reduce the cost by 4 (Lambda, B-Wing)

2) lowest PS ship can't cost less than 12 points. 8 ship maximum per 100pt squad (Tie Fighter, Z-95)

3) lowest PS ship with a turret slot can't cost less than 16 points. 4 ships w/ 360 R2 arcs maximum (HWK-290, Y-Wing)

4) lowest PS ship with a turret primary can't cost less than 26 points. 3 ship w/ 360 R3 arcs maximum (YT-1300)

Those are all nonlinear kludges that do nothing to fix the underlying problems.

Except those "modifying factors" are the result of playtesting. Again, there is likely a formula for a starting point. But the starting point is meaningless once you start tweaking based on playtest data.

Clearly there is a formula. I'm surprised some of our math-wing posters haven't cracked it yet. lol

"Traveling through hyperspace ain't like dusting crops, boy! Without precise calculations…"

If there is a formula, but the point values it spits out are typically modified by playtesting, that's going to look a lot like no formula at all. As MajorJuggler said upthread, it's trivially easy to calculate a linear regression model--any piece of statistical software all the way down to Excel will do it. But that doesn't mean the model that gets spit out will be meaningful or helpful, which is why modeling is at least as much an art as a science.

For instance, I can plug the following values for all ships in Waves 1-3 into Excel, and out pop the following coefficients:

Attack: 3.8

Agility: 5.7

Hull: 3.0

Shields: 2.0

Pilot Skill: 1.1

Pilot Ability (coded 1/0): -0.2

Elite Pilot Talent (coded 1/0): 1.4

Large Ship (coded 1/0): 1.8

Rebel Ship (coded 1/0): 5.1

Intercept (flat value): -22.3

In this formula, each point of Attack is worth 3.8 points on the final cost, each point of Agility is worth 5.7, and so on. And it does an okay job of matching the costs assigned to each ship--the largest residual (difference between predicted and actual cost) is less than 4 points.

But the model doesn't make any sense . Is FFG really increasing the costs of Large ships by almost 2 points? Are they charging us extra points to belong to the Rebel faction? Are they offering a discount for ships with pilot abilities? Are shields really worth less than hull? Probably not, in all cases. This model is "right", in the narrow sense that it was computed correctly according to a well-defined statistical process, but it's not meaningful at all.

There are reasons to suspect that this isn't a good list of values to use to predict the cost (it just happens to be a list of stats I already had put together in a spreadsheet). But I suspect no list of predictors is going to work well in a linear-regression context, because there's too much about this game that just isn't linear.

(Incidentally, in case you were wondering: this quickly hacked-together regression which is certainly and completely meaningless says the "right" value for the TIE Defender is just 22.3 points.)

Vorpal, if I could, I would hug your right now. I'm a logic and philosophy professor. My published works argue for the return of meaning (metaphysics) to science. To once again recognize that math is an art that aids science, not the science itself. You perfectly demonstrated this. If I remember correctly, you stated you are working on your PhD? If that's the case, I am excited to know that the world of science will be gaining minds like yours.

Stone37: I agree completely (and I'm a Mathematics Graduate)

this quickly hacked-together regression which is certainly and completely meaningless

This reminds me of a saying:

"If you torture the data long enough, it will confess to anything."

Every formula I've seen hasn't taken the dial into account. I don't know why. But until that's included of course the formula will be off. The Lambda proves dials influence point costs, as does the A-wing. The Lambda has less manuevers than every other ship in the game, and more reds. That's clearly tweaking a set of data to lower point costs. The A-wing has an amazing dial to the point of redundancy and it was overcosted for what it did. How much did that Green 5 forward nobody ever uses cost?

Stone37: I agree completely (and I'm a Mathematics Graduate)

Fantastic to hear! (sorry to all my fellow X-wing players for temporarily hijacking this thread) Know that the recent history of mathematics and science has been a very empirical one. Aristotle's 4th cause has been sneered at and thrown out the window. Logic and metaphysics has largely been called "myth" and "fantasy". It's good to see the tide shifting. Math is a fantastic tool meant to help those who use it find meaning in nature.

Every formula I've seen hasn't taken the dial into account. I don't know why. But until that's included of course the formula will be off. The Lambda proves dials influence point costs, as does the A-wing. The Lambda has less manuevers than every other ship in the game, and more reds. That's clearly tweaking a set of data to lower point costs. The A-wing has an amazing dial to the point of redundancy and it was overcosted for what it did. How much did that Green 5 forward nobody ever uses cost?

I take the dial into account in the extended math in my Lanchester's thread, but determining dial value is still a work in progress.

Edit: incidentally, with the -2 cost reduction, A-wings (at least PS1 A-wings) won't need the dial and boost to be competitive. Their jousting efficiency will already very high, sitting between the TIE Fighter and the B-wing. The dial and boost at that point will just be gravy.

That's good to hear. I figure the best place to start looking is on a Tie Fighter versus Interceptor as they are the most similar ships in many respects.

If you allow for a couple of modifing factors and playtesting, then it still works up through wave 4. Try this out:

1) if the ratio between White and Red maneuvers is 1:1, reduce the cost by 4 (Lambda, B-Wing)

2) lowest PS ship can't cost less than 12 points. 8 ship maximum per 100pt squad (Tie Fighter, Z-95)

3) lowest PS ship with a turret slot can't cost less than 16 points. 4 ships w/ 360 R2 arcs maximum (HWK-290, Y-Wing)

4) lowest PS ship with a turret primary can't cost less than 26 points. 3 ship w/ 360 R3 arcs maximum (YT-1300)

Those are all nonlinear kludges that do nothing to fix the underlying problems.

How so? These modifiers look at two aspects that can't be calculated in the ship's stats.

A) how much space a squad can cover with its combined firing arcs.

B) does its Maneuver Dial hinder it's movement

Look at the Lambda and B-Wing dials, they are the only two in the game with a 1:1 ratio of Red:White, and the cost 4 less than the formula predicts. What has the current Meta done to adress this? Add Engine Upgrade or Advance Sensors at 4-3 points. That looks like the result of playtesting to me.

The A-Wing looks overcosted, and it has the highest ratio of Green:White on its dial (i forget exactally what atm)

Other movement aspects are too subjective to quantify into points. For my playstyle, a 1 forward is the most important maneuver, for someone else a 5 forward is better.

If you apply that guy's formula with my modifiers to every pilot, it works out to within a point or two on average, for every ship through to wave 4. I'm sure there are factors we're not aware of, but this is too close and consistent to just write off.

Until its on the board and play tested we don't know if its any good or not.

Harsh realism incoming:

One of the most popular arguments for players is: "we'll see how it plays out." I'm sorry, but when someone uses that argument on me, it's basically telling me you're incapable of logical thought and/or lack the experience to make an accurate projection. Any experienced player will be able to judge scale, power, weigh the pros and cons, understand cause and effect, just by reading cards and letting their superior experience, game knowledge and skill guide them. Saying "we'll see how it plays out" is quite simply a pillowfort for lesser players who lack the tools to decipher what's in front of them.

As we get more information on the Defender, we will have a better understanding exactly whether or not this craft is worth the 30 points base. I have a good feeling that we won't, because for 3 more points, you have the Firespray, which has 1 less agility, but superior in every single way from the stats perspective. The only thing that remains is the dial, but I'm not holding my breath.

Well that was quite the Appeal to Authority...

With some No True Scotsman for good measure...

and as much as I like Scotch...

"Realism" he says...

Except, that it has happened more than one time in this game. Take the Lambda. It was pretty much derided once the dial was revealed. Now, people have shown how to make the Lambda work. Hell, a single poster worked really, really hard to make an all Lambda squad work. Experience with these ships can make them work better than you would think at first glance.

We will see. I suspect the Defender will fly just fine when compared to the Firespray. White K-turn more than makes up for the lack of rear arc, and being a small base means it's a bit harder to pin down, thus those HPs will last a bit longer.

Except, that it has happened more than one time in this game. Take the Lambda. It was pretty much derided once the dial was revealed. Now, people have shown how to make the Lambda work. Hell, a single poster worked really, really hard to make an all Lambda squad work. Experience with these ships can make them work better than you would think at first glance.

We will see. I suspect the Defender will fly just fine when compared to the Firespray. White K-turn more than makes up for the lack of rear arc, and being a small base means it's a bit harder to pin down, thus those HPs will last a bit longer.

Except, that it has happened more than one time in this game. Take the Lambda. It was pretty much derided once the dial was revealed. Now, people have shown how to make the Lambda work. Hell, a single poster worked really, really hard to make an all Lambda squad work. Experience with these ships can make them work better than you would think at first glance.

We will see. I suspect the Defender will fly just fine when compared to the Firespray. White K-turn more than makes up for the lack of rear arc, and being a small base means it's a bit harder to pin down, thus those HPs will last a bit longer.

The best use of lambda is not a Fortress build. It's the Buzzsaw and Doom Shuttle builds that allow the ship to put out a ton of damage and draw oceans of fire for minimal points.