New Points Costs – A Statistical Model (Updated for S2 2019)

By ClassicalMoser, in X-Wing

23 minutes ago, gamblertuba said:

Yup, most of that sounds pretty much bang on. Especially 5's and 4's being mostly under and overpriced respectively.

Edit for evidence, here is the trend for initiative of the top 50 pilots from meta-wing so far.

image.png.2fdc878708bc6360b33e9341aa714d41.png

After nerfs to Rebel Beef and Phantoms, Init 4 is being squeezed out. I think I1 and I6 are just pulling from a smaller pool overall.

I3 is seeing an interesting uptick though.

32 minutes ago, gamblertuba said:

After nerfs to Rebel Beef and Phantoms, Init 4 is being squeezed out. I think I1 and I6 are just pulling from a smaller pool overall.

Mostly agree. We see some key cases like Mace Windu and Braylen Stramm, but Mace almost always uses Sense and Stramm would still be as good at I2.

I think regarding I1 it's partly due to the smaller pool and it's partly other factors; I1 named pilots are mostly a little overpriced and talentless generics tend to be less efficient than similar named pilots (on chassis that have I1 generics anyway). Classic example is the TIE/ln. The Academy and Obsidian pilots are almost never used because Inferno is so efficient.

Frankly Howlrunner needs a 5 point hike and most of the rest could come down a point or two (even Iden). This would lead to much more variety in terms of swarm builds. It's always Howl/Inferno/Wampa or similar, and that gets really really old. I've never seen a tie swarm without Howl either. They need to raise her price and make it more of a choice, like the way they did with Drea Renthal (though of course they overcooked that situation).

I always find the quirks interesting.

  • Diamond Boron Missiles to 5 points, while Concussion remaining at 6 seems odd: DBM is strictly better on any ship with two Missile slots.
  • Bomblet Generator to 3 points, while Proton Bomb to 6. Bomblet can wiff to bad dice, but doesn't deserve to be half-price because of it.
  • WTF is up with Munitions Failsafe? It looks like it costs 18 points?
    • To that end, the N-1 entries look strange, like maybe there's some sort of errors. I don't see Padme, and Ric and Naboo Handmaiden don't look like they have the right baseline costs.
  • R2-D2 Astromech down to the same price as R2 Astromech. This is almost surely driven by the fact that huge numbers of Jedi and N-1 will take R2 Astromech, but regen Rebels don't show up as much.
  • The first-season differences with non-Talent Rebel Y-Wing going up, but non-Talent Scum Y-Wing going down, are gone. This seems like a product both of Leia going up and not anchoring VTG swarms, but also VTG and Turrets going up, making the 4VTG+1 Swarm no longer possible, where Rebels clearly had the superior version, but Scum's version would be fine, if the Rebel one got priced out. Without particular common lists to skew things, the two Y-Wings settle into the same price, which is good.

Very cool!

[edit -- yes, I did comment on this already upthread, but it's still cool so gets another thumbs up!]

On 6/12/2019 at 2:18 PM, ClassicalMoser said:

Basically this yields a formula of:

[New Cost] = [Old Cost] + { [Old Cost] x [Max Correction] x [Adjustment] } + { [Old Cost] + [Scarcity Bonus] }

I might be reading that wrong, but did you mean to say:

[New Cost] = [Old Cost] + { [Old Cost] x [Max Correction] x [Adjustment] } + [Scarcity Bonus]

Since it looks like you're adding [Old Cost] twice.

Just a suggestion - I would avoid a formula that adds / subtracts points, and instead would multiply the old cost by some coefficient:

[New Cost] = round( [Old Cost] * [coefficient] )

where the mean coefficient for all ships is something close to 1. The rationale is that you want to change ship efficiency by some percentage. I only briefly skimmed so maybe it already works this way. 🙂

Also, 200 points is still not enough granularity for the really cheap stuff like vulture droids and TIE Fighters. Ah well.

Edited by MajorJuggler
7 minutes ago, MajorJuggler said:

I might be reading that wrong, but did you mean to say:

[New Cost] = [Old Cost] + { [Old Cost] x [Max Correction] x [Adjustment] } + [Scarcity Bonus]

Since it looks like you're adding [Old Cost] twice.

Actually I think I meant:

[New Cost] = [Old Cost] + { [Old Cost] x [Max Correction] x [Adjustment] } + { [Old Cost] * [Scarcity Bonus] }

Where the scarcity bonus is a percentage (max 5%).

7 minutes ago, MajorJuggler said:

I only briefly skimmed so maybe it already works this way.

If I understand you correctly, I'm pretty sure it does work this way.

7 minutes ago, MajorJuggler said:

Also, 200 points is still not enough granularity for the really cheap stuff like vulture droids and TIE Fighters. Ah well.

And my poor pet TIE Interceptors and A-Wings. Alas.

Thank you for the feedback, by the way! You're a legend around these parts so your validation means a lot to me!

Edited by ClassicalMoser
1 hour ago, theBitterFig said:

I always find the quirks interesting.

  • Diamond Boron Missiles to 5 points, while Concussion remaining at 6 seems odd: DBM is strictly better on any ship with two Missile slots.
  • Bomblet Generator to 3 points, while Proton Bomb to 6. Bomblet can wiff to bad dice, but doesn't deserve to be half-price because of it.
  • WTF is up with Munitions Failsafe? It looks like it costs 18 points?
    • To that end, the N-1 entries look strange, like maybe there's some sort of errors. I don't see Padme, and Ric and Naboo Handmaiden don't look like they have the right baseline costs.
  • R2-D2 Astromech down to the same price as R2 Astromech. This is almost surely driven by the fact that huge numbers of Jedi and N-1 will take R2 Astromech, but regen Rebels don't show up as much.
  • The first-season differences with non-Talent Rebel Y-Wing going up, but non-Talent Scum Y-Wing going down, are gone. This seems like a product both of Leia going up and not anchoring VTG swarms, but also VTG and Turrets going up, making the 4VTG+1 Swarm no longer possible, where Rebels clearly had the superior version, but Scum's version would be fine, if the Rebel one got priced out. Without particular common lists to skew things, the two Y-Wings settle into the same price, which is good.

This last one was rather a rush job and there were likely some errors, most notably the MF and N-1 as you mention. Some of the new stuff was probably added incorrectly.

Also this should all be taken with a large amount of salt as the data sample size is still quite small (accounting for Concussions, DBM, R2-D2, and Bomblet, which I assume got near the maximum scarcity bonus as they haven't shown up in competitive play much yet).

7 minutes ago, ClassicalMoser said:

[New Cost] = [Old Cost] + { [Old Cost] x [Max Correction] x [Adjustment] } + { [Old Cost] * [Scarcity Bonus] }

that =

[New Cost] = [Old Cost] * {1 + [Max Correction] x [Adjustment] } + [Scarcity Bonus] }

So yeah that looks like a giant multiply. [Scarcity Bonus] actually ends up being negative if the bonus applies, I assume.

7 minutes ago, MajorJuggler said:

that =

[New Cost] = [Old Cost] * {1 + [Max Correction] x [Adjustment] } + [Scarcity Bonus] }

So yeah that looks like a giant multiply. [Scarcity Bonus] actually ends up being negative if the bonus applies, I assume.

Ah, yes, exactly. I just distributed it anyway because why not. It makes it somehow clearer for me to see it that way. 😛

And yes the scarcity bonus is always negative

Edited by ClassicalMoser

Bump, just for awesomeness.

Edited by Sciencius

I love when you do these. It reminds me that I should get writing point review for Midwest Scrub. Minor goofs here and there.

Leatin Ashera is in a Z-95 for some reason.

All of the N-1's are off. Ric in particular is 35 points with no suggested points change. If Ric had been 35 points, I would have to go to FFG and slap some a smattering of Brooks'. Padme is missing altogether.

Precise Hunter going up by 1 point seems odd as it has a representative sample size of 5.

Hera Crew has a representative sample size of 224? Is this correct? That makes her the fifth most used upgrade which doesn't sound correct.

Munition Failsafe is 18 points with a suggestion to go up a point?

In any case, great work!

2 hours ago, 5050Saint said:

I love when you do these. It reminds me that I should get writing point review for Midwest Scrub. Minor goofs here and there.

Leatin Ashera is in a Z-95 for some reason.

All of the N-1's are off. Ric in particular is 35 points with no suggested points change. If Ric had been 35 points, I would have to go to FFG and slap some a smattering of Brooks'. Padme is missing altogether.

Precise Hunter going up by 1 point seems odd as it has a representative sample size of 5.

Hera Crew has a representative sample size of 224? Is this correct? That makes her the fifth most used upgrade which doesn't sound correct.

Munition Failsafe is 18 points with a suggestion to go up a point?

In any case, great work!

Yeah this last one was a little hasty and those are probably genuine errors. There’s been more data in the past couple of weeks so I’ll be updating this pretty soon and will review those areas and fix them

Updated to add all info since July 31. Note that the Wave V stuff does not have enough data yet to yield a reliable projection. Several things look like they're getting buffs due to the paucity of play they've seen so far. In some cases this may prove to be warranted; we'll wait and see.

Also worth noting that this is strictly algorithmic and a human looking at and tweaking these costs would do better than the algorithm alone. For example, it makes little sense to up both Howl and Iden to 42 when all the power is really coming from Howlrunner herself. More realistically she should go up to 44 and Iden should stay, or she should go up by more and Iden should come down.

There is also a cumulative effect with the adjustments. I doubt that Vultures, Struts, and ESC all need to go up by a point each. A nerf on one or two would probably be sufficient, but this model can't totally account for that.

Enjoy!

https://docs.google.com/spreadsheets/d/1fZqj7rroGGcAPio285FJ7qOnluaUtixxs508FJq1pwQ/edit?usp=sharing

My favorite bit: the algorithm suggests every Starviper is overpriced, needing to go down, mostly by two points. Except for Black Sun Assassins. Why?

Because specifically @Crimsonwarlock wins enough with them to prevent the model from determining they need to get cheaper.

//

Oddnesses:

There are named Torrents your model would *increase* in price. Are there some folks out there secretly winning with them? I'm sure that's a cumulative effect thing, though, that they're brought up by Jedi being good.

Likewise, Rebel Scout HWK is +2. I'd guess that's a Leia effect. And I'd be wrong. The ship shows up more often as a Sabine carrier. Hrm. Odd.

Sear being +0 seems dubious to me. Dude is #1 on Metawing, Belbullab is #1 on Metawing because of mostly him. Maybe that suggests that he's not the really important part of Vulture swarm.

//

Potentially useful information: All Reapers down except Vermiel, who's up. Seems like dude might be sneaky-good and worth flying. I mean, free dice mods can be scary.

31 minutes ago, theBitterFig said:

There are named Torrents your model would *increase* in price. Are there some folks out there secretly winning with them? I'm sure that's a cumulative effect thing, though, that they're brought up by Jedi being good.

Likewise, Rebel Scout HWK is +2. I'd guess that's a Leia effect. And I'd be wrong. The ship shows up more often as a Sabine carrier. Hrm. Odd.

Sear being +0 seems dubious to me. Dude is #1 on Metawing, Belbullab is #1 on Metawing because of mostly him. Maybe that suggests that he's not the really important part of Vulture swarm.

Not sure where you're getting this. The only torrent that went up was the GST, and that only by one. The others came down by 1 each, except for Tucker who stayed the same. Jedi lists don't do as well when they take named clones with them I guess.

The Rebel Scout didn't change.

Sear also went up by one, as did pretty much every single card in the sear swarm list.

Quote

Potentially useful information: All Reapers down except Vermiel, who's up. Seems like dude might be sneaky-good and worth flying. I mean, free dice mods can be scary.

Vermeil is great. He has a phenomenal synergy with Vader, and he's actually a reasonable candidate for juke.

Quote

My favorite bit: the algorithm suggests every Starviper is overpriced, needing to go down, mostly by two points. Except for Black Sun Assassins. Why?

Because specifically @Crimsonwarlock wins enough with them to prevent the model from determining they need to get cheaper.

Again, this is specifically intended for a human to look at and make inferences. It's nice to have an objective starting point, but if you can suggest an alternative plausible explanation for the changes suggested, you're probably better off that way.

I always enjoy seeing how much of a nerf it insists Kylo and Rebel Han need. It's interesting because no one really complains about their pricing. Similarly it makes Fett, Teroch, and Fenn go up a point.

Edited by ClassicalMoser

We can only wish for such a dramatic cost restructuring in January.

Dropping the IG Title to zero is a nice idea but I'd rather keep it at 1 or even price it higher and then drop the IG pilots by more, so a single Aggressor in a list can be useful.

Obviously your model doesn't think so it just bases everything of usage and Win rates, which makes the old aggressor and Defender look like they need some help.

Edited by Scum4Life
Spelling

Tiniest of tiny quibble. Under the 2019 S2 upgrades table you have Resistance Han crew with an actual cost of 12 instead of 6 and things odd from there.

Edited by Pa Weasley
3 hours ago, ClassicalMoser said:

Not sure where you're getting this.

The "2019 S1 Pilots" tab, which is where the spreadsheet opens when I click the link.

Is it possible to change the tab order, so someone first looking at your sheet doesn't immediately get the previous results? Anyhow, my observations seems more to be about what your model thought the price ought to be, and times when FFG changed it more than the sheet would have (and not the sheet suggesting an increase). But it's pretty easy to mistake "Real Price" and "My Price" for what they appear to be.

3 hours ago, ClassicalMoser said:

Again, this is specifically intended for a human to look at and make inferences. It's nice to have an objective starting point, but if you can suggest an alternative plausible explanation for the changes suggested, you're probably better off that way.

Oh, to be sure.

I've always just always found the quirks amusing. It's not criticism, but more me pointing out things that make me smile, and thinking about why the model might have made these decisions. Blade and Blue B-Wings are both at 42, stands out. I can kinda see why the model made that decision: every B except Blade has performed well.

Other thoughts:

  • Moldy Crow keeps going up. 18 had felt really high, was 19 S1, and 20 S2. Dang that ship is strong.
  • Rey's Millenium Falcon and Dauntless have ticked up, after being lower than real cost in S1. This kind of demonstrates that FFG's price adjustments down on the pilots worked, as well as more upgrade options over time. Kind of cool to see an upgrade as an indicator of the ships doing well, when they started off 1e so lackluster.
  • Major Vynder down. Wither Hatchetman? Also kind of interesting to see a list which did well show some age. Did Vynder mainly prey on quad Phantoms?

That said, some possible typos: I think your "Current Cost" for CH-10P is a typo at 6 instead of 7, Proton Torpedoes list current cost at 12 and not 13, and at very least Precognative Reflexes seems to have a S1 cost. Most others are null, so probably just a typo or such.

14 minutes ago, Pa Weasley said:

Tiniest of tiny quibble. Under the 2019 S2 upgrades table you have Resistance Han crew with an actual cost of 12 instead of 6 and things odd from there.

Electro-Proton Bomb also seems off.

1 hour ago, theBitterFig said:

The "2019 S1 Pilots" tab, which is where the spreadsheet opens when I click the link.

Is it possible to change the tab order, so someone first looking at your sheet doesn't immediately get the previous results?

Done. Thanks for pointing that out. For some reason I forgot to do this.

1 hour ago, theBitterFig said:

I've always just always found the quirks amusing. It's not criticism, but more me pointing out things that make me smile, and thinking about why the model might have made these decisions. Blade and Blue B-Wings are both at 42, stands out. I can kinda see why the model made that decision: every B except Blade has performed well.

I noticed that. It actually happens a bit that the two generics get squished to the same price. Obsidian and Academy have always come out to the same amount as well. It's further proof that most jousting-platform ships are initiative-scaled too hard. For dodgy ships with an I4 generic it makes sense. Less so when going from I2 to I3 and a talent.

1 hour ago, theBitterFig said:

Moldy Crow keeps going up. 18 had felt really high, was 19 S1, and 20 S2. Dang that ship is strong.

Surprisingly so. I was annoyed at 18 points but now it actually seems legitimately justified. Adding an arc and a 3-primary is crazy enough without the focus retention. It's certainly the strongest title in the game. Another title for the HWK would be super interesting, maybe based on devices or on cloaking?

But Palob and Dace certainly need some love. And maybe Kyle Katarn. I love the old focus fountain.

1 hour ago, theBitterFig said:

Major Vynder down. Wither Hatchetman? Also kind of interesting to see a list which did well show some age. Did Vynder mainly prey on quad Phantoms?

I was also curious about this. I do think his major success was in some small local circles (St. Louis IIRC?). It could be he's just not getting picked up as much anymore now that the Inquisitors are such a bargain with Jendon instead.

1 hour ago, Scum4Life said:

We can only wish for such a dramatic cost restructuring in January.

I really really really want the devs to just look at this costing model. NOT because I expect them to do it all, but just to see where the major outliers actually are and to have a plan for dealing with them. But of course I know they're busy with all manner of things and don't want to sort through a spreadsheet made by one humble scrub who doesn't even have time to get gud on the tabletop. Still though.

I did this before the last adjustment and it made leaving the YT-2400 alone look so bizarre, plus the way the Scyks and Strikers were left alone.

1 hour ago, Pa Weasley said:

Tiniest of tiny quibble. Under the 2019 S2 upgrades table you have Resistance Han crew with an actual cost of 12 instead of 6 and things odd from there.

1 hour ago, theBitterFig said:

That said, some possible typos: I think your "Current Cost" for CH-10P is a typo at 6 instead of 7, Proton Torpedoes list current cost at 12 and not 13, and at very least Precognative Reflexes seems to have a S1 cost. Most others are null, so probably just a typo or such.

1 hour ago, theBitterFig said:

Electro-Proton Bomb also seems off.

Fixing these now.

Edit: They should be fixed.

Edited by ClassicalMoser
13 hours ago, theBitterFig said:

My favorite bit: the algorithm suggests every Starviper is overpriced, needing to go down, mostly by two points. Except for Black Sun Assassins. Why?

Because specifically @Crimsonwarlock wins enough with them to prevent the model from determining they need to get cheaper.

Hilarious.

I'm the reason the black sun can't have nice things.

Edited by Crimsonwarlock

More cost errors:

IG-88D (crew) is 3 points now, not 4.

Hotshot gunner is 7 points, not 5.

Maul is already overcosted at 12, it's just the only option available for scum to add force, raise to 14 probably prices it out of use.

Looks like the algorithm wants to end boba/fenn(+7 points between the pilots and maul). It does well, but is a symptom of lack of solid alternatives for scum.

9 hours ago, Cerebrawl said:

Maul is already overcosted at 12, it's just the only option available for scum to add force, raise to 14 probably prices it out of use.

This is probably true. Again, this is pure data that a human can make decisions based on.

9 hours ago, Cerebrawl said:

Looks like the algorithm wants to end boba/fenn(+7 points between the pilots and maul). It does well, but is a symptom of lack of solid alternatives for scum.

This is probably not true. I’m not saying they necessarily need to go up by a large amount, but if we really want to see more middle-initiative pilots, we can’t have so many go-tos at I5-6. I4 has solid dodging potential, as long as those true aces are actually that much more expensive.

Fenn and Fett are doing very well. Not just in context of the scum faction, but in context of the game as a whole. They’re a very good bargain at their current price. A small increase certainly wouldn’t kill them (especially when almost all I5-6 goes up by significantly more, which is worth noting), but it would help bring the other pilots on the same chassis more into the game. It would add variety to the viable suite of options.

Again, you have to consider it within the framework of the new costs as a whole.

Ran the numbers again.

I changed the algorithm a little this time so there aren't as many crazy flukes. I'm mostly using arctan functions instead of harmonic functions, so it's less error prone and a little smoother/less flukey. I also tweaked the algorithm to punish lists that are only a little above average less aggressively (a problem I've had with it for awhile, finally solved).

I'm planning on actually going through soon to do real human suggestions with commentary/reasoning, which is really the way this is meant to be used. Especially for the upgrades, which are almost impossible to judge by the success of the lists they're taken in because they compose such a small part of the overall success. I may just post that in a separate thread.

Feel free to check it out, and as always, take it with a grain of salt. As-is, it's basically just a "state of the meta" report, more or less.

https://docs.google.com/spreadsheets/d/1fZqj7rroGGcAPio285FJ7qOnluaUtixxs508FJq1pwQ/edit?usp=sharing

@ClassicalMoser There's still "current" point cost errors in there. First I noticed: of the Aggressors, only A is correctly costed. Correct costs are A) 68, B) 64, C) 65, D) 64.

Hotshot gunner is 7 not 5.

Funny: Black Sun Ace -2 to 39. Cartel Marauder ´+1 to 39. (Now the meme list of 5 kihraxz could be 5 Black Sun Aces with Crack Shot).

Edited by Cerebrawl

Skakoan Ace is also 39.

It would be interesting to see 3xDecimators, 3xVCX, 4xE-wings, 4xSith Infiltrators, 4xSilencers, 5xX-wings, 6xTie-Interceptors...

Edited by Boreas Mun