I am not sure if anyone had posted something like that yet. And I figured it could be informative for interested people.

I am putting all games I like into statistical models. For X-wing, I think the big picture of what I do is similar to Juggler, but I am certainly less prolific at sharing and providing amazing tools and records to the community.

A couple specifications that might differ from other jousting models:

- The offensive value of each ship is estimated from the average number of damage it does per turn to an agility 2 ship - I chose that as 2 green dice is really close to the global average as what you expect in defense.

- The defensive value is estimated from the number of 3-dice shots the ship can handle in average before going down. 3-dice is a decent average of the shots a ship takes. There are more ships with 2 red dice than 4 red dice, but the range 1 bonus is such that 3-dice is very close to the average of what a ship suffers.

A little sophistication is that I used a library of vassal tournament playoff logs to figure out the average number of shots occurring at long, medium and short range, and respectively index the breakdown between those in my model. It turns out that range 2 shots are significantly less frequent than close and long range shots.

I also used these logs to estimate how often a ship uses action-driven mods in defense and offense. And it turns out that Offensive modifications are more common, mainly due to focus firing depriving the target of defensive tokens when taking multiple shots in the turn.

Here are the results using the first list of point values released .

And a few of the conclusions I drew from those :

no major completely unfair outlier. The ships at the bottom of the list indeed offer much more in terms of flexibility than the pure jousters on top.

I have indicated the ships benefiting from broader arc options (in green the ships with 2 arcs with the same fire power, or a 180 arc, in light green the ships with a weaker secondary arc, and in orange the turrets.

Even among ships with no additional arc option, the ones on the bottom have often much more flexibility or power concentration

Winners:

- Swarms are back (on top) where they should be

- Tie bombers are now on the TIE fighter / Z-95 jousting level.

- Several ships don't get taxed much for the options they offer:

- Alpha gunboats would almost be decent jousters as they are, but of course they're way better when decently clothed

- TIE phantoms are not 'that low' in terms of jousting for the great options they provide

- the swarm of naked HWKs was the worst possible V1 list ... well it looks like it is the most cost-efficient turret now, and by far. You really don't pay that much to be able to move their arc

- Auzitucks are still up there: they're almost average jousters, but it does not consider their wide arc and reinforce ...

Losers:

- The E-wing is really priced too high - I think that's the only big mistake of this first path at assigning points

- The Tie advanced is priced as it was in V1, but with a much weaker ability, and it was not seeing play. And even Vader really costs a premium.

- The VCX-100 and the Jumpmaster really took a blow, probably some overcompensation after disgusting many players for the last years of V1

I would say that the selection of playable ships looks very large, provided there is no outlier on the 'too strong' side that's hidden there. Good start!

1 hour ago, Fanfan said:

I am not sure if anyone had posted something like that yet. And I figured it could be informative for interested people.

I am putting all games I like into statistical models. For X-wing, I think the big picture of what I do is similar to Juggler, but I am certainly less prolific at sharing and providing amazing tools and records to the community.

A couple specifications that might differ from other jousting models:

- The offensive value of each ship is estimated from the average number of damage it does per turn to an agility 2 ship - I chose that as 2 green dice is really close to the global average as what you expect in defense.

- The defensive value is estimated from the number of 3-dice shots the ship can handle in average before going down. 3-dice is a decent average of the shots a ship takes. There are more ships with 2 red dice than 4 red dice, but the range 1 bonus is such that 3-dice is very close to the average of what a ship suffers.

A little sophistication is that I used a library of vassal tournament playoff logs to figure out the average number of shots occurring at long, medium and short range, and respectively index the breakdown between those in my model. It turns out that range 2 shots are significantly less frequent than close and long range shots.

I also used these logs to estimate how often a ship uses action-driven mods in defense and offense. And it turns out that Offensive modifications are more common, mainly due to focus firing depriving the target of defensive tokens when taking multiple shots in the turn.

Here are the results using the first list of point values released .

And a few of the conclusions I drew from those :

no major completely unfair outlier. The ships at the bottom of the list indeed offer much more in terms of flexibility than the pure jousters on top.

I have indicated the ships benefiting from broader arc options (in green the ships with 2 arcs with the same fire power, or a 180 arc, in light green the ships with a weaker secondary arc, and in orange the turrets.

Even among ships with no additional arc option, the ones on the bottom have often much more flexibility or power concentration

Winners:

- Swarms are back (on top) where they should be

- Tie bombers are now on the TIE fighter / Z-95 jousting level.

- Several ships don't get taxed much for the options they offer:

- Alpha gunboats would almost be decent jousters as they are, but of course they're way better when decently clothed

- TIE phantoms are not 'that low' in terms of jousting for the great options they provide

- the swarm of naked HWKs was the worst possible V1 list ... well it looks like it is the most cost-efficient turret now, and by far. You really don't pay that much to be able to move their arc

- Auzitucks are still up there: they're almost average jousters, but it does not consider their wide arc and reinforce ...

Losers:

- The E-wing is really priced too high - I think that's the only big mistake of this first path at assigning points

- The Tie advanced is priced as it was in V1, but with a much weaker ability, and it was not seeing play. And even Vader really costs a premium.

- The VCX-100 and the Jumpmaster really took a blow, probably some overcompensation after disgusting many players for the last years of V1

I would say that the selection of playable ships looks very large, provided there is no outlier on the 'too strong' side that's hidden there. Good start!

That 0.35ish discrepancy between the Advanced and the X... what were they thinking

Fang Fighters are made for jousting and they are just middle tier... lol

Also, as the Y-Wing is more efficient than the Firespray, I think I'll the Scum pirate build with Drea Renthal instead of Kath Scarlett

But looks like the Empire will indeed Strike Back in 2.0!

Interesting stuff. One thing I think you may need to adjust is the spread of short/medium/long range shots in 2nd edition, but you'll need more data before you can do that. I would expect that in 1st edition, with double reposition and actionless mods being more prevalent it's often easier to get into range 1 without losing offensive power than it will be in 2nd edition, so the decision to token up versus repositioning will be more meaningful. Also, with how easy it is to set up devastating ordnance shots in 1st edition, range 3 is much more beneficial to certain lists in certain situations than it will likely be in 2nd edition.

This does not consider dials, or does it? As e.g. the G1A ends up high, but the ship has a quite horrible dial.

15 minutes ago, Volkomor said:

Fang Fighters are made for jousting

Nope.

It does not consider dials indeed. A goal of these jousting models is to address : how much do you pay for these dials and actions ?

Hmm. So naked Scyks would be a terrible idea. Torpedoes kinda cost an arm and a leg but seem powerful. I'm not sold on the cannon options available so far. Looks like the Scyk might be a filler, flanker ship to threaten with munitions but the Scyk sucks again?

25 minutes ago, gamblertuba said:

Hmm. So naked Scyks would be a terrible idea. Torpedoes kinda cost an arm and a leg but seem powerful. I'm not sold on the cannon options available so far. Looks like the Scyk might be a filler, flanker ship to threaten with munitions but the Scyk sucks again?

My mistake actually. It looked suspiciously low, and I realized I had entered '1' as its green dice value. I will proofread the list and repost it.

Ewing is indeed the very worst non-turret ship.

Five Kihraxz might actually be a bit interesting...

Seven Quadjumpers!!!!1!!1!!! (Do it.gif)

11 hours ago, Commander Kaine said:

That 0.35ish discrepancy between the Advanced and the X... what were they thinking

3 hours ago, Fanfan said:

My mistake actually. It looked suspiciously low, and I realized I had entered '1' as its green dice value. I will proofread the list and repost it.

Ewing is indeed the very worst non-turret ship.

It is, but I think the basically free double-action first engagement is worth some fraction of that gap.

The problem with this data is that it inevitably does not account for piloting and control over the engagement, double mods and similar abilities, dials, initiative order, let alone unique abilities and synergies. For example, Luke laughs at swarms, particularly at range 3.

14 hours ago, Fanfan said:

- The Tie advanced is priced as it was in V1, but with a much weaker ability, and it was not seeing play. And even Vader really costs a premium.

So I'm always a big fan of efforts like this, but I wanted to cover this specifically because it's false, and very easy to demonstrate that it's false. It feels true, because adding results feels better than changing them, but I'll show my work. I'm just going to assume you have a target lock, but if you have a focus as well ATC 2.0 jumps past 1.0 very quickly.

ATC 2.0, just target lock at range 2: http://xwing.gateofstorms.net/2/multi/?d=AAAAAAAAAAA&a1=MwAAAAAABAA (2.25 damage)

ATC 2.0, "i'm not willing to spend my target lock but didn't bring fire-control system because I'm bad" at range 2: http://xwing.gateofstorms.net/2/multi/?d=AAAAAAAAAAA&a1=MQAAAAAQAAA (1.5 damage)

ATC 2.0, FCS, unwilling to spend target lock: http://xwing.gateofstorms.net/2/multi/?d=AAAAAAAAAAA&a1=MQAAAAAUAAA (1.98 damage)

ATC 1.0, just target lock range 2: http://xwing.gateofstorms.net/1/basic/?q=IAAAAAAAAAAAAQAAAAAAAA (2 damage).

Obviously, when you add force, focus, or range 1 2.0 ATC outshines 1.0 quickly. These are the edge cases where it's even close.

EDIT: I guess I should add - the tie advanced is the same price as an xwing, and strictly better offensively.

Second, Fang Fighters having a low "defensive efficiency" leads me to believe that you've ignored concordia faceoff. While I get this is an early revision, this is incredibly misleading. They have the best defensive ability it the game.

6 minutes ago, Brunas said:

So I'm always a big fan of efforts like this, but I wanted to cover this specifically because it's false, and very easy to demonstrate that it's false. It feels true, because adding results feels better than changing them, but I'll show my work. I'm just going to assume you have a target lock, but if you have a focus as well ATC 2.0 jumps past 1.0 very quickly.

ATC 2.0, just target lock at range 2: http://xwing.gateofstorms.net/2/multi/?d=AAAAAAAAAAA&a1=MwAAAAAABAA (2.25 damage)

ATC 2.0, "i'm not willing to spend my target lock but didn't bring fire-control system because I'm bad" at range 2: http://xwing.gateofstorms.net/2/multi/?d=AAAAAAAAAAA&a1=MQAAAAAQAAA (1.5 damage)

ATC 2.0, FCS, unwilling to spend target lock: http://xwing.gateofstorms.net/2/multi/?d=AAAAAAAAAAA&a1=MQAAAAAUAAA (1.98 damage)

ATC 1.0, just target lock range 2: http://xwing.gateofstorms.net/1/basic/?q=IAAAAAAAAAAAAQAAAAAAAA (2 damage).

Obviously, when you add force, focus, or range 1 2.0 ATC outshines 1.0 quickly. These are the edge cases where it's even close.

Second, Fang Fighters having a low "defensive efficiency" leads me to believe that you've ignored concordia faceoff. While I get this is an early revision, this is incredibly misleading. They have the best defensive ability it the game.

We do disagree on the TIE advanced - not being willing to spend the lock is really a thing. And even then accuracy corrector was arguably a better alternative to advanced targeting computer for many low PS variations of the ship in V1.

I do indeed consider the base version of the V2 advanced as fairly weak. I will test a little more VaderV2 to forge an opinion on him.

But yes, you're right on the Fang fighters, I thought I had Concordia face-off modeled, but I forgot to toggle the option. I would still not consider it a defensive powerhouse. It used to only be vulnerable at range 2 thanks to Autothrusters. But it's indeed better than my current depiction. I will post a correction. That was a lot of ships to model at once and I got a little less careful than the progressive first edition building of the model.

4 minutes ago, Fanfan said:

We do disagree on the TIE advanced - not being willing to spend the lock is really a thing. And even then accuracy corrector was arguably a better alternative to advanced targeting computer for many low PS variations of the ship in V1.

I do indeed consider the base version of the V2 advanced as fairly weak. I will test a little more VaderV2 to forge an opinion on him.

I don't mean to be rude here, but this is incorrect. "Not being willing to spend the lock" with an xwing is also a thing, and it's also wrong. Why are tie advanceds different? This is a problem in which you can enumerate all possible options, and spending the lock is always correct if you're maximizing damage. The real answer if you're chasing efficiency is to just bring FCS, but I'll understand if you aren't going to include it because then everyone will want everything included.

I guess, let me be more clear. Here's our scenario:

TIE Advanced firing at range 2 at a defender: Chooses to not reroll with the target lock for better damage next turn: 1.5 damage.

X-Wing firing at range 2 at a defender: Chooses to not reroll with the target lock for better damage next turn: 1.5 damage.

Why are you treating these separately?

9 hours ago, Jike said:

Interesting stuff. One thing I think you may need to adjust is the spread of short/medium/long range shots in 2nd edition, but you'll need more data before you can do that. I would expect that in 1st edition, with double reposition and actionless mods being more prevalent it's often easier to get into range 1 without losing offensive power than it will be in 2nd edition, so the decision to token up versus repositioning will be more meaningful. Also, with how easy it is to set up devastating ordnance shots in 1st edition, range 3 is much more beneficial to certain lists in certain situations than it will likely be in 2nd edition.

Yes, that's why I did not use the most recent vassal logs, as i felt the 2nd edition would be closer in terms of gameplay to XwingV1 in ~2015, when ordnance did not get its boosts yet.

But I agree my assumptions will probably change as the 2nd edition settles in. I went with my best early guesses.

How are you modeling the various combinations of fire btw? There's a fair amount of subtlety and skew around things like using avg damage as a proxy for "shots to kill" (does not work accurately in general) and considering various combinations of attack vs. defense tokens (the two are actually confounded since the defender has to spend tokens proportionally to the strength of the attack).

This is indeed precisely why the calculator has the ability to chain multiple attacks against the same defender (focus fire) - treating attacks as independent can lead to very unrepresentative numbers, so I'd be curious on your methodology and use of tools/math here.

1 minute ago, Brunas said:

I don't mean to be rude here, but this is incorrect. "Not being willing to spend the lock" with an xwing is also a thing, and it's also wrong. Why are tie advanceds different?

I guess, let me be more clear. Here's our scenario:

TIE Advanced firing at range 2 at a defender: Chooses to not reroll with the target lock for better damage next turn: 1.5 damage.

X-Wing firing at range 2 at a defender: Chooses to not reroll with the target lock for better damage next turn: 1.5 damage.

Why are you treating these separately?

The comparison I am making is with the TIE Advanced in V1, that scores 2 damages at range 2 both in its ATC version and in its accuracy corrector version, the latter not even requiring a lock that you might not be able to get, and allowing an evade for example.

You're not being rude : if I have been missing something and my logic is flawed, I will be just glad to try fixing it based on the feedback I get. I have barely played any V2, so, I certainly misrepresented a couple things.

8 minutes ago, Fanfan said:

The com  pa  rison I  am making is with the TIE Advanced in V1, that scores 2 da  mages at ran  ge 2 both  in its  ATC  version and in its acc  uracy  corrector version, the latter not even requiring a lock that you mig  ht not be able to get, and allowing an evade for example.   

This is a good example of why comparing dice rolls in isolation is not particularly useful. The reason AC didn't get a lot of use later in 1.0 life is because it got to the point where if you were not capable of getting 3 hits you were never hitting certain enemies. So it doesn't matter that your "average of 2.0 hits" is a pretty good average, it might as well have been 0 in those matchups.

Since 2.0 doesn't have AC and the hyper-defensiveness is toned down, I'm not sure the comparison and conclusion that "TIE/Adv with AC aren't seeing play in 1.0 so they are bad in 2.0" is appropriate or supported.

Ignoring AC then, if you're comparing the regular version with a lock, the 2.0 version is clearly better (2.25 hits vs. 2.0). But again to generalize this you really need to be modeling what it is shooting *at*.

10 minutes ago, Fanfan said:

The comparison I am making is with the TIE Advanced in V1, that scores 2 damages at range 2 both in its ATC version and in its accuracy corrector version, the latter not even requiring a lock that you might not be able to get, and allowing an evade for example.

You're not being rude : if I have been missing something and my logic is flawed, I will be just glad to try fixing it based on the feedback I get. I have barely played any V2, so, I certainly misrepresented a couple things.

Ah, I missed that you were comparing to accuracy corrector.

I agree, if we were modeling first edition it would make sense to compare to accuracy corrector, but in second edition the idea the you're modeling an xwing with 3 attack dice and a target lock, and valuing this higher from a 3 attack dice with a target lock AND the hit to crit conversion as separate doesn't make sense.

And, good! I got distracted typing the first one so the tone was short, glad it didn't come off the wrong way. Cool on Concordia - give it a shot next time you have a chance, I think you'll find that you're almost always at range 1 (if you want to be) because of the linked actions, so you're attacking at range 1 but your opponent is attacking at slightly worse than range 2. Couldn't begin to tell you what % of shots you'd be taking at what ranges or anything, though.

20 minutes ago, punkUser said:

How are you modeling the various combinations of fire btw? There's a fair amount of subtlety and skew around things like using avg damage as a proxy for "shots to kill" (does not work accurately in general) and considering various combinations of attack vs. defense tokens (the two are actually confounded since the defender has to spend tokens proportionally to the strength of the attack).

This is indeed precisely why the calculator has the ability to chain multiple attacks against the same defender (focus fire) - treating attacks as independent can lead to very unrepresentative numbers, so I'd be curious on your methodology and use of tools/math here.

In V1, I had a "whole game" model, where each turn, the model would give a shot at determining the odds of being able to shoot at each given enemy ship, at what range, and which would be the likely most profitable target (most often being the previously damaged ship)

But it would require significant revision to have it run with solid assumptions for V2, and would not help that much anyway in determining the relative strength of each ship, maybe just help determining favorable match-ups.

It looks like this tool would be much closer to what you would be interested to see.

I agree a better assessment for the offensive power of each ship, for example, would be to try to determine a representative sample of all the shots a ship is taking, against which target, at what range and with which available tokens/mods in offense and defense, in a competitive setting.

This is obviously quite challenging. The best way to make that feasible (I think) would be to have a formatting of VASSAL tournaments that would allow an easy compiling of this - We're not really there yet. Or maybe a group of dedicated players 'recording' the actions of their games - I would not recommend such an effort.

So, the whole question is 'Which simplification leads to a fairly accurate model anyway ?' 'What assumption considerably reduces the complexity of the model and/or its data collection without harming too much the accuracy of its depiction ?'

And I really do think there are 'efficient shortcuts' that don't degrade too much the accuracy of the results, yet make it possible for a data scientist to get a decent idea of the respective jousting power by ship with a reasonable time involvement in model building.

In V1, this kind of tools felt like it gave me a better comparative knowledge of each ship, less biased by my all-natural mindset related to the last experience playing a couple games with that ship. And almost at all times, there were ships that stood particularly high on this list considering the additional options they offered in comparison to similarly efficient jousters ... they were regularly indeed the most competitive ones, and this numerical analysis would tell us that even before getting the extensive table experience to figure it by regular means.

Any updates to this? Particularly looking at automods like Fang Fighter ship ability and common buffs like Howlrunner. A named academy tie being near the top end of efficiency even without free rerolls confirms a lot of the results I’ve seen so far.

On 7/29/2018 at 9:45 AM, player3010587 said:

The problem with this data is that it inevitably does not account for piloting and control over the engagement, double mods and similar abilities, dials, initiative order, let alone unique abilities and synergies. For example, Luke laughs at swarms, particularly at range 3.

That's the point, it's not supposed to show you that because those are going to be different depending on the players and the match up. They are intangible and variable. The things that stay the same regardless of player skill and matchup is the base stats that each ship brings to the game and the minimum number of points it costs to bring those stats into a list.

On 7/29/2018 at 5:22 AM, Commander Kaine said:

That 0.35ish discrepancy between the Advanced and the X... what were they thinking

Probably how they can put Vader on the ship at PS6 without breaking the game.