Wednesday, 26 July 2017

"Where are those transmissions you intercepted?" -

Community member Haslo has put together a fantastic web tool that builds on the excellent Listjuggler database that Sozin created a couple of years ago.  That web tool is called Meta-Wing, and basically I'm a massive fan.  I use it regularly and it's underpinned some of my recent articles, such as on the ARC-170 or when I explored how TIE/sfs are being used.

That link again:

If Meta-Wing is not already bookmarked in your browser then go ahead and set it right now.  I'll wait until you're done.

Ready?  Ok, good.

Meta-Wing has been been gradually gaining traction and awareness over recent months and I wanted to take the time to properly introduce it to everyone, and clear up some common misconceptions.  What it is, what it does, how to use it, what it's good at doing and what it's bad at doing.

What is Meta-Wing?

To understand what Meta-Wing is really requires understanding what Listjuggler is, so let's take a step back and do that first.  Listjuggler is a website most commonly used by TOs to enter squadlists from their tournaments so that everyone can see what people played.  You can go into Listjuggler and browse through recent tournaments, checking out squads, pilots, loadouts, it's a really handy database.  

Meta-wing is a data aggregator that imports all of the squad lists that have been entered into Listjuggler and lets you sort, search or rank that Listjuggler data.  

Plain English... what does that mean?  

Well If you wanted to know something like "how are people using Roark Garnet?" and were using Listjuggler then you'd have to click into every tournament and look to see if Roark was being played, note all the squads and upgrades down, click on the next tournament and look for Roark again... basically it's going to take ages.  Meta-wing has already done all that for you, though, so that all the squads using Roark are aggregated together and you can look at them all side by side on one page.  

It's all just much more convenient to get into the data and find out what you want to know.

Limitations of Meta-Wing

There's three really important things to note about Listjuggler that should be remembered all through everything you're looking at in meta-wing, though:
  • Not every tournament is entered in Listjuggler.  You're not seeing a complete picture from every X-Wing tournament being played, and indeed it might be weighted towards certain regions, stores or TOs that run tournaments more frequently, or are more rigorous in entering their results on Listjuggler.
  • When a tournament is entered in Listjuggler it's rare for every squad to be entered.  Most commonly you'll get TOs entering only the Top 4 or Top 8 squads and this means you're not seeing every squad or ship that's being played, even from the tournaments that are entered.  Even at the World Championships, for instance, only the Day Two lists were entered and even then it wasn't quite all of them.
  • Meta-wing doesn't add or amend the Listjuggler data at all.  If it's entered wrong in Listjuggler then it's wrong in Meta-wing. Errors are rare but they do happen, and when you start aggregating data up those errors become difficult to spot (thought fortunately they also become a small part of a larger sample size).

So if meta-wing is only getting an incomplete picture that represents just a sample of squads from a sample of tournaments then what's the point?  Is it even worthwhile at all?  Is it just 'noise'?

While Listjuggler and Meta-wing only represent a portion of all the X-Wing squads being played it's a pretty significant portion.  It's not a perfect picture but as a perfect picture is impossible to get hold of it's probably the best we're going to get.  It does mean that there's a couple of things to bear in mind whenever you use meta-wing for analysis of the metagame, though...
  • Meta-Wing is relatively good at saying what is doing well.  Because many tournaments are only entering the Top-8 cut lists meta-wing's sample size is weighted towards the top end of the metagame.  If something keeps doing well at tournaments then meta-wing keeps seeing it and becomes more and more confident that it's actually good.
  • Meta-Wing is relatively bad at evaluating the rest of the metagame.  Because Meta-Wing doesn't see many TIE Punishers or Kihraxz Fighters it's got a lot less data to go on.  There might be TIE Punishers just missing out on the cut to Top-8 all over the place while Kihraxz Fighters are only ever on the bottom tables... but Meta-Wing doesn't know that, it just knows that there's not many of either ship making Top-8.  This also means there could also be some really great rogue strategies that Meta-Wing either doesn't see at all, or assumes is a fluke result because it only sees it once or twice.
  • Meta-Wing can't factor for player skill.  Did a list win because it's good, or because the player using it was good?  Meta-Wing doesn't know, it just knows that the list won.  As sample sizes increase we have to assume that this factor is evened out to some extent - if 100 players all won using Dengar then they can't all have been super-elite players, it's safe to assume Dengar is pretty good.

Meta-Wing Ranking and the 'Magic' formula

There's probably one aspect of Meta-wing that draws more attention, misunderstanding and arguments than anything else: the 'Magic Ranking' that it applies by default to most of the searches you do.  

Before we go any further, one of the most common misconceptions about the Meta-Wing ranking is that it's a measure purely of popularity.  Many people assume that a ship that appears 80 times is ranked above a ship played 60 times, above a ship played 30 times... etc.  

Meta-Wing is better than that, and it attempts to create a ranking system based on how successful the ships/pilots etc have been in the data it gets from Listjuggler.
  • The ranking system isn't perfect.
  • Despite its flaws the ranking system is pretty good - it gets a lot more right than it gets wrong.

The ranking system is based on the 'Average Percentile' score for the thing you're looking at, which Meta-Wing calculates by looking at all the times it sees that thing in the data and where it finished, eg. if I list got played once and won the tournament it played in it would get a 100% percentile score.  If it finished second it scores lower, lose a semi-final it score lower again, etc etc.  Meta-Wing grabs all the results it can find in the Listjuggler data and gives you the average performance for that ship/pilot etc.

But Meta-Wing isn't finished there, because it also understands that not all tournaments are the same.  Meta-Wing takes the 'Average Percentile' score as it's base and then adjusts it for two factors - how many samples there are in the data, and how large the tournaments are that the data is coming from.  

This isn't a perfect system but what it's trying to do is downweight results that are either random outliers, or are only doing well in small tournaments where it's assumed the quality of opposition is a bit lower.  These are both things I tend to agree with.

The result of this algorithm is output as a 'Magic' bar, and the ranking system is based on that.  It's not a ranking of popularity, it's a ranking of success that assumes repeated success in big tournaments is a better indicator of quality than infrequent success in small tournaments.  Meta-Wing's 'Magic' ranking system rewards things that are being repeatedly successful or successful in big events, and suppresses results from things where it's less certain they're representative of true strength.

Meta-Wing rankings reward success, not just turning up
Both these 'Magic' factors are optional and can be turned off, by the way, but they are applied by default.  I find it can actually be quite informative to turn them off and see what pops to the top of the rankings - you tend to find some unusual squads that have been doing well, just in small events or haven't been picked up yet by many players, so it can be a good source of off the wall ideas!

In the end, I find the best way to approach the Meta-Wing ranking system is that it's very helpful but to try not to get too caught up in precisely where everything falls in the rankings.  So you find that your favourite pilot is ranked #4 instead of #1... well before you begin an angry rant take a deep breath and remember where that ranking comes from:
The average results of that pilot.  Of some of the squad lists.  From some of the tournaments.  Played by some players.  Some of whom are good and some less good.  That a computer then either promoted or downvoted based on how confident it was that the samples were relevant.
Meta-Wing ranks things.  It's trying to be helpful.  Usually it IS helpful.  Don't let that get in the way of you using the tool to help you, because Meta-Wing does so SO much more than rank things.

Using Meta-Wing

The front page of Meta-Wing does a great job of introducing the sort of things the website is good for, but I'm going to lead you through a specific example of the sort of things I've used it for, and things that Meta-Wing has taught me along the way.

I blogged recently about my 'Nettling Imps' squad (Quickdraw, Backdraft, Vessery) that I've been flying to good success and which a few players picked up as the basis of their squad.  Well, that squad was entirely born from my use of Meta-Wing.

Let's step back in time a little and set the date filters on Meta-Wing to about what they would have been when I was starting to make this squad at the beginning of May, looking at data from Wave 10 (when Lightweight Frame came out).  

At this time I'd pretty much never played with my two TIE/SF and the perception was that the Imperial fleet had nothing left after the Palpatine and TIE/x7 nerfs had shattered it's best weapons.  However I'd just written a piece about the ARC-170 that I was pleased with (using Meta-Wing to find loads of examples of how they were being used) and wondered if I could repeat it from a standing start with the TIE/SF.

So I turned to Meta-Wing... and found to my surprise that the TIE/SF was actually the highest-ranked Imperial ship in this timeframe (after the Lambda Shuttle and TIE Defender, which had both recently been nerfed).  The TIE/SF was being played about as often as the ARC-170 that I liked, and was actually winning significantly more (34.4% percentile vs 30.4%).  So I clicked the TIE/SF link to drill further...

The dedicated TIE/SF page then had all the information right at my fingertips to explore what was making the TIE/SF tick for players, which was handy because I was coming into this one pretty much cold...

This told me a few things.  Firstly that like 95% of all TIE/SFs being played were one of the two named pilots.  Secondly it told me that Quickdraw was proving to be the stronger of the two, although there wasn't a whole lot in it.  The final take-away I had from here was that the Omega Specialist was hardly being played (in 11 squadrons vs 140 Quickdraws) but when it was getting used it had actually done better than Backdraft in terms of Average Percentile - I couldn't be sure that this was a true reflection (you can see how much Meta-Wing downweighted the 'Magic' score for the Omega Specialist for only having 11 results) but as the Zeta Specalist was SO much worse maybe the EPT slot was the valuable part of the ship, not Backdraft's pilot ability?

Meta-Wing also showed me all the various squads it could find that had used the TIE/SF...

I had two major take-aways from this.  Firstly although the TIE/SF was winning tournaments (indicated by the fact that the 'Cut' score was 1 - it finished first in the Top-8 cut) those tournaments were actually pretty small most of the time.  That's not a good sign so I scrolled down in search of larger events and what I started to find was that in the bigger events there was a lot of evidence that Quickdraw and Colonel Vessery were making for a really powerful combination.

Quickdraw/Vessery/Palp, Quickdraw/Vessery/O.Leader, Quickdraw/Vessery/Sabacc... it seemed like anything you put with Quickdraw and Vessery could be pretty strong.  The second take-away from the data, though, was that in each case the cost of fielding the Quickdraw/Vessery combination was forcing players to compromise a little and bring in a cheaper third ship.

I'd seen enough to start formulating a plan, but Meta-Wing had a little bit more to share to help me out because it also listed & ranked all the upgrades that had been used on the TIE/SF.
  • The average percentile result for the TIE/SF is 34.4%.  Three quarters of all TIE/SF equipped Fire Control System and the percentile score for those is 37.8%.  The percentile for the quarter of ships that didn't take FCS must be AWFUL for it to pull the overall average down to 34.4%!  I need to equip FCS, this tells me, it also me that an FCS-equipped TIE/SF should have been ranked a bit higher in the ship rankings.  There could be an upside here...   
  • Expertise is played by less than 20% of pilots but has a great Average Percentile result.  Having the action economy of free Target Lock (from FCS) and free Focused attacks (from Expertise) was key to success, as was saving your Focus action to defend yourself!
  • Supposedly important cards like Lightweight Frame and Pattern Analyser might be luxuries - they weren't much different from the average 34.4% Average Percentile score.

This was all really good information that helped to set me on the path to my finished list, but checking out the bottom half of the Upgrades list told me more again.
  • Homing Missile/Guidance Chips was played rarely but had been the single most successful TIE/SF archetype when it was played!  This had gone almost entirely under the radar, and still largely has (although you may be about to see a lot of Cruise Missile/Guidance Chips combinations for 2pts less).  Remember how higher up I said that Meta-Wing can miss rogue strategies that work, just because it doesn't see them very often?  Well, this is what that might look like.
  • Sensor Cluster didn't add anything much.
  • Primed Thrusters were as good, or better, than Pattern Analyser.

Knowledge is power.  An hour earlier I'd known nothing about TIE/SFs but now I knew an awful lot more, thanks entirely to Meta-Wing.  

Incomplete data?  Yes.  

Just some tournaments not all?  Yes.  

Just some of the lists not all of the lists?  Yes.  

But despite all those limitations Meta-Wing had still provided me with information that DRAMATICALLY cut down the dead ends and trial and error needed to get to a working squad.

Meta-Wing had told me that TIE/SF are actually good, especially Quickdraw and especially when paired with Colonel Vessery.  This might sound obvious now but at the start of May I was jumping onto a bandwagon many people didn't even realise had started rolling yet.  It had told me the pilots to use and the important upgrades (being able to modify my attack dice without actions).

But while all that was just using it to copy other people's good work (not necessarily a bad thing) Meta-Wing had also suggested to me that there was a weakness that I could look to fill, if only I could find a way to fit in a third tough ship instead of having to scrimp and save points making my TIE/SF scary.

If I only I could find a way to modify my attack dice, but at a low cost... 

And Nettling Imps was born.  I had the finished first cut of my squad list on my screen within an hour of opening Meta-Wing, ready to head to the table and try it out.  Without this tool the amout of work I would have needed to do in Listjuggler to get the same result would likely have taken days, if I'd ever bothered to do it at all.

Thanks Meta-Wing. Thanks, Haslo!

Off You Go!

Right.  Go and play.  Meta-Wing knows what's been working from Wave 11, it's just waiting for you to find out.  Don't be hanging around here.

Also, if you're a TO then it would amazing if you could get into the routine of posting your results on Listjuggler.  The world is watching!

*** Bonus Section: Advanced Meta-Wing ***

Meta-Wing is a little gem in its own right, but I've found that for those who know their way around Excel there can be some more to unlock.  I won't go into too much depth here because I think a lot of the fun is in working out what you want to know how and how to find it out, but here are some quick examples of things I've done in the past by downloading Meta-Wing data and cleaning it up/playing with it.

Upgrade Flexibility (click to embiggen)

Historical Ship Rankings (click to embiggen)

The Changing Metagame - Top-10 Pilots By Month (click to embiggen)

Pilot Differentials - Which Pilots OutperformTheir Ship? (click to embiggen)

Enjoy!  And remember to thank Haslo if you see them around...