Special: SOS UKGE

Special: SOS UKGE 2019

DISCLAIMER: 

This article is an examination of how the finals of UKGE would have looked if they had cut to top 2 using a Strength of Schedule ("SOS" from now on) instead of GW's idiotic tiebreaker system ("GWITS").  It is not meant as a condemnation of any player.  It is not meant to imply that any player did not deserve their spot in the tournament.  It is only meant to point out the failures of GW's system.

The best solution to this would be to change the final cut to include all players with perfect match records.  There were six of these at UKGE, so four players who never lost a match didn't even get a chance to play for 1st.  The two who did get to play were chosen by GWITS - a system that is deeply flawed in that it rewards players for playing against opponents who are much less successful than they are.  SOS, on the other hand, rewards players for playing against stronger opponents, and is widely used in most competitive game/sport formats.

Once again, big congratulations to all the undefeated players for kicking ass in a very large tournament, you should have all been included in the final cut!  However, with a top-2 cut, this isn't going to happen most of the time.  If GW won't fix their top-2 cut (due to time restrictions, a reasonable objection), they really should use SOS.

On the upside, it looks like GW is experimenting with some different tournament formats (including a two-day format that has a larger final cut and more rounds).  Hopefully, this will result in fewer inconsistencies and injustices due to their truly terrible tiebreak and matchmaking systems. 

(For a detailed discussion of why SOS is a better system than GWITS, please see the discussion here.)

The GWITS final round standings:

According to the GWITS, the players were ranked thusly after the final round (player names will not be used, in an attempt to reduce the perception that I am slamming on any given person).

1. Ylthari's Guardians A
2. Ylthari's Guardians B
3. Stormsire's Cursebreakers A
4. Mollog's Mob 
5. Stormsire's Cursebreakers B
6. Godsworn Hunt (what!?)

Tiebreakers for each player - the factors that determined who among the 12 point players would be in 1st-6th place after swiss - are listed below, in this format: Warband - Game Losses - Total Glory Differential.

Warband	Losses	Glory Diff
Guardians A	1	84
Guardians B	1	83
Cursebreakers A	1	82
Mollog	1	72
Cursebreakers B	1	49
Godsworn	2	not calculated

At the end of Round 4, Guardians A and Guardians B played in the final.

A few things of note:  1st through 5th place were decided by glory differential only, and 1st-3rd were separated by only 2 glory.  3rd place missed the cut for the finals by only a single point of glory.  While I can't speak for other players, this would be extremely frustrating for me.

Secondly, while not noted on this chart, it is worth pointing out that of 109 total players, 3 dropped after first round losses.  One might assume that these players would have then finished 107th, 108th, and 109th.  Not so.  Because GW has decided that the first tie breaker is games lost, those players who stayed in and lost all their matches (0-4) finished lower than the 0-1 drop players, because they lost more games.  Seven players finished below the 0-1 drop players.  One of the 0-1 drop players also recorded a loss first round, but recorded zero game losses, so that player finished highest of all the 0-win players.  That sucks out loud.

The SOS final round standings:

If the tournament had used SOS instead of GWITS, the tiebreakers would have been:

1. Opponents' match record percentages
2. Player's game record percentage
3. Opponents' game record percentages

A full explanation of the DCI's SOS system - used by Magic the Gathering, can be found here.

In that case, the tiebreakers for each player would have looked more like this:

Warband	OMW%	PGW%	OGW%
Guardians A	0.7075	not calculated	not calculated
Guardians B	0.5825	0.89	0.5025
Cursebreakers A	0.6540	0.89	not calculated
Mollog	0.5200	not calculated	not calculated
Cursebreakers B	0.5825	0.89	0.5725
Godsworn	0.6450	0.80	not calculated

This would have resulted in the following standings after swiss:

1. Guardians A
2. Cursebreakers A
3. Godsworn (WHAT!?)
4. Cursebreakers B
5. Guardians B
6. Mollog

In this case, the final would have been between Guardians A and Cursebreakers A.

There's some good stuff to be gleaned from here.  First, the Guardians A player would have been first after swiss either way!  That's good, it shows that even GW's poorly designed system can end up rewarding players for beating more difficult opponents - sometimes.  If you've seen my other articles on the subject, you'll note that this wasn't the case at LVO or Adepticon.  Additionally, the 2nd place player would have changed, resulting in a different final match.  Finally, some of the other top 6 got jumbled around.  While this wouldn't have changed the outcome of the final match, it might have mattered to some folks.

Second, Godsworn did even better in SOS than they did in GWITS.  I don't even know what to say about that, except I'm fucking impressed.  Good job, man.

Coda: The Inevitable Counterargument

Predictably, after reading (or not reading) this article, someone will definitely make an argument something along the lines of "If SOS had been used during the whole tournament, it's impossible to predict how things would have gone, because everything would have been different."  Let's address that.

First, SOS is a tiebreaker, not a matchmaking system.  It's entirely possible to use GW's (broken, stupid) matchmaking system and still use SOS as tiebreakers.  So in fact, the matches could have been exactly the same.

Second, even if the matches wouldn't have been the same, there's value in analyzing data through a different lens, even if it can't account for all variables.  Economists use this technique all the time; it's called controlling for a variable.  You can find a detailed explanation of how this works, written by someone much smarter than me, in the works of Steven Levitt.

However, you may want a simpler, less whole-book-long look at how controlling for a variable works.  For that, let's take a look at one of my favorite sayings: "If my aunt had balls, she'd be my uncle."  I used to say this a lot, mostly to illustrate that a particular argument was pointless because it was examining a hypothetical in a way that couldn't account for all the possible outcomes.  In other words, I was making the same argument that I illustrated above.  Because I was young, and kind of dumb.

Since then, I've grown up in a lot of ways, and I look at things much differently.  Most significantly, I realized that just because my aunts' possession or lack of testicles doesn't necessarily determine their genders.  In studies from 2016 and 2011 respectively, it is estimated that 0.6% of the population identifies as trans, and 0.4% identify as non-binary.  So, even if my aunt did have balls, there's still only a 99% probability that she would identify as being male, and therefore, my uncle.  (And I would, of course, respect that person's preference in naming conventions, because I'm not a piece of shit).  I have a lot of aunts, so if they all had balls, there would be a pretty good chance that at least one of them would still be my aunt.  But most of them wouldn't.

So now, I think about that saying very differently.  Now, I think: "If my aunt had balls, she'd probably be my uncle."  You can apply the same thinking to the SOS/GWITS conversion.  Certainly there is a mode of thinking that dictates that changing to SOS would change every thing about the tournament, and the results would be impossible to predict.  But looking at a controlled-for variable (+/- balls) allows us to view a probable outcome (uncle vs. aunt). 

Whereas if we simply decide that a change to a single variable will automatically trigger an Ashton-Kutcher sized ripple in probability every time and result in a totally unpredictable outcome, you've learned nothing, examined nothing, and taken the easy way out.  You have no idea what to call your parent's sibling.  You have brought shame to your whole family.  Meiwaku.