Smash 3DS Smash 4 3DS Tournaments and Swiss Systems

[Amazing Ampharos]

With the release of smash 3DS and the leak and such, I've been thinking a lot about competitive rulesets for smash 3ds and what the unique advantages of the format and the particulars of what we've seen mean for tournament administration. Thinkaman raised a very good point to me about this, and I think it's worth discussing.

Double elimination brackets have been the standard tournament structure for smash in the past, but for 3DS, they're a pretty bad format. The thing is that a 3DS tournament has "infinite set-ups" and an entire round can be played at once. Let's assume each round takes a half an hour to resolve (triple timeout plus six minutes of time to pick stuff in-between games, find your opponent, etc.). A 128 man double elimination bracket has 14 or 15 rounds which would take up to 7.5 hours (highly volatile) and contain a total of 254 or 255 matches. An 8 round swiss tournament (a no elimination format commonly used in games such as chess) would take likely precisely 4 hours and contain a total of 512 matches while having exactly identical predictive power over who the best player is. It's pretty obvious that the swiss format offers a radically superior value to its players while taking way less time and a predictable amount of time at that; it seems like magic, but infinite set-ups is really powerful.

I'm pretty familiar with swiss style tournaments from playing chess in high school, and while it is not a perfect format, it's pretty great and to me would be insane not to use for 3DS given that on average it's significantly faster, gives more information about the skill of the players, and offers a much greater value to the players entering particularly those who are of lower skill levels and are often eliminated early in double elimination brackets. I could go into a lot more detail about how swiss works and how things like swiss style tiebreakers work work practically for smash, but before going into all that, I want to know. What do you guys think about this on a basic level?

 

[The Mayor]

I used to be in the TCG Community, and i have to say, out of all tournament Styles used, Swiss is by far the greatest. It gives everyone the chance to play a full tournament and Has an amazing tiebreaker system. I agree that Swiss would be the best for 3ds and greatly support it, i plan to run smash tournaments soon and i was probably gonna do swiss anyway

 

[UpsilonFox]

Brilliant idea, as this is the first Smash that can really take advantage of a swiss tournament set up it is something that TOs really should consider if they are to run 3DS tournaments.
I really like the swiss format from my experience playing swiss Yu-Gi-Oh tournaments, really great tourny format that gives a result that more accurately reflects the distribution of skill among the competitors.

Now if only enough people in Australia played Smash Bros to have tournaments like in the US, Europe and Japan.

 

[BADGRAPHICS]

I was talking about this in another thread a few months ago. I play competitive Riichi Mahjong, and uses the same format.

As well as the obvious logistical advantages, it also offers players the ability to come back from previous poor matches, where double elimination may occasionally eliminate the best overall competitor due to one error in judgement.

 

[Malcolm Belmont]

I don't know much about competitive tournament but from the way you described it..it seems like a great idea and would make it shorter but easier to manage at the same time.

 

[Raijinken]

The advantage of setups is the greatest point for the 3DS version. Though, if the wireless interference becomes apparent with large numbers of systems in an area, it may have to be played in smaller groups of setups to reduce interference, thus reducing that advantage.

 

[SmasherP83]

I think a few of us members were talking about using the swiss system in another thread a year ago. Overall I think it's an excellent idea!

 

[infomon]

I'm not familiar with Swiss-style tournaments. After skimming through http://en.wikipedia.org/wiki/Swiss_tournament it seems like there are several variants.

Can you propose exactly how you envision a Swiss-style Smash tournament would work? Thanks!

 

[SmasherP83]

I'm not familiar with Swiss-style tournaments. After skimming through http://en.wikipedia.org/wiki/Swiss_tournament it seems like there are several variants.

Can you propose exactly how you envision a Swiss-style Smash tournament would work? Thanks!

Not an expert on it yet but I will say thanks to everyone owning a 3DS version of Smash it'll be a lot easier doing swiss tournaments and everyone has a chance to play!

 

[Amazing Ampharos]

I'm not familiar with Swiss-style tournaments. After skimming through http://en.wikipedia.org/wiki/Swiss_tournament it seems like there are several variants.

Can you propose exactly how you envision a Swiss-style Smash tournament would work? Thanks!

Sure! This is going to be pretty heavy stuff though!

Round one of a swiss is paired just like a traditional bracket for the most part with seeding and such (commonly for 100 players seed 1 plays seed 51 instead of seed 100 such that seed 1 will get on average better tiebreakers than seed 50 if there are no upsets, but for smash, I think seed 1 playing seed 100 is best). After this, round two is paired with winners playing other winners while losers play other losers. Imagine it being made just like round one's pairings within each score group except initial seeding is a low ranking factor on the tiebreakers that determine seeding for this round. This iterates each round with an absolute rule that no one will ever play the same opponent twice and occasional need to pair people with different scores when you end up with odd numbered groups (there are a few ways to decide who has to be the odd duck here, probably best to depend on software to fairly randomly make this decision while making sure the current best tiebreak members of each group don't get picked if possible to avoid).

Tiebreaks are a big thing in swiss both for later round pairings and for final results that are meaningful for all places (in a swiss, getting 17th versus 18th means something). I would ensure all rounds are reported with total game count like "2-0" versus "2-1" being documented. Then tiebreaks go like this:

-Total number of individual games won (more is better, rewards your losses being 1-2 instead of 0-2).
-Total number of individual games lost (fewer is better, rewards winning 2-0 instead of 2-1).
-Median-Buchholz System (average match score of opponents ignoring highest and lowest scoring opponent)
-Solkoff System (average match score of all opponents)
-Initial Seed

Note if these tiebreakers are being used at the very end of the event instead of to pair the next round, smash should use two modifications. One is that if the first two tiebreakers are tied for the rank of first place in the event, there should be a play-off game such that who you are paired with doesn't affect your ability to win overall even with equal direct performance though in the very unlikely event it's more than a two way tie we'd have to pull the top two players from here on overall tiebreaks since time will become a factor. Second is that the final tiebreak inverts such that lower seed gives you a higher rank at the last step; higher seeds generally have easier pairings throughout the tournament so it should be assumed that if a lower seed is so precisely tied in measurable performance with a higher seed that the lower seed performed slightly better (this fifth tiebreak is seldom very important in a deep swiss like smash would always run). Also, if there are byes due to an odd number of players in the overall event, a bye should be modeled as an opponent who loses every game 0-2, and byes should be granted to the seed 50 player if there are 99 entrants (byes are free wins but are generally bad for tiebreaks such that they would likely be upsetting to a top player to receive in an early round but probably welcome by a mid-range player) and in all cases bye is always paired with the lowest score group.

This is just how I would run it; it is very open to suggestions and improvements, and I still haven't found optimal software to pair this. Also, I didn't mention this before, but I would generally run most smash events as an 8-round swiss other than small locals (under 32 players) which I would run as 6 round and national scale events (more than 128 players) which I would run as 10 round. I hope this explained the structure, and I hope the particular decisions I made make sense for smash (notably since smash has access to some considerably less arbitrary tiebreakers than chess and the top few smash players tend to be way, way better than their next closest competitors, giving the weakest opponents to the strongest seeds should produce overall better pairings). If you have any particular questions about what I've laid out, feel free to ask.

 

[infomon]

I appreciate the advantages of the system: more matches for everyone, strong player-ranking information at the end (17th vs 18th place means something). But it also seems quite complicated, and I wonder how robust it is to tournament complications (e.g. someone goes AWOL and gets disqualified).

In your proposal, there is an advantage to winning a set 2-0 more than a 2-1? That's a change, and may affect counterpicking / character choice. Maybe that's ok, I'm just pointing it out -- it may not be appropriate for smash.

I assume TO software is available to help run this style of tournament?

 

[Nintendrone]

Someone give this man a medal; this is genius! 3DS wrecks the limiting factor of any tournament: the setups. The only issue is that I don't know of software capable of doing this format, but it couldn't be that hard.

 

[Amazing Ampharos]

There are a lot of programs out there, but most of what I've found isn't very robust and often sucks when it comes to custom tiebreakers or custom algorithms for pairings (rank 1 plays rank 51 seems hardcoded to every swiss application I've found... which really annoys me because that only makes sense if your tiebreaks primarily rely the other players and not individual match scores). Automating a swiss in general is easy with a variety of software, but getting the precise behavior I believe is optimal may take some work (but this is not a complex algorithm so scripting up something should probably be pretty easy; I'm no programmer, but even I may be able to make something workable). As per relying on individual match scores, that was my invention, but it's not really novel to smash given that every tournament I've entered with round robin pools (the good kind!) did the same thing and it's generally a lot better information than the scores of your opponents.

In terms of ease, I'd point this out. Swiss is generally harder on the TO since it's a less intuitive format from "bracket theory" standpoint. From a player's perspective swiss is about a million times easier. You always know when you have to play because you play in every round and the rounds take about the same amount of time. Wandering off should be a pretty much non-issue; since every player knows a new round starts every half an hour and that every player always has to play in every round, there's just no excuse not to be there so you just issue DQs super fast to people who don't show whereas with a bracket you get ahead in winners and are never really sure when you play again. The format seems intimidating until you try it once, and then you really wonder why you ever thought brackets made sense beyond set-up optimization.

 

[BADGRAPHICS]

It wouldn't be too difficult to put together a spreadsheet to manage these tournaments. It could be custom-tailored to meet the needs of the community, and easy enough to share. I could do it myself, but I can't guarantee I'd have the time.