News Shop
Events Forums

Tournament Theory Crafting


I apparently have not have enough dabbling in the theoretical, so I’m going to take it up a level, and start dissecting tournament designs.

Dimensions of Tournament Design

Match Wins: Best of N
Schedule: Single-day, Weekly
Pool Play: Swiss, Round Robin, None
Elimination Play: Single Elim, Double Elim
Cast: Full Cast, Limited Cast
Who Counterpicks: Winner, Loser, Winner’s Choice, Loser’s Choice
Counterpick Options: Stable, Available Cast, Random
Timer: Fast, Medium, Slow, None
Who Can Play: Open, Invitation

Existing Tournaments, Dissected


Match Wins: Best of 5/Best of 7
Schedule: Weekly
Pool Play: Round Robin
Elimination Play: Single Elim
Cast: Full Cast
Who Counterpicks: Loser
Counterpick Options: Available Cast
Timer: Medium
Who Can Play: Open


Match Wins: Best of 7
Schedule: Weekly
Pool Play: None
Elimination Play: Double Elim
Cast: Limited Cast
Who Counterpicks: Loser
Counterpick Options: Available Cast
Timer: Medium
Who Can Play: Open

Lum’s Lucky Lottery

Match Wins: Best of 7
Schedule: Weekly
Pool Play: None
Elimination Play: Double Elim
Cast: Full Cast
Who Counterpicks: Loser’s Choice
Counterpick Options: Random
Timer: Medium
Who Can Play: Open


Match Wins: Best of 3
Schedule: Single-day
Pool Play: None
Elimination Play: Double Elim
Cast: Full Cast
Who Counterpicks: Loser
Counterpick Options: Available Cast
Timer: Fast
Who Can Play: Open

Summer Smash

Match Wins: Best of 7
Schedule: Weekly
Pool Play: None
Elimination Play: Double Elim
Cast: Full Cast
Who Counterpicks: Loser
Counterpick Options: Available Cast
Timer: Medium
Who Can Play: Open


Of the tournaments that have been run recently, it appears that each deviates from a “standard” format in one (or maybe 2 related) dimensions.


  1. Historically, are there other tournaments that should be dissected this way (I suspect yes, but I wasn’t around, so I don’t know the details).
  2. Are there any dimensions that have obviously bad options in them (like, Winner counterpicks, which likely makes the match even more lopsided?)
  3. Are there dimensions of tournaments that I’ve missed, that either have been experimented with before, or that should be?
  4. Can we come up with any more interesting tournament variations by selecting different subsets of these options?

[Tournament] Lum's Long Odds - Ongoing!

The best tournament

Match Wins: Best of 4
Schedule: Monthly
Pool Play: No kids in the pool
Elimination Play: Yes
Cast: Only characters playable are in this list. If you pick a character not on this list you are disqualified. Argagarg, Bal-Bas-Beta, DeGrey, Geiger, Gloria, Grave, Gwen, Jaina, Lum, Menelker, Midori, Onimaru, Quince, Rook, Setsuki, Troq, Valerie, Vendetta, Zane
Who Counterpicks: Loser’s Choice
Counterpick Options: Unstable
Timer: None
Who Can Play: Everyone except @Copper8642


Persephone’s Pick Your Poison was a very interesting and unusual format. Essentially there was an approved list of match-ups (in this case, a lopsided list and an even list). Players took turns picking which MU they wanted, and then their opponent picked which side of it they would play. So each person picked two lopsided MUs where their opponent would decide to play the side they felt was most advantaged, and then you both picked one even MU, and then one player picked a last even MU which would be the Ace Match, if the set went that long.

There are team events as well, though those can be trickier to organize. I think they all had ‘team’ in the name somewhere. This would be either full cast standard CP, or would involve a draft of characters first.

Yomi draft and reverse draft is another way of building a stable of characters to play.

For events that use stables “losing character is eliminated” leads to the loser of the 1st match receiving a disadvantage, compared to “winning character is eliminated.” For that reason I think ‘stable’ tournaments have tended to move to the Conquest fomat (winning character can no longer be picked).

Persephone’s Pick Your Poison involved drafting MUs, rather than individual characters.

Some of the events that Zqxx has put on had extremely limited casts - only Rook and BBB, for example.

There have been a couple of double-blind events. Instead of switching to standard CP after the results of the first match, both players pick from their legal list of characters in double-blind fashion again after each match is complete. There are pros and cons to this approach. Mathematically, it should lead to less lopsided match-ups being selected, I think? LK4O4 wrote a ton of material about this that went over my head on the old forums. However, you can have situations where players pick into a bad match-up again and again, which is a deal-breaker for some.

One format that Sirlin explained to me at FSX this year was interesting and may be worth a look, but I can’t remember the exact details. It’s essentially a ‘stable’ format (you pick three characters from the cast, and you can only play characters from that limited pool during the event), but instead of being double-blind into standard CP, each match both players randomly roll for their character. Each game involves randomly re-rolling from within each player’s stable, but there was some additional rule about characters being locked out that I can’t remember. It may have been, like, each pair of characters that just played are frozen out for one game, or something? I feel like it was more involved than that…

The stated advantages of the format are that it makes counterpicking largely ineffective, because you will now random roll into lopsided games occasionally, rather than the winner of a match almost always being put into their worst MU in standard CP format. There was an argument that this would then also reduce the degree to which Troq/Geiger/Zane would be picked, since they are often picked as CPs. I’m not sure I would agree with that 2nd argument, but it would still be an interesting format to try.

The difficulty of the format, especially played in the Steam client, is that you don’t have a way to randomly roll for characters in a fair way. So there would need to be a TO involved in each match (or ahead of time in the scheduling conversation) for this format to work.

@Sirlin would you mind outlining the format rules?


Who’s the missing character it’s driving me crazy! :’(


Took me a while, but, :persephone:.


breathing returns to normal


Uh… which tournament was this??? :thinking: Sounds like a one-way trip to endless :bbb: mirrors. :laughing:

For the curious, the tournaments I’ve run have been:

  1. Conquest-style where everyone’s stable is :midori:, :rook:, and :troq:. (Oh, the In-Humanity! [Single Day Tournament])
  2. Standard CP style where only non-human characters were allowed (Rook, Arg, Lum, BBB, Perse, Gwen, Ven, and Troq). (Oh, the In-Humanity! [Single Day Tournament])
  3. Conquest-style where each person chose a stable by picking one character from each pair of Yomi’s siblings (Grave or Jaina, Gwen or Gloria, Menelker or Midori). (Sibling Rivalries [Single Day Tournament])
  4. Double blind random rolls for every single game (SUPER SERIOUS SATURDAY [Single Day Tourney])
  5. A conquest style tournament with pre-made stables that roughly emulated the canonical alliances of Yomi lore (Fractured Factions [Single Day Tourney]).

All have been single day, Bo5, and fast timer.


Oh woops! Sorry for misattributing! It was called ‘A Perfect Idea That Can’t Go Wrong’ or something.


I believe that one was done by @shax


I was thinking about tournament formats that would encourage playing a broader range of characters, and came up with this. Please shoot holes in it (or I might be forced to run it).

Lum’s Long Odds

Match Wins: 4 Games per match
Schedule: Weekly
Pool Play: Swiss
Elimination Play: None
Cast: Full Cast
Who Counterpicks: Loser
Counterpick Options: Available Cast
Timer: Medium
Who Can Play: Open

Unlike most tournaments, for this tournament, the victor will be determined by points. In particular, you earn points for each win based on the matchup rating determined for the matchup (this would be based on the historical MU chart). For instance, if you win a 7-3 match, you get 3 points, but if you win a 3-7 match, you get 7 points. The goal, then, is to pick/counterpick your way into matchups where you think you have an edge over the historical ratings.

Because it’s points based, it’s important that each player play the same number of games (hence the predetermined 4-games-per-set, rather than a Bo7).

EDIT for flavor


I really like the idea of creating tournaments that reward breadth of character knowledge/encourages people to play different characters! But I would have two reservations about this format specifically:

  1. You mention that since it’s points-based, it’s important to play a set number of games. But really, that’s not quite true. Assuming that no Yomi MU is considered worse than 7-3, then if my opponent has more than 7 points more than I do going into game 4, then game 4 is pointless. I cannot possibly make up the difference. Even more problematic, though, consider this scenario:

Game 1: :lum: vs :argagarg:. :lum: wins and gains 7 points. Player 2 has the chance to CP but decides to stay as :argagarg: for whatever reason.
Game 2: :lum: vs :argagarg:. :lum: wins again and gains 7 points, for a total of 14 points

At this point, the :lum: player has already won the set. Even if player 2 switched to a character with a 7-3 MU against :lum: and wins, gaining 7 points, all player 1 has to do is switch into a character whose MU against player 2’s CP is less than 7-3.

  1. I feel like this format is very vulnerable to Matthews effects in the sense that, if I’m falling behind in points, I need to “CP” into bad MUs, but that just increases my chance of losing the next game, setting me even further behind in points. Seems like it basically punishes losing players by making them play even harder MUs.

  2. Personally, I agree with the axiom that “a win should be a win.” That is to say, the quality of the victory shouldn’t matter; all that matters is that I won the game. The idea that I could win as many games as my opponent but still be the loser feels kind of bad.

  3. Arguably this format does not even incentivize playing a broad range of characters per se. Rather, it rewards playing characters who are generally considered bad. After all, if you can play a character with a poor MU spread well, then your wins are worth more and your losses hurt less.


Ah, I think I wasn’t clear. I was proposing that the position in the tournament be evaluated by total points over all matches, rather than each match being evaluated by points. So matches of 4 games would really just be a way of testing whether you beat the historical averages better than your opponent in an effort to score points towards the overall tournament win.

I think that clarification maybe helps address your points 1, 2 and 4. It doesn’t really help with point 3, though, that seems like a legitimate concern.

For more detail on point 4, I whipped up a quick table

That gives you the amount of points you would expect to win given your personal assessment of a MU and the historical value of that same matchup. You can expect to get more points the more you can bend the MU from the historical average, but if you can’t, you actually get fewer points on the ends of the spectrum than you get for more even matches. So, I think this might encourage people to counterpick into some non-typical matchups, because you don’t necessarily want to be pushing towards as lopsided of games as possible (or rather, you want them lopsided because you know the MU better than your opponent).


I just made another table of the relative points gained for each amount of bending the matchup:

I was surprised at the nice banding, but I probably shouldn’t have been. Anyhow, given that even matchups get you between 2 and 2.5 pts per game, and each .5 point that you bend the matchup gets you .5 points more than your opponent on average… a 4-game matches would mean that if you are +0.5 over your opponent, you’d expect to get about 2 points more in that match than against an opponent you were even with (or, about 1 more game). Over the course of an 8-week series, someone that is +0.5 over the field would expect to get about 16 more points over the course of the tournament… I’m not sure if that’s enough to differentiate high-performance players from low, but I think we might just need to play the tournament out to see.


You have given me the sads, Jonny :frowning:


It’s pretty interesting! You’d probably want to take an image of the match-up spreadsheet before the event, so that you avoid results of this and any other ongoing tournament altering the points payout. How would Swiss format work though? You’d have to pair people up based on their points, right? Instead of set wins? Would you rematch #1 and #2 if they’re the same across multiple weeks somehow?


Yeah, I’d do the Swiss matchup based on points, rather than set wins. I would probably stick w/ the standard Swiss setup of tweaking the matches so that players don’t play each other more than once. That said, I’d have to do it all manually (or code something up), since Challonge doesn’t have that level of flexibility for their Swiss implementation.

And yes, I’d capture the historical matchup chart (and round to the nearest .5, probably), and then post that info in the tournament intro thread.


Is there a limit on the number of wins you can assign within a challonge match? You could just put the pt values straight in there, if there isn’t, rather than the game wins. Not sure if that helps or not…


Challonge lets you configure pts per match win and pts per game win, but you can’t base it on pts scored in a game (which would be the natural thing to do). So, I’d have to add 20ish matches (or more, to cover half-points), and it appears they have a limit on how many you can have… :frowning:


Oh! Well, that does make more sense :laughing:.

Another concern comes to mind, however: It seems to me that there is a decently high chance that, towards the later rounds of the tournament, certain players will already be too far behind in points to ever be able to make up the difference. This is a pitfall in any Swiss tournament, but it might rear its head faster here due to the design of the tournament allowing for different rates of point accumulation.

Now to be fair I don’t really know how one would do the math to prove or disprove this idea :stuck_out_tongue:. That is to say, I’m not sure if I’m correct or if I’m being misled by the idea of points and that this tournament is no more prone to lame-duck effects than a normal Swiss. But if true, it could lead to a tournament with a high percentage of lame-duck players, which wouldn’t be a lot of fun.

Additionally, if a lame-duck player decides to drop out of the tournament, how would we assign points to the players who were supposed to play them in future rounds?


Yeah, that’s definitely a concern. One possible out is that in general, people will be playing against those w/ similar skills, and thus should have the ability to regain points… But really, it’s kinda the same problem as with IYL, as far as lame-duck dropouts goes. I think I’d probably lean towards awarding a fixed number of points in the case of a forfeit (probably 10, in the case of 4-game matches, which would be the expected value of 4 5-5 games).

This might just be a case of “Let’s try it once, and hope it doesn’t suck too badly”?


If winning a 7-3 matchup gives you 3 points and winning a 5-5 matchup gives you 5 points, the EV of a game in a 5-5 matchup that the player is 50% to win is 0.4 points higher than the EV of a game in a 7-3 matchup that the player is 70% to win.