Tournament Tiebreakers

Lounge
1

I have a scenario here, and I’m wondering what the general opinion is regarding ordering:

Consider a round-robin pod with 6 players.
Abe goes 4-1, losing only to Beth
Beth goes 3-2, losing to Carl and Frank
Carl goes 3-2, losing to Dave and Abe
Dave goes 3-2, losing to Beth and Abe
Earl goes 1-4, beating only Frank
Frank goes 1-4, beating only Beth

Clearly Abe is ranked first, and one of Beth, Carl, and Dave should be ranked second - I think there is general agreement that “matches won” is the primary sorting method, but comment if you disagree.

Which of Beth, Carl, and Dave should be ranked 2nd, though? Is it more meritorious to have beaten a (4-1) player, or to not have lost to a (1-4) player? Is there some other criteria that should be referenced? What about ranking Earl and Frank?

2

Typically I break ties first by head to head if available, and then quality of wins.

In a game with a “score”, going with points scored is the easy answer.

In a situation where there isn’t, though, I think “quality of wins” being computed by opponent win loss is acceptable, and preferable to quality of losses as a first tie breaker.

In other words, I think Beth is 2nd, Dave is third, Carl fourth, then Earl, then Frank last

3

If there is a point system in the game, total every scores. Else go by Strenght of Schedual.

4

In this case, full SOS is the same for everyone with the same record, since all 6 players played the same 5 other players.

5

In this case the amount of games played is too high for the amount of players.

6

What if these 6 just happened to play against 5 people with matching records, then? Do you think it should matter if Beth lost or won when she played an opponent who went 1-4?

7

Well, I think using MTG tournament tiebreaker system would be the best.

8

One method I’ve seen (I think it might have been for UDE events) was to use the Sum of the Squares of the rounds you lost as your tiebreaker. Obviously that’s kind of meaningless if you’re doing Round Robin, but for a more typical Swiss style of tournament it’s nice because it serves as an estimate of the quality of your opposition (if you lose early you have a low SS and have been fighting lower-ranked people all tournament), and it’s relatively easy to explain and calculate (high numbers are good, the math isn’t that hard, and it’s based entirely on your own performance rather than trying to track down the performance of all of your opponents). There might be situations where it doesn’t do a good job distinguishing between placements near the top of a bracket though, which is certainly a strike against using it as a tiebreak.

9

Well, I think for tournaments there need to be more than just win/lose score to get a tiebreaker. It need a MoV (Marge of Victory) to be able to break tie. In X-Wing, it is squad point destroyed. This also allow to set a timer and have game end due to time without being a simple draw.

In Codex, it would be base hp.
In Yomi, it would be character hp.

Basically, you add or substract (depending if you won or lost) the difference between both opponent to the max value.

Codex example: P1 base has 3 hp left, P2 base has 7 hp left when game end due to time, giving a difference of 4. Base hp max value is 20. So P1 has a MoV of 16 and P2 has a MoV of 24.

10

Quality of wins > Quality of losses, if you’re doing a round robin and need your tiebreaker to look that deep. So FrozenStorm won the thread with the first post.

For Swiss, my favorite Strength of Schedule breaker is Average Opponent Win %, but dropping both the highest and lowest value to remove the effect of randomly getting matched with the best player or the worst player in the tournament round 1. The end result is pretty close to Fry’s suggestion, but it tends to do a good job being very granular and actually useful as a tiebreak between placements at the top of the bracket.

If the game is played in sets (like, Best of 5), game win % is a barely acceptable tiebreaker. While yes, winning 3-0 is better than winning 3-2, both are winning the set, and that’s all that /should/ matter. Maybe somebody has a playstyle that relies on feeling out the opponent first, and those first 2 losses were just scouting data before completely crushing their opponent after the download was complete. It feels bad to punish that kind of player via tiebreakers. But, higher game win % is very highly correlated with higher match win %, so this can still squeak by and give a useful granular tiebreak if head to head or Strength of Schedule don’t cut it.

Margin of Victory tiebreakers are completely unacceptable. Average margin of victory is not even kind of correlated with match win%, and making it a tiebreaker completely changes the range of strong strategies for a game, because you’re adding a secondary goal. Now, you don’t just have to use a strategy that wins matches, you have to use a strategy that wins matches while maximizing your margin of victory. This is especially bad if the design space of the game you’re playing uses whatever you’re measuring as a resource.

In Yomi, Gloria, Jaina, and Gwen all use HP as a resource. From experience, I can say that all 3 of them tend to end matches on less hp than average than most of the rest of the cast. Because in normal Yomi, not caring about Margin of Victory Yomi, all 3 of them are perfectly happy dropping themselves to very low hp to help get a win.

While in Codex, cards like Dark Pact or Lich’s bargain let you sacrifice base hp for current advantage.

I would personally not participate in any Yomi/Codex tournament that planned to use a Margin of Victory tiebreaker.

2 Likes
11

@Wildhorn

The problem with using Margin of Victory as a tiebreaker is that it warps the way people will actually play the game. You don’t want to make more controlling strategies/characters “worse” because they use their hit points as a resource more readily than aggressive strategies.

Personally, I think Average Opponent Win % is a great way to break ties; however I don’t think it’s necessary- or even desirable- to throw out the high and low win %'s from each player’s data set. Those “outliers” still had an impact on the difficulty that player experienced in achieving the record that they did.

2 Likes
12

The thing with counting quality of wins more than quality of losses is that it rewards the less consisyent player.

If Beth beat Abe (who is a better player) and lost to Frank (who is a worse player), then her games had more variance than Carol’s games where she lost to the better player and beat the worse one.

To me, that implies that Carol is the better player, and Beth was relying more on luck (to her success and detriment)

1 Like
13

Counterpoint: What if the game in question is one without randomness? Say Chess for example.
Beth beat a stronger opponent and lost to a worse one. It’s hard to say where that places her, but presumably she didn’t win by luck here.
Carol lost to a stronger opponent and beat a worse one. She is clearly worse than the stronger one and better than the worse one.
How does Carol compare to Beth?
I would think Beth is better, because she beat a stronger opponent.

14

I was actually thinking specifically of chess. Perhaps Beth took some risky unorthodox lines, while Carol played “standard.” While that might be a good strategy against Abe, pursuing it against Frank was a mistake?

15

Perhaps, but maybe that’s something she has to learn. However, simply in terms of ranking, I think Beth is better because she beat a stronger player.

16

Sirlin did a whole podcast on tournament formats and there was a large section on tiebreakers.

The conclusion was that ideally a match win is a match win, regardless, but given that we need some kind of tie breaker, margin of victory (if match = best 2 out of 3 games, not within each game) is the best. As mentioned earlier, you really don’t want to do anything to change the incentives in game to anything other than “winning a game is winning a game.” You really don’t want some situation where winning by being aggressive is worth more than winning by being conservative.