Swiss system - unfair for the best players?

Cheers from Norway!
Me and two friends are responsible for organizing the monthly tournaments in Oslo Pinball Club, along with some standalone events. One of these is a christmas matchplay tournament, which we had last year and again this year,3rd November. Now, we have some disagreement us organizers in between when it comes to the format, specifically swiss system.

We have X rounds of 4player groups matchplay, first round ifpa slaughter seeding, and after that its swiss, with points 7-5-3-1. One of the others are very dissatisfied with swiss because he feels it is unfair for the good players, because if they get to the top group with the best players and hardest competion, and comes last he only gets one point, whereas the winner a group or two below will still get 4 points. So he feels you gets punished for playing well.

I for myself dont really think its unfair, although I see his point, but over 5-6 rounds it evens out and the top 8 or whatever amount of players who advance, is in most cases the ones who have played best and most consistent. Besides, its even more welcome for the mid-levek players who might have a better chance of advancing if the play well too.

He had one suggestion, to weight the points you get in the top groups. I see his point, but dont really know how this will affect the competition,and does not the point of swiss go away then?

I have understood swiss is mostly a european thing. So i wonder, what do you players and organizers in the US and elsewhere think about this format? Is it unfair, why/why not?

Edit: Must add, ive been to several matchplay tournaments in sweden,austria etc and have heard no complaints publically from the top players there at least.

Thanks! :slight_smile:
Morten

1 Like

Swiss works better in 2-player matches, with no one able to complain it’s unfair.

Ask the ‘top player’ if he’d be willing to take fewer points in an earlier round to play in an easier group later. The answer is ‘no’. If it isn’t, I’m sure the player can contrive to lose his earlier matches and try to come back at the end of the tournament; after they try that a few times they’ll take their points where they can get them.

Note that you need to play ‘enough’ rounds for Swiss to work, where ‘enough’ such that the players at the top of the standings aren’t just the single-elim winner. For 2-player matches this is 2-3 rounds more than a single elim bracket at least. (Some tournaments use 2-3 times that many rounds…)

I was just made aware that Pinburgh actually is swiss system?
points 3,2,1,0 for each game, and between each session the groups are shuffled so you meet players with the same win/loss (points) ratio, If I am not wrong.

I’ve always understood “swiss” to be a qualifying round consisting of head to head only, where you play a match against everyone, or everyone in your group. Top X # of players with the most wins, (or top Y # of players with most wins from each group if multiple groups), move on to a finals structure.

I thought that was more “round robin” style, when you meet everybody once Head to head? Anyways I just read Matchplay’s description of Swiss: “Players will be paired against players with the same win-loss record. No attempts will be made to pair players against previously unfaced opponents (adjacent pairing is used).”

Also the other matchplays tournament with swiss-style has been both a variety of 4-player groups and head to head.

I’ve always understood “swiss” to mean “paired with those similar to you in standings/points”

Have never exclusively associated it with any particular format (head to head, 4-player, etc), but maybe this is wrong!

1 Like

In a one day duration, or especially one evening, “local” sized tourney (less than 30 players?), I think that the best version of Swiss is to have players play against opponents of similar total points… but with the constraint that you don’t face the same player again until you’ve faced all the players at least once. This is fairly easy to manage in head-to-head, but quite difficult to manage in 4-player groups.

The first time I used Swiss in head-to-head play, I didn’t use this constraint, and looking at the qualifying rounds data after the event, I noticed that 2 of the players that made the finals had a MUCH more difficult slate of opponents than the other 2 finalist. And there were 1-2 players who missed the finals while playing a much more difficult slate of opponents than the 2 “easy” finalists. It was because the 2 “easy” finalists had unintentionally lost early, and because they were highly skilled players, they ended up easily winning against lower quality players.

Putting in the constraint of not facing the same player again somewhat fixes this. I don’t believe that MatchPlay software provides for this.

For reference, Pinburgh does NOT have this constraint: you are allowed to face the same player(s) again and again if the players’ points and current tier provides for that.

It doesn’t even out: by design, the Swiss system gives some players a more difficult draw than others. I wouldn’t say that it’s unfair for the best players; it’s unfair for those who perform well in early rounds.

Using 4-player group Swiss as a qualifying format - especially when multiple games are played per round - does lead to situations where players are incentivised to throw matches.

With “even out” I guess I meant that in the end, usually the best players still manage to advance, and the least good players dont. And then for the average players in the middle, it might be a bad or lucky beat.

It is still a risky strategy to “lose” matches intentionally in the early rounds, I would say.

Swiss style matchplay has infected the local area here in Toronto since I brought it back after playing it at IFPA Germany. Been a resounding success.

We usually play head to head, not groups of 4 though.

That being said I don’t see it as being unfair, difficult, but not unfair. Weighting the top groups would be unfair as that would essentially just be the rich getting richer and you would end up with a few people running away with the event so it becomes pointless for anyone else to keep playing.

the biggest complaint we get with Swiss system is people don’t like playing the same person over and over again.

But I think Swiss system keeps things more fair because better players are forced to cannibalize each other meaning the tournament remains relevant for everyone throughout the entire qualifying.

4 Likes

It does and it doesn’t. It’s complicated.

In group play no attempt will be made. As noted in the quote above players are grouped by their current points and adjacent pairing is used within the group. This is because computers are hard.

In head-to-head play a simple attempt is made to avoid repeat opponents. It goes like this:

  • Players are grouped by number of wins
  • Players are shuffled within each group
  • First player is matched with an unfaced opponent if possible
  • Second player is matched… and so on. If there are no unfaced opponents a random opponent is chosen

This works “good enough” for most tournaments. I’d love to implement a FIDE swiss algorithm, but it’s very complicated (details: http://www.fide.com/fide/handbook.html?id=167&view=article ) and would be quite time consuming to do.

I try to focus on developing the features people ask for and not many people care enough about pairing algorithms to ask for FIDE swiss pairings. :slight_smile:

3 Likes

It’s often used as a qualifying portion in pinball tournaments, but it originates in Chess as a means of identifying a winner without having to play a full round robin tournament. That’s why the system is “unfair” to players who win in the early round (as @LCM points out).

Having played Swiss system chess tournaments, let me put on my Math User hat for a while. The Swiss system is designed to determine the best player when there are too many people to play a round robin. As designed for Chess, in each round, the best players so far are matched against each other. The general idea is that if you have, say, 100 players, going 6 rounds, most of the better 50 will win game 1 and get paired up for round two; the best 25 of those will win again in round 2; proceeding along, the expectation is that the tournament winner will go 6-0, 5-0-1 or at worst 5-1 with perhaps a two-way tie for first. It’s necessary to match records up in cases like this to prevent having many people go 6-0 or 5-1 and getting an inconclusive result. If everyone plays an “equal average mix” of opponents, with the top players never meeting face to face, that can easily happen. But with Swiss parings, you’ll typically get a 6-0 person, a few 5-1’s, one of whom lost game 6, the others losing earlier, a lot of 4-1-1’s and 4-2’s and a ton of 3-3, 2-2-2, etc., i.e. people around even.

When pinball events use the Swiss system for “strike” tournaments, the principle works as intended provided the number of strikes is high enough. Three or more seems to work pretty well; my observations are that one or two strikes are usually too few to avoid randomness, though time constraints may take precedence. When pinball uses Swiss in other ways, like at Pinburgh, flaws emerge. In the Chess model, if you lose in round one, the best you can do after round two is to be even with the player who beat you in round one if they lose in round two and you win. When the Swiss system is expanded beyond head-to-head, though, it weakens. At Pinburgh, you can end up ahead of everyone who beat you in your round one group if you get a 12 in round two, say by getting placed in an “easy” low group. This is where people like your friend start to take issue with it. At Pinburgh, the large number of games and rounds played - - 4 games per round, 10 rounds - - tends to overcome this weakness and bring things back closer to the intended result by the end of it all. As discussed elsewhere, it’s not perfect, but it works really well.

The key for you is to have “enough” rounds if you’re doing 4-player groups. That’s hard to do given how long a 4-player game takes. It will work reasonably well if you do 5-6 Swiss-style “sorting” rounds, then take the top 8 players for two playoff rounds. If you don’t have time for that, a head-to-head 3- or 4-strike format is better able to sort the players out in a shorter time frame.

3 Likes

We’re nearing the limits of my math abilities, but I do know that the amount of rounds you need to play in head-to-head swiss is the binary log of the player count. So for 100 players you need to play 7 rounds. You can calculate the number here (if inclined): https://www.easycalculation.com/log-base2-calculator.php

As always the solution is to play more pinball :slight_smile:

Pinburgh is Swiss-style, but does not seed the top players against one another until Round 10. In each round there is a “tier” size that is meant to increase the difficulty of the higher groups, but without overwhelming the top players with the highest competition right away. It is a difficult balance to strike, and each year we adjust the tier size in response to player performance and comments.

I agree that a Swiss format that immediately sends the top 4 players into a group can provide an incentive for a player to play “just well enough” to stay away from killer groups.

3 Likes

Also, definitely willing and interested to hear suggestions for making the Pinburgh group assignments more fair. We have some ideas we’re taking into account for next year, but there’s always room for improvement.

Two FYIs:

  • We won’t ever be able to do 2-player head-to-head at Pinburgh; the 4-player groups help immensely to observe each other. A tournament of this size would probably be impossible without 4-player groups.
  • Most pinball tournaments differ from chess in that there are multiple rounds played with the same players, so results are things like 3-2-1-0 instead of just 1-0. This causes difficulty with Swiss because all the 3-0 players “win” the right to beat each other up in Round 2 :wink:
1 Like

Interesting. I didn’t realize that the Pinburgh tier sizes shifted from year to year! This part of the tournament has always impressed me and has also been the hardest for me to succinctly explain to newer players who want to know “how it works.”

Has the tier size adjustment consistently trended in one direction, or has it gone back and forth?

Actually, the Swiss system in chess doesn’t exactly pair the best against the best. In fact, the ratings that players come in with keep the best players apart until it can no longer be done. 1 vs 2 won’t happen until there are few people with perfect scores remaining and due to who has already played each other 1 and 2 must play each other. Much like Pinburgh, the rating gap separating opponents is slowly narrowed until the final rounds when it is all on the line, and you have a round 10 with amazing match-ups.

In fact, I think Pinburgh actually does it better than Swiss does chess, because often the big matches in chess happen 2-3 rounds early in say a 9 round tournament. That is why there are often 2 move draws in chess tournaments in the final round to ensure a nice payout, but in Pinburgh a 4 way tie for a 6 might actually knock all four players out of the playoffs.

It’s generally been a percentage of the field size, but we’ve trended toward making it wider. The original idea was that at the end of both Day 1 and Day 2 the top seeds would be together; recently we’ve opened Day 1 to be wider throughout.

Here’s the way the tier sizes have adapted over the 6 years of Pinburgh.

2016 “tier size” (~700 players, 175 players per division): full / 300 / 128 / 64 / 32 /// 96 / 48 / 16 / 8 / 4
2015: full / 240 / 100 / 40 / 16 /// 96 / 48 / 16 / 8 / 4
2014 (400 players, 100 players per division): full / 200 / 80 / 40 / 16 /// 64 / 32 / 16 / 8 / 4
2013 (~400 players): full / 128 / 48 / 16 / 4 /// 64 / 32 / 16 / 8 / 4
2012 (~270 players), 2011 (~170 players): full / 96 / 48 / 16 / 4 /// 64 / 32 / 16 / 8 / 4

The other change from 2011 to 2012 was an adjustment to how Round 1 was allocated, so that the 3-player groups in Round 1 occur at the bottom end of the pairings instead of the top.

1 Like

Wow, 2-move draws? That’s crazy and wildly collusive. At least a 4-way tie for a 6 is also physically possible :wink: