FakeGeorge
Smash Journeyman
This article was intended to be something of a brief introduction to Swiss, as applied to pools, and then serve as a place to explore its viability. My introduction ended up being rather long and more like a guide. I feel like this can be useful for any TOs interested in using this format, so I'm sticking with it. Still, if you read nothing other than this first paragraph, I ask you, "How well do you think Swiss format pools would work as an alternative to Round Robin?"
My goal here is to explore this format and see if it can be viable. I have my initial thoughts below, but they are by no means comprehensive.
Why Swiss?
In my area, it's difficult to get a venue for a two day event. In fact, the longest that I can get is 13 hours to run a tournament. For a local tournament, this is ample time to hold Teams and Singles. We can even have Round Robin pools before a bracket if we wish. However, if I were to ever try and host a larger regional tournament, this becomes a lot to do in one day, especially when time is needed for registration and cleanup.
Every organizer has to deal with time constraints. Because of this, I think some people may benefit from a possible solution for those who need to cut time while running Singles pools, bracket, and Teams. One solution is to have Swiss format pools instead of round-robin pools.
What is Swiss?
In case you are not familiar with the Swiss format, Wikipedia has a good article. In short, the Swiss system operates by having entrants play a pre-determined number of rounds. The number of rounds is determined by:
Log₂[n] , where n is the number of entrants, rounded up to the nearest integer.
Within these rounds, players are paired based upon their current win-loss record. The two main factors in determining pairing are:
1. Record: Players are paired with players of a similar Win-Loss record if possible.
2. Previous opponent: Within a Swiss tournament, no entrant should be paired with a particular player more than once.
After these two general principles, many specifics could be employed to very specifically determine each round's pairing. Just glimpsing at the FIDE guidelines can give you an idea of how far it can go.
Swiss versus Round Robin
Swiss is a lot faster than Round Robin. For an 8 person pool, there would be 28 RR matches, and only 12 Swiss matches (4 matches per round * 3 rounds). There is extra time involved with Swiss, as pairings for a given round cannot be made until all of the previous rounds' matches are finished. Then, you either have pools sharing some computer program to make pairings, or each pool's pairings must be done manually. For a small pool (~8), this takes only a few minutes. However, more players take more time. Still, if a match averages 15 minutes, then for an 8 person pool, Swiss is definitely faster than a Round Robin.
Round Robin tends to be more accurate, since everyone plays everyone. However, ties can occur, and it takes more time to break them.
Swiss, on the other hand, has some clear-cut divisions, through win-loss record, with established tie-breakers for participants with the same number of wins and losses. Since every matchup hasn't been played, each player's average opponent's win % (let's call this AOW) can be looked at. Basically, we can say player A was paired with better players than player B, so A deserves to be ranked above B. If the players are still tied, you can look at the average of a player's opponents' AOW (i.e. who's opponents played the best people?). If there is only a two-way tie and both entrants played one another, then the result of that match can also be used to break ties.
Problems with Swiss
Swiss isn't without its flaws.
1. Tiebreaks really can seem arbitrary. It would suck to not make it out of pools due to you being paired with the worst player in the pool first round. That is clearly not a fault of the player.
2. Round pairings are difficult. Consider this situation: You have a pool of 8 players with 4 advancing. They are ranked 1-8 in terms of skill (1 is the best, and 8 is the worst). Let's assume there are no upsets. You seed the first round, so that your matches are:
1 v 8
2 v 7
3 v 6
4 v 5
After the first round, players 1 through 4 are 1-0 and 5 through 8 are 0-1. Let's say you pair the next round randomly (within W-L record, of course).
1 v 2
3 v 4
5 v 7
6 v 8
Now you have three groups, 1 and 3 at 2-0; 2, 4, 5, and 6 at 1-1; and 7 and 8 at 0-2. Pairings for the next round are again, random:
1 v 3
2 v 4
5 v 6
7 v 8
The final records for players are as follows:
1: 3-0
2, 3, 5: 2-1
4, 6, 7: 1-2
8: 0-3
Due to random pairing, the 4th most skilled player will not make it out of pools, which is certainly unfair. Rules for pairing are needed. The FIDE (chess people) has rules, and that's for a reason.
3. Running Swiss is complicated. Pairings are complex, and tiebreaks can be a pain. If you don't know what you're doing, you can really screw things up. In turn, this may very well ruin the experience for the players.
4. The numbers aren't nice. That is, a 9 person pool can be a pain. You have to do 4 rounds, with a bye every time. Then, end up with weird breaks for W-L record. The best option is to have a power of two number of members. It's unlikely that this will happen often, so the next best is to have a number of entrants that is less than the nearest power of 2. Something like 6 or 7 can work, but 17 instead of 17 pushes it to 5 rounds. One person can make a huge difference.
Tie breakers can still be implemented to get rankings, but it lacks the consistency and natural breaks provided by pools of 8 or 16 players. In addition, I know from experience that tiebreakers can feel very arbitrary, but hopefully proper seeding will prevent this.
Ultimately, I'm not sure if this is a big problem, but it certainly can complicate things. I think the more opposed the community is to using tiebreakers, to more of a problem this will be. If players understand or at least accept breaking ties based on who played players who did better, then it might not be a problem at all.
Solutions
1. Seeding a pool is all you really need to do to help this out. Ideally, every player will be seeded, but we just don't know in smash. I think the practical solution is to have people advance in groups of W-L. For an 8 person pool, advancing four will mean everyone who went 2-1 or better will advance. For a 16 person pool, you get clear breaks at 15, and 11 people. That's everyone 3-1 and better or everyone 2-2 and better. Tiebreakers would then only determine seeding in a bracket and not whether or not a player makes it. That's not perfect, but it's certainly better.
2. Pairing rules. I'm not really sure what the best set would be, but the first two are pretty much a given:
i. Players should not be paired together more than once.
ii. If possible, players should be paired with players of a similar record.
After this, FIDE uses a system that attempts to "pair players who have played low rated players before with players having high ratings now." So if we were to follow the general idea behind this, we want to pair players who are doing well with players with a low AOW, all within a group having the same record. Since both players should have the same record already, I think using either a player's current AOW or his initial seed is the best representation of a high "rating." Since every player will not necessarily have a seed, I think the AOW should be used.
With this, I suggest a third pairing rule:
iii. Players having a low AOW should be paired with players having a high AOW.
To implement this, just calculate and rank each player by AOW. Pair the top with the bottom and move inwards.
What if there is an odd number of players in the group?
iv. The player with the lowest AOW in a group having the same record is paired with the player with the highest AOW in the next lower record group
Byes are necessary for an odd number of people in any Swiss system. Here are some general rules for byes:
v. The player with the lowest win % (worst record) should be given the bye. In the first round, the lowest seed gets the bye.
vi. No player should receive more than one bye.
For now, I feel that these pairing guidelines would be sufficient to fairly create pairings each round.
3. Pairing is a pain, and so is calculating results. A computer is a huge help. It is a simple matter to create a spreadsheet that calculates a person's win %, given wins and losses. From here, you can calculate a person's AOW if you are given the opponents.
Since I am considering pools, I suggest designating pool leaders. They can record results, call matches, and possibly create pairings. When it's all done, you can get the data and calculate the final results. In addition to needing each player's final record at the end, one also needs the pairings for each round, so that AOWs can be calculated.
Before the tournament starts, you can talk with the pool leaders and make sure they know what to do and how to do it. Pairing can be complicated, so make sure the process is clear. It might help to make a spreadsheet with each pool having its own individual sheet. That way, pool leaders can enter data and easily calculate AOWs for pairing. Then, it's a small step to have pool leaders record each round's pairing right there.
This doesn't make Swiss any less complex, but having and educating pool leaders will help reduce errors and help it run smoothly. I think having a computer is almost necessary, especially for larger tournaments, because of the numerous calculations.
I think it is important to note that there are Swiss tournament programs out there, but I am not familiar with any of them.
4. I'm not really sure how to best address this. Then again, it might not be a big deal. Hopefully my future experiences with Swiss can clarify it. Unless a player cap is set it is impossible to control entrants. Tiebreakers will probably have to be used for something like a 9 person pool, so make sure the actual methods of tiebreaking are listed prior to the tournament.
Final Remarks
For me, this is still a work in progress. I think there is real potential here, but the nature of breaking ties is a lot different than what we see in smash today. Usually, a head-to-head match is the only way to settle stuff, so using statistics might make players feel cheated.
Right now, I think the thing to do is test out the format and get feedback. At the moment, I've run Swiss pools at two local tournaments. I used the results to seed a bracket, not to eliminate any players. The reaction was generally positive, but the players were a bit confused initially. LasT 3 will hopefully grow to be a fairly large regional tournament, and I am using Swiss pools. From this, I should be able to see if it's worth it or at least practical.
Personally, I'm not sure what is the best order to implement tiebreakers. Right now, I'm leaning towards AOW, then opponent's AOW, and finally, the head-to-head. I put the head to head last, because very rarely is there a two way tie, so the first two methods can narrow the field. For pools of 8 or so, I don't think more than the AOW will have be needed.
I'd love to hear other people's thoughts on Swiss pools, as well as feedback on the article/ guide. Thanks!
My goal here is to explore this format and see if it can be viable. I have my initial thoughts below, but they are by no means comprehensive.
Why Swiss?
In my area, it's difficult to get a venue for a two day event. In fact, the longest that I can get is 13 hours to run a tournament. For a local tournament, this is ample time to hold Teams and Singles. We can even have Round Robin pools before a bracket if we wish. However, if I were to ever try and host a larger regional tournament, this becomes a lot to do in one day, especially when time is needed for registration and cleanup.
Every organizer has to deal with time constraints. Because of this, I think some people may benefit from a possible solution for those who need to cut time while running Singles pools, bracket, and Teams. One solution is to have Swiss format pools instead of round-robin pools.
What is Swiss?
In case you are not familiar with the Swiss format, Wikipedia has a good article. In short, the Swiss system operates by having entrants play a pre-determined number of rounds. The number of rounds is determined by:
Log₂[n] , where n is the number of entrants, rounded up to the nearest integer.
Within these rounds, players are paired based upon their current win-loss record. The two main factors in determining pairing are:
1. Record: Players are paired with players of a similar Win-Loss record if possible.
2. Previous opponent: Within a Swiss tournament, no entrant should be paired with a particular player more than once.
After these two general principles, many specifics could be employed to very specifically determine each round's pairing. Just glimpsing at the FIDE guidelines can give you an idea of how far it can go.
Swiss versus Round Robin
Swiss is a lot faster than Round Robin. For an 8 person pool, there would be 28 RR matches, and only 12 Swiss matches (4 matches per round * 3 rounds). There is extra time involved with Swiss, as pairings for a given round cannot be made until all of the previous rounds' matches are finished. Then, you either have pools sharing some computer program to make pairings, or each pool's pairings must be done manually. For a small pool (~8), this takes only a few minutes. However, more players take more time. Still, if a match averages 15 minutes, then for an 8 person pool, Swiss is definitely faster than a Round Robin.
Round Robin tends to be more accurate, since everyone plays everyone. However, ties can occur, and it takes more time to break them.
Swiss, on the other hand, has some clear-cut divisions, through win-loss record, with established tie-breakers for participants with the same number of wins and losses. Since every matchup hasn't been played, each player's average opponent's win % (let's call this AOW) can be looked at. Basically, we can say player A was paired with better players than player B, so A deserves to be ranked above B. If the players are still tied, you can look at the average of a player's opponents' AOW (i.e. who's opponents played the best people?). If there is only a two-way tie and both entrants played one another, then the result of that match can also be used to break ties.
Problems with Swiss
Swiss isn't without its flaws.
1. Tiebreaks really can seem arbitrary. It would suck to not make it out of pools due to you being paired with the worst player in the pool first round. That is clearly not a fault of the player.
2. Round pairings are difficult. Consider this situation: You have a pool of 8 players with 4 advancing. They are ranked 1-8 in terms of skill (1 is the best, and 8 is the worst). Let's assume there are no upsets. You seed the first round, so that your matches are:
1 v 8
2 v 7
3 v 6
4 v 5
After the first round, players 1 through 4 are 1-0 and 5 through 8 are 0-1. Let's say you pair the next round randomly (within W-L record, of course).
1 v 2
3 v 4
5 v 7
6 v 8
Now you have three groups, 1 and 3 at 2-0; 2, 4, 5, and 6 at 1-1; and 7 and 8 at 0-2. Pairings for the next round are again, random:
1 v 3
2 v 4
5 v 6
7 v 8
The final records for players are as follows:
1: 3-0
2, 3, 5: 2-1
4, 6, 7: 1-2
8: 0-3
Due to random pairing, the 4th most skilled player will not make it out of pools, which is certainly unfair. Rules for pairing are needed. The FIDE (chess people) has rules, and that's for a reason.
3. Running Swiss is complicated. Pairings are complex, and tiebreaks can be a pain. If you don't know what you're doing, you can really screw things up. In turn, this may very well ruin the experience for the players.
4. The numbers aren't nice. That is, a 9 person pool can be a pain. You have to do 4 rounds, with a bye every time. Then, end up with weird breaks for W-L record. The best option is to have a power of two number of members. It's unlikely that this will happen often, so the next best is to have a number of entrants that is less than the nearest power of 2. Something like 6 or 7 can work, but 17 instead of 17 pushes it to 5 rounds. One person can make a huge difference.
Tie breakers can still be implemented to get rankings, but it lacks the consistency and natural breaks provided by pools of 8 or 16 players. In addition, I know from experience that tiebreakers can feel very arbitrary, but hopefully proper seeding will prevent this.
Ultimately, I'm not sure if this is a big problem, but it certainly can complicate things. I think the more opposed the community is to using tiebreakers, to more of a problem this will be. If players understand or at least accept breaking ties based on who played players who did better, then it might not be a problem at all.
Solutions
1. Seeding a pool is all you really need to do to help this out. Ideally, every player will be seeded, but we just don't know in smash. I think the practical solution is to have people advance in groups of W-L. For an 8 person pool, advancing four will mean everyone who went 2-1 or better will advance. For a 16 person pool, you get clear breaks at 15, and 11 people. That's everyone 3-1 and better or everyone 2-2 and better. Tiebreakers would then only determine seeding in a bracket and not whether or not a player makes it. That's not perfect, but it's certainly better.
2. Pairing rules. I'm not really sure what the best set would be, but the first two are pretty much a given:
i. Players should not be paired together more than once.
ii. If possible, players should be paired with players of a similar record.
After this, FIDE uses a system that attempts to "pair players who have played low rated players before with players having high ratings now." So if we were to follow the general idea behind this, we want to pair players who are doing well with players with a low AOW, all within a group having the same record. Since both players should have the same record already, I think using either a player's current AOW or his initial seed is the best representation of a high "rating." Since every player will not necessarily have a seed, I think the AOW should be used.
With this, I suggest a third pairing rule:
iii. Players having a low AOW should be paired with players having a high AOW.
To implement this, just calculate and rank each player by AOW. Pair the top with the bottom and move inwards.
What if there is an odd number of players in the group?
iv. The player with the lowest AOW in a group having the same record is paired with the player with the highest AOW in the next lower record group
Byes are necessary for an odd number of people in any Swiss system. Here are some general rules for byes:
v. The player with the lowest win % (worst record) should be given the bye. In the first round, the lowest seed gets the bye.
vi. No player should receive more than one bye.
For now, I feel that these pairing guidelines would be sufficient to fairly create pairings each round.
3. Pairing is a pain, and so is calculating results. A computer is a huge help. It is a simple matter to create a spreadsheet that calculates a person's win %, given wins and losses. From here, you can calculate a person's AOW if you are given the opponents.
Since I am considering pools, I suggest designating pool leaders. They can record results, call matches, and possibly create pairings. When it's all done, you can get the data and calculate the final results. In addition to needing each player's final record at the end, one also needs the pairings for each round, so that AOWs can be calculated.
Before the tournament starts, you can talk with the pool leaders and make sure they know what to do and how to do it. Pairing can be complicated, so make sure the process is clear. It might help to make a spreadsheet with each pool having its own individual sheet. That way, pool leaders can enter data and easily calculate AOWs for pairing. Then, it's a small step to have pool leaders record each round's pairing right there.
This doesn't make Swiss any less complex, but having and educating pool leaders will help reduce errors and help it run smoothly. I think having a computer is almost necessary, especially for larger tournaments, because of the numerous calculations.
I think it is important to note that there are Swiss tournament programs out there, but I am not familiar with any of them.
4. I'm not really sure how to best address this. Then again, it might not be a big deal. Hopefully my future experiences with Swiss can clarify it. Unless a player cap is set it is impossible to control entrants. Tiebreakers will probably have to be used for something like a 9 person pool, so make sure the actual methods of tiebreaking are listed prior to the tournament.
Final Remarks
For me, this is still a work in progress. I think there is real potential here, but the nature of breaking ties is a lot different than what we see in smash today. Usually, a head-to-head match is the only way to settle stuff, so using statistics might make players feel cheated.
Right now, I think the thing to do is test out the format and get feedback. At the moment, I've run Swiss pools at two local tournaments. I used the results to seed a bracket, not to eliminate any players. The reaction was generally positive, but the players were a bit confused initially. LasT 3 will hopefully grow to be a fairly large regional tournament, and I am using Swiss pools. From this, I should be able to see if it's worth it or at least practical.
Personally, I'm not sure what is the best order to implement tiebreakers. Right now, I'm leaning towards AOW, then opponent's AOW, and finally, the head-to-head. I put the head to head last, because very rarely is there a two way tie, so the first two methods can narrow the field. For pools of 8 or so, I don't think more than the AOW will have be needed.
I'd love to hear other people's thoughts on Swiss pools, as well as feedback on the article/ guide. Thanks!
Swiss is one of the formats I've been considering for a few months to run at my next regional.