How to Choose 2 teams in 1 Trials

[barmar]

Obviously, a good seeding system will decrease the chance that this happens. Then again, if you're that confident in your seeding system, there really isn't any reason to have a tournament. Just select your number 1 and number 2 seeds.

Well, if you take that to its logical conclusion you'll never play any games, you'll just decide all tournaments by looking at everyone's seeding points.

 

Seeding is a rough estimate of the ranks of the teams based on history, and then the actual play is used to refine and update it with current data. The basic assumption on which this is based is that players' abilities and team rankings change gradually, and we're mostly concerned with the top-ranked teams. For instance, if the only error in the seeding is 1st vs. 2nd place, and everyone plays according to form, these two teams should meet each other in the final and then the error will be corrected. On the other hand, if the error is between the two lowest teams, it won't be corrected -- they'll both lose their first matches, and there's no playoffs in the early rounds to discover their relative ranks.

 

When I was in high school or college I read a book of essays by Charles Dodgson, AKA Lewis Carroll. Although he's best remembered as the author of the Alice in Wonderland books, he was primarily a mathematician and logician. I don't remember the details, but one of the essays was on how best to structure a tennis tournament. I did some googling, and I haven't been able to find the text of the essay online, it was titled "Lawn Tennis Tournament" and published in 1883.

[JanM]

Wow! Thanks for all the thoughts. If I can figure out how to quote several things in one reply, I'll try to answer as many questions as I can and see whether you can help some more B)

If you are going to do all of this, why not just have two trials with seeding and byes.

Isn't that what we currently have? BB is every other year, and we have a trial each year. The winner the first year is USA1, the winner the second year (which USA1 doesn't play in) is USA2.

Or am I thinking of something else?

In 2008, the USBC winner will be the only US team in the Olympiad. Therefore we have only one USBC to select both teams for 2009.

I would have the losers of the semi final match play while the finals were going on. Winner of finals=USA 1. The loser of the finals plays the winner of the the match between the losing semi finalists. The winner of this match is USA2.

That's how we do it for the Women. But there is fairly strong feeling that we have too many strong teams in the Open to start the Repechage with the Semi-final losers, so we're trying to come up with something that lets earlier losers get back in without making the whole event too long.

I'm sure I don't know what I'm talking about but if we assume that

 

any team beaten by the winner is a candidate for 2nd strongest team (in this event)

any team beaten by any other team is not.

 

Then perhaps the teams that lost to the winners in head to head play, and no one else, should compete amongst themselves for 2nd place.

That makes a lot of sense, except for the problem someone pointed out later that it means we either have to wait to start the USA2 bracket until the USA1 winner is decided or play a lot of irrelevant matches.

Run a double-elimination event. The team that goes undefeated is USA1. The team that ends with one loss is USA2. Constructing this can be a bit complex, but should be possible. Allowing byes actually makes this easier to construct, and it removes the luck factor of playing the best (winning) team in an early round.

That's sort of what we're looking at, but it's not as easy as it sounds - the problem is that we want the two events to run side by side and teams keep dropping into the USA2 bracket, so the matches have to be different lengths, and where do you start (IOW, which losers first get into the USA2 bracket). Before continuing to struggle with numbers of teams and lengths of matches, I was wondering if anyone might come up with a better suggestion :)

With this said and done, I’m not sure whether it would be worth doing the necessary work. My impression of the USBF is that subjective considerations outweigh objective criteria.

I guess that depends on what you mean by subjective and objective. If you mean that we're not likely to be swayed by mathematical models that conflict with our impressions of what has worked in the past, you're right. The reason that you would have a hard time convincing me that a single elimination KO is "better" than a RR followed by a KO is because my experience convinces me that seeding isn't anywhere close to perfect, and the first round of a KO is going to be very strongly influenced by seeding. Maybe I'm wrong, but a lot of people seem to agree with me, even some who disagree with me on almost everything else.

With regard to conventions, I think you've misunderstood me - the reason that we don't use the WBF rules is that most of our players are more familiar with ACBL rules and we aren't willing to force them to comply with a slightly different, not necessarily better, set of rules for the USBC. Of course, the winners (and maybe a second team) will have to comply with those rules for the WBF event, but they'll have a captain and coach to help with that, as well as a few months to work on it at a time when they know it will be relevant for them.

In short, if you can provide a very specific description regarding the set of constraints, I think that folks can identify the most accurate tournament. Absent this type of description, the problem isn’t an attractive one to work on.

I don't know what sort of constraints you mean, but basically:

The USA1 bracket is going to look like this year's USBC (maybe some matches might be shorter because they don't completely eliminate anyone.

USA2 needs to be chosen in an event that runs parallel to USA1 and where USA1 losers drop into it. A double elimination KO if you want to call it that, but some other format might be possible.

The USA2 event would ideally end no more than 3 days after the USA1 finals (because of fatigue for the players mainly).

Runner up & both semifinalists play a long "3-teams-match"

(with carry over for the runner up)

That's how we did it in 2001, and despite the excellent result (USA2 won the Bermuda Bowl that year in case you've forgotten), there was a very strong feeling that this wasn't a good format and I don't think there is any chance that we will use a 3-way again :)

The qualifying should mirror the actual event, so I'm not a big fan of byes. Stamina should be a big factor in team selection, and if a team gets a free ride to the semi-finals, then stamina doesn't play nearly the factor that it should.

I'm also very skeptical about augmentation. If four (or even 5) warhorses grind it through and somehow gain one of the spots, I can see putting the two seats on the open market and letting someone buy their way onto our national team. Isn't that possible? Or does the USBF have a say-so in who is allowed to be added?

As to byes, there's a very strong opinion that the three major KOs during the year should be "part" of the USBC, which is achieved by the bye structure. That gets debated periodically but so far the "Byes" have it. As to augmentation, the USBF has to approve augmentees and we have strict rules and a very strong Credentials Committee who review any proposed augmentees, to avoid augmentation of someone who is not as good as the rest of the players on the team. So far it seems to be working and I certainly hope that will continue to be the case.

It seems for the past few years or decade, the USBF has done a good job in letting us nobodies have a shot, but rewarding the better teams for regular season performance.

I agree with you and I think that our basic approach is a good one. But every fourth year, we have to choose 2 teams, and that's our current issue.

[hrothgar]

I guess that depends on what you mean by subjective and objective. If you mean that we're not likely to be swayed by mathematical models that conflict with our impressions of what has worked in the past, you're right. The reason that you would have a hard time convincing me that a single elimination KO is "better" than a RR followed by a KO is because my experience convinces me that seeding isn't anywhere close to perfect, and the first round of a KO is going to be very strongly influenced by seeding. Maybe I'm wrong, but a lot of people seem to agree with me, even some who disagree with me on almost everything else.

The easiest way to proceed would be to compare two different formats:

 

1. A single elimination with no seeding

 

2. A round robin that reduced to a single elimination (using the RR results to seed the KO)

 

If the single elimination KO wins without seeding...

 

Few parameters that I'd need to create an accurate model.

 

How many teams do you anticipate entering the event?

 

How much time do you expect to be able to allocate for the event?

 

What is the average amount of time necessary to complete a board?

 

Can you make a reasonable guess about the shape of PDF describing the strength of the various teams? (Uniform? Normal? Bimodal? Something exotic?)

[JanM]

Few parameters that I'd need to create an accurate model.

 

How many teams do you anticipate entering the event?

 

How much time do you expect to be able to allocate for the event?

 

What is the average amount of time necessary to complete a board?

 

Can you make a reasonable guess about the shape of PDF describing the strength of the various teams? (Uniform? Normal? Bimodal? Something exotic?)

I think that between 25 and 30 teams will enter, probably closer to 25.

A one-winner event now takes 11 days. I think we're looking at no more than 14 for a 2-winner event, and hopefully 12 or 13.

We allow 9 minutes per board. I know that seems high, but somehow it doesn't seem to be in practice.

That's hard - for one thing, I'm not sure what your different models are, and for another it's very hard to analyze the strengths of teams. Most of them played this year and last, and you can see the scores on the usbf website, but I don't know if that's an appropriate method to evaluate or not.

[mike777]

It is 2007. We need to pick two teams for late 2009? We have picked zero so far.

 

Why not just have two trials if almost everyone does not like the simple suggestion of Justin's?

 

Justin's saves money. Two trials spends more money.

 

One in 2008 and one in 2009?

[cherdano]

Richard, I would be curious about the results of your simulations (but maybe in another thread, as this isn't really relevant for the question of this thread). Why don't you just assume a normal distribution with a standard deviation of s.th. like 0.3 IMPs (or maybe try several values), 25 teams, and 600 boards (the last trials used 60 boards a day).

[JanM]

It is 2007. We need to pick two teams for late 2009? We have picked zero so far.

 

Why not just have two trials if almost everyone does not like the simple suggestion of Justin's?

 

Justin's saves money. Two trials spends more money.

 

One in 2008 and one in 2009?

Because the 2008 USBC will pick our Olympiad team.

[mike777]

It is 2007. We need to pick two teams for late 2009? We have picked zero so far.

 

Why not just have two trials if almost everyone does not like the simple suggestion of Justin's?

 

Justin's saves money. Two trials spends more money.

 

One in 2008 and one in 2009?

Because the 2008 USBC will pick our Olympiad team.

Again so why not two trials in 2008? One early for 2008(olympiad) and one later in 2008 for 2009(usa1)? Have a third trial in 2009 for usa2.

 

Again if money is the main issue then Justin's idea as used already in the Womens seems cheapest and fairest.

 

I am a bit lost on what the main number one priority is....saving money or other?

[cherdano]

Run a double-elimination event. The team that goes undefeated is USA1. The team that ends with one loss is USA2. Constructing this can be a bit complex, but should be possible. Allowing byes actually makes this easier to construct, and it removes the luck factor of playing the best (winning) team in an early round.

That's sort of what we're looking at, but it's not as easy as it sounds - the problem is that we want the two events to run side by side and teams keep dropping into the USA2 bracket, so the matches have to be different lengths, and where do you start (IOW, which losers first get into the USA2 bracket). Before continuing to struggle with numbers of teams and lengths of matches, I was wondering if anyone might come up with a better suggestion B)

Assuming you don't want to touch your rules about byes, I doubt there is a good solution, unless you are willing to have some 3-ways early in the loser's bracket, or round-robin-among-5-with-two-teams-advancing, etc.. E.g. if you would start the losers bracket with the losers of the "round of 8" (usually 6 teams playing), you could have a 3-way among the 3 losers with two survivors, who then join the losers of the semifinals for a 3-day knockout. The winner of that will then play against the loser of the finals for the second berth (who will have had a one-day break after losing the final).

I am sure this is the kind of suggestion you came up with yourself, so...I think it's just a matter of which kind of mess in the loser's bracket to live with. (Oh, and of course some rules about carry-overs in case of replays in the loser bracket.)

[JanM]

It is 2007. We need to pick two teams for late 2009? We have picked zero so far.

 

Why not just have two trials if almost everyone does not like the simple suggestion of Justin's?

 

Justin's saves money. Two trials spends more money.

 

One in 2008 and one in 2009?

Because the 2008 USBC will pick our Olympiad team.

Again so why not two trials in 2008? One early for 2008(olympiad) and one later in 2008 for 2009(usa1)? Have a third trial in 2009 for usa2.

 

Again if money is the main issue then Justin's idea as used already in the Womens seems cheapest and fairest.

 

I am a bit lost on what the main number one priority is....saving money or other?

There are some players who think one Trials every year is too many (we're currently asking for opinions on that from the players). I think there is about a 0% chance of selling the idea of 2, even if that made sense. And I'm pretty sure it doesn't - one of the objections to what we do in the Rosenblum year (select USA1 for the following year's BB in a USBC held later than normal in the year) is that there's too much time between selecting the team and the team playing in the BB. That might be okay when the alternative is not to hold the USBC that year, but not when there's already a USBC that year.

The issue isn't money, it's time - there are already a lot of bridge events on the calendar. There really isn't room for another. Note that we can hold the Rosenblum year USBC late because the World Championships are held early. In 2008, the Olympiad is likely to be in the early fall.

[mike777]

Interesting, thanks for your comments.

 

I am speechless when I hear that someother bridge event in a 52 week year would overtake trying to play as USA 2 in a second trials but ........

I would have guessed people in the USA would drop just about anything to jump at chance to play as USA2. :) Especially if it means no one else shows up....

 

Pros get paid for another event, sponsors have another shot at glory at a cost that is close to zero for them. Nobodies have a shot since no one else shows up..seem perfect. :)

 

I must say no time would have been my last guess as for a reason to not hold a second trial for usa2.

[uday]

OK, what about another uneducated guess?

 

We're already trusting the seeding (or points, or whatever it is that grants byes). Why not trust it further, and split the field into two brackets, one spearheaded by seed1, and the other by seed2, run the brackets independently of each other, and take the 2 winners?

[JanM]

OK, what about another uneducated guess?

 

We're already trusting the seeding (or points, or whatever it is that grants byes). Why not trust it further, and split the field into two brackets, one spearheaded by seed1, and the other by seed2, run the brackets independently of each other, and take the 2 winners?

That's a fascinating idea - I'll "run it up the flag pole" :)

[jdonn]

OK, what about another uneducated guess?

 

We're already trusting the seeding (or points, or whatever it is that grants byes). Why not trust it further, and split the field into two brackets, one spearheaded by seed1, and the other by seed2, run the brackets independently of each other, and take the 2 winners?

I don't think that one would fly at all. I think you were onto something much better with your first idea. It seems an absolute necessity, if using any sort of round robin format, to have every team play every team.

[JanM]

Pros get paid for another event, sponsors have another shot at glory at a cost that is close to zero for them. Nobodies have a shot since no one else shows up..seem perfect. :)

Another Trials would be far from free, for anyone - the players who'd have to take the time out of their schedules, pay for plane, hotel, meals, etc; the sponsors who'd ditto plus pay the pros; the USBF volunteers who are already stretched pretty thin. It's just not going to happen.

[mike777]

Pros get paid for another event, sponsors have another shot at glory at a cost that is close to zero for them. Nobodies have a shot since no one else shows up..seem perfect. :)

Another Trials would be far from free, for anyone - the players who'd have to take the time out of their schedules, pay for plane, hotel, meals, etc; the sponsors who'd ditto plus pay the pros; the USBF volunteers who are already stretched pretty thin. It's just not going to happen.

ok thank you for your comments..it is not going to happen...but this is starting to sound more like a money problem not a time issue but anyway ...

 

If 2 or 20 nobody or seminobody teams only show up for usa2 trials this sounds like a great thing for bridge. The pros and sponsors are too busy with busy lives so why veto the rest of us..oh well....The highest goal is not for usa2 to be the best team..or the team that can win.

 

ok 2 trials out....:)

[TimG]

Suppose 16 teams qualify from a round robin to a USA1 bracket.

• After Round 1 of the KO phase, there will be 8 winners still in USA1 bracket, and 8 losers who go to USA2 bracket.
  
• After round 2, there will be 4 winners in USA1 bracket and 8 one-time losers in USA2 bracket (4 winners from USA2 bracket and 4 losers from USA1 bracket).
  
• After round 3, there will be 2 winners in USA1 bracket and 6 one-time losers in USA2 bracket (4 winners from USA2 bracket and 2 losers from USA2 bracket).
  
• After round 4, there will be one winner in USA1 bracket and 4 one-time losers in USA2 bracket (3 winners from USA2 bracket and 1 loser form USA1 bracket).
  
• The remaining team in USA1 is crowned USA1 and the four teams in USA2 play two more rounds of KOs to determine USA2.
  

When there are teams with byes into the KO phase of the event, the number of teams coming out of the round robin may have to change and some of the teams may need to be sent immediately to USA2 bracket.

 

When one team has a bye into the round of 8, the round robin needs to produce 14 teams for USA1 bracket and 2 teams for USA2 bracket.

 

When two teams have a bye into the round of 8, the round robin needs to produce 12 teams for USA1 bracket and 4 teams for USA2 bracket.

 

When one team has a bye into the round of 8 and one team has a bye into the round of 4, the round robin needs to produce 10 teams for USA1 bracket and 10 teams for USA2 bracket.

 

When two teams have a bye into the round of 8 and one team has a bye into the round of 4, the round robin needs to produce 8 teams for USA1 bracket and 12 teams for USA2 bracket.

 

When teams have a bye into the round of 16, it is a simple matter to adjust the number of teams qualifying from the round robin into USA1 bracket.

 

I seem to recall that the KO matches in the trials are longer than one day. If they are day and a half affairs (6 segments of 16 boards) the above would fulfill the condition that the USA2 result takes no more than three extra days.

[matmat]

i don't really see why you need knockouts... what's wrong with taking the top 7 teams after the early round robin and running a second such competition with 6 rounds, taking the top two teams?

[mike777]

i don't really see why you need knockouts... what's wrong with taking the top 7 teams after the early round robin and running a second such competition with 6 rounds, taking the top two teams?

so no seeding..no byes...no reward for year long bridge performance in the regular season?

 

Add on top of the general usa hatred for round robin formats

[matmat]

i don't really see why you need knockouts... what's wrong with taking the top 7 teams after the early round robin and running a second such competition with 6 rounds, taking the top two teams?

so no seeding..no byes...no reward for year long bridge performance in the regular season?

 

Add on top of the general usa hatred for round robin formats

yeah... and knockouts are great b/c if the best team has one bad session, they can no longer represent the federation. yeah. that's wonderful.

[JanM]

If 2 or 20 nobody or seminobody teams only show up for usa2 trials this sounds like a great thing for bridge. The pros and sponsors are too busy with busy lives so why veto the rest of us..oh well....The highest goal is not for usa2 to be the best team..or the team that can win.

Well, if the best teams don't show up it's impossible for the winner to be the best team :), which is one of the reasons not to hold an extra event. Our primary goal is still to select the two best teams to represent the US in the Bermuda Bowl.

[mike777]

If 2 or 20 nobody or seminobody teams only show up for usa2 trials this sounds like a great thing for bridge. The pros and sponsors are too busy with busy lives so why veto the rest of us..oh well....The highest goal is not for usa2 to be the best team..or the team that can win.

Well, if the best teams don't show up it's impossible for the winner to be the best team :), which is one of the reasons not to hold an extra event. Our primary goal is still to select the two best teams to represent the US in the Bermuda Bowl.

Ya...I strongly disagree with that being the highest priority but agree it is in most countries. Selecting the best team or team with best chance to win.

 

As noted before you need to spell out what the rank of goals are in selecting a team. If the top goal is a level playing field..reward, regular season performance and let the chips fall...that is a very different top goal..than best team win.

[Cascade]

I think the best format would be based on a round-robin. This eliminates at least two problems with knockouts:

 

1. When you start a new round you throw away all of the information you have already gathered (except which team won the previous knockout matches). This is even more of a problem when you are trying to find the second best team.

 

2. Knocking out half of the teams at each stage makes it relatively likely that the best team (or one of the best two teams) are knocked out at an 'early' stage. It is a myth that the best team wins and the best team could easily get knocked out in one of the earlier rounds.

 

A compromise method that I have used in the past is successive round-robins where a relatively small number of teams are eliminated at each stage. This makes it less likely that the best team(s) are eliminated early. You can also carry forward the results from one round-robin to the next. Perhaps you might weight the scores e.g. 100% of the scores against other qualifiers and 50% of the scores against non-qualifiers or some other weighting.

 

A format for 25 teams might look like the following (with many details to be added):

 

Days 1 and 2 : Round-Robin 25 teams 6-board matches - five teams to be eliminated

 

Days 3 and 4 : Round-Robin 20 teams 8-board matches - four teams to be eliminated

 

Days 5 and 6 : Round-Robin 16 teams 10-board matches - four teams to be eliminated

 

Days 7 and 8 : Round-Robin 12 teams 12-board matches - four teams to be eliminated

 

Days 9 and 10 : Round-Robin 8 teams 20-board matches - four teams to be eliminated

 

Days 11 and 12 : Round-Robin 4 teams 48-board matches - two teams to be eliminated

[Gerben42]

If you can tell me what the bye structure will look like I can try and figure out a working double elimination model for this. Since there are 2 more rounds after determining USA1 this can be 90-board matches to be finished in at most 3 days.

 

I kinda like this Round Robin model also (with carryover to the next round). This is similar to the Polish method, except that they use pair qualification.

[awm]

I like Cascade's suggestion. This seems to remove "luck of the draw" in terms of who plays whom when, which is one of the problems with a KO-like structure. It also allows the selection of two teams without substantially increasing the number of days in the event, and it gives teams that might be better than their seeding points would indicate a chance to go deep in the event (i.e. they won't have to play the Nickell team in an elimination match early on). I'd suggest that IMP scores carry-over from day to day whereas victory point totals do not, which prevents a team which is sure to be eliminated from throwing a late round-robin match to help their friends in later days.