[karlson]

I'm sure Jan will also comment, but this was discussed a little bit at the meeting. The fact that the FISU trials are hopefully going to be after the collegiates suggests to me that while the college winners are encouraged to play in them and probably have a good chance to make the team, they are welcome to pick up teammates or partners from other schools.

I do think that promoting the collegiates and encouraging people to participate in the fisu trials is a fine idea. Also, we're apparently sending two teams, which gives some wiggling room. I'm not sure there's going to be an enormous turnout of non-acbl-collegiate people for the fisu trials, so I don't think this is false encouragement even if they are going to have to play in the trials.

[JanM]

The announcement you saw was published before our Junior Committee meeting in San Francisco. The reason for the "wiggle room" was to let us decide what would be most sensible. As Karlson said, at the meeting several people suggested that Intercollegiate players might be happier to form teams with players from other schools than to be locked into the team with which they competed in the Intercollegiates. That's why we're hoping that ACBL can change the dates of the Intercollegiate finals to Thursday and Friday, so the FISU trials can be after the Intercollegiate finals. It is also correct that we plan to send 2 teams to the FISU, so depending on the turnout for the Trials, it is still possible that the Intercollegiate winners could go intact. Which team is designated USA1 and which is designated USA2 is, of course, completely irrelevant - it's just a way of distinguishing between the teams.
As for the "citizen" question, US citizenship is not required for USBF membership, but "residence" is. The definition of resident is a little complicated, combining a requirement of having lived in the US for 50% of both the last 12 months and the last 24 months with a requirement of being eligible to compete in the next World Championship. For Juniors, though, only the residence requirement applies, since the WBF rules on competing for different countries within a short time span do not apply to Juniors. Thus any Junior who has lived in the US for 6 of the last 12 months and 12 of the last 24 months and intends to remain in the US is eligible for USBF membership and to compete on USBF teams. I missed the "citizen" language when I looked quickly at the announcement being posted about the Intercollegiates, although given the complexity I'm not sure it would have made sense to be more accurate .

[MiniMeck]

Thanks for the detailed responses and explanations Jan.

[JanM]

Today's good news
The ACBL Intercollegiate finals will be held on Thursday and Friday, July 17 & 18 (if the days and dates don't match, trust me on the days, not on the dates). Therefore, the FISU trials will be Saturday & Sunday, July 19 & 20 (ditto). Players in the Intercollegiates will be encouraged to play in the FISU trials, either on their Intercollegiate team or on other teams. Unless the turnout for the FISU trials is significantly lower than we hope, it will be used to select both of the USBF's teams for the 2008 FISU event.
More good news - there will be 8 teams in the Intercollegiate finals. Thursday will be a complete Round Robin. I think Friday will be a KO with the top 4 teams qualifying, but that's just my guess.

[kfay]

JanM, on Dec 14 2007, 07:03 PM, said:

Today's good news
The ACBL Intercollegiate finals will be held on Thursday and Friday, July 17 & 18 (if the days and dates don't match, trust me on the days, not on the dates). Therefore, the FISU trials will be Saturday & Sunday, July 19 & 20 (ditto). Players in the Intercollegiates will be encouraged to play in the FISU trials, either on their Intercollegiate team or on other teams. Unless the turnout for the FISU trials is significantly lower than we hope, it will be used to select both of the USBF's teams for the 2008 FISU event.
More good news - there will be 8 teams in the Intercollegiate finals. Thursday will be a complete Round Robin. I think Friday will be a KO with the top 4 teams qualifying, but that's just my guess.

Geeze. Why go from 8 to 4 to 8? This really bugs me due to what happened to my team last year. Glad the ACBL saw the error in reducing the field but this sort of thing has happened to me a million times.

[TimG]

kfay, on Jan 30 2008, 04:36 PM, said:

Glad the ACBL saw the error in reducing the field but this sort of thing has happened to me a million times.

I've heard of the five and six year college plans. But, the million year plan?

[hrothgar]

kfay, on Jan 31 2008, 12:36 AM, said:

JanM, on Dec 14 2007, 07:03 PM, said:

Today's good news
The ACBL Intercollegiate finals will be held on Thursday and Friday, July 17 & 18 (if the days and dates don't match, trust me on the days, not on the dates). Therefore, the FISU trials will be Saturday & Sunday, July 19 & 20 (ditto). Players in the Intercollegiates will be encouraged to play in the FISU trials, either on their Intercollegiate team or on other teams. Unless the turnout for the FISU trials is significantly lower than we hope, it will be used to select both of the USBF's teams for the 2008 FISU event.
More good news - there will be 8 teams in the Intercollegiate finals. Thursday will be a complete Round Robin. I think Friday will be a KO with the top 4 teams qualifying, but that's just my guess.

Geeze. Why go from 8 to 4 to 8? This really bugs me due to what happened to my team last year. Glad the ACBL saw the error in reducing the field but this sort of thing has happened to me a million times.

I echo KFay's confusion

Running a round robin to reduce an 8 team field to 4 doesn't make any sense...

I'm almost positive that running a series of short Round Robin matches will significantly degrade the accuracy of the event. Its much better to allocate two full days to a straight KO. You'll have more than enough time to add an additional round to the event and also lengthen the number of boards played each round.

I'm 99% sure that this can be demonstrated using a Monte Carlo simulation.

In all seriousness. This is (supposedly) a championship event. It shouldn't be designed as a crap shoot.

[helene_t]

hrothgar, on Feb 1 2008, 04:35 PM, said:

Running a round robin to reduce an 8 team field to 4 doesn't make any sense...

I'm almost positive that running a series of short Round Robin matches will significantly degrade the accuracy of the event.  Its much better to allocate two full days to a straight KO.  You'll have more than enough time to add an additional round to the event and also lengthen the number of boards played each round.

I'm 99% sure that this can be demonstrated using a Monte Carlo simulation.

Why? Let me argue the opposite. Suppose there are two strong teams that might as well flip a coin instead of playing a short match. Then what you don't want is those two teams to meet in an early phase of the KO where the matches are shorter. In a round robin they are bound both to end among the best 3 which means that they will not meet before the final. Besides, with an RR you can improve the accuracy of later matches through a carry-over.

I have here made the assumptions that
- relative RR results are adequate surrogates for a head-on
- RR is not more expensive per board than KO
- The final would be longer than the semi- and quarterfinals

Those assumptions could be criticized but it's not obvious to me that they are inferior to some model that would favor a complete KO.

[david_c]

It's nothing to do with me, but I've got to disagree with Richard - the format seems ideal. With a straight KO a relatively weak team could get drawn against a very strong team and their only experience of the event will be getting hammered. Having a round-robin ensures that everyone gets to play at least one day in the main event, and gets to meet everyone else. I don't even think it makes much difference to the "accuracy" of the event - the semi-final and final are only reduced by 25% in length. But who cares about accuracy anyway? It's for the trials that accuracy would be more important.

[Gerben42]

I am happy to do some MC simulations again, of course I will include a parameter that we are talking about junior bridge --> more swings

[mycroft]

I like the double-knockout. Yes, it's a pain, yes, it causes issues with timing and byes, it takes longer, and nobody knows how to run one, but it does do a good job of balancing "one and out" with "draw hit 1v2, which meant the final was the first round".

Making it work is an exercise for the reader.

[hrothgar]

Gerben42, on Feb 1 2008, 06:53 PM, said:

I am happy to do some MC simulations again, of course I will include a parameter that we are talking about junior bridge --> more swings

Hi Gerben

I'd certainly appreciate seeing some real results. (One of these days I'll write the necessary code in MATLAB runs this through the Statistics Toolbox and the Optimization Toolbox and see what spits out).

In any case, from my perspective, the best design criteria is measuring how often the tournament identifies the best team. I'm not really concerned about the rank ordering of teams 2-8.

I'm not sure what folks are planning for the conditions of contest. In general, a two session 7 round Swiss Type format uses either 7 or eight board rounds. If we assume that the event will last for three sessions we could probably increase the round length to 12 boards. (I think that this would be a reasonable starting assumption)...

As for the KO matches, I think that we could have considerable flexibility in designing the length of the different rounds.

I'd start by assuming something like the following

Two Day schedule

Round of 8 = 1x32 board session
Round of 4 = 2x32 board sessions
Round of 2 = 3x32 board sessions

One Day schedule

Round of 4 = 1x32 board sessions with seeding from the Round Robin
Round of 2 = 2x32 board sessions

[FrancesHinden]

hrothgar, on Feb 1 2008, 03:35 PM, said:

Running a round robin to reduce an 8 team field to 4 doesn't make any sense...

I'm almost positive that running a series of short Round Robin matches will significantly degrade the accuracy of the event. Its much better to allocate two full days to a straight KO. You'll have more than enough time to add an additional round to the event and also lengthen the number of boards played each round.

I'm 99% sure that this can be demonstrated using a Monte Carlo simulation.

i'm sorry but an mc simulation won't "prove" anything. I'm pretty certain i can demonstrate that any (vaguely sane) format i happen to prefer is 'best'; it will depend on the characteristcs of the teams concerned.

[helene_t]

FrancesHinden, on Feb 1 2008, 06:22 PM, said:

i'm sorry but an mc simulation won't "prove" anything. I'm pretty certain i can demonstrate that any (vaguely sane) format i happen to prefer is 'best'; it will depend on the characteristcs of the teams concerned.

I think you exaggerate a little. Some tournament forms exist for weird historical reasons or because of other success criteria than accuracy (see e.g. David's post). But of course an MC simulation only settles the issue given whatever assumptions one bases the simulation on.

As a simple model one could use what Cascade used for the PABF final and I used for the Bermuda Bowl: IMPs per board are normal distributed, independent and with constant variance, with a mean equal to the difference between the strength factors of the two teams. I think this would lead to something like a Swiss survivor event.

If one believes that there is a random interaction effect related to each pair of teams, this would favor an RR only structure without a KO phase, or a survivor event with slower elimination than in the simple model. OTOH if one believes that matches between strong teams and weak teams are not so informative, either because of higher variance or because some strong teams are good at crunching weak teams and that is not the quality we are looking for, one should strive to get the weaker teams eliminated ASAP.

[hrothgar]

FrancesHinden, on Feb 1 2008, 07:22 PM, said:

hrothgar, on Feb 1 2008, 03:35 PM, said:

Running a round robin to reduce an 8 team field to 4 doesn't make any sense...

I'm almost positive that running a series of short Round Robin matches will significantly degrade the accuracy of the event.  Its much better to allocate two full days to a straight KO.  You'll have more than enough time to add an additional round to the event and also lengthen the number of boards played each round.

I'm 99% sure that this can be demonstrated using a Monte Carlo simulation.

i'm sorry but an mc simulation won't "prove" anything. I'm pretty certain i can demonstrate that any (vaguely sane) format i happen to prefer is 'best'; it will depend on the characteristcs of the teams concerned.

Hi Frances...

The word "prove" might be a bit strong. Even so, I think that MC simulation is the best way to approach this type of issue.

Formal models focus the decision making process.

You're quite right when you note that the data that you feed into a model will have a significant impact on results. However, I would much rather have a specific technical discussion about variance in the skill between teams, distributions of board results, and appropriate sensitivity analysis rather than some "warm and fuzzy" talk about my "feelings" about how some event "should" be run.

[cherdano]

Well, you certainly won't prove that the selected format is "wrong", since you can't prove that your goals are correct and other goals (choosing the second best team, giving every team at least one day of play) are incorrect.

[hrothgar]

cherdano, on Feb 1 2008, 08:45 PM, said:

Well, you certainly won't prove that the selected format is "wrong", since you can't prove that your goals are correct and other goals (choosing the second best team, giving every team at least one day of play) are incorrect.

Here once again, I'd argue that formal modeling is the way to go:

You're perfectly correct that a formal model won't determine whether its more valuable to select a second place team or guarantee that every team plays for a full day.

However, adopting a formal model forces you to make explict choices of these types of tradeoffs.

[JanM]

I don't have anything to do with designing the format for the Intercollegiate Finals, but I do think that allowing all of the teams to play for at least one day is a very legitimate goal for this event. Of course, one purpose of the event is to choose a champion. But another purpose is to promote bridge, and specifically to promote bridge among college students.
I often hear complaints after the first day of the GNT National Finals from people who have been eliminated and don't like the fact that they "came all this way" to be eliminated after only one day of play. I'm sure that if I hung out more in the Flight B & C areas where teams are eliminated after one session, I'd hear more grumbling. And grumbling is something that the organizers definitely want to avoid in the Intercollegiates (not that organizers don't in general want to avoid grumbling, just that in some events the grumbling is more acceptable because other goals are more important).
In this particular event, I really do believe that it is appropriate that the #1 goal be for the participants to have a good time, not in the sense of being at a party, but in the sense of seeing the activity of participating in a serious bridge competition as something that they enjoy doing and would like to continue doing. We need to do everything possible to see that all of them come back again when it's on their own dime.
This is also an event where the seeding figures to be terrible - most of the players won't have any kind of track record on which to base seeding. That argues against a straight KO. And the short time argues against a double elimination KO (I think). So a combination of Round Robin and KO seems like a good solution to me.

[mike777]

edit