Draws on RealBridge
During the recently concluded HCL Swiss League, and other Swiss Leagues that BridgeFromHome has run on the RealBridge platform, many people have often pointed out how, the draws being made by RealBridge are inconsistent with the draws that would have been made had we been making the draws manually, or using other scoring softwares.
We have been asked many questions on the topic ranging from "Is this legal?" to "Why is RealBridge making draws in such a cock-eyed fashion?"
In this post, we will attempt to answer some of the queries that have been raised for the general education of the player. We would have to stress, though, that first of all, BridgeFromHome has no control over the kind of draw generated by RealBridge. Furthermore, BridgeFromHome has no individual personal preference for either type of draw.
This is a rather long, and perhaps a bit technical post. So, if this doesn't really interest you, do feel free to go on enjoying games on RealBridge as you always have.
How does RealBridge compute draws?
Most softwares use a 'top-down' approach to computing a Swiss League draw i.e. The top ranked team meets the second ranked team and so on, subject to the proviso that the teams have not played each other before.
RealBridge computes draws slightly differently
It uses an algorithm that "creates 'minimal badness', where the more different teams' scores are, the higher the badness would be for them to play each other. (But playing the same opponents again would be _very_ bad.) It considers the entire draw as a whole, rather than starting at the top at creating match-ups from there" (in the words of Graham Hazel, the developed of RealBridge software).
At the risk of appearing foolish, we believe it uses an algorithm which minimizes the square of differences between the scores of the competing teams subject to the proviso that teams should not play each other more than once i.e. it is some kind of regression model.
Is the methodology legal?
In the words of Gordon Rainsford, Chief Tournament Director, EBU,
"These are not matters dealt with under the laws but you could specify them under your regulations if you wanted".
In the words of Laurie Kelso, secretary of the WBF Laws Committee and WBF Director
It is the responsibility of the Regulating Authority to authorise the use of appropriate scoring procedures and software.
That means if your are using RealBridge as the platform to run a 'Swiss' Teams event, then the Supplementary regulations should probably specify the actual pairing method that will be used.
In short, there is no question of legality involved in using a methodology of this type. Rather, it is up to the tournament organizer to specify the methodology that will be used.
And as long as RealBridge offers only this methodology, we are constrained to use it.
Is this methodology desirable?
In the words of Laurie Kelso, (who was also CTD HCL, 2019)
I am reasonably familiar with the RealBridge philosophy regarding 'Swiss' draws (in both Teams and Pairs events). I too have had to field various player questions and complaints about the outcomes from this type of approach.
My (personal) observations would be:
a) Appropriateness
This depends a lot upon the event structure and your ultimate goal. If the objective of the Swiss is to qualify a number of teams to the next stage of an event, then I think this approach has a lot of merit in that it tends to produce relatively 'equivalent' match-ups amongst the upper third of the field and hence those teams in contention for the last few qualifying spots will theoretically be paired against relatively similar strength opponents (i.e., each match-up will have a very similar score differential).
If however, the primary objective is to obtain a single winner from amongst a large number of entries, then the method is very much sub-optimal. This is because less of the top contending teams ultimately get to play one another (e.g., I have frequently observed instances where the current teams running first and second are never drawn against each other). I also think the effects of 'underswissing' are more accentuated - which leads to a greater number of rounds being necessary in order the achieve the same confidence levels in respect to accurately determining a winner.
In Conclusion
This methodology has been developed working closely with senior officials from regulatory bodies in Europe, Australia and experts in the area of Bridge movements like Ian McKinnon.
Therefore, on that score, we should be able to rest assured that this is a "desired" form of computation rather than a sub-optimal one which the bridge world is being forced to adopt thanks to RealBridge.