Fairness of movements online
There's been a lot written about the fairness of movements, especially where the movement isn't all play all (or all N/S play all E/W).
I have been unable to find any that address the fundamental difference with online play - that they are barometer events (so everyone plays every board).
As far as I can tell, this negates most of the arguments as to which movement is fairest, as (assuming no manipulation of seating). they're all pretty much the same.
Yes you can get unfairness (a pair plays a flat board against a strong pair but a "volatile" one against a weak pair), but that's just luck, not anything inherently wrong with the movement.
Comments?
Comments
Balance still has an effect with barometer movements. It is affected by how many times you play in the same direction as each other pair and how many times in the opposite. For example, in the recent 5-table Open trial, it was necessary for one table to arrow-switch half of each round in order to achieve perfect balance.
Ian McKinnon has written about this. https://qldbridge.com.au/wp-content/uploads/2021/12/BRIDGE_DIRECTING_COMPLETE.pdf
...and I always thought the M39 5 table movement was balanced but it is not.
Everybody compares 4 times with pair 10 but taking pair 2 as an example, they compare 3 times with 1, 3 5 and 8 and 5 times with 4, 6, 7 and 9.
Following on from this is it possible and has anybody written a balanced movement without resorting to arrow switches?
Hasn't the narrative and switches on pp 29-30 now been discredited such that we now arrow switch between 1/7 and 1/8?
Research from a while ago:
Hinted that BBO uses a Flower movement; this is not well balanced.
RealBridge has a movements for each number of tables that have been optimised for balance, but it assumes all the rounds are played and the balance is not so good if Rounds are reduced.
This all might be different now
M39 5 table has a Manning SDEV factor of 1.81, which isn't bad. I don't think you can get a perfectly balanced movement with an odd number of tables
Yes - if you analyse comparing two pairs playing (mainly) in the same direction it comes out at just over one in eight boards. This is not the case if you analyse two players playing (mainly) in opposite directions.
(technically the total matchpoints is R. (T-1) and total opponents is 2T-1 giving an average of R.(T-1)/[2(T-1)+1] which is a bit less than a half of R. Since the level of competition is generally (R-2Q)-2Q where Q is the number of boards arrowswitched, you can simplify such that 4Q = R - (a bit less than half R) so Q is (a bit more than R/8).
This of course only applies when comparing two pairs that don't play the same arrowswitched boards i.e. are some distance apart. Those that play the same arrowswitched boards have increased competition.
One consequence of this is that you should aim to play the second best pair in the room on an arrowswitched round. They should be amenable to this of course since they are playing the best pair in the room on an arrowswitched board.
I suppose that if you are playing 12 rounds of 2 then in theory you should arrowswitch the last board of the penultimate round, as well as that of the last round. I don't think EBUScore has that movement in its list though!
Barometer basically does nothing to influence movement fairness – every non-barometer movement in which every pair plays every board can be made into a barometer movement with equivalent balance by simply taking (for each pair) a list of the (opponent, board) matchups they are scheduled to play, then playing them in order of board number rather than in the order they would otherwise have been played.
This means that it's possible to create some very unfair movements by taking an unfair all-players-play-all-boards movement and barometerising it. For example, you can imagine a movement in which half the rounds (i.e. far too many) are arrow-switched (causing players sitting in the same direction to become allies of each other, i.e. one of them doing well helps the others to also do well). If you make that into a barometer movement, it will still have the same very unbalanced balance properties.
In general, when a barometer is unfair, it's generally because it's arrow-switched incorrectly (and arrow-switching a barometer correctly is hard because you have to switch some of the tables but not others on most or all of the rounds – switching an entire round would do nothing because it would be equivalent to just rotating the board being played that round). BBO's "Howell" (not actually a Howell) movement is particularly bad at this, creating a setup where some pairs of players are superopponents who are sitting in the same direction almost every round they don't play each other, whilst some sit in the opposite direction every round and thus have zero competition (or are allies if the round on which they would play each other is truncated); the way it chooses which pairs oppose which on each board is fine, but the directions in which they play are wrong.
Are there any discussions about barometer events. Maybe the arrowswitches should run contrary to the movement of the pairs. let's see for 7 tables and 7 rounds.
Round 1: Table 1 arrowswitched (pair affected = 1 EW)
Round 2: table 7 arrowswitched. (Pair affected = 6 EW)
Round 3: table 6 arrowswitched. (Pair affected = 4 EW)
Round 4: table 5 arrowswitched: (Pair affected = 2 EW)
Round 5: table 4 arrowswitched: (Pair affected = 7 EW
Round 6: table 3 arrowswitcehd: (pair affected = 5 EW
Round 7: table 2 arrowswitched (pair affected = 3 EW)
That seems to work well in some respects BUT the same board is arrowswitched. To get that to change it looks like we will have to arrowswitch jumping tables (1 - 6 - 4 - 2 - 7 -5 -3 ) affecting pairs (1, 5, 2, 6, 3, 7, 4 EW) respectively.
OK - so we could jump in the opposite direction.
Agreed, but if a movment is curtailed, that's not the case.
I take your point though.
If you take a standard 7 table, 7 round Mitchell with an arrow-switch on the last round and, keeping the same [NS, EW, Board] triples but playing as a barometer, you get:
ie, per round, switch on table: 2, 3, 4, 5, 6, 7 , 1
Note that EBUScore doesn't do this when you tick Option "Barometer Pairs", it just changes the boards, rendering the arrow-switch pointless, giving the same balance as the original unswitched mitchell
What about barometerising a Mitchell that was arrow-switched on the first round instead of the last round? I think that might produce an "arrow switch if the round number matches the table number" rule, which is at least an easy rule for the Director to remember (and visit the correct table to tell them to switch).
@ais523 said:
Yes, doing this gives switch where table = round