Playing the same pair twice

[Deanrover]

:) Hi there. As you may know, we haev two broad types of tourneys. Untimed tourneys have no time limit to play the boards. The two fastest tables to finish round 3 say, pair off to play each other in round four. The way things are at the moment, it is possible for pairs to face each other multiple times during a single tournament. Many people have complained about this. Personally, I would like the director to be able to determine whether pairs can face each other more than once or not.

 

i) What do you think of this?

ii) To Uday/Fred - could you do this?

 

DBS

[fred]

:) Hi there. As you may know, we haev two broad types of tourneys. Untimed tourneys have no time limit to play the boards. The two fastest tables to finish round 3 say, pair off to play each other in round four. The way things are at the moment, it is possible for pairs to face each other multiple times during a single tournament. Many people have complained about this. Personally, I would like the director to be able to determine whether pairs can face each other more than once or not.

 

i) What do you think of this?

ii) To Uday/Fred - could you do this?

 

DBS

 

You raise a good point, but the Directors are already

overworked and one of our basic goals is to have the

software handle as much of the tournament organization

as possible.

 

Probably we will implement an untimed movement with

no playbacks at some point, but we will use software

to do the pairings (as opposed to the Director).

 

Fred Gitelman

Bridge Base Inc.

www.bridgebase.com

[QQSV]

I would like to echo this request in a bit stronger fashion, because I think replays seriously damage a movement. Duplicate movements are supposed to give all the pairs a (relatively) equal shot by allowing pairs to play against a bunch of other pairs. But consider a recent tourney experience of mine: I played a twelve round (one board per round) tourney. In that tourney I played one pair five times and another pair three times. If we were stronger than that pair that I played five times, then not only was I able to pile on the IMPs (thus artificially inflating my score), but I also deprived the other strong pairs of the chance to grab IMPs from them. Similarly, if we were weaker, then my IMP score was massively depressed. The result was that the final standings lost meaning. Indeed, my goal should have been to play at the same speed as a pair that was much weaker than we, and to worry less about playing the best bridge possible.

 

By playing this pair repeatedly we were both able to complete play more quickly than we might have otherwise. But if we cared about the results we then had to wait just as long (if not longer) because the slow pairs (who were now paired with other slow pairs, thus making them REALLY slow!) took forever to finish, and the results had to wait for everyone.

 

I agree with Fred that the software (not directors) should do the pairing, but I think that there should be no option here: repeated play against a pair should be unconditionally prohibited by the software. It would be more fun (I got tired of saying "Hi again again again again!" after a while), would be more fair (as I explained), and would get the tourney over quicker (because the slow pairs couldn't get stuck playing against one another indefinitely.)

 

Please consider this as you work on the next version of the movement scheduler. Thanks!

 

--Q

[uday]

I'm all for change here; but consider this (assume unclocked)

 

Right now, the way it works is : As the tourney starts, the field is divided into sections of up to 15 tables. Each section moves independently of the other.

 

A section can have either an odd number of tables, or an even number of tables (sorry, not trying to treat anyone like an idiot here).

 

Even number of tables:

When a table finishes a round, the software waits until another table finishes the same round. Then, the two tables are paired against each other. If a unique fit cannot be found, the software shrugs and just swaps EW pairs.

 

Odd number of tables:

Each pair of tables that finishes is matched up as described above. Evntually, 3 tables are left. These 3 tables now wait for each other and are matched up together.

 

 

 

The whole point of an unclocked movement is to allow fast players to wrap up the tourney without waiting. Typically, on an 8 bd tourney ( normal time: 1 hr ) the fast pairs finish in 40 minutes and the slow pairs take 1 hour and 20 minutes.

 

So: how do we improve this?

 

a) Leave it alone

;) Wait for more tables each time before matching up. Maybe wait for 3 tables to finish, whether in even or odd sections. That will make things better, if not perfect.

c) Leave it alone, but make this modification. If a pairing would cause a playback by force , then wait for another table. The danger here is that it is possible that this will sometimes make everyone wait for the slowest table. Perhaps this could be solved with some smart matchup logic, but i fear that the programming for that would be complicated.

[Cave_Draco]

There are "Games Theory" solutions for incomplete tournaments... My books are in storage, :-[.

"Swiss" was a manual attempt at a solution but I feel sure that there is PD code that addresses the problem, ;D.

 

Maybe, hrothgar is aware of a Games Theory solution? The code shouldn't be that difficult, maybe a wait after the first 2-3 rounds?

[hrothgar]

Simple question here:

 

This movement seems deliberately designed to allow players to make "efficient" use of their time.

The primary goal is to move players through the tournament quickly.

 

In this case, why preserve the notion of "sections".

I would expect that the end goal could be accomplished most efficiently if everyone were lumped in a single pool, and than matched by speed.

 

In contrast, I would think that a section breakdown would be most useful in an environment where the goal was to match every pair against very other pair

[uday]

I preserve sections for two reasons

 

1) We use a lot of older BBO code. This code is retricted to 16 results per board. We could change this but this would be a non-trivial effort.

 

2) Here, in ACBL land, we are aware that maximum masterpoints can be issued with section sizes of 15 tables. True we dont issue masterpoints yet, but someday, perhaps we will

 

 

The goal of the clocked movement is to create a movement that

a) Finishes at a known time. This matters to some directors and probably some players

B) Can run unattended even if a pair is missing (a sitout is generated for an opponent)

 

 

The goal of the unclocked movement is to create a movement that

a) Allows faster player to finsh play quickly

:) Allows a more relaxed playing environment ( people dont get so itchy about slow play)

c) Allows slower players to avoid feeling rushed

 

 

The ultimate goal of any movement would be to create a movement that

a) Does not require a director

B) Is self healing when players vanish or stop responding

c) Allows the insertion of pairs after the tourney starts

d) Ends at a known time

 

I'm nowhere near achieving this goal, i think :) But we will get there someday

 

 

I believe that I can safely swap players between sections easily enough between rounds if i need to.

[mink]

Uday wrote:

The ultimate goal of any movement would be to create a movement that

a) Does not require a director

B) Is self healing when players vanish or stop responding

c) Allows the insertion of pairs after the tourney starts

d) Ends at a known time

 

I described a movement that comes close to this goal early this month, but still waiting for feedback. Here it is again:

 

There is a timelimit for each round, which can be as long as the host desires (I would vote for 6 minutes per board). Next round starts when the timelimit is over AND at least 80% of the tables have finished - in that way, the timelimit is automatically exceeded if there are difficult boards involved. When the next round starts, tables that are not yet ready are allowed to finish the board they are currently playing. But when that board is finished, the round is over for that table, and both pairs are seated for the next round already in progress. This means that there is never an unfinished board due to the timelimit that would require a director to assign a result, but maybe there are boards that could not be played. For these, the software should assign an artificial score, giving average- for a pair that caused the delay and average+ for an innocent pair. A pair which is late for one round and therefore starts the next round later than most others, can try to play faster in order to catch up with the rest and thereby avoid to miss a board. In order to be able to determine which pair is responsible for slow play the software must measure the thinking time of each player. It is possible that both pairs receive an average- for a board that is missed due to slow play of both pairs in a previous round.

 

If there are less then 2 minutes left for the current round, a new board of that round is not started anymore but missed.

 

If a player is not online when a new round starts, that pair is not seated until he reconnects. Software should monitor the time a player is offline and remove the pair from the tourney if this time exceeds a percentage of the elapsed time. If a player is not online for 3 minutes without returning, the board in play should be cancelled and the director should be automatically called to assign a result. There should be no penalties for bad connection, though, and a pair should not be made responsible for a delay caused by bad connection - in such a case both pairs might get a average+ for a missed board.

 

In a swiss movement, of couse you cannot seat the pairs for the next round based on the results of the current round if not all tables are ready. Therefore I suggest to take the results of the second last round as the basis for seating the pairs for the next round. Seating for rounds 1 and 2 is done at random. Only pairs that finish a round in time (this includes those who missed a board as there was less the 2 minutes time) are seated in the swiss way. The slow pairs are seated in the order they finish. Software should refrain from seating 2 pairs if that would cause any 2 of the remaining unseated pairs to play against each other again.

 

As the results of the previous round are known prior to the end of the current round, seating for the next round can be done premature if fitting pairs finish the current round prior to its end. In that way all the moving of pairs is not required to happen at the same time, thereby avoiding a peek of system load. And the pairs could introduce themselves and inform each other about the systems they play. However, giving the cards to these early-fromed tables should be delayed until the start of the next round, as it would be unfair if some had more time for the boards than others.

 

Of couse, all ideas presented here would still make sense if there is no swiss seating but all pairs are seated randomly in all rounds.

 

Advantages achieved:

• no unfinished board
  
• time the tourney takes is known in advance, apart from slow play in the last round
  
• no need to find subs (but still possible)
  
• very few work for directors
  
• swiss seating possible
  
• sitouts caused by odd number of pairs tend affect those who played slow
  
• slow players have a reason to play faster
  

Karl

[andych]

Looking forward to this enhancement not to play with same pair twice.

I have played in some tournaments playing with more than 1 pair twice...

 

Is it possible to setup a poll for the most desired enhancements?

[DrTodd13]

I played an 8 board tourney this weekend. My partner

and I played one pair 3 times, a different pair 3 times,

and two other pairs one time each. This seriously

damages the meaningfullness of the movement. It

also detracts a lot from the fun of the event.

 

Todd

[uday]

I can't completely prevent playbacks.

 

Currently:

 

pairs:

clocked movements have no playbacks

unclocked movements have playbacks

 

Indy:

clocked movements have some small number of playbacks

unclocked movements have more playbacks

 

 

Next server, which we're testing, is better about unclocked movements.

 

The bottom line is that unclocked movements cause playbacks. I am not completely sure why directors seem to prefer unclocked movements, but i'm guessing it is because so many players are unwilling to accept an AVE-MINUS for slow play, and the director doesnt want to deal w/the adjustments

[DrTodd13]

I think an interesting idea could be to give each pair (or player in an indy) a

certain amount of playing time for the whole tourney. Each pair could choose

to use that time when they needed, e.g., particularly difficult hands, etc.

If you exceed your time allotment, there are a variety of things that could be

done. They could be fined a certain number of IMPs/MPs based on how much

they exceeded their allotment. They could also just not get to play the rest

of the hands and get a certain penalty per unplayed board. I guess this could

result in one pair who has no one to play. If this is the case, I guess they get

ave+. If you wanted to get really fancy, you could even compute time used

exactly for the pair rather than just taking the total playing time of the hand

and assigning that to each pair. This would obviate situations in which one

pair just happened to play a bunch of slow pairs.

 

Clocked tourneys: what is frustrating is that sometimes you finish the round and

have 8 minutes of idle time OR you get difficult hands and you are pressed for

time. The playing time of hands could be very bi-modal.

 

Unclocked tourneys: what is frustrating here is that so many people finish

30 or 45 minutes before other people and have to wait that long for the result.

It is also frustrating that you can't join another tourney even though you've

finished play in this one and you have no guarantee how long it will last.

[uday]

I have to think about your track-clock suggestion.

 

In the meantime, what I've done (for the next version of the server, out a few days after the Bermuda Bowl) is to skip boards that start "too late".

 

That is, if there are fewer than 5 minutes in a round, and a new board is started, the server assigns averages for the rest of the round.

 

I'm not sure what effect this will have. I am sure that it will speed up the very slow pairs in unclocked movements. The host will be able to control this, somewhat, by fiddling with the time assigned per board.

 

For clocked movements, this will have the effect of preventing some of those requests-for-adjustments, as well as reducing (somewhat ) the wait at the end of each round.