[modicum]

I'm a newish bridge player who has been playing on BBO quite a bit, especially in ACBL tournaments. I have learned a lot and had a great deal of fun playing here. The great thing about BBO tournaments is that you can play whenever you want, they only take an hour, and there is always good competition available.

One of the big disadvantages of BBO tournaments over going to a real-world club or tournament is that the results seem to be much more erratic. (All of my data here is anecdotal, I would be interested if someone has studied this more closely.) When I play with one of my regular partners at a regular night in the club we tend to get about 45-55%. Anything much higher is a pleasant surprise, and lower is a disappointment. At BBO, I find that I can play with a regular partner and go from 30% one game to 70% the next game with little change in play. This volitility actually helped me as a new player, becuase I would sometimes place in BBO tournaments before I was placing at the club. But if we want winning BBO tournaments to mean something and deserve the same level of respect as winning real-world tournaments, then they should strive to be as exact a measure of skill as possible.

The obvious culprit here is that in online games you only play against four other pairs, where a real-world club or tournment you tend to play against 5-13 other pairs. So a major determinate of how well you do is how good your opponents are. The easy solution to this would be to extend tournaments for two hours so players could play twice as many opponents. This is unappealing, as it undermines the benefits of online bridge: short games with plenty of players.

The better approach would be to stop looking at online bridge games as analogs of real-world games. Much of the process BBO uses for these games is exactly the same as it would be in offline bridge games: play is divided into similarly-sized sections, players play several boards against one set of opponents before advancing to the next one. But online bridge lacks many of the constraints of real-world bridge: players an all play the same board simultaneously, they are always at the right table, and it doesn't take any time to move between tables. As a result, it seems like in an online tournament we can allow players to play more opponents, and opponents of similar skill, without lengthening the tournement.

The process I would propose would be much close to the way that Swiss games are played today:

1. In the first round each pair would be matched against a random set of opponents.
2. In the second and later rounds, pairs would be matched based on their performance in the earlier rounds. So the winning pairs would play more winning pairs, and the losing pairs would play more losing pairs.
3. Rounds would last for 1-2 boards, and there would be 5-10 rounds.

So by the end I will have played more different opponents, and my later opponents should be at a similar level to me.

It would be very different from real-world bridge, and matching winners probably means that the results would be quite flat. But neither of those are bad things. It would be a game that keeps the things I like about online tournaments, but makes them a better game of skill.

Thoughts?

Aaron

[Gerardo]

BBO supports swiss (kinda: it allows for opponnent repetition) tournaments already.

[helene_t]

Hi Aaron, welcome to the forums! I hope you will enjoy our company!

What you suggest is similar to Swiss movement. It's a great format for clocked tournaments. There are not so many of them but look for them, if you find one you can ask the tournament host when he expects the next one to run.

The greater volatility is partly due to the smaller number of boards. Swiss movement does not address that problem, but it does address the larger heterogeneity of the field. I also think there are more wild gambles online than in real life.

[TimG]

modicum, on Feb 17 2009, 12:16 AM, said:

The obvious culprit here is that in online games you only play against four other pairs, where a real-world club or tournment you tend to play against 5-13 other pairs.

I think the obvious culprit is the reduced number of boards played. This has something of a snowballing effect: players think they need a very big game to win such a short event, so they take more gambles in an effort to win. Playing a solid 58%-63% game will never win, so some people don't try to play solidly, they play to make action in hopes of scoring 65%-70% (or higher).

[Rain]

Isn't it true that in most of these big tourneys, results are compared against a far larger field? That would average results more, no?

I've always felt the reason for the fantastic results (I had a tourney about 2 weeks ago where the winners won with 80ish%!!) is, ahem, the fantastic style of tourney goers. People are chilling, relaxing, not being that serious about the game.

[hotShot]

Imagine a one board tourney, you would obviously need 100% to be the only winner. If you have a shared top you can't be the only winner.

In a 2 board tourney a top and 1 average board produce 75%

In a 4 board tourney a top and 3 average boards produce 62.5%,
while 3 tops and a zero will get you to 75%.

In a 25 board club tourney a top and 24 average boards produce 52%.

As you can see having a tops is very helpful in short tourneys. In a bigger field the usual scores are often shared scores. So their %score gets lower.

People adapt their strategy to win those events by taking more risks to get a top. If they get tops in the first rounds, they might play a little more careful.

If they get zeros in the first round, some just quit the tourney and try to win the next tourney.

[helene_t]

Rain, on Feb 17 2009, 04:29 PM, said:

Isn't it true that in most of these big tourneys, results are compared against a far larger field? That would average results more, no?

Yes but it doesn't solve the problem since no matter how large the number of tables is, it can never reduce the number of rounds required to reach the same accuracy by more than 50%, relative to a 4-table tournament. Relative to a larger tournament, the scope is even smaller.

Assuming everyone "deserves" close to 50% matchpoints and that there are no ties on individual boards, the squared standard error of a 50% score with n tables is 0.25 /(n-1) while the variance between the scores is 1/12. IOW for a 4-table tourney there is twice as much noise as an infinity-tables tourney and you would need twice as many rounds to compensate for it. For a seven-table tourney you would need 50% more rounds than the infinity-tables tourney.

[modicum]

Thanks for the thoughts everyone.

Clearly the number of boards affects the volatility of the result, and if players would play more boards the results would be flatter and fairer. But there is no way people could squeeze more boards into an hour, and the one hour tournament format works well. So I think playing 10-12 boards in an hour is a good format.

Also, the number of opponents you play would affect the volatility of your results. If you played all your boards against the same opponent, then your result would have a lot to do with how good your opponent is. And if you played 12 different opponents the results would be fairer. Given that it doesn't take any time to move between tables why do you only play against four opponents in an hour? This seems like a leftover from playing bridge off-line, where it takes time to move between tables so you'd better play several boards while you are there.

Helene - when you say that BBO supports a 'Swiss Movement', do you mean that it supports a Swiss Team game (with 4-person teams), or it uses a Swiss-style movement in a pairs game? If it is the second, that sounds like a great way to reduce the volatility of our results. While I like Swiss Teams, I don't think they are a good format for quick online games. The teams have to be larger, and you have to trust your teammates. But what I like about playing Swiss in a large field is that they match you against against appropriate opponents based on your results in earlier boards. This helps make sure everyone is challenged. So what I wonder about in our pair games is why does it seem like all of your opponents are randomly selected? Selecting opponents who had performed at a similar level would help to flatten results and makes everyone's game a challenge. This also seems like a leftover, because in an off-line bridge tournament you don't have time to compute new pairing between each round.

[TimG]

modicum, on Feb 17 2009, 12:27 PM, said:

Given that it doesn't take any time to move between tables why do you only play against four opponents in an hour?

While it doesn' take time to change opponents and tables, you do have to wait for everyone to be done. One three-board round will have less variance time wise than three one-board rounds. A tournament with 12 one-board rounds will take longer than a tournament with 4 three-board rounds. The boards in a 3-board 14 minute round might take 7, 4, and 3 minutes and everything finishes on time. If you make it 3 one-board rounds, this pair is going to be a couple minutes late one round.

As to the question about Swiss, yes, I believe BBO supports Swiss Pair movements. Both clocked and unclocked.

[helene_t]

modicum, on Feb 17 2009, 06:27 PM, said:

Helene - when you say that BBO supports a 'Swiss Movement', do you mean that it supports a Swiss Team game (with 4-person teams), or it uses a Swiss-style movement in a pairs game?

Yes, it's a pairs movement. Some TDs will anounce at the beginning of each round "at table 1 we now have USA/Japan Vs Turkey/Bulgaria" or some such, that creates a nice competitive atmosphere

[Gerardo]

On the other hand, I think ACBL tourneys on BBO are not Swiss but fixed movement.

[mtvesuvius]

Gerardo, on Feb 18 2009, 01:57 AM, said:

On the other hand, I think ACBL tourneys on BBO are not Swiss but fixed movement.

ACBL tournements are a Fixed Miitchell movement. I'm not sure if ACBL would allow tournements with a Swiss Pairs movement, but I don't think so. 1 Board tournements are really hard for directors to run, especially speedballs...

[TimG]

mtvesuvius, on Feb 18 2009, 09:23 AM, said:

Gerardo, on Feb 18 2009, 01:57 AM, said:

On the other hand, I think ACBL tourneys on BBO are not Swiss but fixed movement.

ACBL tournements are a Fixed Miitchell movement. I'm not sure if ACBL would allow tournements with a Swiss Pairs movement, but I don't think so. 1 Board tournements are really hard for directors to run, especially speedballs...

I rather think ACBL would allow Swiss pairs as long as they fit within the prescribed masterpoint award parameters. ACBL is generally pretty flexible when it comes to club games.

[barmar]

The reason you tend to play against so few pairs in the online games is due to one of the advantages you cited: the tourneys are only an hour long. Most club games are 24 boards, so you get to play against 8-12 opponents depending on whether it's 3 or 2 board rounds, but these games take about 3 hours. Online games tend to max out at 12 boards, and they use 3 board rounds so they can finish in 1-1.5 hours (speedball vs slowball). If you want the quick games, you have to put up with the scoring disadvantages.

[awm]

First, it's not clear to me why the long rounds. If we are playing twelve boards in an hour, why can't we play 12 rounds of 1 board? So often the ACBL tourneys give me 4 rounds of 3 boards each. I suppose it might take a bit more time to play 12 rounds of 1 board, but it's not like we're moving from table to table or anything so it shouldn't be all that much more time.

Second, I'm not a big fan of swiss movements especially with repeated play. Look at it this way... you'd like in the long run (as number of boards becomes large) for the rankings to reflect skill level. Now suppose that we have a swiss event where the current ranking is fairly accurate. In the next two rounds, you expect to play one team that's better than you, drop a bit, and play a team that's worse than you (or the same in the opposite order) and move back up. Your expected matchpoints is essentially 50% (unless you are the best or worst pair in the entire thing). So rather than reinforcing the current ordering, the result is that almost everyone is pretty average in expectation and you smudge the lines between pairs. Another way to see the same thing is that everyone is playing against a different "strength of field" with weaker pairs playing a weaker field -- tending to imply that everyone should do roughly the same. Swiss events are fairly good for picking the top pair (and bottom pair) and not so good for ranking the rest of the field.

[hotShot]

The swiss movement was invented to find the best player while meeting more other good players than bad players, without playing a complete round robin. In a swiss movement to best players should be in the top few, while the worst player are in the bottom few. You sacrifice a reasonable ranking of the big majority of teams for saving many rounds.

If you play a table based movement like you do offline, each pair is only playing a very small number off opponents. Assume Pair A is much better than Pair B. Pair A unfortunately hits the pairs Worldclass1 to Worldclass4 while Pair B plays the pairs Novice1 to Novice4. Guess who has a chance to score well? The results of such a tourney are completely random. The winners are most likely lucky people who meet weak opponents, that messed up badly.

Given the shortness of online tourneys Swiss is the better format, because most of the good and the bad pairs will score well and most of the bad pairs score bad.

[uday]

Quote

First, it's not clear to me why the long rounds. If we are playing twelve boards in an hour, why can't we play 12 rounds of 1 board?

We could but it would take more than 1 hour -- playing more than 1 bd per rd definitely improves total throughput, and also reduces the impact of thinking for a bit, minor disconnections, Q&A sessions with the opps or TD, and the like.

I think 1 bd per rd would be ideal but i can't see how to squeeze that into an hour. And if I dont the schedule gets ugly real fast ( 13:00 , 14:10, 15:20, etc ).

Two bds per round might be manageable. Maybe i'll try some experiments with that.

[awm]

In fact it's interesting, it seems like sometimes the ACBL tourneys play 6 rounds of 2 and sometimes 4 rounds of 3. It appears to depend on the total number of competing pairs... but there are always way more than 12 pairs so it's not like we couldn't always play 6 rounds of 2... it has something to do with section sizes etc.

[barmar]

I play in the Saturday night speedballs most weeks, and they're always 4x3, never 6x2. Maybe the 6x2's are the slowballs?

[helene_t]

hotShot, on Feb 18 2009, 11:18 PM, said:

The swiss movement was invented to find the best player while meeting more other good players than bad players, without playing a complete round robin. In a swiss movement to best players should be in the top few, while the worst player are in the bottom few. You sacrifice a reasonable ranking of the big majority of teams for saving many rounds.

I might be wrong, but I think swiss is fair for everyone because the field will be relatively balanced. At each table the EW pair and NS pair will tend to be equally strong, so it is not an advantage being compared to EW pairs rather than to NS pairs.

On the other hand, the weaker you and your opponents are, the more random the result. Ideally everyone would play against world class opponents, the the comparison of their results would be the most reliable. So the evidence for number 1 being the best is stronger than the evidence for number last being the worst.