[helene_t]

We play teams-of-8 in regional competition. I.e. two fourhanded teams play against two fourhanded teams from another club, each play two 2x6 board matches against each of the opp teams. At the end the IMPs from the four 2x6 matches are added up and converted to VPs (the 0-20 English scale, not the 0-25 WBF scale).

We have two good pairs and two mediocre pairs, so now the question how to team them up.

If we expect to win we should probably maximize the gain by letting a good pair team up with a mediocre one because two good results at the same board is diminishing return if it happens in the same match (added before the total points are converted to IMPs) but not if they happen in different sub-matches. OTOH maybe volatility is smaller if the two mediocre pairs are teamed up, for the same reason.

I thought that by argument of symmetry it shouldn't matter in terms of expected VPs if we expect to tie with the other team, but maybe that is not necessarily true. Maybe it could depend on the "structure" of the other team. As it happened yesterday they had three good pairs and one really bad one.

Any thoughts?

[FrancesHinden]

This is an interesting question. Made more so because you are using the VP scale in an incorrect fashion.

If I'd understood what you are doing right, then you have teams A and B, the opposing club has teams X and Y. Your team A plays 12 boards against team X and 12 boards against team Y, your team B does the same thing. Each match is imped seperately, and then the total IMPs are added up and the English 20-0 VP scale is used.

The EBU White Book recommends using the VP scale for a match of twice the number of boards i.e. for this 2x24 board match, use the 48-board VP scale. Is that what you do, or do you use the 24-board scale?

If you use the former, it's not really right because it compresses things. If you A team were net 45 imps over 24 boards, and your B team -5 imps, then converting each of these results using the 24-board VP scale you are 19-1 and 8-12 (or maybe 15.5 - 4.5 if you halve the answer), but a net 40 imps is a 19-1 win, while it's only 17-3 on the 48-board scale.

I think the answer to your original problem depends on exactly how you are scoring this.

If the problem is purely to maximise total imps, I think you are right that you should split your 'good' pairs between the teams. As you say the risk is that your good pairs get two good results on a board but this is dampened by the imps scale, while one good result and one mediocre result twice is better.

[helene_t]

Thanks, Frances.

I suppose we use the 48 boards scale.

[cherdanno]

FrancesHinden, on Nov 7 2009, 10:15 AM, said:

If the problem is purely to maximise total imps, I think you are right that you should split your 'good' pairs between the teams. As you say the risk is that your good pairs get two good results on a board but this is dampened by the imps scale, while one good result and one mediocre result twice is better.

But if the two bad pairs form a team, you are also dampening their common bad results. I suppose if you are the better team, you should split the good pairs, and if you are the worse team, you should have them play together.

[helene_t]

As it happened, we split the good pairs. So both teams lost with a handful of IMPs to the good opp team and massacred the bad opp team. This was as expected and I thought that our chances would have been worse if we have teamed up with the other good pair (OK now you all LOL at me for considering myself part of one of the good partnerships, but Manudude03 sometimes makes up for my moronic play).

[FrancesHinden]

cherdanno, on Nov 7 2009, 04:47 PM, said:

FrancesHinden, on Nov 7 2009, 10:15 AM, said:

If the problem is purely to maximise total imps, I think you are right that you should split your 'good' pairs between the teams. As you say the risk is that your good pairs get two good results on a board but this is dampened by the imps scale, while one good result and one mediocre result twice is better.

But if the two bad pairs form a team, you are also dampening their common bad results. I suppose if you are the better team, you should split the good pairs, and if you are the worse team, you should have them play together.

Yes, this is possibly right.

If everyone is playing the same boards, and you split your good pairs, you also want to think about whether they should play in the same direction in each team, or opposite directions (the former will similarly up the volatility, the latter dampen it).

[helene_t]

Good point Frances, which raises the issue of who has seating rights. When we play internal teams competition we throw a coin for the seating rights for the first segment and then reverse seating rights for the second segment. What usually happens in external competition is that the guests will take seats first so people do as if the host team has seating rights, but that may be wrong. I haven't checked the rules.

[pooltuna]

how is seating selection done?

[Mbodell]

Out of curiosity why aren't you playing the same boards in both matches and then just adding up the 4 scores to then get an IMP result (divide IMP by 2 possibly)? Wouldn't that work? So our teams scored +620, -170, +170, -170 and when we do the sums we end up with +450 for +10 IMPs but since it is "two IMP results" convert that to +5 IMPs. Wouldn't this mitigate the seating question since all pairs are compared on each board?

[FrancesHinden]

Mbodell, on Nov 7 2009, 09:06 PM, said:

Out of curiosity why aren't you playing the same boards in both matches and then just adding up the 4 scores to then get an IMP result (divide IMP by 2 possibly)? Wouldn't that work? So our teams scored +620, -170, +170, -170 and when we do the sums we end up with +450 for +10 IMPs but since it is "two IMP results" convert that to +5 IMPs. Wouldn't this mitigate the seating question since all pairs are compared on each board?

You can't score this way*. It multilates the imp scale. If your four scores were +620, +620, -170, -170 then you have scored two game swings, but you only get 14 imps (not 2x10 = 20 imps). If you insist on adding all four scores up, you probably should do something like halve your total points result before imping.



*I know some teams-of-eight events do, but they shouldn't

[Mbodell]

FrancesHinden, on Nov 8 2009, 03:19 AM, said:

Mbodell, on Nov 7 2009, 09:06 PM, said:

Out of curiosity why aren't you playing the same boards in both matches and then just adding up the 4 scores to then get an IMP result (divide IMP by 2 possibly)?  Wouldn't that work?  So our teams scored +620, -170, +170, -170 and when we do the sums we end up with +450 for +10 IMPs but since it is "two IMP results" convert that to +5 IMPs.  Wouldn't this mitigate the seating question since all pairs are compared on each board?

You can't score this way*. It multilates the imp scale. If your four scores were +620, +620, -170, -170 then you have scored two game swings, but you only get 14 imps (not 2x10 = 20 imps). If you insist on adding all four scores up, you probably should do something like halve your total points result before imping.



*I know some teams-of-eight events do, but they shouldn't

You're right, should divide the points in 2. It is sort of like IMP pairs I guess, although even there it isn't the same since if your scores are +2220, 0, 0, 0 or what not the one huge swing would be more muted in pairs.

[gnasher]

Shouldn't you should IMP each result against each other, then add up the swings?

It seems to me that IMPing half of the aggregate is equivalent to scoring against a datum in IMP pairs.