Shanghai observations

[Gerben42]

After 9 of the 21 rounds the score is:

 

1. US1 180

2. POL 166

3. NWY 161

4. ITA 158

5. AUS 151

6. NTH 148.5

7. CHN 148

8. BRA 147

...

14. US2 129

 

The 4 top European teams are well in the qualifying zone so far, although Italy is struggling for their standard. In the European championship they usually did well in boards 17 - 20, now the matches are only 16 boards...

 

No huge surprises in the zones outside Euro / US. China, Australia and Brazil are well-established bridge-nations. But... who thought US1 would dominate the table like this with the Nickell team 18 VP behind the last qualifying spot? Even without Soloway I'd expect them much higher. We've seen their comebacks before, and still a long way to go.

[Aberlour10]

After the disaster in Estoril not many expected such a strong performance by Team Poland (playing without the absolute top pair) Even if they will not enter the top 4 of the event, it seems, polish bridge is "back"B) at the international "open" stage.

 

Robert

[hrothgar]

If there is anything to be learned from past competitions, its the following:

 

Early round results mean next to nothing. There is a lot of disparity between the skill levels of the different teams playing. Early results often mean that average teams were beating up on the weaker sisters. The system needs a bunch more time to converge...

[helene_t]

Maybe it would be better to compute the Lehman rating of the teams, instead of total VIPs?

[matmat]

If there is anything to be learned from past competitions, its the following:

 

Early round results mean next to nothing.  There is a lot of disparity between the skill levels of the different teams playing.  Early results often mean that average teams were beating up on the weaker sisters.  The system needs a bunch more time to converge...

so i was bored. (well, more like procrastinating, really)

 

USA1 181 -0.02

POL 166 0.4

NOR 161 -3.15

ITA 158 0.85

AUS 151 -0.02

NED 148.5 2.68

CHN 148 0

BRA 147 0.33

JAP 144 0.04

ARG 143 0.19

SA 138.5 -0.53

INA 133 -0.02

PAK 130 0.18

USA2 129 -1.15

EGY 128 -0.28

NZ 126 0.39

SWE 121 0.01

TAI 105 -0.52

CAN 101.5 -0.32

TT 101 0.76

IRE 96.5 0.26

IND 96 -0.06

 

i may have entered a few scores with minor errors into the spreadsheet. anyhow

the second column is obviously the VP, the third column is relative, first order, strength of schedule. (i can give the expressions i used to whoever cares, they're not very thought out or complicated, for that matter) -- positive numbers imply a tougher schedule, negative numbers an easier one. it seems clear that most teams are close to 0, with the notable exceptions of Norway, who has played a substantially easier schedule thus far, and the Netherlands has had much tougher opposition.

[vuroth]

You seem to have 2 CANs in there mat. I'm pretty sure we've only sent one team....

[Gerben42]

The top Canada is USA2 in disguise. By the way the IOC uses INA for Indonesia, IND for India, not sure if this is helpful information B)

[matmat]

You seem to have 2 CANs in there mat. I'm pretty sure we've only sent one team....

can't ever have enough CANs

[helene_t]

After ten rounds, the correlation between the coefficients in a linear regression model, and the total VPs, is:

http://www.geocities.com/helene_thygesen/bermuda10.jpeg

 

Norway ranks lower in LR coefficients than total VIPs. This is consistent with Matmat's observation that Norway had been favored with easy opposition in the beginning.

 

Btw, it is not necesary to type in stuff in the spreadsheet. At least if you use Firefox and OpenOffice you can just paste the web-pages into the spreadsheet and fit the regression model, very easy. (Dono about Excel, maybe you can ask Shubi for advice with that one).

[matmat]

Btw, it is not necesary to type in stuff in the spreadsheet. At least if you use Firefox and OpenOffice you can just paste the web-pages into the spreadsheet and fit the regression model, very easy. (Dono about Excel, maybe you can ask Shubi for advice with that one).

i must be using the wrong site to get my data from. i don't really see anything easily cuttable-pastable on the unmentionable site. i could be wrong.

[helene_t]

http://www.worldbridge.org/tourn/Shanghai.07/Results.htm

[helene_t]

After some surprisings in the beginning, most of the teams now rank about as I would expect, give or take a handful of positions.

 

Some exceptions: USA 2 open team has already been discussed. But also the Irish open team and the Dutch women's team are disappointing, while the South African open team and the Danish women's team are doing very well. I haven't seen any vugraph (awkward timezone), anyone who has seen those teams who can comment?

[paulg]

The Irish Open team seem to have been out of form since their excellent result in finishing 2nd in the Europeans. In fact they had been playing consistently well for 2-3 years up to that point.

 

Since then they've finished 7/8 in the 2006 Lederer, lost relatively early in the Sping Foursomes, performed poorly in the Nashville NABC (both in the Spingold and other events), qualified for the final at the EBU Summer Congress (Brighton) teams but then only finished 5/8.

 

Unsurprisingly all three pairs are below average in the butler but Hanlon/McGann, easily their best pair, are lowest.

 

Paul

[matmat]

looks like the poles may have had dinner at some lousy restaurant last night

[whereagles]

helene, what are you trying to correlate there? names to scores???

[matmat]

helene, what are you trying to correlate there? names to scores???

actual scores to actual true performance, so to speak...

 

 

if i told you that sored 30 points in a basketball, and so did kobe bryant, you'd immediately suspect that my game was against weaker opposition than kobe's... right? she's just trying to normalize the actual scores to the opposition that the teams have faced, thus far.

[helene_t]

helene, what are you trying to correlate there? names to scores???

The model is that the IMP result (HOMEIMP-VISITORIMP) is

HOMECOEF-VISITORCOEF+epsilon

where epsilon is normal distributed with mean zero and some unknown variance. The problem is singular so the software imposes the constraint that USA2 must have coef zero (because USA2 is last in alphabetical order). That is a formality and doesn't matter.

 

The residuals are in fact nicely normal distributed, but whether the expected IMP result is additive I haven't checked (it would clearly not be the case in a very heterogenous field, but in this case I think it's a reasonable assumption).

 

The diagram shows the coeficients in the model on the x-axis and the total VPs on the y-axis.