[bluecalm]

Ok I know lol sample size and lol variance but I used the script I wrote some time ago to calculate those stats:
http://www.bridgebas...r-in-the-world/

To do the same for Bermuda Bowl 2011 hands.
Here are the best cardplayers (both defense and declarer play is compared to dd result after first lead is made, so if you make a game which wasn't makeable after 1st lead you are +10imp, same goes for defense):

Stats for Levin are unreliable as his and her name were merged by my not very intelligent script.
To qualify you need to have at least 75 hands of cardplay on vugraph (not being dummy).

BOCCHI       : 234    , 243    , avg: 0.96
MADALA       : 234    , 247    , avg: 0.95
BURAS        : 75     , 90     , avg: 0.83
RUSSYAN      : 79     , 103    , avg: 0.77
MARTENS      : 51     , 75     , avg: 0.68
HERBST       : 145    , 215    , avg: 0.67
PACHTMAN     : 67     , 101    , avg: 0.66
KOWALSKI     : 78     , 126    , avg: 0.62
MORSE        : 63     , 107    , avg: 0.59
ROMANSKI     : 72     , 123    , avg: 0.59
GINOSSAR     : 53     , 97     , avg: 0.55
PALMER       : 64     , 120    , avg: 0.53
PIGANEAU     : 72     , 141    , avg: 0.51
Senior       : 107    , 212    , avg: 0.50
ZUR-CAMPAN   : 42     , 84     , avg: 0.50
BESSIS       : 61     , 122    , avg: 0.50
LAURIA       : 110    , 221    , avg: 0.50
ROBINSON     : 55     , 114    , avg: 0.48
D'OVIDIO     : 55     , 117    , avg: 0.47
WEINSTEIN    : 105    , 227    , avg: 0.46
CASTELLO B   : 34     , 77     , avg: 0.44
AUKEN        : 36     , 82     , avg: 0.44
DEWI         : 93     , 219    , avg: 0.42
GOSNEY       : 38     , 91     , avg: 0.42
VERSACE      : 84     , 211    , avg: 0.40
MANOPPO      : 30     , 77     , avg: 0.39
HURD         : 112    , 289    , avg: 0.39
SEMENTA      : 88     , 234    , avg: 0.38
MULLER       : 61     , 163    , avg: 0.37
FALLENIUS    : 80     , 217    , avg: 0.37
SADEK        : 35     , 97     , avg: 0.36
WOLFF        : 39     , 110    , avg: 0.35
WOOLDRIDGE   : 103    , 291    , avg: 0.35
Michielsen   : 40     , 116    , avg: 0.34
BOYD         : 42     , 128    , avg: 0.33
WILLARD      : 58     , 179    , avg: 0.32
FREDIN       : 70     , 225    , avg: 0.31
FISHER       : 37     , 119    , avg: 0.31
AHMADY       : 31     , 100    , avg: 0.31
CRONIER      : 58     , 198    , avg: 0.29
DEAS         : 28     , 97     , avg: 0.29
DHONDY       : 54     , 195    , avg: 0.28
ARONOV       : 21     , 77     , avg: 0.27
ZAGORIN      : 28     , 106    , avg: 0.26
STEFANOV     : 19     , 75     , avg: 0.25
BATHURST     : 26     , 103    , avg: 0.25
LASOCKI      : 27     , 107    , avg: 0.25
Levin        : 89     , 358    , avg: 0.25
WIJS         : 36     , 148    , avg: 0.24
BALDURSSON   : 24     , 110    , avg: 0.22
BRENNER      : 17     , 78     , avg: 0.22
SEAMON-MOL   : 21     , 102    , avg: 0.21
JONSSON      : 22     , 108    , avg: 0.20
HAMILTON     : 19     , 97     , avg: 0.20
FINKEL       : 16     , 85     , avg: 0.19
TUEJE        : 39     , 210    , avg: 0.19
DUBOIN       : 41     , 221    , avg: 0.19
BROCK        : 22     , 123    , avg: 0.18
Drijver      : 35     , 203    , avg: 0.17
Gaviard      : 22     , 147    , avg: 0.15
LALL         : 40     , 305    , avg: 0.13
NEVE         : 19     , 147    , avg: 0.13
LEENHARDT    : 18     , 154    , avg: 0.12
BOJOH        : 24     , 214    , avg: 0.11
Brink        : 22     , 200    , avg: 0.11
LASSERRE     : 15     , 158    , avg: 0.09
MURNIATI     : 21     , 223    , avg: 0.09
POIZAT       : 17     , 182    , avg: 0.09
MEYERS       : 10     , 109    , avg: 0.09
NYSTROM      : 9      , 142    , avg: 0.06
NARKIEWICZ   : 5      , 93     , avg: 0.05
UPMARK       : 7      , 136    , avg: 0.05
SMITH        : 4      , 103    , avg: 0.04
ZACK         : 4      , 111    , avg: 0.04
BAREL        : 3      , 109    , avg: 0.03
GRENTHE      : 3      , 144    , avg: 0.02
STANSBY      : 3      , 278    , avg: 0.01
Verhees      : 1      , 155    , avg: 0.01
KLUMPP       : 0      , 75     , avg: 0.00
SCHERMER     : -2     , 149    , avg: -0.01
PROOIJEN     : -4     , 143    , avg: -0.03
ROSENBERG    : -3     , 104    , avg: -0.03
BERTHEAU     : -5     , 154    , avg: -0.03
Dekkers      : -4     , 103    , avg: -0.04
BROWN        : -5     , 122    , avg: -0.04
JORGENSEN    : -5     , 118    , avg: -0.04
SHI          : -14    , 174    , avg: -0.08
CHAMBERS     : -13    , 147    , avg: -0.09
Schwartz     : -10    , 103    , avg: -0.10
KASLE        : -9     , 78     , avg: -0.12
GRUE         : -35    , 292    , avg: -0.12
EDGTTON      : -13    , 101    , avg: -0.13
HOU          : -13    , 92     , avg: -0.14
Simons       : -19    , 121    , avg: -0.16
VILLAS-BOA   : -16    , 89     , avg: -0.18
KAMIL        : -26    , 136    , avg: -0.19
STOCKDALE    : -26    , 118    , avg: -0.22
PASMAN       : -26    , 115    , avg: -0.23
VANHOUTTE    : -36    , 148    , avg: -0.24
FLEISHER     : -35    , 140    , avg: -0.25
MARTEL       : -48    , 184    , avg: -0.26
CULLIN       : -28    , 105    , avg: -0.27
CAMPOS       : -22    , 76     , avg: -0.29
GAWRYS       : -29    , 92     , avg: -0.32
EINARSSON    : -30    , 94     , avg: -0.32
KALITA       : -37    , 94     , avg: -0.39

Say what you want but this guy:



Is always on top!

[Fluffy]

card playing has little to do with this stat, this has to do more with bidding 90% of the time.

[bluecalm]

Quote

card playing has little to do with this stat, this has to do more with bidding 90% of the time.

Why do you think so ?

[Fluffy]

I missread, though you win IMPs for being defender to a game contract when the cards are in a way where you can make grand.

But now I reread it seems like your script is more sofisticated.

[bluecalm]

It takes double dummy result at the moment afte 1st card is already played and compares it to actual score at the table.
True, you get +10 if you are sitting there with stiff K of trumps behind AQ but it happens to all defenders (and you get -10 as declarer) so it's just variance depending on where you are sitting.
Actually I calculated separate stats for declarers and defenders and then combined them. If you take only defenders Bocchi-Madala are +1.6imp and next pair is +1imp

I did the same for European Team Championship and again, Bocchi Madala who topped the butler also topped cardplay stats by about 1 imp/hand.

[hrothgar]

Don't suppose you have a normplot of the scores?

I'd like to see how close this is to being normally distributed...

[bluecalm]

You mean the scores for any hand ?
This is not normal distribution as big scores like +10 and -10 are frequent. This is probably not a problem though because usually those distributions with bumps on tails have properties of normal distribution (like in poker for example where you have bunch of small pots and allins but not so many hands for half a stack).
I calculated standard devs for various stats in my old post I linked in OP.

As example for those stats I had 6926 hand recorded during vugraph of BB2011. And the ditribution was:

[0, 1, 0, 16, 2, 6, 17, 19, 93, 23, 47, 5, 7, 36, 36, 112, 42, 112, 166, 553, 4365, 541, 180, 81, 40, 119, 41, 28, 3, 5, 54, 24, 96, 34, 13, 2, 1, 4, 2, 0, 0]

(First is number of occurances of -20imp swing, 2nd is -19, 0 in the middle (the one with 4365 occurances) and the last one is +20imps)

Note that avg is not 0 (I posted in in my old post) because you are twice as often on defense. I posted avg's in my old post.

[hrothgar]

 bluecalm, on 2012-January-10, 06:34, said:

You mean the scores for any hand ?

The distribution of player strength

[bluecalm]

Quote

The distribution of player strength

There is huge selection bias in vugraph hands as good players tend to be shown more often so they have more hands.
This way you have bunch of Bocchi's and Versace's with huge amount of hands and hundred of weaker players who made their only appearance in vugraph when their team played Italy or USA.
I posted only results for players with 75 hand which was arbitrary number chosen by me. In my old post I posted players with 500+ hands on vugraph.
Interesting thing is that when you compare dd results to actual results (just that, without going into declarer play etc.) you usually arrive at scores similar to butler on big tournaments which makes me believe that this is actually quite a good way to measure performance.

[gnasher]

 bluecalm, on 2012-January-10, 07:20, said:

Interesting thing is that when you compare dd results to actual results (just that, without going into declarer play etc.) you usually arrive at scores similar to butler on big tournaments which makes me believe that this is actually quite a good way to measure performance.

Or that both methods have the same flaw: they don't reflect differences in the quality of the opponents. If you tend to play against the stronger of the opposing pairs, you will suffer both in the Butler and in the comparison against double-dummy.

[JLOGIC]

 gnasher, on 2012-January-10, 08:07, said:

Or that both methods have the same flaw: they don't reflect differences in the quality of the opponents. If you tend to play against the stronger of the opposing pairs, you will suffer both in the Butler and in the comparison against double-dummy.

Yeah obv I'm just bitter, but my hands were presumably the trials + the bermuda bowl and nothing else, and when I was on vugraph in those events it was probably vs the best teams. I think the Italians are on vugraph a lot regardless of who they play because of their star power, thus more hands against weak opps. The women/seniors facing much easier competition should also do well, it is more a measure of how good you are relative to your opponents.

[bluecalm]

Quote

Yeah obv I'm just bitter, but my hands were presumably the trials + the bermuda bowl and nothing else,

This BB2011 ONLY no trials, no other hands.
Obviously very small sample size and obviously players from USA2 and Netherlands played majority of their boards vs very tough opponents.

Quote

I think the Italians are on vugraph a lot regardless of who they play because of their star power, thus more hands against weak opps. The women/seniors facing much easier competition should also do well, it is more a measure of how good you are relative to your opponents.

Yeah, it's for sure have huge match selection bias, no doubt about it.
On the other hand I saw my share of vugraph (as I prefer it to movies) and imo Bocchi's cardplay is in fact out of this world.
(him topping about every butler in major national team event in this century is another factor contributing to that opinion)

[JLOGIC]

You do not find it coincidental at all that most partnerships are pretty close to each other? Madala Bocchi, Versace Lauria, Driver Brink, Ron Eldad, Kalita Gawrys, Hurd Wooldridge, Fredin Fallenius, Fleisher Kamil, Bathurst Zagorin etc etc? I think it is largely because they played against the same opponents.

Edit: Or maybe it's because defense is counted the same. I do not really understand how the defense part works.

[JLOGIC]

BTW I always thought a good way to do a butler would be to adjust for performance of your opps when they did not play you. For instance, if you are +1 imp/bd vs people who are -1 imp a board otherwise, that should be worse than being +.5 imps a bd vs +.1 imp a board opps. It would probably be impossible for your results as many pairs would have no other data though.

[bluecalm]

Quote

BTW I always thought a good way to do a butler would be to adjust for performance of your opps when they did not play you. For instance, if you are +1 imp/bd vs people who are -1 imp a board otherwise, that should be worse than being +.5 imps a bd vs +.1 imp a board opps. It would probably be impossible for your results as many pairs would have no other data though.

It would be possible. The way to do that is to model our beliefs in such a way that at the beginning every pair has some probability distribution when ti comes to its winrate and adjust it with every match played.
This way pairs with not so many hands will be treated as more or less "generic opponents" and the ones with many hands will gravitate towards their true winrate. Then it's enough to do few iterations of that to arrive at stable point and voila.

Unfortunately my math and programming skills suck to much to implement this awesomeness

[gnasher]

Adjusted Butlers still aren't perfect, because the sponsors tend to play each other, but they're certainly better.

It's doesn't sound hard to implement. Just keep recalculating them, adding the opponents' Butler from the previous iteration to each swing. Stop when none of the numbers change between interations. If it never stops, you'll know that it's not as easy as I claim.

[bluecalm]

Quote

It's doesn't sound hard to implement. Just keep recalculating them, adding the opponents' Butler from the previous iteration to each swing. Stop when none of the numbers change between interations

There are sample size issues for pairs which don't appear on vugraph too often. Just because some pair lost 36 imps in on 12 board segment vs average opposition doesn't mean they are -3imp/hand pair (and it doesn't mean that the pair who beat them should only get credit for beatnig some -3imp/hand droolers). The way to solve this is to make some bayesian adjusting and this is the part which is kinda difficult. I agree though that for some decent programmer it should be a breeze
I attempt it when I have some more motivation but it's not one day thing for me.

[FrancesHinden]

 gnasher, on 2012-January-11, 05:15, said:

Adjusted Butlers still aren't perfect, because the sponsors tend to play each other, but they're certainly better.

It's doesn't sound hard to implement. Just keep recalculating them, adding the opponents' Butler from the previous iteration to each swing. Stop when none of the numbers change between interations. If it never stops, you'll know that it's not as easy as I claim.

I did this with the full results from the European championships a few years ago. It's not that hard. I didn't find the results particularly illuminating, however, they told me that the people I already thought were very good players, were very good players.