[Tomi2]

Hi,

there have been two or three recent threats about the meaning of such statistics as:
avg IMPs when declaring etc.

I have started recording deals I played on BBO since february and try to add deals I play in live tournaments manualy (oh dear my last tourn will not improve my stats...)

my aim was to compare the stats to JEC because he plays lots of boards, always timps and always vs reasonable opps.

now I found this nice source
http://www.rpbridge.net/9x65.htm

made by Richard Pavlicek

there you see stats for world class pairs but only from vugraph matches, so blitzes against weak opponents say in RR of Bermuda Bowl, do not improve the stats so much

I want to start a new discussion on that stats, what is you opinion on these facts:

Some players are more often declarers as their partners. What is the difference where you would say "this is not random, those two players try to make one declarer more often"?

Duboin 26.35% / Sementa 22.28% on 520 deals
Balicki 25.59 / Zmudzinski 20.92% on 1250 deals
Fred is also 23.85/26.29 behind in his partnership, maybe he can answer from his experience

then there are same stats about beeing on lead / not beeing on lead
-do you think this is random?
-do you think top players, who know their opponents well try to make one specific player make the opening lead more often?

I can tell that in one mixed event, when I had a good partner, we found out, that the usual mixed strategy "letting the guy declarer" is not a good idea but "letting the woman make the opening lead" worked well, so the one who had a female oponent adjusted his play (sry if this sounds sexistic, but we played against lots of partnerships like Helness-Helness and we expected Tor to be stronger than his wife...)

of course there are expected and unexpected differences in their avg. IMP stats
Dubion playing with Sementa mas 1.1 IMPs per board declaring!! thats 0.6 more as playing with Bocchi, any resons? or just random?
I would say on ~500 deals he was declarer ~150 times, so with his Bocchi stats Duboin should have made ~75 IMPs, now he scored ~150.
75 IMPs can obviously come from lucky slams / grands making 6 times, or does he really play (bid?) better now?

Muller has +1.81 IMP/board on hands where he makes the opening lead, any ideas why?

Meckwell are declarer in 53.74% of the hand and defendes in 46.16% (they passout one of 1000) but both players have better stats in defending than declaring - so should they change their methods and bid less often on borderline hands?

Jansma-Verhees were declarers only 45.13%

Brink/Drijver are better declaring than defending, should they practise this more / work on their defense agreements etc.?

Chagas is defending (on lead) 2 imps better than declaring and 1.8 imps better than his partner on lead


hope there will be some nice aspects that help me undestand my stats (beeing worse declarer than my pds but better on opening leads) better

[Free]

Tomi2, on May 18 2010, 11:42 AM, said:

Meckwell are declarer in 53.74% of the hand and defendes in 46.16% (they passout one of 1000) but both players have better stats in defending than declaring - so should they change their methods and bid less often on borderline hands?

I read the entire post, and I'm not sure how to respond to all of this. I think it's important to realize that it's not as simple as you might think. As an example I quoted the Meckwell part, because you make a conclusion (you ask a question) which is imo incorrect. They win a lot of imps defending because they force opponents to make mistakes. This can result in opps declaring at a level which is too high. Obviously, by forcing opponents into mistakes, sometimes Meckwell bid too high and go down themselves. This can be an explanation of why their defensive scores are better than their declarer scores.

It's a balance you need to be comfortable with. Suppose you bid very solid and bid most of the sharp games as well. You won't go down a lot. So you expect to gain imps whenever you're declarer (frequent case will be that you stay in 3M and other side goes down in 4M). On the other hand, you also don't force your opponents into mistakes. You don't push your opps an extra level higher, so your defensive gains will be a lot less as well.
Turn this all around, you play a very agressive style with which you'll force your opponents to make mistakes in bidding AND play. Now you'll go down a lot, but you'll make some impossible contracts. The net result will be less than the solid style. Also, your opps will overbid a lot, so they'll go down more. Result: your defensive score goes up, your declarer score goes down.
The end result: adding the declarer and defense scores. The ultimate goal is to maximize the total of those 2 scores together, and not one of them in particular.

[hrothgar]

Tomi2, on May 18 2010, 01:42 PM, said:

Some players are more often declarers as their partners. What is the difference where you would say "this is not random, those two players try to make one declarer more often"?

Duboin 26.35% / Sementa 22.28% on 520 deals
Balicki 25.59 / Zmudzinski 20.92% on 1250 deals
Fred is also 23.85/26.29 behind in his partnership, maybe he can answer from his experiencebut better on opening leads) better

Life is pretty hectic right now and I don't expect to have as much time to answer this as I would like.

The best suggestion that I can give you is the following: Conceptually, this is the same as testing whether a coin is "fair". If I flip a fair coin a large number of times, I expect to see roughly 50% heads and 50% tails.

This is a "classic" example in applied statistics (none too surprising, since Probability and Statistics evolved out of individuals who where studying games of chance)

Wikipedia has a decent article which will help get you started...

http://en.wikipedia...._a_coin_is_fair

[Tomi2]

this was only a suggestion (the meckwell thing) - I just want to hear impressions how to read those stats

another thing I saw:

Balicki - Zmudzinski both win over 70% (thats great!) of their declared contracts but Zmudzinski has 0.3 imps per board more. Can this be, because Balicki makes more safetyplays that lose one imp each while his partner goes down in games 1 per 20 times but saves the IMP 19 times?

[hrothgar]

Tomi2, on May 18 2010, 01:42 PM, said:

Some players are more often declarers as their partners. What is the difference where you would say "this is not random, those two players try to make one declarer more often"?

As a side bar, lets assume that we do comprehensive statistical analysis and determine that we are 99% certain that that there is a statistically significant difference in the frequency with that Fred declarers as opposed to Brad.

This doesn't mean that the partnerships methods are specifically designed to increase the frequency with which Fred declares. There are a number of other possible hypotheses that are also consistent with this data...

In general, Statistical tests work best when you

1. Start by formulating a hypothesis
2. Collect a data set to test said hypothesis

(There is an silly conversation occuring on rec.games.bridge about the perils of starting with a strange looking data set and then trying to test it)

If you are seriously interested in a study of this sort, I'd recommend something like the following:

1. Start by formulating a list of well know pro / client pairs. (A priori, it seems reasonable to assume that one or more of these partnerships might skew their style to maximize the chance that the pro declares)

2. Test this specific set of hand records

3. If you want some real fun, advance the claim that Players X and Y are using different systems...

[mgoetze]

Tomi2, on May 18 2010, 11:42 AM, said:

Meckwell are declarer in 53.74% of the hand and defendes in 46.16% (they passout one of 1000) but both players have better stats in defending than declaring - so should they change their methods and bid less often on borderline hands?

Alternate hypothesis: because Meckwell enter the bidding so aggressively, they have a much better idea what their partner holds when they are defending (particularily, there are more negative inferences - "no, he can't have 7 points with KJxx in hearts, he would have overcalled 1♥ with that" ).

[mikeh]

I don't know a lot about Fred's approach to preempting, but I have read comments about Moss's tendencies to make very aggressive preempts. If he makes more preemptive calls, due to personal style, than Fred does, he will declare more hands than Fred. Similar tendencies may apply in other WC partnerships. It is not uncommon for the two members of a partnership to differ slightly in their approach to opening/overcalling light or preempting. These differences, in the long run, would (I assume) manifest themselves in frequency of being declarer.

I would be astounded if in any good partnership there was any tendency to try to make one partner declarer!

I will leave it to the mathematicians amongst us to opine on whether the sample sizes cited in the OP are large enough to render the results significant in and of themselves.

The earlier comment about the trade-off between declarer and defence stats was very well put, imo.

[awm]

I've also read comments that Brad is more likely than Fred to make a "funny" 1NT opening (off-shape, upgraded for spot cards, etc in addition to straight-out psychs). This probably also increases Brad's relative odds to declare.

I agree with others that it's hard to read much into these statistics directly, and that it's unlikely any top partnership is really trying to "turn the hands." You do see this type of thing in pro-client pairs sometimes though.

[CSGibson]

mgoetze, on May 18 2010, 05:59 AM, said:

Tomi2, on May 18 2010, 11:42 AM, said:

Meckwell are declarer in 53.74% of the hand and defendes in 46.16% (they passout one of 1000) but both players have better stats in defending than declaring - so should they change their methods and bid less often on borderline hands?

Alternate hypothesis: because Meckwell enter the bidding so aggressively, they have a much better idea what their partner holds when they are defending (particularily, there are more negative inferences - "no, he can't have 7 points with KJxx in hearts, he would have overcalled 1♥ with that" ).

There is also the fact that Meckwell plays a pressure system, often jumping straight to game after their partner's limited opening. That puts opponents in a position where they have to make a guess as to whether to come into the auction, which I imagine increases Meckwell's imp expectancy on defense (ie, sometimes they guess wrong). Ad that to the frequency with which they enter auctions, which tends to decrease the ability of opponents to bid accurately, and the fact that some opponents, even world class opponents, see a precision 1 club as an invitation to preempt anything and everything, and I see a lot of gain simply by pushing people to bad contracts.

[hotShot]

I think a lot of these numbers describe random noise. Look at these numbers of players that are in the list with different partnerships.
Duboin
Semeta 26.35% declarer
Bocchi 24.59%
Hamman
Soloway 26.52%
Zia 23.27%
Wolff 26.15%
Zia
Hamman 24.17%
Rosenberg 25.61%

[hanp]

It's not all random noise but the numbers are quite small for many pairs. The 4654 hands for Meckwell is by far the largest number, and I think that we can reasonably conclude from their +0.49 per board that they are extremely good and from their 27.8% and 25.9% declaring percentages that they are very aggressive.

For the pairs that have played much fewer hands (such as Muller - De Wijs), some of these numbers should be taken with a grain of salt. Each IMP score applies to a little over 100 hands, not a whole lot.

One might expect that a pair that preempts very aggressively does better on defense (their opponents will be in the wrong contract more often) but worse as declarers (they'll get hammered more often). Taking this a step further one might conjecture that aggressive pairs score better on defense, and conservative pairs score better when declaring. This is certainly true for Meckwell (who do quite a bit better on defense) as well as for Becker-Schwartz and Jansma-Verhees (two of the least aggressive pairs who both do much better when declaring).

I wouldn't dare to say anything about the declaring skills. For example, if player A bids very well when their partner B has opened 1NT, the declaring score of B is likely to go up. Bidding and playing are so mixed up that it is hard to conclude anything about one of the factors without having isolated numbers.

[hanp]

hotShot, on May 19 2010, 03:11 AM, said:

I think a lot of these numbers describe random noise. Look at these numbers of players that are in the list with different partnerships.
Duboin
Semeta 26.35% declarer
Bocchi 24.59%
Hamman
Soloway 26.52%
Zia 23.27%
Wolff 26.15%
Zia
Hamman 24.17%
Rosenberg 25.61%

Why is this random noise? Hamman plays a different system with Zia than he did with Soloway. Also his style might have changed.

[hanp]

Tomi2, on May 18 2010, 07:27 AM, said:

Balicki - Zmudzinski both win over 70% (thats  great!)  of their declared contracts but Zmudzinski has 0.3 imps per board more. Can this be, because Balicki makes more safetyplays that lose one imp each while his partner goes down in games 1 per 20 times but saves the IMP 19 times?

Surely bidding differences will have a far larger impact on these percentages than their eagerness to make safety plays. For example, underbidders will make many more contracts than overbidders. It isn't clear to me that making over 70% of your contracts is something you need to be proud of.

On the other hand Duboin, Becker and Jansma, the three declarers with the highest making percentage, all win more than 1 IMP per hand they declare.

[hotShot]

hanp, on May 19 2010, 11:33 AM, said:

hotShot, on May 19 2010, 03:11 AM, said:

I think a lot of these numbers describe random noise. Look at these numbers of players that are in the list with different partnerships.
Duboin
Semeta    26.35% declarer
Bocchi      24.59%
Hamman
Soloway 26.52%
Zia        23.27%
Wolff      26.15%
Zia
Hamman    24.17%
Rosenberg  25.61%

Why is this random noise? Hamman plays a different system with Zia than he did with Soloway. Also his style might have changed.

I know that you can do the required math yourself. I had a look at the quoted website and noticed that in partnerships that are registered with more than 2000 boards, players have a smaller deviation from 25% than in those with less than 600 boards. Often this is an indicator for random noise. The quoted player have results both above and below the expected 25%, another indicator for the presence of random noise.
Since the expected value is 25% and almost all of the values are within the interval [23,27] and for larger numbers within [24,26]. So I estimate any effect to be of smaller or equal size to the noise.

My estimation could be wrong, but from 400 boards played 25% would be 100, 23% would be 92. Since you are the math expert, you can tell me if that is significant enough.

[Tomi2]

my calculations were, that the prob that in a partnership like bal / zmud the likelyhood, that the declarerplays are split so or worse is 1.8% and for Grue-cheek its 0.2% (for any 29%-20% or more extreme break of playing the hands)

[hanp]

Since for all pairs the chance that any given player will declare is about equal to 25%, the standard deviation for one hand is about 10.8%. That means that for the Hamman-Soloway partnership (2225 hands) the standard deviation is about .23% and for Hamman-Zia (666 hands) the standard deviation is about .42%.

Given that Hamman declared 3.25% more hands when playing with Soloway, the chance that the percentages would be this far off if they were equal for both partnerships would be about 2%, i.e. quite small.

Of course, if you start studying a large data set, then the chance that you find something for which the probability that it occurs is less than 2%, is very high. But the above should give you an indication of how large the expected errors are, more than a percent would be unusual for 600 or more hands played.

[Jlall]

lol at some of these guesses. Hamman and Soloway played strong club with 13+-16 NT, and indeed they opened all 3325 13 counts 1N.

Zia and Hamman play natural with a stronger NT.

Hamman and Wolff played similar to Hamman and Soloway.

Zia and Rosenberg played weak NT.

The lighter you open and the more often you open 1N, the more likely you will be to play the hands (see Meckwell, Grue, etc).

[bluecalm]

Quote

my calculations were, that the prob that in a partnership like bal / zmud the likelyhood, that the declarerplays are split so or worse is 1.8% and for Grue-cheek its 0.2% (for any 29%-20% or more extreme break of playing the hands)

I am pretty sure it's not random for Bal - Zmud. Bal is the captain there, the one to open offshape/off points 1NT, aggressive preempts or other stuff. Zmud is very very solid and always has his bids.

[MarkDean]

hanp, on May 19 2010, 07:02 AM, said:

Since for all pairs the chance that any given player will declare is about equal to 25%, the standard deviation for one hand is about 10.8%. That means that for the Hamman-Soloway partnership (2225 hands) the standard deviation is about .23% and for Hamman-Zia (666 hands) the standard deviation is about .42%.

Given that Hamman declared 3.25% more hands when playing with Soloway, the chance that the percentages would be this far off if they were equal for both partnerships would be about 2%, i.e. quite small.

Of course, if you start studying a large data set, then the chance that you find something for which the probability that it occurs is less than 2%, is very high. But the above should give you an indication of how large the expected errors are, more than a percent would be unusual for 600 or more hands played.

Where did that 10.8% come from?
I get 43.3%.

[minimonkey]

As far as my poor undergrad stats knowledge gives variance of 1 outcome of a mutinomial

=number of hands* prob playing the hand* prob not playing the hand

This makes the rough 95% confidence interval

25% + or -

sqrt(number of hands* 18.75%)

I am sure this is wrong

If you want to see if 2 partners are different the 18.75% should be replaced with

(% of hands played by the partnership/2)* (1- (% of hands played by the partnership/2)

Anyway I am sure this sheds light on nothing but my ignorance.