[bid_em_up]

kenrexford, on Feb 7 2007, 06:09 PM, said:

The anecdote about the friends match was to point out how I sort-of had a "scouting report" available to me of my friend, and how this gained. Had I instead not known this person at all, but watched his bidding and play, reviewed his bizarre committee actions, and gained access to his beliefs about me, then I'd have the same information. A scouting report.

It did not gain. It tied.

Had you just played bridge and not bid the slam (as you stated that your gut feeling was the slam would fail), then you would have gained.

[kenrexford]

bid_em_up, on Feb 8 2007, 11:33 AM, said:

kenrexford, on Feb 7 2007, 06:09 PM, said:

The anecdote about the friends match was to point out how I sort-of had a "scouting report" available to me of my friend, and how this gained.  Had I instead not known this person at all, but watched his bidding and play, reviewed his bizarre committee actions, and gained access to his beliefs about me, then I'd have the same information.  A scouting report.

It did not gain. It tied.

Had you just played bridge and not bid the slam (as you stated that your gut feeling was the slam would fail), then you would have gained.

Sure, on THAT hand I would have "gained."

However, here was the problem. We won the match (and the event) by a slight margin, and I anticipated that we would win that way anyway.

Had I not bid the, say, 30% slam, the results would have been that 70% of the time we'd win by a large margin and 30% of the time we would lose. I hedged and ensured the win.

[cherdano]

By the way, since I made a negative comment above, I should maybe add that some of the statistics linked to by winkle above are indeed quite impressive. Lauria-Versace gain one imp on average in a deal where they are defending the same game contract as opponents at the other table. Since this is a sample of 205 deals, it's absolutely statistically significant, and quite an impressive figure in my opinion. (Of course, the fact that they usually have competent declarers as teammates will help that figure, it would be more useful to have a CrossIMP score for this in a large field.)

[FrancesHinden]

Your point shows why it's so hard to do this analysis for pairs rather than teams.
Lauria / Versace are great players... but I promise you my imps/board would look pretty good if I could score up with Bocchi & Duboin or Fantoni & Nunes!

[FrancesHinden]

cherdano, on Feb 7 2007, 12:40 AM, said:

winkle, on Feb 6 2007, 06:03 PM, said:

- What are their Achilles' heels relative to their peers?

    - This is the most striking observation of all.  In spite of all their
    successes, both Duboin/Bocchi and Fantoni/Nunes are awful (relative to their
    world class peers, mind you) in slam bidding decisions.  On boards where
    either those Italians or their counterparts are declaring in slam (but not
    both), Duboin/Bocchi lose 0.93 imps per board, while Fantoni/Nunes lose 2.12
    imps per board.  Just think how much better the Italians would be if they
    fix this.

Except that your number are not even close to being statistically significant.

In 40 slam vs non-slam deals, you get 40 swings of at least 10 IMPs either way. If you assume two equally good pairs bidding these against each other, you have to expect a total swing of sqrt(40)*10 either way, i.e. roughly 60 IMPs. Only if you had one pair losing/winning more than 120 IMPs on 40 slam deals could you tell with some statistic certainty that they are better or worse than the average of their opponents.

You would need a much larger number of deals to get statistically significant numbers for specific questions like this, unfortunately.

Arend

I'm not sure that this isn't significant, because on the majority of "slam vs non-slam" deals it's not random whether you bid slam or not. If (and it's a big if) one pair have worse judgement in the slam zone, then they will bid the wrong ones: this total swing you expect it not random, it's a reflection on their bidding ability.

Let's say that
i) All potential slam hands are 100% or 60% each equally likely and each with nothing to the play.
ii) Pair A, who are too conservative, bid all the 100% slams, but not the 60% ones
iii) Pair B bid all of them
iv) The swing for slam off/game making or both making is 10 imps either way

In 40 potential slam deals under these conditions, you expect
20 flat boards (excluded from winkle's analysis)
12 10 imp swings to pair B
8 10 imp swings to pair A

or a total expected swing of only 1 imp per board

In fact, if you then add in the condition that F&N are more likely to make 12 tricks than their opponents on average, and more likely to take 2 tricks in defence on average, it makes things look even worse (particularly if they are over-aggressive rather than conservative in the bidding).

[helene_t]

FrancesHinden, on Feb 9 2007, 11:49 AM, said:

I'm not sure that this isn't significant, because on the majority of "slam vs non-slam" deals it's not random whether you bid slam or not. If (and it's a big if) one pair have worse judgement in the slam zone, then they will bid the wrong ones: this total swing you expect it not random, it's a reflection on their bidding ability.

But this non-randomness is an alternative hypothesis. Of course, there are lots of alternative hypotheses that can't be rejected. But Arend's point is that the zero hypothesis, namely that a slam swing is equaly likely to go one side or the other, can't be rejected.

The slam swings are "random" in the sense that if 50% of all (in the long run) slam swings go this way and 50% go the other way, then the number of slam swings in the 40-sample that go this way is (40,0.5) binomial distribution, i.e. one slam swing is independent of another. This will normally be true unless playing goulash. (It could also be violated if the team who's lacking behind is going for swings, but that would make the dispertion higher and thus make the findings even less significant).

[hotShot]

Values that could be of interest:

% of bid contracts made
% of doubled contracts made

% of doubled opps but they made

big numbers here might show "solid bidding style"

declarer play:
% of the possible tricks made in bid contracts with reference to the double dummy result for the bid contract.
Example:
You play 2♥, dd solver says you can make 9 tricks and you got only 8 or perhaps you got 10.

defense:
% of possible defense tricks taken compared with dd solver result.
Example:
Opps play 2♥, dd solver says you can make 5 tricks and you got only 4 or perhaps you got 6.

sacrifices over opps game/slam:
Opps bid game/slam and your side played:
% of sac's that made a good score

If you opps have a low number here you should X them more often.

[cherdano]

FrancesHinden, on Feb 9 2007, 03:49 AM, said:

I'm not sure that this isn't significant, because on the majority of "slam vs non-slam" deals it's not random whether you bid slam or not. If (and it's a big if) one pair have worse judgement in the slam zone, then they will bid the wrong ones: this total swing you expect it not random, it's a reflection on their bidding ability.

Let's say that
i) All potential slam hands are 100% or 60% each equally likely and each with nothing to the play.
ii) Pair A, who are too conservative, bid all the 100% slams, but not the 60% ones
iii) Pair B bid all of them
iv) The swing for slam off/game making or both making is 10 imps either way

In 40 potential slam deals under these conditions, you expect
20 flat boards (excluded from winkle's analysis)
12 10 imp swings to pair B
8 10 imp swings to pair A

or a total expected swing of only 1 imp per board

In fact, if you then add in the condition that F&N are more likely to make 12 tricks than their opponents on average, and more likely to take 2 tricks in defence on average, it makes things look even worse (particularly if they are over-aggressive rather than conservative in the bidding).

Various comments:
1. I am not a statistician, so I don't really know which confidence analysis would be most appropriate here. (If I wanted to do this carefully, I would use Baysian formula to compare the a-posteriori likelyhood of, say, "FN are as good slam bidders as their opps" with "FN expect to lose 1 IMP per board on slam bidding swings" assuming they have equal a-priori-likelyhoods.) However, I think generally the "physicist's approach" -- compute the standard deviation and claim it is significant if "result +- 2*std deviation" is above or below what the null hypothesis predicts -- works well enough.
2. Your example above would get a loss of 2 IMPs per board in winkle's computations, note 1.
3. Maybe we are playing a different game but I seem to see many slams that are pretty close to 50% (finesse plus or minus small chances).
4. I guess Helene has explained it better than me...

Arend

[Rain]

Quote

Duboin/Bocchi and Fantoni/Nunes are awful....in slam bidding decisions

[MickyB]

I believe that your expected score when opening a strong club depends very much on your hand.

Minimum-ish balanced hands are happy to open 1♣ and then pass in competitive auctions. They rate to do well if partner describes his unbalanced hand.

Stronger balanced hands may have to take an uncomfortable call in competition, but a standard 2NT opening isn't dealing with these hands too well either.

Three-suiters are happy to pass if the opposition settle in one of their suits, and double for takeout if not.

Offensive single-suiters, particularly with a major, are well described by opening a strong club and then rebidding their suit at whatever level is necessary.

Of course, there are exceptions - maybe the auction will be at 5♦ by the time it gets back to you, and you will feel unable to introduce your massive club suit. Maybe the opponents will manage to play in your fit, giving you +350 against a slam, but those tactics can certainly backfire.

Other hands tend not to cope well in interference, as their second call won't describe the hand nearly so well.

Credit to DavidC - he covered these sorts of issues on his bridge blog much better than I could.

[hotShot]

I've been playing around and made me a program that is reading lin-files and analyzes the card play. At the end I calculate a number: unlucky played cards / board.

An unlucky played card is a card that reduces the number of tricks for your side.
This could be just unlucky as in not fishing a stiff K or some serious error.

A few results:
Star (world class) 0.22 (070 boards serious competition)
GIB 0.3 (about 1000 there are 3 GIB's playing each board )
Junior champion 1 0.33 (150 boards serious competition)
Junior champion 2 0.38 (200 boards training and serious competition)
Junior Star 0.4 (570 boards fun team games)
Senior Star 0.48 (330 boards fun games with other stars)
Me 0.5 (10000 boards with hundreds of pickup partner)
Intermediate 0.64 (700 boards)
Beginner Level 0.83