[1eyedjack]

Not having BridgeBrowser myself I would be interested to know if its statistical sampling and analysis is capable of undertaking the following two functions:

1) Work out the average net imps as a function of the number of bids to reach the contract.

The purpose would be to evaluate the "four-handed" issue of whether accurate bidding to find the best theoretical contract (using a large number of bids to reach it) loses more by disclosing information to opponents than it gains. In other words you might poll all contracts that ended in 3NT, and work out the net IMP effect if you get there in only two bids (1N-3N) or in 4 bids (1N-2C-2D-3N).

I appreciate that you might need to take some care in phrasing the problem to BridgeBrowser. You would only want to include in the 1N-3N population those hands which had the potential for alternative treatment (ie responder has 4 card major).

2) Assign a weighting to the IMP gain/loss on a particular hand depending on the average IMPs of the opponents over a period leading up to that hand.

Thus, if in one hand within the population, N-S, whose average IMPs over the last month was break-even, gain 2 IMPs against opponents whose average IMPS over the that period was -2 IMPs, then you might rate the hand as neither gain nor loss for statistical purposes, and put down the +2 IMP absolute gain as down to a disparity in the levels of skill of the opposing players, rather than as a result of a feature of the cards.

The Lehman rating system in OKB did something like this (using an algorithm that was about as easy to understand as Schedule 23, Finance Act 2004). I am NOT advocating publishing individual players' ratings on this basis, only using it to evaluate and weight the significance of individual results in a sample.

I have a horrible suspicion that if BB is capable of doing this then it would take until the resolution of the Middle East conflict before it finishes one set of calculations.

[inquiry]

1eyedjack, on Aug 18 2006, 02:56 AM, said:

Not having BridgeBrowser myself I would be interested to know if its statistical sampling and analysis is capable of undertaking the following two functions:

1) Work out the average net imps as a function of the number of bids to reach the contract.

The answer to question one appears to clearly be no. I say appears to be, since I see no obvious way to do it, but I have found non-obvious ways to do several things. The bidding analysis tool, which is what you typically to go through openers rebid effectively. That is, you can study the statistical advantage of opening bid, overcall, response, advancers bid, and openers rebid. Then, of course, final contract. You can combine these bids with ohter functions (vul conditions, hcp, suit legth, suit quality, controls, overall hand shape, etc).

As for question 2, ratings are overrated. We can talk about what you can do and not do, none of which takes all that long.

[tlgoodwin]

As I recall it, Jean-Rene Vernes studied this issue. He looked at World Championship deals where the same contract was played at both tables, but after a different number of bids by the declaring sides. The result was not at all surprising: the side that used fewer bids did significantly better on these deals than the more talkative side. I don't recall just how much each extra bid "cost," according to Vernes. The study was published either in the "IBA Annales" or in Pierre Collet's "Bridge de Competition" magazine, late 1960's or early 1970's.

The result of the study wasn't surprising: extra bids tend to convey extra information, and extra information tends to be useful to the defenders in the same-contract situation. The flip side -- and the side that is perhaps harder to study -- is that extra information may be useful to the declaring side when it points them to a different contract.

TLGoodwin