|Note: for those with a "Real" bridgebrowser disk, you can do all of
this on your local data. It is highly recommended that you get the updated
Day 1 can be found here
Day 2 can be found here
Day 3 can be found here
What is the distribution of contracts, and what are their expected
|1. Fire up BRBR Online. The latest version is 2.3A-h-4|
|2. Select S(can R(eset All (only need to do this once, just make sure you havent restricted all your searches on some previous visit)|
|3. Click on the tab about halfway down, BELOW "BRidgeBRowser (tm)" which says "Date/Scoring".|
|4. Adjust the right hand calendar to 6th July. The idea is to get a
full week of data, in this case will be 2674 boards including IMP and MP.
NOTE: if you stop a TIME search, you need to RESET and then set the
parameters again, as the "From" calendar always gets updated by searching
|5. If you are lazy or don't like messing around with spreadsheets you can (instead of 4.) instead go to the "View Tab" and check the hand limit box, and set the hand limit to 100,000 (default is 10,000). No need to enable "View" as an auxiliary term in this case. This will give you approximately round numbers by simple shifting of the decimal point - but unfortunately they will be mostly IMP. To defeat THIS problem you should do only IMP or only matchpoints in two separate runs, there is a radio button in the middle of the Date/Scoring tab to control this behaviour.|
|6. Click on the tab marked Bid Analysis, check the "Enable as auxiliary term" box. Check the "No listing" box.|
|7. Click back to "Date/Scoring" (the calendars) and press "Search". You can and should watch the data accumulate by clicking on the Bidding Analysis tab. If you stay on the tab with the calendars, you will spend lots of CPU time filling up the results window with data you don't need, and possibly even crash if you go on long enough. If the Bidding Analysis is frontmost when accumulating, the effect of the "No listing" box is to suppress the main display.|
1. Here is what the "Bid analysis" tab looks like on my screen. If your screen is big enough, stretch the window downwards by grabbing the bottom, thus:
Note that the "slider" above the table (not visible in the graphic)
to "CtrOS" (Contract Opener's Side) as in all the previous examples, since
we are most interested in contracts (moving it will change the contents
of the right 4 columns and can be done any time WITHOUT upsetting the cumulated
results). However today we will be looking into this a bit more closely.
|2. Notice the overwhelming number of 3NT contracts in the CtrOS column.
In the CrtNO (when the opponents of opening bidder play the contract),
perhaps it is not surprising that the biggest numbers are for 4H and 4S.
(I added together some numbers using Excel for the next statement).
Not counting partscores that count as game when doubled, 51.8% of all
contracts are Game (3NT,4H through 5NT), and a further 5.6% are slams (6C
and up). Never let it be said that Bridge is boring!
|3. The other aspect you may wish to consider is which contracts make a profit on average. Whilst there are some anomalies with low frequencies that would need to be checked out with lots more examples (eg 6C,7H,7S) the consistently profitable contracts are 1S,1NT,2H,2S,3NT,4H and 4S.|
|4. Check "Detailed" and then press the button marked "Write Report
(CSV)". You wil be prompted for the name of an output file. The result
is suitable to be read in Excel or any Spreadsheet program (Quattro, Lotus,
Works, etc). Included in the report are columns showing the Standard Error
of the mean (as a rough guide, the mean + or - one SEM covers 66% of the
distribution, I think, I am sure some statistician will correct me).
All of these numbers can be viewed interactively by holding the mouse over the cell in the table, if you don't feel like messing around with a spreadsheet. However the curious of you will surely want to do just that, mess around.
For example if some action (eg 3NT contracts) has an IMP average of 0.51 with an SEM of 0.03, you can be pretty sure it isnt a statistical fluctuation. But if the results (eg CtrOS 7H at matchpoints, 48% with SEM 5.2) diverge more, then you cannot place so much weight on it.
In the latter case, you would likely want to go back and look at (only)
grand slams, and get a bit more data.
|The next table is imported from my spreadsheet, and I have combined
all the contracts played by either opener or non-opener to get a single
figure, but distinguished MP from IMP.
NOTA BENE: the relative frequencies are probably accurate, but the result
scores on the low frequency contracts need to be done using the same number
of contracts or at least until the SEM falls to something closer to that
of the most frequent ones.
|IMP #||IMP%(freq)||IMP result||MP #||MP%(freq)||MP Result||Contract|
We establish the norms for percentages of contracts and their matchpoint results. This data was from a single week on okbridge!
Some possible assertions (feel free to challenge these!):
a. +0.5 imp is approximately equivalent to a 2.5% gain at matchpoints. Same for losses.
b. Playing 1 of a suit seems to be profitable most of the time, the exception being 1C at matchpoints. A related assertion which we have made is that you should always open in 4th seat holding more than your fair share (10) of HCP, and that Pearson points are simply incorrect.
c. 5NT is (probably) the worst contract of all time, followed by the (MUCH MORE FREQUENT) less well-known swamp contract, 2NT. You should really try not to play 2NT! This ties in with David Stern's original question in Day 1 about 12+12.
d. more data on slams is needed than presented here.
e. 3NT accounts for 18% of all contracts!
f. These norms of matchpoint results may be interesting because, for example, the success of a player or pair can be judged in terms of their ability to exceed them.
I hope this is useful. Please write to me and tell me what you didn't understand.
Here's the email again
|1. You should (always) get the latest client.
that address again: