What happened to "Bridge Analyser" (correct spelling) computer program?

[Douglas]

The Internet just leads me circles to nowhere.

Douglas

[Bertel Lund Hansen]

Post by Douglas
The Internet just leads me circles to nowhere.

Is it this one (an app)?

http://bridge.esmarkkappel.dk/

[Douglas]

12:35 PM (4 hours ago)

Post by Douglas
The Internet just leads me circles to nowhere.

Is it this one (an app)?
http://bridge.esmarkkappel.dk/
--
/Bertel

Unfortunately not. Thank you for the suggestion.

Douglas

[Peter Smulders]

Post by Douglas

12:35 PM (4 hours ago)

Post by Douglas
The Internet just leads me circles to nowhere.

Is it this one (an app)?
http://bridge.esmarkkappel.dk/
--
/Bertel

Unfortunately not. Thank you for the suggestion.
Douglas

Try this one:

https://bridge-analyser.software.informer.com/download/

[Douglas]

6:07 AM (4 hours ago)
https://bridge-analyser.software.informer.com/download/

It is a good download link. Thank you for the info.

However its program registration email is no longer
available. Therefore the program will fail after 28
trial days. And I know from experience it is not
renewable until I get a new computer to install it
on (for only another 28 days.)

Why this is important to more than myself is that its
bridge dealing outputs nearly optimal random bridge
deals in my current 60 99-each deal hand record stat
format; 0.29 mean, and 0.99 standard deviation.

This kind of verified output simply obsoletes all
other known bridge dealing programs currently
available now by some magnitude.

This means there is nearly no inherent average bias,
which in turn means to odds-playing bridge players
their plays will comport, on average, more often to
the calculated expected values than what happens
currently.

I believe the Bridge Analyser program needs to be
preserved in some fashion, commercially or free,
so that it available to those who might want access
to the best of what is available in current bridge
dealing practice.

It could also well lead to other bridge dealing
sources, commercial or free, to update their
particular random generator source.

Douglas

[Douglas]

Here is the trade off currently being made for the convenience
of computer dealt bridge deals.

When cards are shuffled and dealt, all 53,644,737,765,488,792,
839,237,440,000 possible deals are possible for the players.

When a computer currently deals, only about a trillionth of
that 29 decimal digit number of deals are possible for the
players.

The only current exception to that trade off seems to be Bridge
Analyser. Which program, of course, is disappearing from
practice.

I tell you this in an attempt to let you visualize the
vast difference in actual playing potential.

Douglas

[Peter Smulders]

Post by Douglas
Here is the trade off currently being made for the convenience
of computer dealt bridge deals.
When cards are shuffled and dealt, all 53,644,737,765,488,792,
839,237,440,000 possible deals are possible for the players.
When a computer currently deals, only about a trillionth of
that 29 decimal digit number of deals are possible for the
players.
The only current exception to that trade off seems to be Bridge
Analyser. Which program, of course, is disappearing from
practice.
I tell you this in an attempt to let you visualize the
vast difference in actual playing potential.

The source and executables of BigDeal are in the public domain.
It can generate every possible bridge deal with equal probability.
https://sater.home.xs4all.nl/doc.html

[Douglas]

Post by Peter Smulders
The source and executables of BigDeal are in the public domain.
It can generate every possible bridge deal with equal probability.
https://sater.home.xs4all.nl/doc.html

You are looking at a classic example of perfectly well
intentioned people relying on someone else's untested
theory.

I refer you to page 5 and 6 of Knuth's "The Art of
Computer Programming, Volume 2."

The output of the deals is the eating of the pudding. It is
simply not what it is claimed to be.

I know this is hard information for you to take in.

I've tested another run of 30 99 deal hand records. The
first run produced a mean of 0.84. The second of 0.87.
Then I did not have a scale established, so I assumed
that those were in-range values. When I get to integrating
those two runs into a single 60 results I expect to have
a mean somewhere between those two values.

But now I have a scale established. I finalized it last
evening. I know the optimal value (you would call it E.V.)
is 1.0. I have 60 instances. The square root of 60 is 7.746.
When I divide the uniform standard deviation of 0.288675 by
that value I get 0.037267782249... Multiply by 1.96, and
I get 0.073. This is the 95% (two-sided) confidence interval.

1.0 minus 0.073 = 0.927. 0.8 something is way outside this
confidence interval, and therefore these 60 hand records
are far too biased to be unbiased random deals.

In case you haven't noticed, I've succeeded in "operationally
defining" unbiased randomness.

My final evaluation steps enumerated above are known as
"standard error" calculation.

Douglas

[Peter Smulders]

Post by Douglas

Post by Peter Smulders
The source and executables of BigDeal are in the public domain.
It can generate every possible bridge deal with equal probability.
https://sater.home.xs4all.nl/doc.html

You are looking at a classic example of perfectly well
intentioned people relying on someone else's untested
theory.
I refer you to page 5 and 6 of Knuth's "The Art of
Computer Programming, Volume 2."
The output of the deals is the eating of the pudding. It is
simply not what it is claimed to be.
I know this is hard information for you to take in.
I've tested another run of 30 99 deal hand records. The
first run produced a mean of 0.84. The second of 0.87.
Then I did not have a scale established, so I assumed
that those were in-range values. When I get to integrating
those two runs into a single 60 results I expect to have
a mean somewhere between those two values.
But now I have a scale established. I finalized it last
evening. I know the optimal value (you would call it E.V.)
is 1.0. I have 60 instances. The square root of 60 is 7.746.
When I divide the uniform standard deviation of 0.288675 by
that value I get 0.037267782249... Multiply by 1.96, and
I get 0.073. This is the 95% (two-sided) confidence interval.
1.0 minus 0.073 = 0.927. 0.8 something is way outside this
confidence interval, and therefore these 60 hand records
are far too biased to be unbiased random deals.
In case you haven't noticed, I've succeeded in "operationally
defining" unbiased randomness.
My final evaluation steps enumerated above are known as
"standard error" calculation.
Douglas

This story is scattered over many postings now, so would you care to
recapitulate what you are doing?
As far as I understand you take a set of 99 deals and from those extract
"a number". Precisely how do you do this? Please give enough information
so that one can reproduce the results.
This number should have an average value of 1.0 (correct?) but in the
case of BigDeal the average was found to be 0.84 and 0.87 which was far
outside the standard deviation. Is that a correct interpretation of your
findings?
What does this mean for the way in which the deals produced by BigDeal
show an observable bias?

[Douglas]

Post by Peter Smulders
This story is scattered over many postings now, so would you care to
recapitulate what you are doing?
As far as I understand you take a set of 99 deals and from those extract
"a number".

From a set of 99 deals (396 hands) I derive six (6) probabilities. I
transform each to a standard normal z_value. Now I have six (6)
z_values. Next I calculate their standard deviation. That is your
"a number" I presume.

Post by Peter Smulders
Precisely how do you do this? Please give enough information
so that one can reproduce the results.

I sincerely hope this is precise enough so far.

I now repeat the same procedure for the next 99 deal hand records
for 29 more times. In the case of the "gold standard" and Bridge
Analyser deals, I continued on to 60 of these cycles for each
before I finalized my two numbers for each.

Post by Peter Smulders
This number should have an average value of 1.0 (correct?)

At last, I get to pick a nit. 1.0 is the optimal value for both
the "a number" mentioned above, and the final mean value upcoming.
There is no should to it, it is simply a neutral central measuring
point.

Post by Peter Smulders
but in the case of BigDeal the average was found to be 0.84 and
0.87 which was far outside the standard deviation. Is that a
correct interpretation of your findings?

Essentially correct. Since your posting, I found time to measure
how far out it is in (of course) a z_value. It is minus 3.17.

Post by Peter Smulders
What does this mean for the way in which the deals produced by BigDeal
show an observable bias?

That 3.17 amount IS the observable bias.

I also calculate the standard error for the standard deviation
of the 60 accumulated z-values.

These are my final two numbers for all 60 hand records. Mean and
standard deviation just like Stat 101. My teacher would be so proud.

Douglas

[Lorne]

Post by Douglas
The Internet just leads me circles to nowhere.
Douglas

The company that was providing the web server service suddenly closed
the server and the first attempt to relocate was a disaster and the new
server (suggested by the old provider) kept going offline or blocking
the page.

However it is now up and running again here:
https://www.cantab.net/users/bridge.analyser/Bridge/index.htm

If you have a current version of the program then do a new full install
and it will create all the links to the new site for future upgrades.

Lorne