Join Bridge Winners
All comments by Thomas Andrews
1 2
You are ignoring the author of this comment. Click to temporarily show the comment.
Technically, I never checked on Beasley's claim - he emailed me that he had published this hand, and I took him at his word.
Dec. 4, 2016
You are ignoring the author of this comment. Click to temporarily show the comment.
The origin if my “fake” nickname, “Binky,” was that I was in a show, and was told my bio for the show could include any nickname. I realized I didn't have a nickname, so I picked “Binky” and put it in the program, taking the name after one of the characters from Matt Groening's cartoon, “Life in Hell.” Binky was a rabbit in that cartoon.

That was way back in 1990 or so. There's a few people who still call me “Binky” occasionally.

I've been calling my hand evaluation method “Binky Points” for at least a decade, possibly as long as 15 years (which is how long they've existed, although I'm not sure when I named them.)
Dec. 11, 2015
You are ignoring the author of this comment. Click to temporarily show the comment.
(Ignore - why can't delete?)
Dec. 11, 2015
Thomas Andrews edited this comment Dec. 11, 2015
You are ignoring the author of this comment. Click to temporarily show the comment.
Nope, “Binky” is just a “fake” nickname that I acquired a long time ago. Too long and irrelevant a story to tell here.
Dec. 11, 2015
You are ignoring the author of this comment. Click to temporarily show the comment.
Binky Points are just my double-dummy-based hand evaluation tools. I called them “Binky” because I didn't want people to take them too seriously. (Double-dummy undervalues queens, for example.)

The average balanced Work Point 15-count has a Binky Points value of about 5.66 in both suit and notrump valuations. (This is an oddity for 15 - it is not often that the suit and notrump values are the same.)

The average for 14-count balanced is about 5.10 in notrump and 5.30 in suit (that's off the top of my head. Usually, a change of one point in Work usually corresponds to a change of about 1/2 BP in notrump, and about 1/3 BP in suits.)
Dec. 11, 2015
You are ignoring the author of this comment. Click to temporarily show the comment.
Here are some sample hands that Binky Points considers better than the average balanced 15-counts for both suit and notrump contracts:

S: AT84
H: KT
D: AT94
C: K96

S: A942
H: A8
D: J953
C: AJT

S: A2
H: 974
D: AQ5
C: AT985

S: 98
H: AT6
D: AK5
C: KT862

S: A54
H: T52
D: AJ7
C: AJT9

S: AQT7
H: 876
D: AT86
C: A3

S: A972
H: AJ92
D: A4
C: JT7

S: T5
H: KT94
D: AK6
C: AT84

S: KT2
H: T2
D: AJT
C: AQT85

S: QT63
H: AT5
D: A9
C: A863

These are the “clearest” good 14 counts.
Dec. 10, 2015
Thomas Andrews edited this comment Dec. 10, 2015
You are ignoring the author of this comment. Click to temporarily show the comment.
One of my favorite cases of restricted choice: After a 1N-3N auction, an opponent leads from four small hearts. Later, you can treat as lower the possibility that he started with four small spades, as well, helping you get count, or maybe giving you an idea where spade honors are.
March 14, 2015
You are ignoring the author of this comment. Click to temporarily show the comment.
You can't really do anything except (1) review the source code, if you have it. Dealer4 might well be using the *exact same* software as the ACBL does in tournaments. (2) Statistically analyze the deals.

It is certainly arguable that “human shuffling” should be emulated by the computer, rather than going for the purer randomness. I prefer the purer randomness of the computer hands.

Whenever a bridge book states the odds of an event, unless it explicitly states it is analyzing actual hand-dealt deals, it is calculating the odds based on the assumption of the purer randomness. It is virtually impossible to determine the odds of anything with much precision about human-shuffled boards, and those odds could change over time as shuffling habits change.

True story. I was playing in a team game in the late 90s, and we were assigned to a table where we had played the previous round. We shuffled the boards and dealt, but we apparently missed one, because I got exactly the same hand as last round, and the auction started identically, so we agreed to re-deal the board. I took the cards and shuffled the deck seven times, per a result from the 1980s, much to my RHO's impatience. I explained the theory as I finished the shuffle. The auction ended up in a slam doubled, making, played by my RHO, with my RHO holding some 8-5-0-0 pattern. After the play, he said, “That was like one of those computer-dealt hands!”

July 16, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Yeah, but there are 39 random numbers being generated. Since the bias is always towards the lower number results, I think there is a cumulative risk. Also, not sure, but it could well be 32 bits rather than 64. I'd have to check the code.
July 12, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Whoops, meant this for this discussion:

Last time I asked, it was more than enough to ensure that all possible deals could be dealt, but not enough to ensure that all possible tournaments of 36 deals could be dealt.
July 11, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
July 11, 2014
Thomas Andrews edited this comment July 11, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Greg, I'm not sure the modulo trick keeps some deals from being belt, it only makes some deals more likely.

On the other hand, if your random number generator uses a seed with fewer than 96 bits, you certainly can't get all deals. Indeed, for a proper tournament, you'd want about 2^{36*96} bits of seed to ensure that you could get 36 independent deals. But nobody actually needs that level of randomness.

The swapping that you describe is essentially what I do, of course, while populating each hand.
July 11, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
I once joked we could greatly increase the number of people who play if we reduced it to 10 cards per suit, so people can count on their fingers…

You don't need a calculator to prove that 52^52 is not divisible by 52!. 52! is divisible by 3, 52^52 is not…
July 11, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Yes, Greg, that was for speed. Each hand is dealt only when information about it is needed, to cut down on random number generation. When I made that change, it was because I found via profiling that a huge amount of time was spent in the RNG. Quite a lot of the time, you only have a condition on one or two hands, so while deal all 52 cards if you are ditching most of the deals?

There are still problems. When you generate a random number from 1 to N, it is rarely uniform for N anything other than a power of 2 - most stock RNGs compute a number from 1 to 2^M and reduce modulo N. That can cause some skew. Certainly not enough to affect the odds visibly without a huge simulation - I'd guess billions of deals or more, and it is not clear to me which way the deals would skew. Certainly, this is still closer to a “pure” simulation than physically shuffling a deck 7 times.

Funny story. I was playing in a team game in the late 90s, and we were assigned to a table where we had just played the previous round. We shuffled the boards and dealt, but we apparently missed one, because I got exactly the same hand as last round, so we agreed to re-deal the board. I took the cards and shuffled the deck seven times, much to my RHO's impatience. I explained the theory as I finished the shuffle. They ended up in a slam doubled, making, with my RHO holding some 8-5-0-0 pattern. After the deal, he said, “That was like one of those computer-dealt hands!”
July 11, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
I'm a mathematically inclined person. I'm certainly capable of writing wrong code, but usually such impressions are due to faulty intuitions of the actual probabilities of bridge, Without an actual example of a simulation, hard for me to defend my software and/or correct any existing problem…
July 10, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Bully for you. Please read the word “might” and try to understand the point I was making, rather than getting defensive. If your comment was out of context, and you didn't mean to imply that bidders should assume a 2NT opener is likely 19 points, that is why I used the word “might.” But then, your comment is out of context.

The *context* of your comment is not, by definition, your comment, the context is the discussion you entered. Please learn the meaning of the word “context” before you use it. Good luck defending your easily bruised self-esteem.
April 19, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Consider the context of your comment, and note the use of the word “might.” And don't be so touchy.

We were discussing how to respond to a 2NT opener. Your comment indicates that you think we should act based on the fact that most 2NT openers are actually 19 points. If you do so, then you certainly *might* be misinforming opponents.
April 19, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Then you may be misinforming your opponents.
April 19, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
If your partner upgrades a 19-count, it is because he thinks it plays as well as an average 20-count. If it doesn't, then he was wrong.
April 19, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Never mind - ignore that original post, my script was broken.
April 18, 2014
1 2
.

Bottom Home Top