An update on measuring bridge dealing randomness in two dimensions

Skip to first unread message


May 14, 2021, 1:24:46 PM5/14/21
I finally got around to looking at the spring 2021 NAOBC tournament results.

I have definitive bridge dealing measurement now. It converges to a useful result with a minimum number of hand records, and is easy to record by hand; very handy for the minimal ACBL record keeping practices.

I finalize this measurement with two values; bias, and volatility.

Bias is ideal if it is zero. Anything less than plus, or minus, 1.96 is acceptable. I think of this as the qualifying value.

Volatility is how much measurement jumps around. It is the essence of how I visualize randomness-in-general.

ACBL, BridgeComposer, Dealer4, and (I think with minimal evidence) BBO are using BigDeal as their dealing algorithm these days.

Left to its standard output, it will produce in short order around plus 5 volatility.

The NAOBC Baldwin Flight A Pairs produced minus 1.29 bias, and plus 5.62 volatility, values. Ho, hum; nothing unusual to see here.

The Open IMP Pairs produced minus 2.51 bias, and plus 2.09 volatility, values.

The Premiere Pairs produced plus 0.69 bias, and plus 1.40 volatility, values.

So, what happened to the bridge dealing in these last two events?

I think they are designed to attract elite bridge players. Numerous ordinary bridge deals will not do for such august luminaries. So the regular flow of random deals is “cherry picked” to exclude some number of ordinary deals. In these two instances, I would rate them as moderate cherry picking.

The now famous 253 bridge deals from last year’s USBF Invitational 1 tournament have minus 1.28 bias, and minus 0.92 volatility, values. I rate them as thoroughly cherry picked. It is a very long statistical way from plus 5 to minus 1; immense even.


Reply all
Reply to author
0 new messages