[all loomis]

in 1930's, milton work counted the number of times a picture card won a trick, in 1000 hands. divide by 100, and you had the original work point count. but they not convenient, so they were rounded off to the nearest integer, et voila. so nothing sacred about '1,2,3,4' unfortunately. useful at nt, but you have to learn to count tricks when evaluating a suit contract.

[all loomis]

 all loomis, on 2017-November-08, 21:57, said:

in 1930's, milton work counted the number of times a picture card won a trick, in 1000 hands. divide by 100, and you had the original work point count. but they not convenient, so they were rounded off to the nearest integer, et voila. so nothing sacred about '1,2,3,4' unfortunately. useful at nt, but you have to learn to count tricks when evaluating a suit contract.

btw, there was a bissell point count which you should look at, as a grounding in evaluating shape hands. it's easy and clever.

[bravejason]

I've read that the Work point count was developed for use with Whist. Given the similarities between the two games, applying the point count idea to bridge bidding would have been quite natural. How the sytem was developed for use in Whist, I've no idea.

There were many point count systems invented for use in bridge. 4-3-2-1 is the one that survived. I think its simplicity was the reason why. There are more accurate systems, but the improved accuracy was apparently never enough overcome the additional complexity except for the more serious players.

I've briefly experimented with alternate systems, but I'm not very experienced and I found that it is too much effort to apply a more complicated system and still have enough brain power left to think about other aspects of bidding like should I raise or show a new suit or overcall or pass or do something else. Give me a simple way to judge the strength of the hand and I can go from there. Give me a complex system and I get bogged down in the minutiae of the point count and miss the critical big picture understanding of if I need to be part-score, game, or slam and in what suit should I be in.

[Zelandakh]

This looks to be a candidate for "Necro of the Year", at just over 7 years. Nice work!

 blackshoe, on 2010-September-16, 21:15, said:

AFAIK, Milton Work invented the 4-3-2-1 point count, probably before the invention of Contract Bridge.

Just a small point. Work popularised the point count method that bears his name but did not invent it. In fact he was strongly against all point-based methods until the 1920s, whereupon he did a U-turn and supported the 4-3-2-1 model in a 1923 publication.

The real inventor appears to be Bryant McCampbell, who published the 4-3-2-1 model for balanced hands in 1915, reputedly not for bridge but for a game called auction pitch. Sometime around the end of 1915 or start of 1916 (sources vary) he then published a book on bridge. If anyone has this book they could perhaps confirm whether the Work count is incorporated. In any case, it does seem somewhat unfair that McCampbell's name has faded into obscurity with Work getting all of the credit.

The credit for confirming the validity of the 4-3-2-1 values should be given to a Canadian called William Anderson. Not only did he come up with (more) accurate values for the honour evaluations but he also went further and formulated the 3-2-1 distributional point model later publicised by Goren.

Most modern studies I have seen suggest that, to the nearest half a point, the 4-3-2-1 model is optimal for NT contracts but the 3-2-1-x model (or variants thereof such as 4.5-3-1.5-1, 4.5-3-1.5-0.5, or 6-4-2-1) are more accurate for suit contracts. There are plenty of further refinements that have been proposed too, some down to such detail that only a computer could practically use them. In reality, most players just seem to note their base hcp score and then mentally add or subtract according to the features of the hand and the auction.

To be honest though, how the numbers were derived does not seem to me to be of any practical use in playing the game itself. Surely in the end all that really matters is making the correct decisions as often as possible?

One last point. As has already been mentioned in this thread, there have been plenty of alternative point count systems suggested in the past. One of these comes from the Vienna system: 7-5-3-1. By my reckoning this is actually more accurate than 4-3-2-1 in terms of the relative values. Has anyone here ever tried experimenting with this at all?

[gordontd]

 Zelandakh, on 2017-November-09, 08:37, said:

One of these comes from the Vienna system: 7-5-3-1. By my reckoning this is actually more accurate than 4-3-2-1 in terms of the relative values. Has anyone here ever tried experimenting with this at all?

As a teenager I sometimes played with members of the Jain community in Mombasa, who all played a variation of the Vienna system that they called Stern and which used the 7531 Bamberger point count. I remember comments like “I couldn’t open, I only had 16 points.”

[Zelandakh]

 gordontd, on 2017-November-09, 10:34, said:

As a teenager I sometimes played with members of the Jain community in Mombasa, who all played a variation of the Vienna system that they called Stern and which used the 7531 Bamberger point count. I remember comments like “I couldn’t open, I only had 16 points.”

16 of their points being, of course, functionally identical to 10hcp in MW (AKQJ, AAJJ, AQQQ or KKQQ, although replacing any queen with 3 jacks would make a possible 11 along with KQQQJJ and QQQQJJJJ is a 12 count, albeit one I think almost everyone would downgrade). Interestingly, if the jack were bumped up to 1.5, all of these anomalies would disappear but the ace and king would still be worth more in relation to the quacks compared with MW (but not as much as in the 3-2-1 count and equivalents). I guess 14-10-6-3 (in whole numbers) would be asking quite a lot of the average bridge player though! Interestingly, the normalisation around K=3 gives 4.2-3-1.8-0.9, which seems like adjustments close enough to MW that they could practically be done in the head, retaining (more or less) familiar point count ranges for bids.

Out of interest, how did they do? And were you ever tempted to play around with the method yourself?

[gwnn]

7531 for NT? I don't think three jacks should only be equal to a queen. If we change it to 1.5, then we're pretty much back to Work (7-5-3-1.5 vs Work's 6-4.5-3-1.5 when scaled up 50%).

edit: oops you said as much too. pre-coffee posting strikes again.

[gordontd]

 Zelandakh, on 2017-November-09, 17:52, said:

Out of interest, how did they do? And were you ever tempted to play around with the method yourself?

It’s so long ago and I was too inexperienced to have a clear impression of how they did, though I do remember that they seemed to open 1C far too often.

My only similar experiments were with BUMRAP which awards 4.5, 3, 1.5, 0.75, 0.25, which is similar but I think a bit better.

This is much simpler in practice than it seems written down, since aces and queens cancel each other out, as do jacks and tens. So you just count points as usual, then adjust by 0.5 for each ace more than queens, by 0.25 for each ten more than jacks (and of course subtract if you have more queens or jacks than aces or tens).

[barmar]

 gordontd, on 2017-November-10, 04:01, said:

This is much simpler in practice than it seems written down, since aces and queens cancel each other out, as do jacks and tens. So you just count points as usual, then adjust by 0.5 for each ace more than queens, by 0.25 for each ten more than jacks (and of course subtract if you have more queens or jacks than aces or tens).

This is similar to the Adjusted Losing Trick Count -- you count losing tricks (missing AKQ in each suit, capped by the length of the suit), then subtract 1/2 loser for every ace more than queens, or add a loser for the opposite.

[Zelandakh]

 gordontd, on 2017-November-10, 04:01, said:

It’s so long ago and I was too inexperienced to have a clear impression of how they did, though I do remember that they seemed to open 1C far too often.

1♣ in Vienna shows a semi-limited opening with fewer than 5 spades, hearts or diamonds. You can perhaps see in this opening the reason why Vienna is regarded as the pre-cursor for all other intermediate club systems, such as Polish, Swedish, etc. I am not sure of the frequencies but as natural systems generally open 1♣ much too infrequently, it might well be that Vienna is closer to the theoretical optimum than, for example, Acol (without interference anyway).

 gordontd, on 2017-November-10, 04:01, said:

My only similar experiments were with BUMRAP which awards 4.5, 3, 1.5, 0.75, 0.25, which is similar but I think a bit better.

This is much simpler in practice than it seems written down, since aces and queens cancel each other out, as do jacks and tens. So you just count points as usual, then adjust by 0.5 for each ace more than queens, by 0.25 for each ten more than jacks (and of course subtract if you have more queens or jacks than aces or tens).

BUMRAP is another variant of the 3-2-1 method. As already pointed out, most authorities think these valuations are better than MW for suit contracts these days and I use a similar procedure, mentally storing both numbers (as the real evaluation tends to lie somewhere in between in my experience.

The suggested normalisation for 7-5-3-1.5 - 4.2-3-1.8-0.9 - works similarly, just with a modification of 0.2 instead of 0.5 for aces versus queens and, if desired, 0.1 instead of 0.25 for jacks-tens. For NT purposes, this in theory ought to be more accurate than either 3-2-1 or 4-3-2-1 schemes. In practice I personally feel that there is a big difference between supported and unsupported quacks, which ought to be factored in to the evaluations. Few methods do this despite many pointing ought the benefits of honours in combination.

It would be nice if someone went through the same procedure as McCampbell one day with this distinction. It would be very convenient if we got numbers like 2 for a supported queen and 1.5 for an unsupported queen; unlikely but convenient. If anyone has access to PK, perhaps he could obtain some figures from his database...

[gordontd]

 Zelandakh, on 2017-November-13, 06:59, said:

1♣ in Vienna shows a semi-limited opening with fewer than 5 spades, hearts or diamonds. You can perhaps see in this opening the reason why Vienna is regarded as the pre-cursor for all other intermediate club systems, such as Polish, Swedish, etc. I am not sure of the frequencies but as natural systems generally open 1♣ much too infrequently, it might well be that Vienna is closer to the theoretical optimum than, for example, Acol (without interference anyway).

This may be true, but my memory was that some of the players were so used to opening 1C with most opening hands that they sometimes forgot to open something else even when they had the hand for it!

[johnu]

Reserved for comments to be posted in 7 more years, in 2024.