[smerriman]

 Dumoti, on 2018-July-15, 11:13, said:

The point of the example is to debunk the claim that Aces + upgrades is superior to Banzai Points when the hands are the right shape.

I thought you were implying that your example was a reason you should use Banzai points instead of normal points, because normal points got you to a bad 3NT. How do Banzai points get you to a better contract? If you're not talking about opener, are you saying you shouldn't have accepted with responder's three kings? Because I'm pretty sure simulations will again disprove that.

So obviously I was wrong, and you were just showing an example hand where both methods fail equally. That doesn't debunk anything.

To show Banzai points may be worthwhile you'd need to provide a hand where the use of Banzai points gives a better result on average. Stephen Tu has already provided a hand (your own, AJxx Axx xxx xxx) where statistics support the opposite.

And again, asking to provide *two* hands has no relevance. Only the average for the hand where a decision is being made. It may even be true that for every single 4333 vs 4333 hand that fails, Banzai points said you shouldn't be in game. That still wouldn't demonstrate in any way that you should use them in the 100% of cases where you don't know the other hand.

[Stephen Tu]

 Dumoti, on 2018-July-15, 11:13, said:

Yes, but when partner opens 1NT his possible shapes are 4-3-3-3, 4-4-3-2, and 5-3-3-2. That means that 67.4% of the time he will have a hand that is the right shape for applying Banzai Points. So if responder also has a 4-4-3-2 or 4-3-3-3 hand, then applying Banzai Points might be worth the trouble and that is something that I wish to investigate. What's the problem?

The problems:

• Other people (Andrews, at least helene & rhm on this forum) have already investigated this thoroughly and have concluded that they are aren't worth the trouble, based on their computer studies, and common sense. You dismiss these results as worthless with no evidence of your own, just some conviction that Cowan is right, even though you have not actually tried applying this at the table yourself, and that Cowan analyzed suits in a vacuum rather than entire deals which is far less convincing than people who analyzed whole deals.
• You are positing some hybrid problem where opener opens using work points and then responder responds using Banzai. I don't see how this can let responder bid to an accurate Banzai total, because a 15-17 hcp range for 1nt can be as few as 19 Banzai, and as many as 30 (though high totals like 27-30 are exceedingly rare, practically nil in practice; 30 would require a quite specific 17 count with all tens and no spot cards, a kind of inverse-yarborough). How do you suppose to solve that? Just go with the average spread which is around 21-24 Banzais, and require 16+ Banzai to FG and 14-15 to invite?
• These days people to avoid rebid problems also open some non insignificant # of 6m322 and 5m422 shapes 1nt.

 Dumoti, on 2018-July-15, 11:20, said:

Well, if it's so easy, then why didn't you do it when I asked you to?

Anyone can invent numbers. What I asked you to do was to invent two 4-3-3-3 hands in which the 4.25-3-2-1-0.5 system was superior.

Can you do it? Or can't you?

OK, here are 6: First three Banzai says bid game, HCP says no. Second 3 HCP says bid game, Banzai says no:

---

---

---

---

---

---

4333+4333 is the worst case for Work count (and for Cowan for that matter), on min cases you are going to need luck to make, but when you invite/GF you really have no idea whether partner is some other shape with 5 cd suit or 6 cd minor or 4432 with good honor mesh which will make a large difference in overall results. If you are conservative to cater to 4333 hands then you might well be winning on 4333 vs. 4333 but lose too much net on the other possibilities. Plus when you have 26 hcp and bid game with flat vs. flat and it's bad you have field protection because no one else can really avoid game, you have to be pretty sure of things to stay out and hang your result all on the auction and not rely on superior declarer play. And really most of these cases Cowan is likely bidding game also, it's not like people have tools to figure out 4333 vs. 4333 and can figure out to stay low in reasonable time.

If you want to say that "be conservative with 4333 hands, especially lacking spots", I don't think anyone disagrees. But if it gets down to "treat Q+J or J+J+T as equal to an ace for opening or responding" or "don't accept invites holding 3-4 aces because they are overrated", that's far fetched for the rest of us.

[helene_t]

You often hear beginners say something like "I only had three aces so I didn't open". Suggesting that you need 4 1/2 tricks in order to have half of the combined strength required to make 3NT.

Suppose partner has four queens. With four aces in your own hand, you expect six tricks in total as you can lead towards his queens, and on average 2 of the 4 finesses will work.

Now suppose you have four kings instead. In each of the suits, an opp will have the ace over either of the honours so you expect four tricks in total.

In that scenario:

4 aces = 4 tricks

4 aces + 4 queens = 6 tricks

4 kings + 4 queens = 4 tricks

4 aces + 4 kings = 8 tricks

4 kings = 2 tricks

4 queens = 1 trick

Of course, this is hopeless oversimplified. Sometimes opps will help you, sometimes you have tricks that you don't have time to cash as opps will cash their tricks first, etc etc.

Such oversimplified scenarios are not convincing. Their are thousands of scenarios which you could chose, so if you want to advocate Vienna Points, Work Points, Banzai points or whatever, you can always find suitable scenaria.

The only way to investigate the accuracy of hand evaluation methods objectively is to do a statistical analysis thousands of hands, finding out how many tricks are actually taken in practice for different combinations of honours. This is easy to do, since data are available, for example in BridgeBrowser or the VuGraph Archive.

It turns out that Work points (4321) underestimate aces slightly for the purpose of notrump contracts. 4.2 rather than 4.0 is more accurate. Presumably this is among other things due to the fact that with aces, you can control tempo.

Compare Axx to QJx. You may think that both will give you one trick in a notrump contract so they should be worth the same. But Axx is more flexible. You can take the first trick if you fear a switch, or you can hold up twice if you want to break opps' communication in the suit. With QJx, it will be opps that control the tempo. They can cash their A and K and then switch. Or they can duck one trick and then cash the entire suit later.

[Tramticket]

The assumptions under-pinning this whole thread are full of holes. The discussion seems to assume that "if only I could find a more accurate way of counting points, my bidding problems would be solved" and "if the valuation were more accurate I wouldn't need to make little adjustments". In real life, the value of a hand is not just determined by the number of honour cards that you hold, but by their placement in the whole hand. Furthermore, the valuation should not be static, but instead should be continuously changing and evolving as the auction progresses. Any point-count assessment is only the starting point for understanding and valuing the hand.

Let me give you three holdings: (i) ♥K3 ♦Q642 (ii) ♥63 ♦KQ42 (iii) ♥63 ♦KQ98. They each have the same point count using Banzai Points or the standard Milton Work Points. But I would judge that (ii) is a better holding than (i) and that (iii) is a better holding still. Honours are working together in the second holding which is a positive adjustment. Honours are supported by better spot cards (the 9 and 8) in the third example - again a positive adjustment.

Now imagine that the auction develops with partner showing a strong two-suited hand with hearts and clubs. Now our KX in hearts is looking like a very useful feature and hand (i) looks much more promising. Hands (ii) and (iii) are looking much less promising and the ♦KQ may be of little use to partner.

If you want to improve your bidding accuracy, I suggest you concentrate on understand how valuations change and evolve rather than focusing on counting points. The Milton Work count (4-3-2-1) has stood the test of time and is a reasonably accurate starter - there is no need to re-invent the wheel.

[Dumoti]

 Stephen Tu, on 2018-July-15, 15:02, said:

The problems:

• Other people (Andrews, at least helene & rhm on this forum) have already investigated this thoroughly and have concluded that they are aren't worth the trouble, based on their computer studies, and common sense.

As I told you before, I have already read all the threads regarding Banzai Points in BBOs forums. Helene's conclusion was that it would earn you 2 IMPs per 1000 boards. She then sniffed disdainfully at the concept of writing a book about Banzai Points and said " Buying a copy of BridgeBrowser would constitute a small fraction of the costs of publishing a book."

Andrews' work has already been debunked.

Quote

You dismiss these results as worthless with no evidence of your own, just some conviction that Cowan is right, even though you have not actually tried applying this at the table yourself, and that Cowan analyzed suits in a vacuum rather than entire deals which is far less convincing than people who analyzed whole deals.

You are positing some hybrid problem where opener opens using work points and then responder responds using Banzai.

No. I came on here asking 3 specific questions:

1. If partner opens a 15-17 1NT how many Banzai Points should I figure he has?
2. How many Banzai Points would someone need to invite?
3. How many Banzai Points would someone need to bid game?

Instead the whole thread has been hijacked and turned into a let's-slam-the-new-guy fest.

Quote

I don't see how this can let responder bid to an accurate Banzai total, because a 15-17 hcp range for 1nt can be as few as 19 Banzai, and as many as 30 (though high totals like 27-30 are exceedingly rare, practically nil in practice; 30 would require a quite specific 17 count with all tens and no spot cards, a kind of inverse-yarborough). How do you suppose to solve that?

Yes, I am aware of the problem. That's why I came on here and asked people what they thought. My tentative solution was to assume that the 15-17 NT opener would probably have 22-26 BZP, average 24. I came up with these numbers by multiplying 15 by 1.5 and 17 by 1.5 to come up with 22.5 to 25.5, which I widened to 22 to 26. Could that become problematic? Sure. That's why I was going to investigate the matter to determine how well it worked.

Quote

Just go with the average spread which is around 21-24 Banzais, and require 16+ Banzai to FG and 14-15 to invite?

How did you come up with the average of 21-24 when I came up with an average of 22-26?

Quote

OK, here are 6: First three Banzai says bid game, HCP says no. Second 3 HCP says bid game, Banzai says no:

---

---

---

---

---

---

All right. I'm not going to sift through all of them. I'm simply going to focus (for now) on the first one.

The first question we need to ask is: How many BZPs are required to bid game? It is generally agreed that, using a standard 40-point system, one needs 26 to bid game, though this number can be shaded to 25. Accordingly, it seems to me that with a 60-point system, the new number should be 37.5 - 39 BZPs to bid game. In fact, I am pretty sure I said something along those lines in my initial post.

So, the first thing I notice about your first example is that it contains only 37 BZPs. This puts the contract at 1-2 points shy of the number I came up with for bidding game.
The second thing I note is that the contract fails only because you have carefully arranged the spades to be 4-1 offsides. If we switch the E-W hands, we might well get a diamond lead and pick up both spades honors for an easy overtrick.

I am also amused by your example 4. You have given yourself 26 HCPs, but the hand only makes 9 tricks because the club hook is on. If we switch the E-W hands, the contract will likely fail (unless you make the double-dummy play of a low club off of the board, planning to put in the jack). And if we use Andrews' method of 4.25-3-2-1-0.5 we get a whopping 27.25 points when supposedly you only need 25.75 to make game.

Not very convincing, I'm afraid.

Quote

If you want to say that "be conservative with 4333 hands, especially lacking spots", I don't think anyone disagrees. But if it gets down to "treat Q+J or J+J+T as equal to an ace for opening or responding" or "don't accept invites holding 3-4 aces because they are overrated", that's far fetched for the rest of us.

No, I'm not saying ANYTHING. I asked 3 questions:

How many BZPs should I figure if partner opens a 15-17 NT?
How many to invite?
How many to bid game?

Instead of getting an answer, I have gotten sucked into defending a system that I have yet to try out over the board!

Give me a break, dude! Take a chill pill and just answer the questions!

[Dumoti]

 Tramticket, on 2018-July-16, 02:17, said:

The assumptions under-pinning this whole thread are full of holes. The discussion seems to assume that "if only I could find a more accurate way of counting points, my bidding problems would be solved" and "if the valuation were more accurate I wouldn't need to make little adjustments". In real life, the value of a hand is not just determined by the number of honour cards that you hold, but by their placement in the whole hand. Furthermore, the valuation should not be static, but instead should be continuously changing and evolving as the auction progresses. Any point-count assessment is only the starting point for understanding and valuing the hand.

Let me give you three holdings: (i) ♥K3 ♦Q642 (ii) ♥63 ♦KQ42 (iii) ♥63 ♦KQ98. They each have the same point count using Banzai Points or the standard Milton Work Points. But I would judge that (ii) is a better holding than (i) and that (iii) is a better holding still. Honours are working together in the second holding which is a positive adjustment. Honours are supported by better spot cards (the 9 and 8) in the third example - again a positive adjustment.

Now imagine that the auction develops with partner showing a strong two-suited hand with hearts and clubs. Now our KX in hearts is looking like a very useful feature and hand (i) looks much more promising. Hands (ii) and (iii) are looking much less promising and the ♦KQ may be of little use to partner.

If you want to improve your bidding accuracy, I suggest you concentrate on understand how valuations change and evolve rather than focusing on counting points. The Milton Work count (4-3-2-1) has stood the test of time and is a reasonably accurate starter - there is no need to re-invent the wheel.

Yeah. I've played bridge before. I'm certainly aware that ♠xx ♥AKxxx is better than ♠AK ♥xxxxx.
Thanks for restating the obvious. We all appreciate it.

[helene_t]

 Dumoti, on 2018-July-16, 11:19, said:

Helene's conclusion was that it would earn you 2 IMPs per 1000 boards.

Thanks for the citation

As I remember, it was 2 IMPs per thousand boards by replacing 4321 with "optimal" AKQJ counting which is something like 4.2-3.1-1.8-0.9. I don't think it had anything to do with Banzai points.

[Stephen Tu]

 Dumoti, on 2018-July-16, 11:19, said:

Andrews' work has already been debunked.

By whom? Certainly not by you.

Quote

1. If partner opens a 15-17 1NT how many Banzai Points should I figure he has?
2. How many Banzai Points would someone need to invite?
3. How many Banzai Points would someone need to bid game?

1. 21-24 ish. There are a decent number of outliers. Average Banzai between 22 and 23, median 22.

2./3. 14-15, 16+ GF if you go by the Klinger/Jackson book recommendation of 37+ Banzai for game. I don't think anyone here has really studied what the threshold is supposed to be, but Klinger is likely about right. If you think it's more, why don't you go out and try it, report back?

If these were the only questions you wanted answered, you shouldn't have ended your initial post with the rather open-ended.question "Thoughts?". Though this being an open forum, you can't really control what people post.

Quote

Instead the whole thread has been hijacked and turned into a let's-slam-the-new-guy fest.

That's because of your displayed attitude to the responses, further explanation below.

Quote

How did you come up with the average of 21-24 when I came up with an average of 22-26?

It's trivial to instruct a dealing program to deal out random deals, select any 15-17 hcp balanced hand that comes up for say south, count up using Banzai point alternative, and report the results in a table of hcp vs. Banzai along with the average. My sample results looked like:

Frequency hcp vs banzai:
   	19 	20 	21 	22 	23 	24 	25 	26 	27   Sum
  15	38	392	834	644	233 	46  	3  	0  	0   2190
  16	0  	2	191	557	578	262 	53  	6  	0   1649
  17	0  	0  	0 	83	349	443	220 	57  	9   1161
Sum	38	394	1025	1284   1160	751	276 	63  	9   5000

Your estimate of scaling by 1.5 doesn't really work, because it's generally easier to get dealt this large # of hcp using fewer high value high cards (aces/kings) than multiple lower value high cards, which take up more of partner's 13 slots, leaving fewer combos of the small cards. Easier to get dealt 2 aces than all 4 queens. So this favors lower Banzai counts being held.

Quote

The first question we need to ask is: How many BZPs are required to bid game? It is generally agreed that, using a standard 40-point system, one needs 26 to bid game, though this number can be shaded to 25. Accordingly, it seems to me that with a 60-point system, the new number should be 37.5 - 39 BZPs to bid game. In fact, I am pretty sure I said something along those lines in my initial post.

If you want to say 38 or 39 then fine, I could come up with counterexamples also. And it would become even easier to find HCP games below this higher Banzai threshold that are actually decent percentage rather than lucky to make. But singular counterexamples are silly and prove nothing, I only really did it to appease you since you demanded a 4333 x 4333 counterexample. I just had the computer select random deals that met the constraints, I wasn't trying to construct hands intending to prove anything, because singular hands don't really prove anything.

Quote

And if we use Andrews' method of 4.25-3-2-1-0.5 we get a whopping 27.25 points when supposedly you only need 25.75 to make game.

Andrews claim is not that "if you have 27.25 hcp between the hands, game makes", or "if 25.75 pts are held, game makes". Or if that if you have 4333 vs. 4333 and this point total game makes or is 50%. His claim is that if you hold 26.25 combined holding this method (25.75 is for 40% games, i.e. you want to bid at IMPS vul), not knowing whether partner is also mirror flat or not, over all hands, you'll make about half the time, so if you have more it's probably a good MP game. Note it's not game 100%, it's game 50%, that means half your games are going down at this threshold, and at MP you'd be indifferent about bidding game at the threshold.


People have been playing bridge many decades longer than you have. I think if raising 1nt to 3nt on 10 balanced was not working out well enough on average they would have figured it out. In ancient days books wanted you to have 26 to bid game (maybe this is rubber bridge influence, where you can make game on the next hand and carryover the partial has value, so being more conservative is OK), players started bidding games on 25 because it works out, not all the time, but more often than not.

Quote

Instead of getting an answer, I have gotten sucked into defending a system that I have yet to try out over the board!

Why the hell are you defending a system you have never tried, nor done any sort of simulation to decide if it is accurate or not? That's why you are getting pushback.

If you just said "OK I know you guys are skeptical, but I want to try it out, what values should I be using for these things", you would have gotten the answers you seek instead of blowback. But instead your posts come off like you are stomping your foot on the ground, construct a sample deal or two that constitutes "proof" that Cowan is right and everyone else is wrong, that you know better than everyone else despite no experience and no statistical data. The reality is *you don't know*. So act like it, take the attitude of "I don't know if Cowan is right or not", come back after you have tried it for awhile and report your findings.

It's easy to have a dealing program spit out a bunch of deals that have Banzai recommend one thing while HCP (adjusted or not) recommend something else. You can load them into BBO and play them out, or use the BBO deal generator with the same constraints, given that you are skeptical of DD analysis. Do it and let us know your opinion, after you have actually played hundreds of these deals, carefully recording your success rates under the various combinations.

[Dumoti]

 helene_t, on 2018-July-16, 15:10, said:

Thanks for the citation

As I remember, it was 2 IMPs per thousand boards by replacing 4321 with "optimal" AKQJ counting which is something like 4.2-3.1-1.8-0.9. I don't think it had anything to do with Banzai points.

No. You said:

Replacing MW by the new method would only earn you 2 imps per 1000 boards or something like that.

So it is possible that something giving more credit to lower honours would be better for 1NT-3NT auctions.

Buying a copy of BridgeBrowser would constitute a small fraction of the costs of publishing a book.

Arguing such a case using made-up examples and case stories instead of data analysis is sooooo 18th century, if not even more outdated.
=============================================
That's exactly what you said.

[helene_t]

Yes but the new method I referred to was not Banzai. On the contrary it gives higher value to aces and kings. Sorry if this wasn't clear.

[mikeh]

Every now and then we get a thread about some different point count system and almost invariably the discussion gets very heated, with lots of debate about how accurate or useful the new idea (new in the sense of not being the 4321 method...some of the variants are actually quite old) is compared to the old Work method.

My take is always the same. Focusing on developing or using a new arithmetical approach to hand evaluation is a sucker's game, that will serve to delay or prevent improvement at the game.

No serious player of my acquaintance uses any conscious arithmetical process to precisely quantify the value of the hand. Such an exercise is silly.

Given that no purely arithmetical approach can work perfectly (for reasons that I will, to some degree, touch upon but which I hope would be mostly plain to all serious students of the game), it suffices to have a robust, simple starting point, and 4321, even if imperfect, has been shown to be pretty good after more than 70 years of widespread use.

But no good player picks up a hand and adds to, say, 15 and then thinks that this hand is worth as much, and no more, than the myriad of other 15 point hands he or she has held in the past.

One needs to consider what high cards one has: as others have noted, generally speaking the 4321 method undervalues Aces, and overvalues Jacks (it also, imo undervalues Kings and overvalues Queens, tho not as markedly).

We look also at where they are in relation to other high cards and our suit lengths.

We look at interior spots...10's are valuable but a 98 combination is more valuable, in a long suit, than a 65 combination and so on.

All of this gives us a preliminary but imprecise sense of the initial starting valuation, and I stress the 'initial' aspect.

Once one knows the 'point count' my suggestion is that one learns to think of it as a 'good' 15 count, or an 'average' 15 count or as a 'bad/soft' 15 count, and so on.

Who cares if Banzai or Zar or Robertson gives different values? We aren't bidding the hand based on arithmetic.

Then listen to the auction, and at every turn you have you should be thinking in terms of how that bidding is affecting your hand...not in terms of points, but in terms of how useful your cards are, whether on defence or offence, based on the information you are getting.

You open 1D, partner responds 1H and RHO overcalls 2S. You hold AQx in spades. Originally this was worth about 6 hcp. Now it is worth at least 7...firstly the King is almost certainly on your right and secondly LHO will lead spades even if he holds the King.

Conversely, you open 1D and LHO overcalls 2S...your AQx has gone down in value, but it hasn't gone down to 4 points, because not only is it possible that LHO lacks the King but you may later be able to endplay him, and, in any event, your knowledge of the likely location of the King gives you an edge in declarer play.

I don't care if you can invent an arithmetical system that allows a calculation of how this sort of information alters the numerical value of a hand. Any such method will involve some assessment of the reliability of inferences (how likely is that LHO would overcall 2S on a Jack high suit?), and even without that would require far too much thinking and calculating at the table.

It is far easier to learn to value cards and to estimate degree of fit with partner, likely location of important cards based on the bidding, and choose your level of aggression or conservatism in the auction based on that dynamic process. Yes, it takes time. Yes, using a numeric valuation system can speed up the learning process.....but never forget that the system is just a tool, and never let it become a crutch.

Good players learn how to bid good slams on 22 hcp or to stay out of bad slams on 32 points, and so on....not by using arithmetic but by listening to the auction and understanding the likely trick taking potential of their cards based on the auction, and not an artificial number generated by some arcane method.

Counting points, no matter what method one uses, should never be more than an approximation of the value of one's cards...it is a starting point and a tool. Efforts to make points the be all and end all of valuation have never worked and will, imo, never work except, perhaps, if one has a powerful enough computer and, so far lacking, powerful and well-written software.

I confess that I do almost always 'count points' in terms of adding up my face card values at the beginning of the auction. This is because we, in our methods, use hcp to define certain actions. So I need to use this metric if I am considering one of those defined actions. I never use adjustments such as adding points for length or shortness, and other than a quantitative notrump auction, I never consciously estimate our combined points, based on the notion, for example, that one needs, say 25 point for a major suit game or 33 points for a slam and so on.

Now, I have played quite a lot and have had the good fortune to have played with some superb players. I understand the attraction of exploring a numerical metric....it appears to offer a shortcut to good bidding but there is no shortcut, and no easy or free lunch. Pursuing unusual metrics serves only to detract from and interfere with learning how to bid.

A closing point: children learn to ride bicycles by having training wheels on the bicycle. Nobody is going to make a fundamental improvement in how children learn to ride bicycles by designing a different training wheel. A truly useful gadget might slice a week or so off their learning, but once they know how to ride, they don't need the training wheels at all. The analogy isn't perfect, because in bridge the basic point count remains a useful tool...but it is not anywhere near as useful as beginners were taught it was.

[Dumoti]

 Stephen Tu, on 2018-July-17, 00:45, said:

By whom? Certainly not by you.

Uhh… yeah. By me. As I pointed out, the original Banzai/Cowan points were developed using an exhaustive analysis of all the combinations involved in hands containing 4-3-3-3 and 4-4-3-2 shapes. Even adding in a shape like 5-3-3-2 involves three new possibilities, which were not covered in the original analysis: 5-3, 5-4, and 5-5 possibilities. No one is arguing that A-x-x-x-x opposite K-x-x-x-x is about the same as Q-J-x-x-x opposite K-x-x-x-x.

Now when Andrews to run a simulation that permits pretty much any hand shape but rules out stiff kings(!), he may have supplied some data. But for you to pretend that said data refute anything having to do with Banzai points is pretty shameless on your part.

Quote

1. 21-24 ish. There are a decent number of outliers. Average Banzai between 22 and 23, median 22.

Fair enough. I figure someone with 15 points has 37.5% of the face cards in the deck, so redone in a 5-4-3-2 method would show 20.625 on the low end (plus the chance of having a ten) vs. 42.5% for a 17 count, so 23.38 on the high end (plus the chance of having a ten).

Quote

2./3. 14-15, 16+ GF if you go by the Klinger/Jackson book recommendation of 37+ Banzai for game. I don't think anyone here has really studied what the threshold is supposed to be, but Klinger is likely about right. If you think it's more, why don't you go out and try it, report back?

But why should I go for that number? By figuring in a 5-4-3-2-1 + points for a 5-card suit, Klinger has shown that he never bothered to understand the initial system nor the reasoning behind it. 5-card suits are not to be considered in such a system because they are out of scope of the original analysis.

An average hand contains 1 ace, 1 king, 1 queen, etc. A 4-3-3-3 shape should be a king better than average to merit an opening bid. Sure, 5-3-3-2 hands might be shaded to 12 because you figure that you have some distribution regardless of whether you count short suit points or long suit points. Even if you subscribe to the rule of 20 or whatever your preference, 4-3-3-3 still needs 13+ to qualify — a king better than average.

But in Banzai Points, a hand that is a king better than average would boast 19 BZPs. So, 19x2 = 38. That's the way I see it, at least.

Quote

If you want to say 38 or 39 then fine, I could come up with counterexamples also. And it would become even easier to find HCP games below this higher Banzai threshold that are actually decent percentage rather than lucky to make. But singular counterexamples are silly and prove nothing, I only really did it to appease you since you demanded a 4333 x 4333 counterexample. I just had the computer select random deals that met the constraints, I wasn't trying to construct hands intending to prove anything, because singular hands don't really prove anything.

Well, you'll forgive me if I find your counter-examples unconvincing. It's easy to construct a normal slam hand that goes down on a double-dummy cross-ruff defense (two voids). That doesn't mean that the contract was bad — just unlucky. The point is to construct two hands in which it is clear that the hand has little chance of making game.

Quote

People have been playing bridge many decades longer than you have. I think if raising 1nt to 3nt on 10 balanced was not working out well enough on average they would have figured it out. In ancient days books wanted you to have 26 to bid game (maybe this is rubber bridge influence, where you can make game on the next hand and carryover the partial has value, so being more conservative is OK), players started bidding games on 25 because it works out, not all the time, but more often than not.

Oh, well, I guess there's no scope for innovation in bridge anymore. Maybe we should all stop playing the game. Still, perhaps I have had different experiences than you have. In my club, I have found people who have been playing the game for 5 decades and they are still horrible at the game!!

Quote

If you just said "OK I know you guys are skeptical, but I want to try it out, what values should I be using for these things", you would have gotten the answers you seek instead of blowback. But instead your posts come off like you are stomping your foot on the ground, construct a sample deal or two that constitutes "proof" that Cowan is right and everyone else is wrong, that you know better than everyone else despite no experience and no statistical data. The reality is *you don't know*. So act like it, take the attitude of "I don't know if Cowan is right or not", come back after you have tried it for awhile and report your findings.

Pot
Kettle
Black.

You know zero about the system -- less even than I do. Yet, you are here attacking me for saying that I want to try the system out. Heck, you can't even defend why 37 points is the magic number for games. As far as you know, no real research of any kind has been done on that matter. Nevertheless, you are here telling me that it's a waste of my time based on nothing more than a firm jut of your jaw as you issue your edicts like some Pharaoh of old? Give me a break.

You remind me of the guy at http://www.bridgebas...980#entry646980 who assures us that a 40-point BZP hand is worth a 6NT contract.

Or of the woman at http://www.bridgebas...699#entry647699 who understands the system so poorly as to think that it claims that AKxx is equal to QJT9 because they both net two tricks. Obviously, the first holding is 9 BZPs whereas the 2nd is 6 BZPs.

Whoever that lady is, she obviously didn't know what she was talking about. Perhaps now she has a more enlightened understanding of the system. Who knows?

[helene_t]

 Dumoti, on 2018-July-24, 14:57, said:

Or of the woman at http://www.bridgebas...699#entry647699 who understands the system so poorly as to think that it claims that AKxx is equal to QJT9 because they both net two tricks. Obviously, the first holding is 9 BZPs whereas the 2nd is 6 BZPs.

Sorry, I suppose I should have used an example with equal number of Banzai points.

But my point is that the Banzai point system was constructed by counting number of tricks taken for various honour combinations. This is the "exhaustive analysis of all the combinations" that you mention.

That is not the correct way addressing the problem. It may be true that KJx-xxx and Axx-xxx both generate on average one trick if played from the xxx hand and no information is available from the bidding and we have tempo to catch any trick(s) set up. But how do we quantify the likelihood of departure from those assumptions? Sometimes KJx is better because information from the bidding or the possibility of an endplay or a defensive mistake allows KJx to make more than one trick on average. Sometimes Axx is better because we can hold up the suit twice or because all we need is a quick trick or because all we need is a 100% single stopper. Or because we have to play from the KJx hand so we usually won't make any tricks.

It may be possible to augment the "exhaustive" analysis to take all these things into account. But it is much easier just to run a regression analysis on BridgeBrowser and see how many tricks are actually taken in real play.

I agree with your point that there may still be room for innovation. It wouldn't surprise me if in some areas, conventional expert wisdom is wrong. Bird/Anthias's analysis suggests that conventional wisdom w.r.t. opening leads is too aggressive and doesn't lead aces often enough. This may be true.

It is also possible that 4321 points is not the best simple evaluation method. But extraordinary claims require extraordinary evidence. Conventional wisdom says that 4321 undervalues aces. The statistical analysis confirms this. So a claim that 4321 overvalues aces is extraordinary and requires a strong argument. An argument that doesn't involve analysis of the play of whole hands, but only looks at suits in isolation, is not good enough.

[Stephen Tu]

 Dumoti, on 2018-July-24, 14:57, said:

Uhh… yeah. By me. As I pointed out, the original Banzai/Cowan points were developed using an exhaustive analysis of all the combinations involved in hands containing 4-3-3-3 and 4-4-3-2 shapes.

You can't debunk something when you have zero experience, zero data, zero logical arguments to back your conclusion. Multiple people have done many computer simulations and tabulated the general accuracy of estimating via work points vs. actual tricks taken and concluded it is more accurate than using Banzai points vs. tricks. You have basically taken the position "I declare Cowan is right", and will ignore all data or studies by any other people, having no experience, no data of your own, and no data from Cowan on whole bridge deals, only suit combination analysis. What does that prove? I might as well do something like say "Kyrie Irving claims the Earth is flat", I assume he is right and ignore all other evidence and arguments to the contrary, then claim I have therefore debunked the idea that the Earth is not flat.

Cowan analyzed suits in isolation, not having tools to look at whole hands. It has been explained to you why this is flawed. You have provided neither any reasoning nor any data why analyzing suit combos in isolation gives better conclusions than analyzing entire hands. Suit combos in isolation is inferior, because of tempo issues. Bridge hands are constrained to have 13 tricks. If the opponents can take 5 tricks before you, you can no longer get 9, even if looking at the suit combos in isolation not having to make discards on the opp's winners you'd eventually add up to 9 tricks.


You keep on complaining about Andrews restricting stiff Kings initially, when trying to get establish a "pure" value. You ignore that he specifically notes the controversy, and also provides data *without* that restriction, and shows the conclusions are not significantly different.


Come back when you have actually played hundreds of hands using Banzai, instead of pulling a conclusion out of your gut/rear end that Cowan sounds convincing to me, therefore everyone else is a bozo so I will ignore them.

Quote

But why should I go for that number? By figuring in a 5-4-3-2-1 + points for a 5-card suit, Klinger has shown that he never bothered to understand the initial system nor the reasoning behind it. 5-card suits are not to be considered in such a system because they are out of scope of the original analysis.

You can hypothesize whatever number for game you want. If you want to wait for 38 or 39 or 40 Banzai's to bid game, go ahead. I haven't done the computer studies so I won't claim 37 is ideal. But it looks approximately right to me from the sample deals I generated for the normal amount of HCP for game, and I assume Klinger published 37 based on experience. I have high confidence that if you wait for near guaranteed 38 Banzai's to raise partner's 1nt to 3nt to game, that you will underbid far too many hands and will show a huge net loss at either matchpoints at IMPs. Come back with data!

Quote

Well, you'll forgive me if I find your counter-examples unconvincing.

Of course singular examples are unconvincing. It's about what works on average, over many thousands of hands. That's what everyone doing computer simulations is doing. It's impractical to print out 10000 bal vs bal hands for you here and show that x% of them with 25 hcp combined make 3nt.

Quote

Oh, well, I guess there's no scope for innovation in bridge anymore. Maybe we should all stop playing the game.

You are free to try out Banzai if you think it will work out better. But for Pete's sake go ahead and actually do it for at least a few hundred 1nt openers, where using Banzai leads you to do something different to what the field using Work does, landing in a different contract, so you can argue from actual experience, rather than just making your argument from gut feeling without having played a single hand this way!

Sure some players in your club have played for decades and still suck. But the general class of player here on the forums is much higher, and we have low tolerance for B.S. arguments with zero evidence & experience behind them. You are doing an argument purely from authority, anointing Cowan as an authority and dismissing all others as bums who don't know anything about bridge or statistics. This will convince no one here. Come back after having played Banzai with whatever threshold you imagine is correct, and carefully recording # of hands, total work pts, total Banzai's, tricks taken, MP score, IMP score, etc. Continuing to argue just from pulling numbers out of your ass is just a waste of time.

There's a ton of innovation in bridge in terms of system design. There are lots of artificial bidding systems around, and in natural systems tons of new twists being introduced all the time in terms of transfers, switches/swaps, relays, puppets, asking bids, multi-way bids, competitive bidding techniques etc. That's where good players are putting their time in to gaining edges over the opposition. Zero of the actual good players think that optimizing point count to the second decimal point is worth any effort, because we know about estimates and error bounds and how much this is potentially worth (e.g. helene's 2 imps over a thousand hands result). And less than zero, if that were possible, would think it a good use of time to try a point count that multiple computer analyses have shown to be actually *worse* than the normal Work count. Some people study point count for curiosity's sake and publish some data and are done with it. A few people become completely obsessed with point count schemes, apparently believing that if they finally work out THE perfect formula for evaluation that their bidding will become so accurate that they slay all in their path, and post pretty much nothing else about bridge. But I've only ever seen this last group of posting on bridge groups/forums (here, bridgewinners, rec.games.bridge), I never see them out in real life winning/high placing in tournaments.
If you think you are going to revolutionize hand evaluation by a new point count for balanced vs.balanced, get better results than the field because of this, and start winning tournaments left & right, based on a debunked study from 1987, good luck with that. We who have played decades longer than you, some at the highest levels, are warning you against that. But you are of course free to do as you wish. But please don't claim you have proved something or know something when you haven't even tried the method in practice!

Quote

Yet, you are here attacking me for saying that I want to try the system out.

I did not attack you for wanting to try the system out. In fact I said you would *not have been attacked* if you just said you wanted to try the system out. You are being attacked for casually dismissing other studies and results as bogus and claiming Cowan is right, with no data or experience of your own, just your say so.

Quote

Heck, you can't even defend why 37 points is the magic number for games. As far as you know, no real research of any kind has been done on that matter. Nevertheless, you are here telling me that it's a waste of my time based on nothing more than a firm jut of your jaw as you issue your edicts like some Pharaoh of old? Give me a break.

I'm saying 37 is probably about right, because Klinger settled on that for his book and hopefully did so based on experience. Also we have data that shows 25 hcp work count is a good target for games, and 15 vs. 10 work points, 15 opposite an average 10, is going to be about 22 opposite 15 for Banzai, so that's 37. If you want to use 38, go ahead. You are making the claim for Banzai, not us, so use whatever you think will work best, report back. But please f*ing play some actual boards first before you post again about this.

[helene_t]

 Stephen Tu, on 2018-July-24, 18:15, said:

But please f*ing play some actual boards first before you post again about this.

I think this is a bit harsh. I haven't played any boards, f*ing or otherwise, using any formal point count methods other than 4321 with various adjustments, and somehow I feel entitled to voice my opinion anyway.

Playing a few thousand boards using 4321 and compare to a few thousands played using some other formula won't settle the issue anyway. Too many confounders and too small sample size.

What might be helpful would be to hack WBridge5 to use Banzai and let Banzai WB5 play a 10,000 board team match against 4321 WB5. I am not volunteering for setting up that experiment, though.

[apollo1201]

 helene_t, on 2018-July-24, 18:39, said:

What might be helpful would be to hack WBridge5 to use Banzai and let Banzai WB5 play a 10,000 board team match against 4321 WB5. I am not volunteering for setting up that experiment, though.

Probably the funniest and most temping post of this overall slightly boring mine is bigger than yours thread. No surprise it comes from a woman then🤣