[Gerben42]

Quote

I think Raptor was part of WJ2000, I know for sure that it is not anymore in WJ2005.

Right. In the poll shown at the end ofhere you can see that this was point 38. The natural 1NT overcall won by 58% to 42% in the votes. I guess that's why the Polish NT was taken out of WJ2005. Such a result is not surprising since natural meanings will always score well in such polls unless the gain of the non-natural meaning is compelling.

I guess this has always been a close vote and if people really would not like it I'd expect it to be less close by means of "are you still playing that?"

Anyway personally I like it that way and even if I might lose some points on this as Fred suggests I am sure I lose much more in other areas so I will work on that instead.

[hrothgar]

Fred advanced the argument that the failure of any top pairs to play a Raptor style 1NT opening suggests that this method may be flawed:

I was just skimming through a number of my World Championship books. I noted something rather interesting.

The French Bermuda Bowl teams often employ very similar methods (5 card majors with, with a strong artificial and forcing 2♣ and 2♦ opening). While this strong 2♦ opening is in widespread use in France, its rarely seen used by top pairs outside this country. The methods haven't spread far despite some great success on the part of the French team...

In a similar fashion, the Polish teams almost always play some version of Polish Club (once again, with great success). With the exception of the occasional emigree pair, I almost never saw these methods used by any top North American pair. (Molson - Baran is the only exception that somes to mind).

I can point to a number of other examples where individual methods are extremely popular within a given a geographic region, but have failed to diffuse very far...

While I agree with your assertion that bridge players are constantly looking for an edge, I also believe that this is counter-balanced by a herd instinct. The top pairs within any given region normally prefer to go with the field, relying on superior declarer play and defense to produce a good card. Its far from clear that weaker pairs should indulge them by playing "their" game.

Kit Woolsey has written some very good books on bridge. From my perspective, one of the most interesting suggestions that he made involved optimal strategy playing against a superior team. Woolsey noted that the optimal strategy for winning the match might involve taking (slightly) inferior lines. For example, holding an eight card fit missing the queen, you might prefer to play for the drop rather than a finesse. The odds of playing for the drop aren't that much worse. And, if you get lucky, this might be enough to compensate for all those overtricks that the "good" team is picking up...

Curious what you think of Woolsey's line of reasoning (The link between this and Raptor is pretty obvious...)

[FrancesHinden]

It seems to me that Hrothgar and Fred are not saying dramatically different things.

Suppose (amazingly) that you are not as good as your opponents/not the best pair in the field.

Hrothgar: if you want to win _now_ you should play a high variance/anti-field system
Fred: if you want to win consistently long term you should improve your standard of bridge, and a playing around with a high variance/anti-field system is not the way to do that.

Both of those seem true to me.

I'm in the 'Fred' camp for one main reason:

You don't have to play anti-field methods to play a high variance or anti-field game. Just look at the way Meckwell play when 60+ imps down before the last segment without changing their system. You can mix up the results by going ultra-conservative or ultra-aggressive (depending on your natural tendencies) as well as psyching sometimes, or putting in the odd 'distortion' (canape opening bids, off-centre 1NT openings, stretched pre-empts with a 4-card major).

What's more, if you take that approach then you can also play a straight-down-the-middle sound game when you are the better pair or team without being 'handicapped' by strange methods, or without changing your methods.

One thing I found in Tenerife (EBL pairs) is that there ain't no such thing as "playing with the field" when the field is sufficiently diverse.

Playing to preserve a big lead and playing to make up a big deficit are two aspects of the game that aren't given a huge amount of attention. We've had team discussions on what to do at aggregate scoring when big leads and deficits are quite common; rather less so at imp scoring.

[Sigi_BC84]

hrothgar, on Feb 8 2006, 07:33 PM, said:

Fred advanced the argument that the failure of any top pairs to play a Raptor style 1NT opening suggests that this method may be flawed:

I've just read in a different topic that double dummy analysis is indeed very close to the real world performance (RL declarers take 0.1 tricks more on average than DD declarers).

So why not do a double dummy analysis of Raptor versus natural notrump? At first thought this looks pretty realistic to me, or does anybody see a reason why it might not work (not every convention can be DD analysed in a meaningful way).

--Sigi

[FrancesHinden]

The DD analysis referred to was of the play only, not the auction.

To do a DD analysis of Raptor, you would need to define your entire competitive bidding methods to the DD program, because you have to include all the auctions you compete due to the negative inferences from playing/not playing raptor. Then you have to teach the DD program to bid properly against raptor.

No-one has yet taught a machine to bid that well.

[Sigi_BC84]

FrancesHinden, on Feb 8 2006, 08:03 PM, said:

You don't have to play anti-field methods to play a high variance or anti-field game.[...]

What's more, if you take that approach then you can also play a straight-down-the-middle sound game when you are the better pair or team without being 'handicapped' by strange methods, or without changing your methods.

One thing I found in Tenerife (EBL pairs) is that there ain't no such thing as "playing with the field" when the field is sufficiently diverse.

OK, so what is actually the case in your opinion? If there is no playing with the field because the field is diverse enough, then, by definition, there are no field-methods and hence no anti-field-methods as well.

Also you're talking about being "handicapped by strange methods" as if it can be assumed that the methods are as strange to the pair who is applying them as to anybody else. I don't think this would be the case at all (probably the pair playing the methods will know them well enough, or no?).

If the methods are a handicap, that makes them bad methods by any means. Just being "strange" in the eyes of the majority doesn't impose a handicap in a general sense (except maybe for badly prepared opposition).

Following the discussion I get a stronger and stronger impression that this is more about beliefs, preferences and feelings than about actual performance of this-or-that method (this doesn't have to apply to Raptor, but see my posting about double dummy analysis above).

While top-experts won't be prone to use inferior methods, this does not mean that most unknown or rarely-used methods are inferior. Richard makes a strong point when he says that geographical location and local regulations play a big role here. So why not use tried-and-true methods if you can be reasonably sure that you will outplay most of the opposition after the contract has been declared? Those methods will be the ones predominant in your country/area, that's why american top pairs do not play Polish Club and almost all Poles do.

Now take a method like Keri, which many people here are raving about. So far I haven't met a pair in real life who are playing Keri. Same applies to Scanian NT and Heeman -- they all play Staymanic systems, which are proven to be less effective than the abovementioned methods. So just because a method is unknown and the field prefers something else doesn't mean it's necessarily better.

Something else is nagging me as well: the possibilities for bidding systems are endless, but only a tiny proportion of these is actually in use (most of them being natural even, now why should this be optimal after all?). This is not a constructive argument but hints strongly towards the fact that there must be countless methods out there which are better than the "Standard(s)" (and still playable). OK, some of these are outlawed, but that opens an entirely new can of worms.

This will conclude my ramblings, maybe somebody here has a comment on this or that...

--Sigi

[Sigi_BC84]

FrancesHinden, on Feb 8 2006, 08:28 PM, said:

The DD analysis referred to was of the play only, not the auction.

To do a DD analysis of Raptor, you would need to define your entire competitive bidding methods to the DD program, because you have to include all the auctions you compete due to the negative inferences from playing/not playing raptor.  Then you have to teach the DD program to bid properly against raptor.

No-one has yet taught a machine to bid that well.

Of course you can only DD analyze deals (with a given contract), and not auctions.

What I'm having in mind is to construct typical deals where Raptor and natural notrump overcalls would apply, and see which one loses or gains more and if it will average out in the end.

This will require some work in specifying the hands but it should be possible. There are tools which allow you to generate deals based on complex specifications (probably you know this but I'm mentioning it anyway just in case).

Alternately one could analyze hands from real world play, looking for cases where Raptor and natural notrump had been used and see which one loses out. I don't have access to BridgeBrowser but maybe somebody will bother to actually do it. Filtering out the Raptor hands shouldn't be too hard...

--Sigi

[FrancesHinden]

The point I was making on the DD analysis thing is that it's not enough to take hands where you would have a Raptor 1NT overcall or a natural 1NT overcall. You have to take pretty much _all_ competitive auctions, because of the inferences available from the lack of use. I've never played Raptor, nor against it, so I'm not certain what these are, but I bet there are loads of them.

Then the opening side will (might) have a different defence to a Raptor 1NT overcall than a natural 1NT overcall, and might end in a different contract. Particularly relevant to DD analysis, if the opening side declarer the hand they will have different information about the opposition layouts: a natural 1NT overcall gives a blueprint to the declarer play that DD analysis won't give Raptor the benefit of (and vice versa but in different final contracts and possibly to a different overall extent).

I agree it could be done, but I think it's a seriously large amount of effort.

[FrancesHinden]

Sigi_BC84, on Feb 8 2006, 08:43 PM, said:

FrancesHinden, on Feb 8 2006, 08:03 PM, said:

One thing I found in Tenerife (EBL pairs) is that there ain't no such thing as "playing with the field" when the field is sufficiently diverse.

OK, so what is actually the case in your opinion? If there is no playing with the field because the field is diverse enough, then, by definition, there are no field-methods and hence no anti-field-methods as well.

Yes, that was a completely random rambling side issue comment. Sorry. I was just staggered at the time how very, very few flat or flattish boards there were.

Quote

Also you're talking about being "handicapped by strange methods" as if it can be assumed that the methods are as strange to the pair who is applying them as to anybody else.  I don't think this would be the case at all (probably the pair playing the methods will know them well enough, or no?).

If the methods are a handicap, that makes them bad methods by any means.  Just being "strange" in the eyes of the majority doesn't impose a handicap in a general sense (except maybe for badly prepared opposition).

I wasn't talking about Raptor in particular. I was addressing Hrothgar's suggestion (which now I can't find, am I in the wrong thread?) that _even_ if a method is (marginally) inferior, the best way to win is to do something different from the herd and so unusual methods are worth playing.

[DrTodd13]

Gerben42, on Feb 7 2006, 08:07 AM, said:

Sorry I guess that was a bit undifferentiated, some things are of course silly (one can think of a 2NT opening showing 10 - 12 balanced for example - no one would even try that).

What I meant to say is that it might mean less than many think. Anyway, playing Raptor NT or natural 1NT overcalls won't make you or break you. Being able to make less mistakes than the opponents will.

On a side note, I think we will see more of these "extremely natural" Fantunes-type systems around in the near future.

P.S. I know that there are downsides to Raptor 1NT, especially when you hold the 15 - 18 balanced with stopper hand type. However it has much more upside than conventions that I would characterize as horrible.

I ran into a pair playing a 7-8 pt balanced 2N in a tournament. Of course, they were only doing it because people were unprepared to defend against it. If this tactic got popular then defenses would be created and this ridiculous bid would be punished.

[Myon]

Raptor does have big effects in competitive bidding. Firstly, you probably have to alert any 1any-2m overcall because that effectively rules out a 4M holding (unless overcaller is very strong), and secondly, you pass with the 15(16)-17bad balanced hand where you would normally have overcalled 1NT (with 17good-18 you'd double and rebid NT), and jump out of the dark later when Pd reopens (we had "weird" biddings like 1any-p-p-X-p-3NT).

Having said that, I know that this won't stop Sigi from implementing a dealer file for it

[han]

FrancesHinden, on Feb 8 2006, 02:59 PM, said:

The point I was making on the DD analysis thing is that it's not enough to take hands where you would have a Raptor 1NT overcall or a natural 1NT overcall. You have to take pretty much _all_ competitive auctions, because of the inferences available from the lack of use. I've never played Raptor, nor against it, so I'm not certain what these are, but I bet there are loads of them.

Then the opening side will (might) have a different defence to a Raptor 1NT overcall than a natural 1NT overcall, and might end in a different contract. Particularly relevant to DD analysis, if the opening side declarer the hand they will have different information about the opposition layouts: a natural 1NT overcall gives a blueprint to the declarer play that DD analysis won't give Raptor the benefit of (and vice versa but in different final contracts and possibly to a different overall extent).

I agree it could be done, but I think it's a seriously large amount of effort.

No, I think that you were right the first time, it can't be done.

Sure, you can run computer simulations that might say something interesting about the convention, but they will never be as decisive as a double dummy analysis of a play contract.

[Chamaco]

Myon, on Feb 8 2006, 08:12 PM, said:

Raptor does have big effects in competitive bidding. Firstly, you probably have to alert any 1any-2m overcall because that effectively rules out a 4M holding (unless overcaller is very strong), and secondly, you pass with the 15(16)-17bad balanced hand where you would normally have overcalled 1NT (with 17good-18 you'd double and rebid NT), and jump out of the dark later when Pd reopens (we had "weird" biddings like 1any-p-p-X-p-3NT)

This is a matter of style, I guess.
While I won't argue with the style you describe, I prefer to stick to the following:

15 balanced = frequent offshape t/o double (unless too much wasted), passing any minimum response

16-17= double then 1NT

18-20 = double then 2NT (regardless of whether it's a jump)


I won't claim it never backfires, but not a disaster either, especially at MP.

At IMPS Team, I confess that sometimes I miss the natural 1NT.

[fred]

FrancesHinden, on Feb 8 2006, 07:59 PM, said:

The point I was making on the DD analysis thing is that it's not enough to take hands where you would have a Raptor 1NT overcall or a natural 1NT overcall. You have to take pretty much _all_ competitive auctions, because of the inferences available from the lack of use. I've never played Raptor, nor against it, so I'm not certain what these are, but I bet there are loads of them.

Then the opening side will (might) have a different defence to a Raptor 1NT overcall than a natural 1NT overcall, and might end in a different contract. Particularly relevant to DD analysis, if the opening side declarer the hand they will have different information about the opposition layouts: a natural 1NT overcall gives a blueprint to the declarer play that DD analysis won't give Raptor the benefit of (and vice versa but in different final contracts and possibly to a different overall extent).

I agree it could be done, but I think it's a seriously large amount of effort.

Frances is right - the "raptor vs. 1NT" issue is FAR too complex to expect much help from present day software. In principle I believe that a program to answer questions like this could be created. However, I do not expect to see this happen unless either a team of super-smart people works on this for a several years or unless there is some kind of fundamental breakthrough in AI research.

If I was setting the line I would make the over-under number something like 20 years.

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

[Sigi_BC84]

Hannie, on Feb 8 2006, 09:12 PM, said:

Sure, you can run computer simulations that might say something interesting about the convention, but they will never be as decisive as a double dummy analysis of a play contract.

No they won't; apparently you took my remark regarding DD accuracy the wrong way. I'm not saying that you can easily (or at all) setup a simulation that will yield a quality hypothesis regarding the value of the Raptor convention.

I'm only saying that given you find a good simulator setup, DD accuracy will not be in your way when it comes to determining the actual trick value of a hand -- regarding real world play of that hand.

--Sigi

[Sigi_BC84]

fred, on Feb 8 2006, 09:52 PM, said:

Frances is right - the "raptor vs. 1NT" issue is FAR too complex to expect much help from present day software. In principle I

You and Frances are both correctly hinting at the complexity of this problem. Let me say though that I was not suggesting that you will find a "42 is the answer"-like result to the Raptor vs Natural question by using a deal generator and double dummy analysis.

I was thinking about solving some of the problems related to the subject, e.g. how dangerous is a natural NT overcall with regard to penalties, and how often do you find a playable partscore using Raptor that you would not have found otherwise (because of supposedly hidden major suit fits that come to light).

Maybe I'm being too naive about the problem and the complexities and inferences involved are too many so that answering questions like the above won't be of much value anyway. Hopefully I've brought my idea across at this point ;-).

--Sigi

[hrothgar]

Sigi_BC84, on Feb 8 2006, 10:50 PM, said:

What I'm having in mind is to construct typical deals where Raptor and natural notrump overcalls would apply, and see which one loses or gains more and if it will average out in the end.

For what is worth, I did graduate work in game theory...
I also think that i have a passing knowledge regarding different bidding systems...

While I find discussions like this one an interesting way to pass time, I'd strongly recommend against trying to do any "serious" work in this field. The topic is WAY too complex to reach anything resembling a definitive answer...

Francis has already alluded to three of the key problems in formal analysis of bidding systems:

The first is the relationship between the "bidding game" and "card play" game. All the different bids that you make (or don't make) have an important impact on declarer play and defense.

Equally significant, you can't analyze individual bids in isolation. Case in point, you can't hope to analyze Ben's Misery preempts without also consider the impact on the major suit opening style.

Finally, you need to consider the opponent's best responses. Unfortunately, many people who vreate a new toy often have significant blind spot's regarding different ways to exploit the methods.

To this list, I'd add one other issue. From my perspective, this is the real deal-breaker in that it increases the complexity of the problem by a couple orders of magnitude. As I've noted in the past, we have no reason to suspect that bidding systems exhibit transitivity. Assume that

2/1 Game Force is a "better" system than Acol
In a similar fashion, Precision is better than 2/1 Game Force
If transitivity holds true, then Precision is better than Acol...

My gut says that transitivity assumptions are not warranted, in which case we aren't looking for an optimal bidding system, but rather, some a population consisiting of some optimal ratio of bidding systems. Throw in the potential for cyclical equilibria and you have a modelling problem that I wouldn't touch with a 10 foot pole (unless, of course, I was being paid by the hour)

[pclayton]

Hannie, on Feb 8 2006, 12:12 PM, said:

FrancesHinden, on Feb 8 2006, 02:59 PM, said:

The point I was making on the DD analysis thing is that it's not enough to take hands where you would have a Raptor 1NT overcall or a natural 1NT overcall. You have to take pretty much _all_ competitive auctions, because of the inferences available from the lack of use. I've never played Raptor, nor against it, so I'm not certain what these are, but I bet there are loads of them.

Then the opening side will (might) have a different defence to a Raptor 1NT overcall than a natural 1NT overcall, and might end in a different contract. Particularly relevant to DD analysis, if the opening side declarer the hand they will have different information about the opposition layouts: a natural 1NT overcall gives a blueprint to the declarer play that DD analysis won't give Raptor the benefit of (and vice versa but in different final contracts and possibly to a different overall extent).

I agree it could be done, but I think it's a seriously large amount of effort.

No, I think that you were right the first time, it can't be done.

Sure, you can run computer simulations that might say something interesting about the convention, but they will never be as decisive as a double dummy analysis of a play contract.

When I first started playing Overcall Structure, I kept pretty detail logs on my MP results every time a systemic call would come up and tried to stay objective about whether or not the actual method created the result or just good (or poor) judgement.

With Power Doubles, its pretty easy to see where we pick up; when the strong hand is put on lead and the lead comes around to the responder of the power doubler; when we get to nail the opponents at the 1 level for +800, and the like. These results have NOTHING to do with judgement; they are borne out of the use of the convention.

With 1NTO, results are more difficult to judge. If we get murdered in a misfit, it must be in the context of "what happened to those that made a takeout x (or passed)"? Or, if we are able to jam their auction and more effectively judge the starin / level, it must be in relation to what a takeout doubler did as well.

Nevertheless, these results are at best anecdotal.

With Raptor, many times the 4-4 (supposedly) side fit usually gets buried, unless overcaller has substantially extra values. I think it might be easier than you think to run Bridgebrowser and pull out 1N overcalls that contain 4M (or other major) / 5+ minor (or other minor), and determine how effective the convention was. A side singleton / void would also be a telltale sign that a pair was using the convention.

Even though the IMP / MP result can't be directly correlated to the mere presence of the convention, it would still be useful to know what the aggregate result would be using these methods.

[fred]

pclayton, on Feb 8 2006, 11:02 PM, said:

Hannie, on Feb 8 2006, 12:12 PM, said:

FrancesHinden, on Feb 8 2006, 02:59 PM, said:

The point I was making on the DD analysis thing is that it's not enough to take hands where you would have a Raptor 1NT overcall or a natural 1NT overcall. You have to take pretty much _all_ competitive auctions, because of the inferences available from the lack of use. I've never played Raptor, nor against it, so I'm not certain what these are, but I bet there are loads of them.

Then the opening side will (might) have a different defence to a Raptor 1NT overcall than a natural 1NT overcall, and might end in a different contract. Particularly relevant to DD analysis, if the opening side declarer the hand they will have different information about the opposition layouts: a natural 1NT overcall gives a blueprint to the declarer play that DD analysis won't give Raptor the benefit of (and vice versa but in different final contracts and possibly to a different overall extent).

I agree it could be done, but I think it's a seriously large amount of effort.

No, I think that you were right the first time, it can't be done.

Sure, you can run computer simulations that might say something interesting about the convention, but they will never be as decisive as a double dummy analysis of a play contract.

When I first started playing Overcall Structure, I kept pretty detail logs on my MP results every time a systemic call would come up and tried to stay objective about whether or not the actual method created the result or just good (or poor) judgement.

With Power Doubles, its pretty easy to see where we pick up; when the strong hand is put on lead and the lead comes around to the responder of the power doubler; when we get to nail the opponents at the 1 level for +800, and the like. These results have NOTHING to do with judgement; they are borne out of the use of the convention.

With 1NTO, results are more difficult to judge. If we get murdered in a misfit, it must be in the context of "what happened to those that made a takeout x (or passed)"? Or, if we are able to jam their auction and more effectively judge the starin / level, it must be in relation to what a takeout doubler did as well.

Nevertheless, these results are at best anecdotal.

With Raptor, many times the 4-4 (supposedly) side fit usually gets buried, unless overcaller has substantially extra values. I think it might be easier than you think to run Bridgebrowser and pull out 1N overcalls that contain 4M (or other major) / 5+ minor (or other minor), and determine how effective the convention was. A side singleton / void would also be a telltale sign that a pair was using the convention.

Even though the IMP / MP result can't be directly correlated to the mere presence of the convention, it would still be useful to know what the aggregate result would be using these methods.

I think you are suggesting that you measure:

0) The expected IMP/MP gain or loss when a Raptor 1NT overcall is made

This will only be meaningful if you also measure:

1) The expected IMP/MP gain or loss when natural 1NT overcall is made
2) The expected IMP/MP gain or loss when a pair playing Raptor is dealt a hand suitable for a natural 1NT overcall
3) The expected IMP/MP gain or loss when a pair playing natural 1NT overcalls is dealt a hand for Raptor
4) The relative frequency of hands appropriate for Raptor versus hands appropriate for natural 1NT overcall

I suspect that 2) and 3) would be hard to extract from BridgeBrowser.

To really do this properly, it would likely be appropriate to consider all dealer/vul combinations separately (and perhaps conclude that, for example, Raptor is a winner only if you are not vulnerable).

In addition, you should make an effort to study the average skill levels of pairs playing Raptor versus those that play natural 1NT overcalls.

Even then, it is hard to say if this would really prove anything.

Impressed by your dedication to research! Great that you take notes and study the gains/losses of your conventions.

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com