negative freebids

[cherdano]

Who cares what was in their minds?

 

Who cares if they were on the bottom end of one range or the top end of another?

 

There is more to making a result than the two players and what they thought each other's bid means. These results are what happened under the stated conditions. I've already built a bidding system from this and many other results like them.

 

At any moment in time, a player takes an action in a similar situation based on his hand and a set of circumstances. Playing 2NT contracts is a heavy loser. I don't need to explain WHY in order to believe that it is. However armed with my belief I will attempt to play to better scores by avoiding it except when there is no alternative.

 

Correct me if I am wrong - but your argument appears to be that any treatment, if played correctly is capable of the best score?

 

Stephen

Stephen, your logic is fundamentally flawed. How many of the partnerships playing in MBC do you think have discussed what to expect from a WJS? (Assuming, they have agreed whether they are playing WJS or SJW at all....) My guess would be around 1%, maybe even less. Do you really think the results these partnerships achieve on average with a 7pt WJS are convincing to show that 4-8 hcp WJS with good discussion are better than 0-6 hcp WJS with good discussion?

 

Do you really still believe your methodology when it seems to indicate a treatment (0-8 NFBs) that no top player at all is using?

 

Han's point about credibility is exactly on the mark.

 

Arend

[whereagles]

There's some logic in trying transfer responses after overcalls. I myself devised a couple of transfer gadgets after 1x (1y), 1x (dbl) and even up to 1M (2M-2).

 

However, it is my experience that natural methods are perfectly fit to deal with most situations. The occasional gains of transfers do not compensate for the memory load.

[inquiry]

Who cares what was in their minds?

 

Who cares if they were on the bottom end of one range or the top end of another?

 

There is more to making a result than the two players and what they thought each other's bid means.  These results are what happened under the stated conditions. I've already built a bidding system from this and many other results like them.

 

At any moment in time, a player takes an action in a similar situation based on his hand and a set of circumstances. Playing 2NT contracts is a heavy loser. I don't need to explain WHY in order to believe that it is. However armed with my belief I will attempt to play to better scores by avoiding it except when there is no alternative. 

 

Correct me if I am wrong - but your argument appears to be that any treatment, if played correctly is capable of the best score?

 

Stephen

Stephen, your logic is fundamentally flawed. How many of the partnerships playing in MBC do you think have discussed what to expect from a WJS? (Assuming, they have agreed whether they are playing WJS or SJW at all....) My guess would be around 1%, maybe even less. Do you really think the results these partnerships achieve on average with a 7pt WJS are convincing to show that 4-8 hcp WJS with good discussion are better than 0-6 hcp WJS with good discussion?

 

Do you really still believe your methodology when it seems to indicate a treatment (0-8 NFBs) that no top player at all is using?

 

Han's point about credibility is exactly on the mark.

 

Arend

To be fair to stephen, he is not using main bridge room hands here, but rather some very old okb data because there is a lot more hands in that database. I think the data is from 1997 to 2001 time range.

 

I think there is a reason that higher hcp on 1x-p-2y is worse. Opener has 11+, lets say average of 14 (guess) and responder has some number. If he has "too much" his preempt prevents his side from finding the best spot when the hand BELONGs to his side. I think a study might be to hold the suit to some minimum quality and lenght and check the results with 4,5,6,7,8,9 pts including bids like pass, simple bids and jump to different levels.

 

this does address a lot of issues (which is the best winning strategy, for example, weak jump shits, strong jump shift, fit jumps, interemediate jump shifts), but it will demonstrate trends on what happens with minimum quality suit and minimum points based on what you bid. And this is easy with bridgebrowser.

[Codo]

.... ( for example, weak jump shits)

I am quite interessted in this new tool, may you explain me your new concept please?

[hrothgar]

Who cares what was in their minds?

 

Who cares if they were on the bottom end of one range or the top end of another?

The short answer to your question is "Anyone with a halfway decent understanding of statistics"

 

There is a big difference between statistical analysis and data dredging. Unfortunately you don't seem to want to understand the distinction.

 

As an analogy: If you applied the same methodology to studying a Precision style 1C opening I'd by 90% sure that your analysis would show that no one should ever play a strong club. After all, any time that you open a strong club your expected score is significantly worse than any other bid in the system. Equally significant, the strong club opening often isn't as accurate as the "natural" bidding styles that most players use. Needless to say, this analysis would be badly flawed. The poorly defined 1C opening is a systemic cost of the light / limited openings bids that comprise the rest of the system. You can't analysis one piece without the other. Unfortunately, your making precisely this same type of mistake.

 

Please understand: I'm not knocking BridgeBrowser. Its a great product. Used correctly, you can shed light on a lot of interesting questions. But your not doing so here. Moreover your blind insistence that you are undertaking valid analysis doesn't do you or your product any good.

 

From my perspective, I think that you should starting by analyzing some very basic questions. For example, you might want to consider looking at board results are seeing whether there is a relationship between the expected score that a given player achieves on boards and the variance of their scores.

[hotShot]

I know in general the advantages of transfers and how to bid them B) But I do not know the specific requirements in this special situation (after intervention). It is easy to see (even for me) that it must be good with a weak 6° suit, but how does the bidding proceed when responder is intermediate/strong with 4-6 cards?

Well the idea is, if partner did not open GF, our auction end with this bid.

 

This whole discussion is meaningless, without the context you play them in.

If you don't have limited openings or no GF opening in your system, responder does not have sufficiant information do finalize the auction.

 

So without a given system the question 0-6 or 5-8 makes no sence at all.

Now e.g. playing precision openers hand is limited to 15 HCP. If responder holds 0-6 HCP they won't miss game very often with a combined 11-21 HCP.

 

The picture is different when your bid has 5-8 HCP. With a combined 16-23HCP

and a little distribution game might be in reach. Here opener can bid on with maximum and good distribution.

 

If we call the first NFB, the others should be called constructive NFB's.

[sfbp]

I think there is a reason that higher hcp on 1x-p-2y is worse. Opener has 11+, lets say average of 14 (guess) and responder has some number. If he has "too much" his preempt prevents his side from finding the best spot when the hand BELONGs to his side. I think a study might be to hold the suit to some minimum quality and length and check the results with 4,5,6,7,8,9 pts including bids like pass, simple bids and jump to different levels.

 

I think it's a bit simpler than that, Ben.

 

Each player is an actor, and he has a decision. I can hear the chuckles about "Captain Theory" coming on! All the other uncertainties (did partner remember, do the opponents get shut out or do they have countermeasures, was it a good field, was someone cheating 17 tables away, did the pub across the hall just close creating extra noise and thereby breaking declarer's concentration, was declarer DEAF?) can be reduced to a statistical measurement of "what happens if?" All of the above influences affect the result. The only question is whether they do so in a systemic way. To use the Baseball/cricket metaphor, how hard should I hit this pitch? (that is really my only choice).

 

For example, getting the agreement wrong will tend to cancel out in the long term, as some will make mistakes in the opposite direction to others (on average overbidders cancel out underbidders too).

 

Presuming that the majority of people bid according to agreements, testing the point count that this bid works on should reflect whether or not this is a good agreement, including the effect it has on opponents and also its effect in/on the field. If I let StephTu rubbish this argument on the basis that "someone might have thought", next one we will have is "The best data obviously come from world championships".

 

Well, I have news for you. They don't. People in world championships make mistakes, perform badly under pressure, etc etc. And where is the virtue of a 6-table movement like the Cavendish where 5 tables have an identical result, perfectly played by world class players against world class players, and on the sixth there is an odd result because Zia psyched? We play systems, adopt treatments, and take actions based on their effectiveness in a multitude of situations. So in fact their overall effectiveness is measurable. Perhaps you need to play different systems in world championships - but this is a second order effect. First you need to play well, and reach your own par.

 

Any standard action will tend asymptotically towards 0.0 IMP/ 50% mp. Think about that for a while. Bridge is about getting the edge. If everyone played weak jump shifts exactly the same way, and everyone defended the same way against them, the data would guarantee to converge on "no profit". In fact simple (1-level) opening bids do just that.

 

It may well be that the best data doesn't even come from OKbridge, or from the tourneys here as opposed to the Main Room, no matter what we think of the general behaviour and ethics of play in each venue. "The idiots and the cheaters cancel each other out" is one way of looking at this.

 

Anyhow back to your contention about looking at different point counts one by one. Yes, this is exactly what I did. Y'all can do it too (as Ben frequently has). If the effect is real it should be reproduced even in the more random environment of the MBC. I've seen this many times.

 

We can come up with all sorts of theories about why some treatments work - but really all those theories reduce to mnemonics for what to do in a given situation. Sorry if I just insulted all the "intelligent design" that went into the creation of all those fancy (or not-so-fancy) bidding systems. It's a frequency thing - why waste available bids on infrequent strong hands when you can have one bid for strong hands and a multitude of different bids for the more frequent weaker ones?

 

Most of us learn to play bridge without ever questioning the point of a treatment like "change-of-suit forcing". People do what works, and that collective wisdom is assembled into "knowledge" and passed down. This is no different from what BRBR does. It just does it quicker.

 

Stephen

[hrothgar]

Any standard action will tend asymptotically towards 0.0 IMP/ 50% mp

Have you ever tested whether bridge scores are normally distributed?

[cherdano]

For example, getting the agreement wrong will tend to cancel out in the long term, as some will make mistakes in the opposite direction to others (on average overbidders cancel out underbidders too).

There is no reason at all that this will cancel out. Maybe many will expect a very weak hand when partner makes a WJS, but when they make them themselves, they won't be able to resist the temptation to make a WJS with 8 hcp? Maybe these partnerships have never discussed how to make a game try opposite a WJS? Maybe average bridge players just aren't very good at evaluating game potential opposite a WJS (since it's a question of shape and controls, not of hcp)? Maybe the actual standard for WJS is actually 0-6 hcp, and thus all those 8 hcp WJS are actual errors and thus to blame for the result?

 

About the "quality of the data": You have frequently made the point that the level of the players doesn't matter, as long as you have enough data. Go watch random tables in the MBC, see what goes wrong. If you still think the average of these results can tell a world class partnership what the best agreements are, I have run out of words.

 

Arend

[hrothgar]

For example, getting the agreement wrong will tend to cancel out in the long term, as some will make mistakes in the opposite direction to others (on average overbidders cancel out underbidders too).

You are confusing two VERY different concepts here.

 

If I make a mistake and mis-apply a convention or a treatment the result is (effectively) unbounded. As my description becomes (progressively) less and less accurate, the impact on our score will (probably) increase.

 

Compare this to the case where I apply a convention/treatment correctly: Here, the impact on our score is bounded because the definition of the bid is also bounded.

[inquiry]

Let me try to see if I can share with you what BRIDGEBROWSER can do on questions like this. Using 1997 to Late Dec 2001 data from Okbridge, I took a simplistic approach.

 

1) Dealer is opener is in first seat

2) Opener has 11 to 20 hcp and opens 1♣, 1♦ or 1♥

3) Second hand does pass (other than pass, 2nd hand is unlimited)

4) Responder has 6 spades to at least the queen and not 6-5, 6-6, or 7-6 distribution

5) Responder hcp range varied from 4 to 8 hcp.

 

A couple trends show up. First, pass is a clear loser, even with 4 hcp, when holding six spades. Second, bidding 1♠ is more or less "average" except whne holding exactly 4 hcp, where bidding 1♠ (instead of the other larger choice which was pass) was a clear winner. Third, when limited to a six card suit, at least in 1997 to 2001, bidding 3♠ and 4♠ did not occur enough to be taken into account (too few bids give too high variability). And Pass with six spades and 8 hcp was also too rare to be used. Here is two table of results.

 

Bid     4hcp     5hcp    6hcp    7hcp    8hcp

Pass  43.01    40.62    40.32    38.49    Not A

1S      51.36    50.81    50.6    50.72    50.31

2S      50.92    52.53    51.68    46.77    50.92

 

Bid    4hcp    5hcp    6hcp    7hcp    8hcp

Pass   -2.77    -0.20    -2.34    -3.31    Not A

1S      0.24    0.05    -0.03    0.01    -0.04

2S      0.59    -0.46    -0.33    1.36    -0.52

 

The data also shows some anamolies, suggesting that all is not as it might first appear. For instance, often a bid is good on average for MP, but bad on average at imps, or vise versa. And while stephen suggestign preempting with less than 6hcp was good, This was not the case at imps with 5 and 6 hcp, but was for 7 at imps, the date is EXACTLY reversed at matchpoints.

 

Note, for this analysis, I stopped at the first responder bid. I did no subsequent follow up on the meaning of the 1♠, 2♠ or pass (that is, I did not just evaluate when opener passed this compared to bidding on (also easily done). This is why 1♠ response is always "average" in these tables=== most people bid 1♠ so the good and bad boards come from subsequent bids. The weak jump shifts introduced a wild card into the hand creating more swings. The only clear loser was PASS with six spades.

 

Now stephens analysis went further into the auctions, based upon what opener rebid. For the parameters I set, not one opener passed the 2♠ jump shift when the 2♠ bidder held 7 hcp (that one occurred 66 times). Three nt and 4♠ picked up a lot of imps for the 2♠ bidder's side as final contracts, but where both large net losers at matchpoints (this is frequency gain versus loss compared to net gain, typical of difference between mp and imps).

[sfbp]

Have you ever tested whether bridge scores are normally distributed?

 

Clearly they are not. But the averages of bridge matchpoint scores are so in the majority of cases. Not sure exactly how to do this, please give me some help! The Cavendish example, might have a Poisson distribution, one of the known shapes for small numbers of events. I actually wish BBO had more tables than 16, from a statistical viewpoint. 52, the OKB standard since about 1997 is probably better in this regard, as more tables tend to spread out the results.

 

Theres was a lot of discussion about ratings being non-normalised on okbridge. You could see the obvious effect on the graph. When I recalculated ratings using more sensible assumptions, they appeared to be normalised. It's obvious how they differed from the chart produced by okb management, and how much like a Bell curve they were, that seeing is believing. Maybe some plotting of matchpoint scores would be enough - although I'd be dredging my schoolday memories to remember whether it was Student, Xi-squared, or some other one of these you use to answer your question.

 

I believe abilities are normalised - this is the basis of all educational testing. So bridge rating at least, is a measure of ability. Maybe not THE measure. So matchpoint averages should be.

 

Stephen

[inquiry]

For fun, I change the hcp requirments to 20+ and 17-19 with the same 6 card spade suit and compared the 1x-1♠ to the 1x-2♠ (strong jump shifts).

 

The results were, with 20+ hcp, the average imp score was a respectable 0.89, while the average MP score was dead on average (49.82). This suggest that slam bidding was improved by the strong jumpshift and looking at contracts after the jump shift supports this hypothesis. Note, after the 20+ jumpshift, slam was bid 88 times out of the 135 matches to this specific auction with this specific requirements (dealer opened, etc).

 

Reducting the strong jumpshift to 17-19 hcp, however gave competely different results. Now the 2♠ jumpshift averaged a minus -1.97 imps and only 47.85 matchpoints. There were 1644 hands on which these "strong jumpshifts" could be made (only 863 did use jumpshift) so deciding exactly where they went wrong (too many five level contracts was one problem) will take more study.

 

This doesn't mean jumpshifts are bad things, just that you need to work on them if you are going to use them effectively.

[david_c]

Reducting the strong jumpshift to 17-19 hcp, however gave competely different results. Now the 2♠ jumpshift averaged a minus -1.97 imps and only 47.85 matchpoints.

I wonder how many of these "strong jump-shifts" were passed out. :) Do the results become positive if you factor those out?

[inquiry]

Reducting the strong jumpshift to 17-19 hcp, however gave competely different results. Now the 2♠ jumpshift averaged a minus -1.97 imps and only 47.85 matchpoints.

I wonder how many of these "strong jump-shifts" were passed out. :) Do the results become positive if you factor those out?

Responders 2♠ with 17-19 after a first seat 1C to 1H openeing bid holding 6S (and not other five card+ suit) occurred 863 times, no OPENER passed this jump to 2♠ at his next opportunity.

 

I think the OKBridge has less pickup partners and when they do, they play a more formal system than what occurs in BBO. We have almost as many members logged on at any given time as OKBridge has total members (well at our peaks).

 

For what it is worht, slam or grand slam was bid 578 times when holding 17-19 hcp (67%)... note, the slam where responder held 20+ hcp was bid "only" 65% of the time. The slams did worse when responder held 17-19 than when he held 20+.

[helene_t]

I have no experience with these specific data (bridgebrowser) but as a statistician, my immediate thought is that it just can't be possible to answer such complex questions as the ones adressed here, using these kind of data. Unless one has a very detailed understanding of the stocastic proceses going on.

 

If someone gets a good result using some particular gadget (say NFB), it may be because it's a good gadget or it may be because he's a good player, or because he confused the opps by misexplaining what the call meant, or because he was playing with his trusted partner who understands excatly what that call meant.

 

Now you say that all those residual effects cancel each otrher out, but they might not. It might be that certain conventions (say Blackwood) are more often used by less experiences players. This will result in a negative bias on the estimate of the utility of Blackwood. It might be that non-standard treatments are more often played by regular partnerships, which would cause a positive bias for the utility of non-standard treatments.

 

Soms treatments (I think negative freebids is one) are often not based on agreement but rather mistakes.

 

Then there's the problem with negative inference. When someone doubles an overcall, the meaning of that double is influenced by whether the partnership plays nfb or not, but you can't allways tell that from looking at the cards, even if you assume that the double was correct.

 

Even if it were possible to make an unbiased estimate of the utility of nfb, what would it mean? Obviously, some agreements are more suitable for regular partnerships, some are more suitable for beginners, some are more suitable when playing against beginners etc. What you're aiming at is some over-all utility for nfb. This won't tell you whether you (or I, or Meckwell, or Gozilla) should be playing nfb.

[sfbp]

Even if it were possible to make an unbiased estimate of the utility of nfb, what would it mean? Obviously, some agreements are more suitable for regular partnerships, some are more suitable for beginners, some are more suitable when playing against beginners etc. What you're aiming at is some over-all utility for nfb. This won't tell you whether you (or I, or Meckwell, or Gozilla) should be playing nfb.

I don't argue (from the data) whether or not it is right to use NFB, and I don't argue whether it is right to use WJS.

 

What I say is, if you are going to use WJS, you better be aware that it will bring you the maximum profit at the lowest end of the range, say 0-6. Most experts when asked will actually insist on this from their practical experience, so the result i found is in accord with expert expectations.

 

Similarly, if you are going to use NFB, I observe, based on this data, that they work best at the lower end, well below any hand that might conceivably construed as constructive. My experience is that whenever partner raises my NFB, whether it was a good hand or a bad one, we get into trouble. Not just because we get too high, but also because opponents now can do the arithmetic and deduce their own fit from knowing that we have one.

 

So i *think* that we can only be sure that these sequences are NFB if someone passed them. I prefer to think of the 9 point variety as "constructive free bids" and I agree totally that it is impossible to tell whether those are forcing free bids below the end of their agreed range, or non-forcing freebids above what I consider to be the upper end of the range.

 

Ben and others said that some people raise an NFB on a good hand and therefore it's unfair to filter out the NFB's by taking only the sequences where opener passed. I can't tell you about the hands where that happened, sigh. But I can tell you that if you always pass a NFB, for whatever reason you decide to, that works best if the NFB has no more than 8 points (maybe less).

 

Is that useful? Maybe not. It's what I do, and it works quite well from the final scores at the table. I also suspect that putting a lower limit on NFB isn't such a good idea, because (say) 7 points guaranteed can be widely scattered and not useful to declarer. This is why most play that the bid promises the values are concentrated in the bid suit. This leads naturally to the bext issue.

 

How much do we overload double with the remaining hands? Several people have pointed out to me (not here) that there is another bid it is perfectly permissible to overload, and that is PASS. Personally I don't find doubling to be a problem - on most strong hands and on most hands with perfect shape. The remainder.... ah well, there'll always be some hands that fall in the crack. At least if you play your crack in a different place from the field you will tend to be less at the mercy of what your particular opponents decided to do on that board. So as long as NFB is not the majority treatment, I suspect it gives us a possible edge.

 

Stephen

[inquiry]

I have no experience with these specific data (bridgebrowser) but as a statistician, my immediate thought is that it just can't be possible to answer such complex questions as the ones adressed here, using these kind of data. ....

 

 

Even if it were possible to make an unbiased estimate of the utility of nfb, what would it mean?

Let me start off by saying that I agree with almost all of your comments. In fact, I (and others) have already pointed out the flaws in the appoach for negative free bids... see my post in this very thread, for example, at http://forums.bridgebase.com/index.php?sho...ndpost&p=133179

 

And like you said, EVEN IF the data could be 100% assigned to negative free bids, and even if the results as stated by stephen where true (negative free bids only work with 6 or fewer hcp), which btw, I don't believe, for the sake of argument lets say it was true), figuring out what that would mean is still not useful. All it would mean is when you have a negative free bid hand, you would get better than average (on average). But what happens when you ahve a hand too good for a negative free bid. In stephens example, it would be 7 hcp and a five card suit. You have to double then bid a new suit to show this hand. Presumably that is forcing, or you have to complicate it by having to cue-bid the second time then show your suit on the third round if you want to accomendate as few as 7 hcp in the double and new suit.

 

So to analyze the "effectiveness" of a negative free bid structure, you would need to evaluate these other strategies as well. Needless to say, this is a hopeless task, at least in my opinion.

 

But what BridgeBrowser can do is tell you something about a number of hand types. For instance, should you pass if your partner opens and you hold six spades and 4 hcp. The data clearly shows the answer is no. (you knew that didin't you?).

 

That 2NT and 5NT contracts, on average are bad (5NT is worse than 2NT BTW).

 

You can probe other examples. You open 1m holding 11-14 hcp balanced, partner bids 1♥ and you have 4♠ and 2 ♥, what worked better, a 1♠ bid or a 1NT bid? Or if partner opens 1NT (15-17 range, balanced) and you hold balanced 9 (4333 no four card major), are you better off pass, bidding 3NT directly, or inviting somehow?

 

These type of questions, where you place known limits on both hands (balanced 11-14, 4♠, 2♥ opposite a 1♥ response, or balanced 9 opposite 15-17 balanced) tend to give the most reliable data. But even here, maybe the 1NT bidder had 15 but his range was 12-14(he downgraded), or maybe he had 16, but his range was 16-18. But you could see how a balanced 9 fairs in say, 3NT opposite a balanced 14, a balanced 15, a balanced 16, etc and then figure out what you do (regardless of the bidding sequence). You can also investigate what a sixth heart is worth on certain auctions, for examples. These types of analysis are easily done.

 

Let me address, however, your point about good players versus bad players, where you said

If someone gets a good result using some particular gadget (say NFB), it may be because it's a good gadget or it may be because he's a good player, or because he confused the opps by misexplaining what the call meant, or because he was playing with his trusted partner who understands excatly what that call meant.

 

Stephen's position on this is that when you look at over 24 million hands, these things all "Average" out. I am not certain this is necesarily so, but I do know why he feels comfortable saying this. It is based upon a study of CARD PLAY by Peter Cheung using the same database. He compared tricks won on the hands (he used the 23 million dataset and another 7 million hand dataset), and he found...

 

The overall total number of tricks taken by the declarer is 9.21 (9.22 for imp and 9.20 for mp).  The double dummy analysis of the same deal produce 9.11 (9.12 for imp and 9.11 for mp).  So actual play by OKBridge player takes 0.1 tricks more then the double dummy analysis result. This is based on 30 million plays on Okbridge

 

I think we can all agree we are none so good that we can take the maximum number of douuble dummy tricks on a hand (left to our own devices). However, it is also true that while declarer can underplay a hand (compared to double dummy), so can the defense, So as a result, sometimes the defense screws up, sometimes the offense, sometimes the double dummy result is obtained. When you add the plus, with the minuses (and the normals) it all averages out over millions of hands. He draws the same "averaging" effect for bidding over a large number of hands. In fact, this maynot be as true as you take smaller and smaller subsets from the millions of hands, but that is the assumption.

 

I prefer to frame questions in the context of hand patterns and strategy rather than "negative free bids" or mozilla, etc. The simple reason being you can not separate into "systems" so easliy. Just becasue someone opens 1♣ with 17 hcp doesn't mean he is playing precision. If he opens 1♣ with 17 points, 1♣ and seven good spades, well, you can begin to draw a conclusion (maybe not precision, but a forcing club for sure).

[jdonn]

Trying to draw conclusions from data drawn this way is laughable. I don't even know what else can be said about it. It may be fun, and it may point out interesting things, but it simply doesn't lead to any reliable conclusions.

 

Ben's methodology is getting a lot warmer than Stephen's. But it's still an impossible task.

 

If you could interview every pair you collected data on about what their exact methods were (or whether they had any at all) and then only use the one's playing a particular system and being on the same page, then we would be in the ball park, though not there yet.

[sfbp]

All it would mean is when you have a negative free bid hand, you would get better than average (on average). But what happens when you ahve a hand too good for a negative free bid. In stephens example, it would be 7 hcp and a five card suit. You have to double then bid a new suit to show this hand. Presumably that is forcing, or you have to complicate it by having to cue-bid the second time then show your suit on the third round if you want to accomendate as few as 7 hcp in the double and new suit.

 

Ben, I'm not actually suggesting NFB's should be 6 or less. My current thinking is that given that 6 or less seem to show a profit (played non-constructively, vide supra) it's about right to divide the hard hands between a normal positive free bid (good 10+ or 11) and there in two, and lump the 7-8 in with the bottom fish - and the data seems to show that you make a small loss on those. Any hand with non-perfect shape that has 9 or more I tend to double and then backpedal whilst bidding. Sure, I get nailed in that side of the crack too. But it seems better than having to pass so many hands because they don't add up to one of the forcing free bids that are standard and endorsed by so many here. Face it, most of those guys are doubling on all 9-point hands anyway, so where's the big loss? Noone says that doubling and bidding guarantees a game force. I tried to deal with that in the prior post I made.

[joshs]

It actually doesn't matter what the response shows.

 

I checked.

 

If you have more than 6 points when you make a weak jump shift (over pass by opp), you get a negative score. Period.

 

Stephen

Thats very interesting. And doesn't have any bearing on the question what is the best range for the response. If your partner expectes a 3 count and you have an 8 count, you will get bad results. If your partner expects a weak jump shift and you have a strong jump shift you will have very bad results. Etc.

 

I am still guessing at your methodology, since you haven't stated it. My hypothosis is that you are looking at:

All hands in which someone opened 1m (regardless of the meaning of the bid) and all 2M responses (regardless of the meaning of that bid or the agreed range, or the lack of agreement of the range). Please correct me if I am wrong about your methodology.

[sfbp]

Trying to draw conclusions from data drawn this way is laughable

 

Laugh all you want. But there's one element here that ought to make sense to even you.

 

The data on WJS are (laughably) clear. It's also clear that bidding very bad hands with NFB works too, when partner passes.

 

The element you missed is one of inductive (or analogous) reasoning. It may be fallacious, but I'm suggesting you could do a lot worse than discard the died-in-the-wool thinking that says "free bids are forcing, therefore anything else is nonsense". Or "I had a bad experience with NFB, and I dont like them any more, the rest of the sheep are right after all". I'm trying to suggest exactly WHY that bad experience, and how, with this very limited and flawed data, it might have occurred.

 

It's backed up by what I have seen at the table. Frequently I got lousy results precisely because we didnt have the absolute agreement that NFB should be passed, and the result we WOULD have got from passing spent the rest of the week annoying me. We now have that agreement.

 

Stephen

[inquiry]

Trying to draw conclusions from data drawn this way is laughable. I don't even know what else can be said about it. It may be fun, and it may point out interesting things, but it simply doesn't lead to any reliable conclusions.

 

Ben's methodology is getting a lot warmer than Stephen's. But it's still an impossible task.

 

If you could interview every pair you collected data on about what their exact methods were (or whether they had any at all) and then only use the one's playing a particular system and being on the same page, then we would be in the ball park, though not there yet.

What I am saying is that making asking questions that are as system independent as possible, or at the very least, easily determined systemically, is the best approach. For instance, the opening in fourth chair with less than 15 pearson points question. There, I only EVALUATED pass compared to bid. It is true, that one could argue that any pair who opens a balanced 10 count in first seat might pass with a better hand in fourth seat and get higher results than the field who passes opposite partners who will not open such hands. But the shear volume of the data there makes that not useful.

 

Also, you could try out the browser on some very specific hands. Let's take what to do with 6 hcp and six card spade suit with at least spade queen showed earlier (after first seat opening bid). Here, I ignored possibilty of a 5-3 heart fit (for instance). If partner opened a heart and I have 3 hearts and 6 spades and 6 pts, is 2♠ better than 2♥? So you could subdivide those and see which strategy seems to work the best.. .i don't object to that.

 

But let me give you and exmaple. I like reverse flannery by responder, so that 1m-P-2H is 5+♠ and 4+♥ and a weak hand. I could easily set up a search strategy where opener bids 1m, next hand passes, and responder has 4-6 hcp, 5+ spades, 4+ hearts, and see how 2H response (for the few enlightened people who play this) fair compared to a 1S bid. Since the field bids 1♠, I can already tell you that bid will be roughly AVERAGE, and since few people bid 2♥, that bid will not be average and will have a high variability. But even if a lot of people bid it, and it showed good results (or baddish results), it will not tell you of the gains on other auctions for people playing this method simply because the there is no way to measure the effect of the removing these hands from auctions where you do not make this jump 2HE bid using bridgebrowser.

 

So use bridgebrowser to ask questions that can get reasonble answers. Should there be a minimum suit quality to make a weak jump overcall? Is it better to open a xxxxx heart suit at matchpoints or a three card minor to AKx or stronger? How effective is a weak raise to 4M with only 4 card support after 1M--DBL-? at different vul's. Does it matter if the major is spades or hearts?

 

These can be examined. You can even get anticipated number of tricks if you hold certain hand types in suits an nt (irregardless of the level of the contract) using bridgebrowser.

 

But as you say, to use it to develop system (based upon statistics) is crazy. But you can call up hands that fit certain requirements and see how your bidding system would handle them (can you spot the weakness for slam for instance).. I used bridgebrowser to find 1000's of misiry hands, and used those hands to modify the system to try to maximize the responses. It is very useful at those things. It is also useful for self-examination. For instance, how well do your preempts work? Using bridgebrowser, this is easy to determine.

 

For instance, if you find over time your average preempt earns you minus 2+ imps per board, you might want to take a look at your preempting style. Maybe it is too aggressive or not aggressive enough. A review of the hands might tell you (ture you will have to hunt for the hands you could have but diddn't preempt on).

[sfbp]

BTW i don't know what it means, but the weak jump shifts to 2M are WAY more successful on BBO. Over 1.25 imp with an SEM of only 0.11

 

Of course there are negligible matchpoint results, and I havent looked at tourneys... yet.

 

I would guess WJS is more common percentage wise than on okb, but I invite you to check. Why should I do all the work?

 

Stephen

[Chamaco]

About negative double: when we (Ben, Han, Matt, me) were toying with various ideas about transfers in competition, we learned to be very reluctant to give up negative double (especially when there is exactly one unbid major). You can't afford not to find 4-4 major fits, but bidding (or transferring into) the suit directly on 4-cards is very dangerous when it consumes space. Negative double is really a great bid!

I agree completely.

 

I hate to give up neg X and use X as a transfer.

 

That's why I play that a new suit at the 1-level is NATURAL, whereas transfers are on from 1NT through the suit below opener's suit.