BBO Discussion Forums: Strong nt with 1nt-2nt as invitational -- what's your minor-seeking structure / what do 2S and 3C mean? - BBO Discussion Forums

 rmnka447, on 2017-May-10, 15:08, said:

No.

 JLilly, on 2017-May-10, 14:39, said:

It's not something I've noticed. All contracts have a variability in the number of tricks taken depending on unknown factors, and the purpose of bidding is to make your assessments more accurately. If not, you bid without meaning. Knowledge of whether partner is 15 or 17 hcp puts you in a position to be in the right contract more often than you would be without that knowledge. I think that is incontrovertible.

The main argument is whether - with a limited gamut of potential bidding sequences - you find a better and more profitable use for the bid that could be invitational.

If you do have such a use, consider increasing the accuracy of pass or blast by restricting the range of the 1NT open.

 rmnka447, on 2017-May-10, 15:08, said:

 Vampyr, on 2017-May-10, 16:01, said:

To expand a bit, yes, you can play 2♠ as "bailout in a minor" (with whatever strong options get thrown in. A lot of people do. I don't like it, but I play Keri, so what do I know?)
However, it is NOT A TRANSFER, and calling it a transfer is prima facie misinformation, at least if you do it to the opponents. It is rare that the opponents will be damaged by this misinformation, but it can happen (lots of clubs and no bid, or majors and make the wrong cuebid, or even we bid and play at the 3 level and get the club or the diamond suit wrong because we "know" where the clubs are). And the TD will get called, and the TD will adjust.

If it's a transfer, it shows a suit (potentially AND, but not OR). If it's a relay, that's different; but if it's a relay showing a limited number of hands, it is better to describe those hands than "partner will bid, and then I will show what I have". So, in this case, "to play 3 of a minor, or slam try with clubs" (or whatever it is).

To answer the OP, I play 2♠ is "2NT" (plus a bunch of rare hands), 2NT is clubs, and we have really weird ways to show diamonds (as well as the hands that the superaccept would normally take care of). Playing Stayman in this structure, usually I play 3♣ is diamonds, and we just bail-or-blast the "if you have honour-doubleton, the suit runs" invitational hands.

 fromageGB, on 2017-May-12, 03:42, said:

Sure, all contracts have variability even with the most precise bidding systems. My question is whether NT contracts are generally higher-variance than suit contracts, ceteris paribus. In NT contracts, for instance, it seems that the intermediate holdings in a longish suit often matter. In a suit contract, you know you can limit your losses in short suits. In a suit contract, a stopper in opps' suit that's poorly positioned is unlikely to cost you more than one trick; in NT it may cost you a few.

 fromageGB, on 2017-May-12, 03:42, said:

I fully agree. And I would add that the wider the opener's range is, the more profit comes from the increased accuracy.

But don't forget the downside. The downside is: When you invite and find the opener with 15 or a bad 16 points and opener declines, you have raised the level by one trick, and some of these games just aren't made. That's when the invitations backfire. It's not an automatic win. It's a trade-off depending on the opening range.

M1cha, Agreed. My main point, though, is that NT contracts at the game-level have greater variance than do suit contracts at the 3- or 4-level. At IMPs, overtricks are worth little and games are worth a lot. Five hands at 1NT and five at 3NT hugely outscores ten hands at 2NT. So there's little value in a 2NT contract.

 Zelandakh, on 2017-May-04, 05:44, said:

i send this to my reg p ty

 JLilly, on 2017-May-18, 02:03, said:

But this is not a fair comparison is it? If we are using an invite then we would expect our bidding to be right more often, not less. Say we have 10 hands suitable for an invite, 2 making 7 tricks, 4 with 8 tricks and 4 making 9 tricks. With pass or blast we might decide on a policy of not missing any games and blast on the 8 making more than 7 tricks, giving 2x +90, 2x -50, 2x -100, 2x +400 2x +600. The invite loses on the 2 7 trick hands but sorts out most of the 8 versus 9s. Let's give one mistake - bidding a vulnerable game when 8 is the limit - giving 1x -50, 2x -100, 3x +120, 2x +400, 2x +600. Totals: blast +1880, invite +2110. Of course these are all made-up numbers but your concern about missing games when inviting is misplaced. Ending in 8 making 7 is much more of an issue but it happens infrequently enough to offset the gains made from being in 2NT making 8 rather than 3NT down one. A policy of "invite rarely, accept often" can help here. As far as I know practically every study into this has concluded that 2NT invites show a small plus when the 1NT range is 3 hcp. That might not make it the most effective use of the call but I think I/A pairs should think long and hard about alternatives before removing invite sequences like this.

Ah, okay. Thanks for the examples. I think on a forum somewhere (here, or maybe Bridge Winners), someone referred to a study that recommended pass-or-blast for 1NT ranges of 3.2 HCP or less. Where are these studies? Thanks, JL

 JLilly, on 2017-May-09, 03:59, said:

That's the style I like. Easy to remember.

Once an opponent reopened on air and went for 1100.

 JLilly, on 2017-May-18, 02:03, said:

I agree. Except that I think it's my point.

 Zelandakh, on 2017-May-18, 06:33, said:

If you come across something, I would be really interested in seeing what they did.

 JLilly, on 2017-May-20, 02:14, said:

That was by someone called "yunlin" here on BBO, the link is in post #12 but he didn't give the details. We would have to write to him and ask (if he's still active here).

Meanwhile I did my own small study partly because I was interested in the effect of pass-or-blast, partly because some very good players in my environment just do the opposite, they use both 2♠ and 2NT to invite to 3NT with different strength. I'm going to report here. (Not sure if today.)

 jogs, on 2017-May-20, 07:29, said:

Nice . This point is not in my study yet.

I'm playing pass-or-blast with two of my partners. Some friends now are going the opposite way using both 2♠ and 2NT for a 2-stage invite which seems very strange to me but they are good players. So I wanted to know and made a small calculation.

Assumptions:
- 3NT with 25 HCP has a success rate of 50 %.
- Every 3 HCP up or down make up 1 trick. (40 HCPs make 13 tricks ...)
- Every 1 HCP increases/decreases the success rate by 10 %. (It's a value I have read somewhere, and I understand it as 60 % for 3NT with 26 HCP, 40 % with 24 HCP etc. JLilly, if it's wider at notrumps, this would be good for pass-or-blast.

From this it follows that we make the following number of tricks with 25 HCPs at the probability shown:
10 tricks (20 %)
9 tricks (30 %)
8 tricks (30 %)
7 tricks (20 %)
or with 24 HCPs 10(10), 9(30), 8(30), 7(30);
with 23 HCPs 9(30), 8(30), 7(30), 6(10).
I'm neglecting the chances of the number of tricks being far off in order not to make it too complicated, but I believe it's a sufficiently good approximation.

If I play 3NT with 25 HCP I get the following scores (with probabilities), non-vulnerable:
+430 (20), +400 (30), -50 (30), -100 (20).
If I play 2NT with the same hands, I get
+180 (20), +150 (30), +120 (30), -50 (20).
If I play 1NT with the same hands, I get
+180 (20), +150 (30), +120 (30), +90 (20).
Similar for the other HCPs.

I'm inviting with all 8- or 9-point hands, I'm not considering hands with 7- or 10+ points. (Actually it doesn't make a difference if points are HCPs or else.) I'm combining them with 15, 16, or 17 points from the most frequent 1NT range, that's 6 combinations. (I also checked a 14-17 1NT, 8 combinations.)
I fill up a table: all scores for all point combinations with boards played on the levels given by one method, then another method. I calculate the score differences between the methods, value by value, and turn the differences to IMPs. Add up the IMPs weighed by the percentages. Simple math on an A4 sheet.

Here's the result, first for the 14-17 1NT:
pass-or-blast vs. the 2NT invitation comes out +2.9 non-vulnerable, +3.3 vulnerable in 8 boards, that's +0.36 IMPs/board non-vulnerable, +0.41 IMPs/board vulnerable.
I was quite surprised to see that pass-or-plast still has an advantage at a 1NT range of 4.

For a 15-17 NT, let me start by comparing the 2NT invitation against the double invitation. It will tell you something about the errors of the approach. My first result was +0.3/+1.3 for 2NT (6 boards) but this is too positive. I had assumed opener would always accept with 16 points. That makes 4 instances of 3NT and only 2 instances of 2NT bid with this method while it's 3 and 3 with the other two methods. This gives this method an advantage because, as we know, at IMPs we should bid game lightly particularly if vulnerable. In order to correct for this, I changed the method such that opener accepts or rejects the invitation with 16 points randomly 50/50. Now 2NT turns out negative with -0.4/-0.65 (in 6 boards) but this value is slightly too negative because of the randomness. An actual opener would rather decide based on "good 16" and "bad 16" points and get a better success rate. This is difficult to quantify. I belive the result would still be negative, so the above value can serve as a negative limit. Anyway, if we end up at -0.3 or -0.45, it's a gain of 0.05 or 0.075 IMPs per board advantage for the double-invitation approach. It's very close to nothing.

For pass-or-plast with 15-17, let me compare it to the double-invitation because it does not contain the uncertainty of the preceding paragraph. This one ends up +2.4/+2.8, or +0.40 IMPs/boards non-vulnerable and +0.47 IMPs/board vulnerable. For pass-or-blast against the 2NT invitation we can use a guideline of 0.5 IMPs/board. Those +1100s not included . And it does not consume a bid, it frees one.

Your criticism is appreciated.

 m1cha, on 2017-May-21, 18:37, said:

Richard Pavlicek has made a detailed analysis here. My take from this is that it is only with balanced hands and 24 hcp distributed relatively evenly between the two hands, IMPS scoring and vulnerable, that if you find yourself in 2NT you are better to blast 3NT.

Working on better hand evaluation methods would probably yield better dividends.

Thanks a lot. That is great data. Though I admit I had to read the introduction several times to understand what the tables mean, particularly the light colors and the overbids.

It shows that my assumptions about the tricks were not that bad. With 25 HCP I had put the probabilities for making 10(+)/9/8/7(-) tricks at 20/30/30/20, actually it is 19.53/34.59/30.33/15.55 for 9 vs. 16 HCP. It also shows an interesting mistake I made. I had assumed that the probability of making 3NT is 50 %. But I actually knew it is > 50 % while for 24 HCP it is < 50 HCP, so I should have assumed 55 % and calculated with 25/30/30/15 even better to the true values.

 Trick13, on 2017-May-23, 23:26, said:

Well that has nothing to do with pass-or-blast. I found that a very interesting detail in itself. It means that if you hold 15 HCP and get invited with 9 HCP, you have a total of 24 HCP and your optimal contract is 1NT BUT now that you were raised to 2NT you are already on the bad side and you should now raise again to 3NT because, although 3NT is worse than 1NT, it is still better than 2NT.

You can see from this data that the invitation itself costs ~ 1 IMP/board which must be recovered in order to score positive. I'm still looking at the 24 HCP table in the balanced section focusing on vulnerable boards. The overbid column shows how much you lose if you blast to 3NT rather than play 1NT. As long as this loss is 0.81 or less, you are supposed to raise 2NT to 3NT and suffer the loss in the overbid column. Only when the overbid loss is 1.10 or more, you're better off passing 2NT and swallow the invitation loss.

Another detail can be read from the table immediately. If vulnerable opposite a 15 - 17 1NT, responder with 9 HCP should always blast to 3NT, never invite! Because opposite 9 HCP opener should always accept an invitation: With 16 or 17 HCP, 3NT is the optimal contract, and with 15 HCP 3NT is an overbid but still better than 2NT. So if you ever invite (vulnerable), it should be with 8 HCP only.

I redid my part of my former calculation to see how much the result changes when I use the probabilities from Richard Pavlicek's study. I used the 14 - 17 1NT opener since there was no discussion about when to accept an invitation and when not. The former values for pass-or-plast vs. inviting with 8/9 was +0.36 non-vulnerable, +0.41 vulnerable. The improved values are +0.23 non-vulnerable, +0.22 vulnerable, all in IMPs/board.

Now that we know that inviting with 9 points is not a good idea, I re-did the calculation with the most positive scenario for invitations I could think of. Responder only invites with 8 HCP, opener accepts with 16 or 17 HCP, rejects only with 15 HCP. The result now is +0.4997 non-vulnerable and -0.0111 vulnerable. So finally we end up with a small arithmetic advantage for the invitation: ~ 1/100 of an IMP/board only for the very special case that resopnder has a more or less balanced distribution with exactly 8 HCP, no 4-card major, opposite exactly 15 HCP with opener, vulnerable. You really want to reserve a bid for that? And we haven't even taken into account the disadvantages such as giving opponents a hint about the lead or telling them opener's exact HCPs.

 Trick13, on 2017-May-23, 23:26, said:

Oh yes, I definitely agree, but this is not mutually exclusive.

Comparing invite to blast with 9 hcp opposite 15-17 at NV, you will only see a difference opposite 15 from opener (you are always in game opposite 16-17 since opener accepts the invite). In this case for blasting:

-3 7.97%
-2 19.49%
-1 36.28%
= 27.16%
+1 7.87%
+2 1.13%
+3 0.09%

Lose 2 (both contracts fail): 27.46%
Lose 5 (2NT makes exactly): 36.28%
Win 6 (3NT makes, or with overtricks): 36.26%

Expected IMPs: -0.188

So it seems definitely better to invite here.

At vulnerable you should blast with 9, but inviting will show an advantage over blasting with 8 (at NV it is probably best to pass with 8).

 awm, on 2017-May-25, 19:26, said:

Yes, thanks. Sorry, I don't know why I missed that. I had the same figures but perhaps I was confused by the light pink filling in the table which said 3NT is better than 2NT but that, of course, was for the vulnerable case only.

 awm, on 2017-May-25, 19:26, said:

At vulnerable you should blast with 9,

Yes. Same calculation as above, but lose 3, lose 6, win 10 vulnerable, ends up in
Expected IMPs +0.6244.
Well actually expected IMPs/board is 1/3 of this in both cases because you invite opposite 15, 16 or 17 but profit (or lose) only with 15. Thus in IMPs/board (9 HCP): -0.0617 nv, +0.2081 vul.

 awm, on 2017-May-25, 19:26, said:

but inviting will show an advantage over blasting with 8 (at NV it is probably best to pass with 8).

Expedted IMPs for 8-15: +1.4546 nv, +1.8961 vul,
expected IMPs for 8-16: +0.9837 nv, +0.5090 vul,
expected IMPs for 8-17: -0.9392 nv, -2.4386 vul.
Average expected IMPs/board with 8 HCPs: +0.4997 nv, -0.0112 vul.

(If you accept the invitation with 17 points only, the second line becomes
expected IMPs for 8-16: +0.5548 nv, +0.8322 vul, with
average expected IMPs/board with 8 HCPs: +0.3567 nv, +0.0808 vul
which means that inviting with nv hands come out a little better than before but any advantage for inviting with vulnerable hands disappears.)

Let's sum it up:
9 HCP vulnerable: bid 3NT;
9 HCP non-vulnerable: invite (accept with 16 or 17);
8 HCP vulnerable: rather invite (accept with 16 or 17), but the profit is next to zero; (EDIT: with good 8 HCP profit is OK, see next post)
8 HCP non-vulnerable: pass 1NT.

A final remark. We have assumed that responder points are integer numbers which is the case with HCP, and it's use by Richard Pavlicek's table. In 'real' bridge you will have 'good' 8 HCP or 'bad' 9 HCP which we may want to count as 8.5 points. If we'd simulate this we might find that the range for inviting may be wider or narrower than 1 point, probably narrower: If we simulate 8.5 points by averaging the above 8-point and 9-point results, the profit is clearly on the side of pass/blast at any vulnerability. (Edit as to the last half-sentence: wrong method → wrong conclusion, sorry.)

Okay, I really wanted to know, so I did the following. I looked at Richard Pavlicek's tables, constant opener HCPs and constant tricks, responder HCPs varying from 7 to 10. I made a regression analysis fitting these 4 values with a parabola which gave me non-integer values of responder HCPs. The agreement with the table values was quite good, most probabilities are reproduced within 0.3 % (absolute). Former expected IMPs for 9 HCP and 8 HCP were reproduced at 9.04 HCPs and 7.97 HCPs, so it seems I got a precision better than 0.05 HCPs. The aim was to determine the range where invitations are useful, or where 1NT should be passed or raised to 3NT. Invitations are accepted with 16 or 17 HCPs, rejected with 15 HCPs (these are still integer). Here's the result.

Vulnerable invite from 7.97 to 8.75 HCPs,
non-vulnerable invite from 8.32 to 9.16 HCPs.
The range for an invitation is ~ 0.8 HCPs.

Note (1): The fractional HCPs are derived from the table, the treatment does not tell us how to calculate them from long suits, sequences, middle cards or whatever.

Note (2): The calculation assumes that everything else is equal. This is not strictly true because the invitation process gives information to the opponents and may help them to find a good lead or deduce declarer's HCPs precisely.

(Finally, if anyone is interested if it makes sense to have an invitation which is accepted with 17 HCP only, the answer is no. Although a positive range exists for such an invitation, it goes
from 8.15 to 8.21 HCPs vulnerable and
from 8.51 to 8.63 HCPs non-vulnerable,
with a range of 0.1 HCPs and overlapping the range of the 16/17 invitation.)

 m1cha, on 2017-May-27, 18:00, said:

In practical terms: Vul invite with 8-8.5, bid game with 9 (and pass with 7.5); NV invite with 8.5-9; bid game with 9.5 (and pass with 8).