[pescetom]

smerriman, on 2019-December-14, 20:43, said:

Assuming that:
- North will accept an invite with any 16+

Is that a legitimate assumption?
I would decline an invite with quite a few 16 HCP hands.

[rhm]

As I said people have different opinions when to superaccept.

Assume the 1NT opener has 4 or 5 spades.
The 1NT opener is still much more limited than responder.
When responder passes your superaccept you will often go down and this is much more frequent than responder having just enough to bid game over a superaccept and game makes, where responder would otherwise have passed at the 2-level. It is not that under those conditions game will always make or even be a good proposition.
It is true that quite often opponents can balance profitably if you do not superaccept and responder is weak, but do they know? Balancing is inherently quite risky
If they do balance you can still bid one more for the road.

Overall I think superaccepting makes only sense if you have a good 1NT opener and at least 4 card support.
You should be slightly more aggressive superaccepting hearts than spades and of course vulnerability for both sides come into consideration as does scoring.
Otherwise superaccepting is a losing proposition, particularly at matchpoints, though you will get your occasional triumph.

Let's make some simple assumptions for superaccept, which are easy to simulate
You superaccept with 5 spades and you superaccept with 4 spades with 16-17 HCP, except when you have 16 HCP and 4♠333

Again I simulated 1000 random deals with the South hand excluding superaccepts for North:

If North declares double dummy North makes at spades

10 tricks or more on 479 deals (47.9%)
9 tricks on 843 deals (84.3%)
8 tricks on 974 deals (97.4%)

If South is declarer

10 tricks or more on 451 deals (45.1%)
9 tricks on 821 deals (82.1%)
8 tricks on 963 deals (96.3%)

North makes double dummy at notrumps

9 or more tricks 259 deals (25.9%)
8 tricks on 497 deals (49.7%)
7 tricks on 788 deals (78.8%)

Compare this to my previous run.
The difference is marginal, given that any simulation run has some statistical variations.

Rainer Herrmann

[pescetom]

rhm, on 2019-December-15, 11:45, said:

As I said people have different opinions when to superaccept.

Assume the 1NT opener has 4 or 5 spades.
The 1NT opener is still much more limited than responder.
When responder passes your superaccept you will often go down and this is much more frequent than responder having just enough to bid game over a superaccept and game makes, where responder would otherwise have passed at the 2-level. It is not that under those conditions game will always make or even be a good proposition.
It is true that quite often opponents can balance profitably if you do not superaccept and responder is weak, but do they know? Balancing is inherently quite risky
If they do balance you can still bid one more for the road.

Overall I think superaccepting makes only sense if you have a good 1NT opener and at least 4 card support.
You should be slightly more aggressive superaccepting hearts than spades and of course vulnerability for both sides come into consideration as does scoring.
Otherwise superaccepting is a losing proposition, particularly at matchpoints, though you will get your occasional triumph.

Let's make some simple assumptions for superaccept, which are easy to simulate
You superaccept with 5 spades and you superaccept with 4 spades with 16-17 HCP, except when you have 16 HCP and 4♠333

Again I simulated 1000 random deals with the South hand excluding superaccepts for North:
...
Compare this to my previous run.
The difference is marginal, given that any simulation run has some statistical variations.

There are bound to be some hands that work out badly with superaccepts and this looks like one, as your simulations suggest.

I fully agree with you about looking at zones and being more aggressive with hearts than spades, but in general I have found superaccepts to be a winning proposition, including an automatic accept with 4+ cards and 15-16 HCP when denoted by a specific bid. When this did send us down it often turned out to be a decent MP score nevertheless.
Ultimately I think this boils down to Law of Total Tricks and the extent to which you trust it. Opponents with half the HCP in the pack will balance almost automatically over a Jacoby transfer at MP, so I see no clear reason not to anticipate them. However weak responder is, the superaccept should be protected by a 5-4 fit.

[smerriman]

rhm, on 2019-December-15, 11:45, said:

Compare this to my previous run.
The difference is marginal, given that any simulation run has some statistical variations.

Just to reiterate though, you're looking at the wrong numbers in terms of deciding whether to actually invite in this situation.

[rhm]

pescetom, on 2019-December-15, 15:11, said:

There are bound to be some hands that work out badly with superaccepts and this looks like one, as your simulations suggest.

I fully agree with you about looking at zones and being more aggressive with hearts than spades, but in general I have found superaccepts to be a winning proposition, including an automatic accept with 4+ cards and 15-16 HCP when denoted by a specific bid. When this did send us down it often turned out to be a decent MP score nevertheless.
Ultimately I think this boils down to Law of Total Tricks and the extent to which you trust it. Opponents with half the HCP in the pack will balance almost automatically over a Jacoby transfer at MP, so I see no clear reason not to anticipate them. However weak responder is, the superaccept should be protected by a 5-4 fit.

Of course everything can happen and will eventually happen in this game.
The Law of Total Tricks is fine when you or your opponents can judge total trump length round the table.
Bridge is a game of incomplete information and anyway I do not consider the LAW a substitute for judgement.
In this case opponents do not know your degree of fit and only the last opponent will get a chance to hear responder pass after the transfer.
I very much doubt that opponents with half the HCP in the pack will invariably balance over 2♠ at MP, particularly when vulnerable.
If they balance you can always bid 3♠ and sometimes they will balance and get doubled or go down when nothing makes, because neither side has a nine card fit.
Jumping to 3♠ when opponents are not in the bidding, when only you know your side has a 5-4 fit, does not look to me to be winning MP tactics.
You may sometimes get a decent MP score going down in 3♠, but I bet your MP score would have been better making or going down less in 2♠.

Rainer Herrmann

[rhm]

smerriman, on 2019-December-15, 16:06, said:

Just to reiterate though, you're looking at the wrong numbers in terms of deciding whether to actually invite in this situation.

I doubt your numbers.
My numbers say that there is almost a 50% a priory chance 4♠ will succeed and more than an 80% chance that 3♠ will succeed.
When there is so liitle to choose between passing 2♠ and bidding game matchpointwise, you can bet that inviting will handsomely win.
Assuming you invite, I guess the 1NT opener will be right most of the time he accepts and also when he declines. This is not a random guess for someone, who knows how to evaluate a bridge hand.

For example if holding ♠Kxx ♥Ax ♦xxx ♣AKQx I would accept, but holding ♠Kxx, ♥Kx ♦KQx ♣KQxx I would decline.

If responder simply bids game he would beat you on almost 48% of the hands (479 delas)
The middle of the road action of inviting will not beat you quite as often, because opener will sometimes decline and game would have made.
But far more often there will be far less wins for you and far more draws when opener declines and 3♠ will make.

For example, say we accept on 600 deals the invitation and we will make game on two thirds of them gives us 400 wins and 200 losses, on some of these deals going down more than one.
On the remaining 400 deals we decline and we will still make 9 tricks or more 75% of the time gives us 300 draws.
On the remaining 100 deals even 3♠ proved too high and we can make only 8 tricks or less, a win for you.

So you would win on 300 deals and my strategy wins on 400 deals.

In other words your numbers are wrong.

Rainer Herrmann

[pescetom]

rhm, on 2019-December-15, 17:11, said:

In this case opponents do not know your degree of fit

I can't see why you say this: we alert (announce, in our RA) the transfer and alert the superaccept, if they had any doubts about our agreements they can ask or just read the CC. So everyone at the table should know we are 5-4 by agreement. Granted there are some agreements that might not distinguish the occasional 5-5, but that would be more reason for opponents to compete anyway. If anything it's one or both opponents who might have inferences about their own fit and thus better vision of the Total Tricks situation.

[vsmague]

I was in front of Antony for that deal
In our system, I have 3 options :
- strategy 1 : transfer then pass, except if super accept (showing max hand with 4 spades) where I bid 4
- strategy 2 : transfer then invitational 3S. I assume there partner will decide to bid 4 with maximum hand or medium hand with nice fit
- strategy 3 : stayman (3 answers) then 2S on a 2D or 2H answer (and 4S on a 2S answer), showing 5S in an invitational hand. Maximum, partner can bid 3NT (I'll sign off at 4S), with minimal hand he stops at 2S, and in between partner relays to 2NT to ask for shortness and I bid 3D to show short clubs. Partner can then decide to bid 3 or 4S according to the hand fit (no lost points in Clubs).

I have just simulated the 3 strategies using double dummy on a sample of 2000 deals. Criteria are :
- Strategy S1 : opener plays 4S if 17H and 4S, else he plays 2S
- Strategy S2 : opener plays 4S if he has 4S, or a maximum hand, or a non balanced medium hand with 3 spades (or Hx), else he plays 3S
- Strategy S3 : responder plays 4S if opener is maximum, or has 4S, or has a medium hand with 3S and no K or Q in C - sometimes opener plays when he has 4S and no 4H. If medium hand with lost point in Clubs, responder plays 3S, else he plays 2S if opener is mini

I have compared the results for the 3 strategies and here are the findings
S1 versus S2 : S1>S2 34% - S1=S2 39% - S1<S2 27%
S1 versus S3 : S1>S3 18% - S1=S3 63% - S1<S3 19%
S2 versus S3 : S2>S3 14% - S2=S3 62% - S2<S3 24%

So seems on the sample that S3 slightly better than S1, both are above S2

Downside of S3 is that the hand is played by respondant in most cases, but the advantage is that declarer is more able to judge his hand given the known club shortness

[smerriman]

rhm, on 2019-December-16, 04:04, said:

I doubt your numbers.
..
For example, say we accept on 600 deals the invitation and we will make game on two thirds of them..
..
In other words your numbers are wrong.

I doubt your numbers as well.

Even if opener has a 17 count, game is making less than two thirds of the time, so I am extremely skeptical that you could achieve a 2/3 success rate over all of opener's hands.

I would expect a reasonable success rate opposite a "normal" invitation, but lower opposite this particular one.

[rhm]

smerriman, on 2019-December-16, 14:01, said:

I doubt your numbers as well.

Even if opener has a 17 count, game is making less than two thirds of the time, so I am extremely skeptical that you could achieve a 2/3 success rate over all of opener's hands.

I would expect a reasonable success rate opposite a "normal" invitation, but lower opposite this particular one.

You should read more carefully.
Game succeeds roughly on 50% of all deals, precisely on 479 deals of the simulation. .
When opener gets invited, opener will not always choose the right 479 deals, but he should get this right far more often right than wrong.
If opener accepts on 600 deals and gets this right two thirds of the time he will succeed in game on 400 deals, while if opener would choose perfectly he would succeed on 479 deals
That is 40% of all the deals.
Two third of the time is a rather conservative estimate if opener can judge a bridge hand.
Note, that if opener chooses the 600 deals completely randomly where he accepts, that is without looking at his hand, game would still succeed on nearly 300 deals and opener will certainly do much better by looking at his hand.

Rainer Herrmann

[smerriman]

rhm, on 2019-December-16, 16:45, said:

You should read more carefully.

I didn't misread anything; I'm not sure what you think I did.

As mentioned above, I am disagreeing with your estimate of opener being able to make the correct response to an invite on 2/3 of the deals opposite this particular hand.

You say it is conservative with no evidence of why. My example was that unless you are declining the invite on some maximums, you're already below 2/3 on those cases, so have ground to make up on the harder cases.

Even the hand that you said you would accept on - ♠Kxx ♥Ax ♦xxx ♣AKQx - requires getting the trump suit right, so while you would like to be in game on that hand, it's not contributing 100% to your success rate, only the 58% of making, dragging down your average further from the 2/3 mark.

My original numbers - while perhaps not the best approach to accepting - resulted in opener making the correct decision 57% of the time, but that wasn't sufficient to make inviting worthwhile. It seems vsmague's acceptance criteria were not sufficient either.

Are you able to provide acceptance criteria that actually work to support your hypothesis?

Perhaps you meant that 2/3 of the time, opener will be able to decline "bad" games and accept "good" games. Sure, that might be achievable, but you're not beating the passers on all of those occasions; the number of times you go down in a "good" game contributes heavily to the final results.

[rhm]

smerriman, on 2019-December-16, 17:17, said:

Are you able to provide acceptance criteria that actually work to support your hypothesis?

For you I ran the following simulation (1000 deals) when North accepts the invitation with the right hands:

South hand as given.

North criteria for accepting the invitation (always balanced):

1) 17 HCP, exactly 2 spades, at least 6 controls
2) 16-17 HCP, exactly 3 cards in spades, but not 3♠433 if 16 HCP, at least 5 controls
3) 15 HCP with exactly 4 spades or 16 HCP if 4♠333, as otherwise North would have already superaccepted.

These are at least some of the North hands, with which I think the invitation should be accepted.

Result of this simulation:

If North declares double dummy North makes at spades

10 tricks or more on 686 deals ---> 68.6%
9 tricks on 938 deals (93.8%)
8 tricks on 994 deals (99.4%)

You see?

Maybe you learn something how good hand evaluation improves your chances of making the proper choice. HCP alone will not do!

Rainer Herrmann

[hrothgar]

Might be easier if you both just posted your code

[vsmague]

HI Rainer
Very interesting - what about the hands that have been passed (so plays 3S), how many are only 8 tricks ?

[rhm]

hrothgar, on 2019-December-17, 08:05, said:

Might be easier if you both just posted your code

I do not write the code
I use Dealmaster PRO

Rainer Herrmann

[rhm]

vsmague, on 2019-December-17, 10:45, said:

HI Rainer
Very interesting - what about the hands that have been passed (so plays 3S), how many are only 8 tricks ?

I have made already 3 simulations.
The last one is not directly comparable, since I only generated deals with North holding a 1NT opening where he should accept.
But if you look at the second one (december,15th 12:45) you can deduce from 1000 deals where North does not superaccept
you have 843 deals where you can make 9 tricks or more.
Therefor on 157 (15.7%) 3♠ will fail (16 deals will already fail in 2♠)
So 141 (14.1%) deals will make exactly 8 tricks in spades

Rainer Herrmann

[Stephen Tu]

This kind of hand is one where I tend to be a bit suspect of double dummy sim generated data. I feel like this is one where declarer's omniscience in the trump suit will tend to outweigh the opps not blowing a trick on lead. There is quite a difference from being able to pick up the AJTxxx vs Kxx for zero losers ~95% in a sim vs only 58% in real life. Or going from 40% vs Kx to 74%.

[smerriman]

Thanks, Rainer. I stand corrected.

[MrAce]

The fallacy of all your simulations on this topic (and others as well) might be that your criteria is HCPs. I specifically mentioned not to super accept with "Aceless" hands (1 or no ace) . Basically the quality of hcps makes a huge difference. For example to me a hand with 4 card support and 2 aces +2 kings with 14 is more than enough to super accept while i may not do that with a hand full of qwacks even if it is 16 hcp. No need to mention the simulations favor the defenders more than it does declarers. Huge majority of players (unless you are playing only in very top level bridge) are much more skilled in declarer play as a person than they are in defense as a pair.. This is a known fact.

Another fallacy in the forums is all of us debate as if we are about to play the BB or top level MP event, and our debates are for training purposes to this match or pair event. Other than Mike (if there is anyone who plays only at these levels please forgive me) your success rate as declarer will be much higher in real life bridge than the simulations suggests for the most debated topics. And if you are not paying in those very top events, the success rate of your opponents when defending against your contracts (that are shown low % in simulations for declaring side) will be extremely low compared to what simulations suggests for the defending side.

[rhm]

MrAce, on 2019-December-18, 22:06, said:

The fallacy of all your simulations on this topic (and others as well) might be that your criteria is HCPs. I specifically mentioned not to super accept with "Aceless" hands (1 or no ace) . Basically the quality of hcps makes a huge difference. For example to me a hand with 4 card support and 2 aces +2 kings with 14 is more than enough to super accept while i may not do that with a hand full of qwacks even if it is 16 hcp. No need to mention the simulations favor the defenders more than it does declarers. Huge majority of players (unless you are playing only in very top level bridge) are much more skilled in declarer play as a person than they are in defense as a pair.. This is a known fact.

I certainly agree with you in principle that HCP are not the sole issue, when looking at game in a trump contract. Controls, particularly aces, distribution (side doubleton, good 5 card suit etc.), length in spades all play a more significant role than the odd HCP in an already tightly limited hand.

Quote

Another fallacy in the forums is all of us debate as if we are about to play the BB or top level MP event, and our debates are for training purposes to this match or pair event. Other than Mike (if there is anyone who plays only at these levels please forgive me) your success rate as declarer will be much higher in real life bridge than the simulations suggests for the most debated topics. And if you are not paying in those very top events, the success rate of your opponents when defending against your contracts (that are shown low % in simulations for declaring side) will be extremely low compared to what simulations suggests for the defending side.

This is debatable.
Of course weaker players tend to be better at declarer play than defense, but the overall level of play tends to cancel out, as long as both sides (defense and declarer side) have similar levels.
The question is really what tactic a good pair should employ when playing against a weaker pair in a matchpointed event.
Time and again I have seen the stronger being overconfident about their abilities, getting them a poor score against weak pairs.
When 4♠ requires expert play and a slip in defense to make there is no need to bid the game, 170 will beat the field handsomely.
The danger is going down in 4♠ when the field is making 8 tricks in 2♠.
Time and again I have seen goods pairs scoring badly against weak pairs that way.

Rainer Herrmann