[Cyberyeti]

johnu, on 2025-February-11, 03:48, said:

My simulations say that upgrading is very reasonable and can lead to good game for the reasons I have given in this thread. My simulations also show that this hand is better than a number of hands that nobody really questions about opening 1NT.

Is K1098 better than KJ32 for example ?

[Cyberyeti]

I would actually open the first hand a 12-14 or 15-17 1N. I would not bid game with the hand opposite and would play 2N.

Second hand is difficult and methods dependent.

Third hand 6♠ is a terrible bid, if partner has AKxxxx, ♥J or xxx then 7 is almost laydown, as it is, if you know partner has 2 of the top 3, 6N is laydown so you should bid it, but blackwood will tell you whether he has AK.

[smerriman]

 johnu, on 2025-February-11, 03:48, said:

My simulations say that upgrading is very reasonable and can lead to good game for the reasons I have given in this thread. My simulations also show that this hand is better than a number of hands that nobody really questions about opening 1NT.

Can you please share your parameters, like I shared mine? Like I said, just calculating game opposite a random 10 has no relevance, so I want to know what you actually simulated.

[johnu]

Cyberyeti, on 2025-February-11, 05:26, said:

Is K1098 better than KJ32 for example ?

Interesting comparison. K&R evaluates both as 3.85 points so exactly equal (I had to restart the evaluator to make sure it was working correctly as that surprised me). In NT, or maybe trumps on defense, K1098 is always at least 1 trick, and may be 2 tricks. KJxx may not be worth a trick (or 1 or two tricks).

In most point counts, KJxx is more points than K1098, so all things being equal, one would probably want KJxx. But does it make a difference? You don't have KJxx, you have K1098.

[johnu]

smerriman, on 2025-February-11, 12:04, said:

Can you please share your parameters, like I shared mine? Like I said, just calculating game opposite a random 10 has no relevance, so I want to know what you actually simulated.

I actually said "random flat hands". Probably should have said random balanced hands, so any distributions with 4333, 4432, 5332. Also 4321 HCP, no restrictions on distribution of honors, or specific honors.

[smerriman]

johnu, on 2025-February-11, 23:49, said:

I actually said "random flat hands". Probably should have said random balanced hands, so any distributions with 4333, 4432, 5332. Also 4321 HCP, no restrictions on distribution of honors, or specific honors.

OK, South has flat hands only, assume they *never* upgrade any 5332 9 counts, no other assumptions:

produce 1000
predeal north SKT98, H943, DA92, CAK7
condition shape(south, any 4333 + any 4432 + any 5332) and hcp(south)>=8 and hcp(south)<=10
action printoneline

My homemade simulation code:

$bidgame = $hcp==10;

$trump = 'N';
if ($spades>=5) $trump = 'S';
else if ($hearts>=5) $trump = 'H';
else if ($spades==4 && ($hearts!=3 || $diamonds!=3 || $clubs!=3)) $trump = 'S';

$level1ntopener = ($trump=='N' ? ($bidgame?3:2) : ($bidgame?4:3));
$level1copener = ($trump=='N' ? 1 : 2);

$delta = getimps(getscore($level1ntopener,$trump) - getscore($level1copener,$trump));

Frequency of IMP differences over 1000 hands:

Partner bids game:
+10 IMPs: 130
-5 IMPs: 7
-6 IMPs: 131
-7 IMPs: 35

Partner invites and we decline:
0 IMPs: 453
-3 IMPs: 51
-5 IMPs: 193

Game makes 43% of the time it was bid, but the average IMP score over the 1000 hands is -0.884 IMPs / hand; 1NT is a very clear-cut loser. Why do our simulations differ?

[johnu]

smerriman, on 2025-February-12, 00:51, said:

Game makes 43% of the time it was bid, but the average IMP score over the 1000 hands is -0.884 IMPs / hand; 1NT is a very clear-cut loser. Why do our simulations differ?

Like I wrote, I put no restrictions on partner's distribution. Looks like you are eliminating hands that could play in spades, or hearts. And including 8 HCP hands?

[Cyberyeti]

johnu, on 2025-February-11, 23:45, said:

Interesting comparison. K&R evaluates both as 3.85 points so exactly equal (I had to restart the evaluator to make sure it was working correctly as that surprised me). In NT, or maybe trumps on defense, K1098 is always at least 1 trick, and may be 2 tricks. KJxx may not be worth a trick (or 1 or two tricks).

In most point counts, KJxx is more points than K1098, so all things being equal, one would probably want KJxx. But does it make a difference? You don't have KJxx, you have K1098.

My point was that K1098 is 3 points, KJ32 4 points, everybody would open 1N with KJ32 as that would be 15 in the original hand, I was interested to see what happened if you ran the same sim with KJ32 rather than K1098.

[smerriman]

 johnu, on 2025-February-12, 02:38, said:

Like I wrote, I put no restrictions on partner's distribution. Looks like you are eliminating hands that could play in spades, or hearts. And including 8 HCP hands?

You stated random balanced hands, 4432, 4333, or 5332, so that's what I adjusted my code to; I didn't exclude anything - plenty of those play in a major (thus the code to determine what the trump suit will be in each case). My original sim didnt impose a balanced hand and got the same result.

You don't invite vulnerable with 8 HCP at IMPs? That would definitely change things; double dummy usually says you should invite aggressively in that case. I suspect you'd be in the minority there.

[johnu]

smerriman, on 2025-February-12, 04:19, said:

You don't invite vulnerable with 8 HCP at IMPs? That would definitely change things; double dummy usually says you should invite aggressively in that case. I suspect you'd be in the minority there.

There have been a number of discussions on Bridgewinners about when to invite over 1NT with 8 HCP (ie expected total HCP 23 min, 25 max). There seemed to a solid consensus that 4333 was not a good distribution to invite. With 4432, some would invite but hope to get to a better 4-4 major fit and live with inviting 2NT when no major fit was found. Some would just pass, and some would invite. I don't remember much about 5332, but certainly for minors a decent 5 card suit is more important than for a 5 card major since you'll probably end up in the major with 3+ card support, while the minor won't even be mentioned in the bidding.

I would invite with a "non" 4333 8 HCP that was upgradeable to 9 HCP. So, good spot cards like 10's, honor sequences, etc. For practical purposes, I'm guessing they play about the same as a mediocre 9 HCP, and I don't have an algorithm to automatically model them, so I leave them out.

[smerriman]

Ha, I was specifically thinking of BW threads - I thought I remember taking away that the consensus was always pass 8 at MPs, but almost always invite at IMPs - I'll need to reread them as it was a while ago.

I still get a net negative score for a 1NT opener even if every 8 HCP response is passed though - those 9 HCP invites where you lose 5 IMPs due to going down 1 in a partscore just add up to too much.