Antrax

[hv=pc=n&s=stht9852dt2ckjt43&w=sj732h743dak653cq&n=sk9854hk6dj84c962&e=saq6haqjdq97ca875&d=w&v=e&b=16&a=pp1c(Nat.)p1h(%21S)p2n(18-19)3c3d(*)4c4s5cppdppp]399|300[/hv] ACBL. 3♦ was alerted by E as showing 4+♥. The person supplying the problem didn't bother stating whether this was east/west's agreement, let's assume the agreement was 3♦ was natural. No screens were used. S sacrificed expecting heart shortness in N and went for 1400. How do you rule? (All players are experts if it matters)

Not really. GIB E shouldn't have a rule to bid 5♥ if GIB W doesn't have replies to it in its book. It's not like a human confusing the bots.

I played exactly like kenberg, which is bad news for kenberg. Would love to understand the correct LOP from both declarer's and east's perspective.

I was sure this was fixed. At least, we've had many more reports of it a while back.

I think I got it, Zel. OP was south and intended 4NT as a quant invite. His partner took it as blackwood and they had a fight.

Should probably ask mobile users. My 27" screen can deal with whatever you throw at it.

That's what I meant, except 2NT should ideally require a diamond stop, otherwise with most hands you can FSF, give a preference, raise the second suit or rebid your own.

I'm not imaginative enough to raise suits on Hx so to me your bidding looks blameless. Singleton in GIB's suit rightfully kept you from jump-shifting and I'm not sure what 3NT should mean but opener shouldn't have to cater to responder inventing a third heart.

Poor GIB, looks like it's stuck for a rebid. A basic GIB would've probably passed you out. Interesting to see faking a 3 card raise rates better in simulations than treating ♦xxxx as a stopper, which is what I suspect most humans would do.

Isn't 5N here a quant bid with 23 HCP or so?

Yes, there is, it gives the opponents a chance to double for the lead. That's why it's important to know how often a real 2NT opener reaches slam off cashing aces, to gauge that against the times the opponents were able to find the killing lead thanks to the partnership making sure, first.

Hey AK1Aleen and welcome to the forums. If you click on the yellow bids the bot makes, you can find out what they mean. In this case you'd see 3♠ is 10-12 HCP with 3 spades, and could judge to raise to game.

On one hand, you're not wrong that checking aces along the way seems right on these hands. On the other, it sort of looks like you want GIB to use Gerber as a sort of control for your 2NT psyches. Do you have an example where you had a 2NT opening and GIB leaped to slam off cashing aces?

You take the derivative of the error relative to the weight and go in the direction opposite the gradient. This is known as Gradient Descent. It's not impossible from a technical perspective, it just might not be the best use of the common BBO development efforts.

This is about how to tune the learning algorithm. It says you and I are great. We play, then you tell it "he's a miserable bastard and I hated every moment of it", so it corrects the weight to not assign people like me to you again. That's just an easy way to describe, in reality you can just mine HMFAG users to tune the algorithm.

The way I see it, a compatible partner means one of two things: someone you have an enjoyable game with or someone you do well with. The second is fairly easy to measure and noise (in the form of non-uniform level of the opposition) will cancel itself out over the long run. The former can be addressed by a simple survey in the end, or just measuring how many hands a partnership played together, and whether that partnership played together more than once. So, despite the strangeness of this claim considering the English meaning of the words, I do think this is something that can be objectively measured. Human input can/should come in the form "do you prefer winning or having a good time?", and probably most people would reply "both". Have you ever worked with customers? In my experience being in the right isn't worth as much as one would expect. Users can just ignore the compatibility rating. Nobody's forcing anyone to play with who the computer finds compatible.

Trucks are stupid

Why on earth not use objective measures, instead? The only thing that might work better is asking people to rate how compatible the partner was, instead of just looking at IMPs won. Other than that, this seems like a computer's job through and through.

Really? I sort of agree with him. Chess grandmasters don't consciously know the weights to assign to different piece configurations, which is why nobody hires them to tell them which weights to use, and instead learning algorithms are used. This seems like a very similar scenario. People don't really know whether the country is 0.7 or 0.4 important to them, so suggesting to give them control just seems like Luddism more than anything.

I disagree. If it works, in the sense you find you consistently have good games when playing with randoms with a high comparability rating, then people will use it without regard for how it works or the desire for control.

I realize it's not what you're asking, but would you prefer E to overcall? If anything, I'd expect a TO double.

Looks like the old issue where once a suit is agreed it's trumps forever. The fact 4NT is keycard for ♦ is a strong indication for that, when there's no suit agreement GIB plays 4NT as keycard for the last naturally-bid suit. That being said, isn't it better to play in your 4-4 fit?

Bummer, I expect he could give the BBF team a hand with the JEC matches.

I'd hoped the joke would be easy enough to get, seeing how Fred is a world-class player and I'm a low intermediate.

It would help directors determine BIT?