[smerriman]

Throughout the Gib challenge events I've noticed people (good players) using a lot of non-systemic bids. This is obviously to fool Gib in some way - but I can't make head nor tail or why, or why this won't lead to your own partner putting you in the wrong contract - and I know from experience that once Gib has an idea of a contract in its head, it's very hard to convince it otherwise.

I can understand perhaps a strange bid now and then if you're losing and need to generate a swing, but that didn't seem to be the case here.

A number of recent examples:

opening 2H in second seat with xx JTxxx Kxxx Jx

opening 2NT with a 5332 18-count

opening 1C first seat with 5332, or 5242, or a balanced 10-count

opening 1NT in third seat with Txx K AQxxxx QJx

opening 1NT in fourth seat with Jxx AQxxxx AQx K

responding 1S to 1C with xx Kxxx Txxxxx x

making a capp 2C overcall with Kx AQxx Txx AQJx

making a capp 2C overcall with Jx Txxxx Txxx Ax

After P P P 1C, overcalling 1D with xxxx xx KJxx Jxx

What's the general idea here?

[1eyedjack]

smerriman, on 2016-October-08, 22:20, said:

I can understand perhaps a strange bid now and then if you're losing and need to generate a swing, but that didn't seem to be the case here.

Yes this is the starting point, of where robot tourneys vary from F2F.

Say you expect, in the long term, to score 55% to 60% by bidding and playing down the line. A bit above average because, heck, you are a better than average player.

Would that long term expectation satisfy you? Would you prefer to come out on top of the heap one time in three, say, at the cost of perhaps bombing out on the other two? Coming in third place time after time would probably rank you in first place in a long term league, because those coming in the top two places will swap around, but I suspect that most players don't see it that way, and would rather hit the podium on occasion, with third place and last place ranking similarly in satisfaction.

If you are a master point chaser, the non-linear MP scale favours wild actions because the rewards for success outstrip the risk that you are committing.

So, suppose you judge that you have an opportunity on a particular hand to take some wild action that rates to net you 90% if it works, 40% if it fails, with roughly equal chances of either result. The math suggests going off piste. And not just "now and again". Each and every hand with a robot is hand 1 of a pickup. There is no memory of past peccadillos, so the concept of "now and again" is pure fallacy. If the math stacks up on that hand, it is good enough to do it every time.

There are three robots at the table all of whom will place absolute faith in the honesty of your bids, represented by opponents over your side in proportion 2:1. Each time you consider, for example, bidding a major suit in search of a fit, you gain when you have the fit, but the amount that you lose in information leakage when you don't have that fit, combined with the confusion that you sow in the process, is disproportionately beneficial compared with a face to face game, an effect that is repeated on each and every hand without the possibility of a concealed partnership understanding developing. This is further aided by the robot's Bird/Anthias mentality to leading short suits against NT. So many times I have concealed a major resulting in a missed major fit, only to have those tricks returned to me by misdefence, often by leading that suit at trick 1.

When, in a tournament, is the best time to "swing", anyway? If you start out with a desire for a 70%+ session, you might randomise hand 1. If it works you can settle down for the rest and play down the middle. If it fails, well there are only 7 more hands to go and you can have another go. Or if you are doing relatively poorly toward the end of the tourney (perhaps because of your earlier wild actions) then you have nothing to lose by randomising toward the end. This seems to favour upping the variance of actions pretty much at any time.

I am not convinced that these principles fully extend to one on one challenge events, but there is a crossover. In particular, in a K/O event the margin by which you lose by coming second is utterly irrelevant.

[nige1]

smerriman, on 2016-October-08, 22:20, said:

Throughout the Gib challenge events I've noticed people (good players) using a lot of non-systemic bids. This is obviously to fool Gib in some way - but I can't make head nor tail or why, or why this won't lead to your own partner putting you in the wrong contract - and I know from experience that once Gib has an idea of a contract in its head, it's very hard to convince it otherwise.I can understand perhaps a strange bid now and then if you're losing and need to generate a swing, but that didn't seem to be the case here.A number of recent examples:

Interesting topic. smerriman's examples in bold italic. My tentative comments in plain.

• opening 2H in second seat with xx JTxxx Kxxx Jx. I confess I did this. A bit dangerous because Gib tends to overbid.
• opening 2NT with a 5332 18-count. IMO, not an enormous distortion
• opening 1C first seat with 5332, or 5242, or a balanced 10-count. IMO 5422 is sometimes OK.
• opening 1NT in third seat with Txx K AQxxxx QJx. IMO this is safer when the singleton honour is in a minor suit
• opening 1NT in fourth seat with Jxx AQxxxx AQx K.. Like this
• responding 1S to 1C with xx Kxxx Txxxxx x. Seems dangerous to me.
• making a capp 2C overcall with Kx AQxx Txx AQJx. Weird.
• making a capp 2C overcall with Jx Txxxx Txxx Ax. A bit frightening. Gib has a habit of taking such bids seriously
• After P P P 1C, overcalling 1D with xxxx xx KJxx Jxx. I suppose there might be an argument for this, at favourable, opposite a passed hand
• What's the general idea here? It probably depends on scoring method, vulnerability, and position. Gib is poor at diagnosing psychs and deviations? 1eyedjack makes a good point: In the old days psychs were more common. Nowadays, regulators are suspicious of them. But no problem with Gib, who won't field a psyche of develop a CPU? But I don't know. I'd be grateful for advice on tactics v Gib.

[nullve]

smerriman, on 2016-October-08, 22:20, said:

opening 2NT with a 5332 18-count

On board 7, set 4, in my R16 BFF Challenge match against you (which I lost by 1 IMP), I opened 2N with

♠AJTxx ♥AQT ♦KTx ♣Ax.

What is this hand really worth? Surely more 18, but maybe not as much as 20? The element of gaming here was that I was expecting a favourable (Bird/Anthias-style) lead if I just didn't mention spades, so the choice was really between 1♣(!) and 2N. This time, that decision was enough to lose the match as partner understandably forced to game with

♠x ♥xxxx ♦xxxx ♣KJxx

and getting a spade lead from ♠7x just wasn't enough.

smerriman, on 2016-October-08, 22:20, said:

making a capp 2C overcall with Jx Txxxx Txxx Ax

That was board 9, set 1, NV vs. V opposite a passed hand. A bit childish, perhaps, but I didn't expect partner to get too excited without a fit.

smerriman, on 2016-October-08, 22:20, said:

After P P P 1C, overcalling 1D with xxxx xx KJxx Jxx

That was board 3, set 2. I have a lot of experience with opening 2♣(!) on hands like this against GiB, and the truth is that it's a HUGE net winner. But since I would then just be exploiting a bug in GiB's bidding and I've wanted to test my (other) bridge skills in these BBF Challenge events, my strategy has been to not do it. But overcalling 1♦ NV vs. V opposite a passed partner seemed completely harmless.