[virgosrock]

---

guess what the mouse-over would show for 4D?

---

what is 5c? why is 5c?

vrock

[Stephen Tu]

I'm not quite sure why you keep on blaming simulations for GIB's bidding problems. Most of GIB's problems come up because of bad *rules* based bidding, not sims so much. Mostly it is following its rule book as to what the bids mean, making the "book bid", sims only come into play on some final bids of the auction, deciding between two close options. The problems mostly come when the definitions and/or priorities for plausible bids are poor, leaving it poor choices at some point of the auction and bad ideas of what people have so inaccurate input into the sims.

1st board:
- why don't you just tell us what 4d was defined as, and what you think it ought to be. What's the point of making us guess? 4d instead of 3nt, or 4d running from the double?
If you think 4d definition is poor, then just say what you think it ought to be, programmer may improve it eventually. Or maybe we disagree with you.

- no one forced you to bid 3nt on 10 count and pass 3nt x, you could have passed or doubled 3s.

2nd board:
I'd blame this more on you than GIB. It's pretty dodgy to double and bid on 17 only when 3 of your pts are stiff K in opp's suit. GIB is expecting stronger, 19+, more working pts, fewer spades (because strong spade one suiters are supposed to favor strong jump 3S/4S overcalls over 2D, rather than doubling than bidding the suit), possible 5-3 heart fit, you can have slam on some constructions so I have some sympathy for 5c though it's kind of pushy and probably should be a tad stronger.

[smerriman]

The first board has nothing to do with simulations; all non-human bids are book bids. 4♦shows 5+ ♦, 3- ♥, 3- ♠, 13+ total points. Looks like a perfectly good description to me? It's not like you want to be playing in 4♦ knowing you both have minimums. 3♠ down 1 looks the best spot (and that's where GIB would have reached). Only yourself to blame here.

In the second board, GIB without simulations would pass your double and not bid 4♥. 4♥ looks like a superior bid to me. I ran a rough simulation of 500 hands myself - giving east a weak 2 in diamonds, west any 4 diamonds, and south any normal takeout double (excluding off shape very strong hands) - and 4 hearts makes 75% of the time. Absolutely clearcut win for GIB's simulations.

5♣ definitely looks pushy, but don't forget you've lied about your strength by bidding 4♠ with that hand. Suppose you really did have 19+ total points. Another simulation of 500 hands showed that there was a slam making in either clubs or spades over 50% of the time - and that's not even taking into account the fact that you could sign off at the 5 level. I'm siding with GIB here.

[virgosrock]

Stephen Tu, on 2017-October-27, 12:28, said:

I'm not quite sure why you keep on blaming simulations for GIB's bidding problems. Most of GIB's problems come up because of bad *rules* based bidding, not sims so much. Mostly it is following its rule book as to what the bids mean, making the "book bid", sims only come into play on some final bids of the auction, deciding between two close options. The problems mostly come when the definitions and/or priorities for plausible bids are poor, leaving it poor choices at some point of the auction and bad ideas of what people have so inaccurate input into the sims.

1st board:
- why don't you just tell us what 4d was defined as, and what you think it ought to be. What's the point of making us guess? 4d instead of 3nt, or 4d running from the double?
If you think 4d definition is poor, then just say what you think it ought to be, programmer may improve it eventually. Or maybe we disagree with you.

- no one forced you to bid 3nt on 10 count and pass 3nt x, you could have passed or doubled 3s.

2nd board:
I'd blame this more on you than GIB. It's pretty dodgy to double and bid on 17 only when 3 of your pts are stiff K in opp's suit. GIB is expecting stronger, 19+, more working pts, fewer spades (because strong spade one suiters are supposed to favor strong jump 3S/4S overcalls over 2D, rather than doubling than bidding the suit), possible 5-3 heart fit, you can have slam on some constructions so I have some sympathy for 5c though it's kind of pushy and probably should be a tad stronger.

I am off the "simulation-based bidding" bandwagon for now. and thanks, Stephen, for your comments.
FIRST HAND
4D was defined as 18+ so I was stuck. In real life 4D is running away from 3NTx since one has doubts and is to play. We absolutely take chances on such hands. "If partner takes out to a supported suit after 3Nx it is to play" should be the rule. I have experienced GIB even jumping to 5 of the minor after such doubles without any significant extras. How could rule-makers not know this? Are they not bridge players? 3NTx run to minor is pretty common knowledge in the bridge world that one does not want to take the risk and is to play.

SECOND HAND
GIB blurbs consistently show double then suit different from partner's suit OR passed partner is 18+ "points". I think I have 18 points. However after your response I think a safe 2S is better.
BUT 5c is just plain silly unless, since partner has shown re-biddable spades, GIB wants to show 4+C en route.

vrock

[virgosrock]

smerriman, on 2017-October-27, 15:08, said:

The first board has nothing to do with simulations; all non-human bids are book bids. 4♦shows 5+ ♦, 3- ♥, 3- ♠, 13+ total points. Looks like a perfectly good description to me? It's not like you want to be playing in 4♦ knowing you both have minimums. 3♠ down 1 looks the best spot (and that's where GIB would have reached). Only yourself to blame here.

In the second board, GIB without simulations would pass your double and not bid 4♥. 4♥ looks like a superior bid to me. I ran a rough simulation of 500 hands myself - giving east a weak 2 in diamonds, west any 4 diamonds, and south any normal takeout double (excluding off shape very strong hands) - and 4 hearts makes 75% of the time. Absolutely clearcut win for GIB's simulations.

5♣ definitely looks pushy, but don't forget you've lied about your strength by bidding 4♠ with that hand. Suppose you really did have 19+ total points. Another simulation of 500 hands showed that there was a slam making in either clubs or spades over 50% of the time - and that's not even taking into account the fact that you could sign off at the 5 level. I'm siding with GIB here.

Again, am off the simulation bandwagon.
FIRST HAND
Would you agree most experts would bid 3NT and run if doubled? The problem is 3N-x-4D is 18+ and there lies the problem.
SECOND HAND
Agreed 2S would be a better bid.

thanks for the responses(s).

vrock

[Stephen Tu]

virgosrock, on 2017-October-27, 16:12, said:

"If partner takes out to a supported suit after 3Nx it is to play" should be the rule. I have experienced GIB even jumping to 5 of the minor after such doubles without any significant extras. How could rule-makers not know this? Are they not bridge players?

A fair number of non-programmers come into this forum make this kind of comment, thinking the programmers who worked on GIB don't/didn't know how to play bridge. They think that the issue is that they examined every auction, made rules for each one, and on purpose wrote down some ridiculous rule that pulling 3nt-x to 4m shows 18+.

This is simply not the case.

The problem is that the number of possible bridge auctions is ginormous, especially when you include competitive auctions. It is simply impossible to examine each auction and define it, write down every possible auction and define all possible bids. Instead the programmers write thousands of rules to TRY to cover as many auctions as possible, with some fallback rules as defaults. Inevitably, some competitive auctions are going to fall through the cracks of the rules, fall back on some default that might make sense in some other context, but doesn't make sense given the full context of the auction. Humans are intelligent enough to handle these auctions. Computers at this stage generally are not. It's really hard to transfer a human's bidding knowledge to a machine, because machines are simply really really dumb. They follow rules very reliably if they know the rules for a particular sequence, but give it a slightly novel situation and they have no idea how to use analogies and common sense to figure out sensible things to do. So when the computer does something bad, your assumption should not be that the programmer doesn't know how to play bridge, or what the bid should mean, but simply that they didn't anticipate this super-rare auction would need to be covered, which is understandable given the practically infinite number of competitive sequences. Imagine being some expert bidder bridge player, world class, but you are tasked with writing down bidding flow chart for 7 year old to bid, who has never played bridge, and starts every day as a blank slate having forgotten everything from the day before. Even you know every bid perfectly, it is really hard to write down rule set to cover every conceivable situation.

This kind of auction, where someone voluntarily bids 3nt, and South has a penalty double, and then West wants to run, is just like really really rare. So it's not particularly surprising that a run to 4d is poorly defined.

Remember computers run on very simplistic, dumb rule following, like following a flow chart. If it gets into a situation where it's off chart, it's very hard for it to recover. There's no such thing as "common sense" for the computer, like it can't reason that "oh if it was 18+, it wouldn't have bid only 3nt earlier, therefore 4d is just changing its mind that 3nt was a good idea", and "figure out" what 4d ought to mean. Someone would have to specifically think to put in a rule for auctions like this, inverted minor followed by a preempt, followed by a balancing 3nt, followed by a penalty double.

[johnu]

Stephen Tu, on 2017-October-27, 19:17, said:

Remember computers run on very simplistic, dumb rule following, like following a flow chart. If it gets into a situation where it's off chart, it's very hard for it to recover. There's no such thing as "common sense" for the computer, like it can't reason that "oh if it was 18+, it wouldn't have bid only 3nt earlier, therefore 4d is just changing its mind that 3nt was a good idea", and "figure out" what 4d ought to mean. Someone would have to specifically think to put in a rule for auctions like this, inverted minor followed by a preempt, followed by a balancing 3nt, followed by a penalty double.

In general, that would be an incredibly bad way to program, but it seems that much of GIB bidding is programmed that way. A more general rule - your side bids 3NT to play and there is a penalty double by opponents. Taking out to a suit does not show extras.

If there is time and resources, more specific sequences could be programmed. On the other hand, so many of GIB's descriptions are inaccurate to say the least that it's not clear what would happen if West takes out to 4♦. Does East think there is enough to bid 5♦? The singleton spade opposite a NT try should warn that there are wasted values in spades.

[johnu]

virgosrock, on 2017-October-27, 09:16, said:

what is 5c? why is 5c?

My free advice is don't make offshape takeout doubles when playing with GIB. More often than not, it's the prelude to a major disaster, and very rarely results in a better result than a simple overcall. GIB doesn't understand how to proceed after one of these takeout doubles so it is foolish to make them.