[gszes]

---

Is it possible that the programmers have decided arbitrarily to convert 4 level contract to penalty merely because they do not think they can MAKE anything they bid
when partner makes a TOX --- the x is described as take out? This needs to be fixed and the computer should NEVER leave a tox in w/o defensive trump tricks (which they
will almost never have. That should solve these stupid (yes stupid) passes with no defense. If I am wrong to balance with this type of hand sobeit BUT the computer
should NEVER EVER convert any self described TOX w/o trump tricks PERIOD (well at least through the 5 level anyway).

[virgosrock]

gszes, on 2018-June-23, 12:31, said:

---

Is it possible that the programmers have decided arbitrarily to convert 4 level contract to penalty merely because they do not think they can MAKE anything they bid
when partner makes a TOX --- the x is described as take out? This needs to be fixed and the computer should NEVER leave a tox in w/o defensive trump tricks (which they
will almost never have. That should solve these stupid (yes stupid) passes with no defense. If I am wrong to balance with this type of hand sobeit BUT the computer
should NEVER EVER convert any self described TOX w/o trump tricks PERIOD (well at least through the 5 level anyway).

I have been yettttttttttttttttt-ing about these defenseless passes for ages. Jdonn had hinted they are working on this about 4 to 8 months ago.

[Stephen Tu]

Never ever convert without trump tricks is an exaggeration at high levels. Sometimes you just hope partner has enough side tricks to set them, as you think your contract has no hope/will get doubled, and it might be percentage to sit. This is where simulations should be able to work out what works on average.

On this hand, I think the problem is it may not be properly taking into account that E as passed hand voluntarily raised to 4H, which should show some degree of trump fit/side shortness/hope of making, which also makes South's hand more shapely on average, and favor pulling. But the description of 4H just says 1+ H.

[steve2005]

Virgo is right he has pointed out numerous examples of bad doubles (takeout or penalty)
In any case it is very difficult to make a penalty double.
Gib will leave in a penalty double on unsuitable hands and pull penalty double when you are destroying the opponents.

And as Virgo has pointed out you can not trust a penalty double by Gib.

There is no simple fix. This would be a major fix to Gib's program.

[virgosrock]

steve2005, on 2018-June-23, 15:47, said:

Virgo is right he has pointed out numerous examples of bad doubles (takeout or penalty)
In any case it is very difficult to make a penalty double.
Gib will leave in a penalty double on unsuitable hands and pull penalty double when you are destroying the opponents.

And as Virgo has pointed out you can not trust a penalty double by Gib.

There is no simple fix. This would be a major fix to Gib's program.

a workaround which has partially worked to decode these asinine doubles is e.g. if they are in 4S in a competitive auction and you have 3S pull it!!! It is adding "points" to its stiff spade.
or if you have AQT9 in diamonds with diamonds bid on your right and supported, DONT double because it has a stiff or void and will pull and mostly you will get doubled in suit it pulls.

[smerriman]

Stephen Tu, on 2018-June-23, 14:18, said:

On this hand, I think the problem is it may not be properly taking into account that E as passed hand voluntarily raised to 4H, which should show some degree of trump fit/side shortness/hope of making, which also makes South's hand more shapely on average, and favor pulling. But the description of 4H just says 1+ H.

But surely the simulations are based on hands where GIB east would bid 4♥, whatever shapes they may be, vs basing them on the human descriptions? In which case the process should be flawless. Clearly some major bug in there somewhere.

[Stephen Tu]

smerriman, on 2018-June-23, 19:17, said:

But surely the simulations are based on hands where GIB east would bid 4♥, whatever shapes they may be, vs basing them on the human descriptions? In which case the process should be flawless. Clearly some major bug in there somewhere.

One would hope so, but one would have to have access to the console/text-mode output of the program to take a look at what it thinks sample hands look like. Would be really nice if they offered the current database as upgrades for us standalone GIB users.