These are the answers that were published:
, 25 years of tears and aspirations at the bridge table.
Declare as often as possible. The robot is a bad declarer. You can declare more often by:
* Opening 1NT and 2NT outrageously often.
* Open a 4-card major with one 4card M and 11-14 HCP
* If the robot opens 1S and the auction continues: 1S — 2H — 3H, never show spade support if you have it.
* If you open 1M and the robot responds in the other M, avoid showing support if possible and if the robot subsequently raises your suit, always play the hand.
Dont forget you have the strongest hand. If you have opened on 11-12 HCP, you can pass partner’s forcing calls if you have a fit or semi fit for hsi new suit.
Safe Finesses. The robot will assume you will make the correct play in a suit. For example in a 2-card ending, you hold AQ in your hand and xx in dummy. The opponents hanve the king and an outside winner. LHO will never stiff his king (assuming you will drop it) so it is always safe to take this finesse.
Exploit robot NT defense. The robots defend exceptionally poorly against NT. Light NT games will often make on bad defense.
Bidding fake suits. Sometimes you can engineer an auction to 3NT where you make a 2/1 in your weaker minor. The robot never adjusts and figures out you have xx in the suit you bid.
Accept all his 5M slam invites. The slams generally make.
, recovering bridge addict
* As you mentioned, open NT a lot. I open most balanced 14 counts and many semi-balanced 15-17 counts (6332s, but even some 4441 and 5431 shapes) with NT. Reasoning: in the 12-board ACBL tournaments, you have 55 minutes to play 12 hands. This is a lot of time. Far too often the robot declarers take a 0% line and go down because their simulations just happened to be lame. You have the time to think about the play properly, so it’s good to declare as much as possible. Plus, the whole point (for me) is that it’s fun to practice my declarer play in a semi-controlled way.
* Lie. The robots have poor judgment, especially in slow, constructive auctions. Even if it violates system, an early jump to 3NT or 4M can keep you from getting too high, or get you to good games. (This is really a corollary of Michael’s suggestion to use the “best hand” feature: your partner is often implicitly limited, even though the robot itself does not take this into account.)
But beware: I’ve found that many jumps to 5 of a minor that “should be” to play (and have vague descriptions when you hover over the bid) are taken as slam-going by the bots.
* Underbid on (some) competitive auctions. In my experience, CHO (center-hand opponent) GiB is too eager to make penalty doubles when, for example, you should really just sell-out to their 4S contract. If you’ve underbid, you can get rich passing partner’s penalty double. But it you lied about your strength previously, as per the previous point, be sure to pull! Letting the opps make doubled overtricks is no fun, and GiB often doubles on “values” you may or may not have actually had.
* Relatedly, be aware that GiB does not understand the Law of Total Tricksvery well. (Although it alerts some of its bids as “The Law: 9 trumps = 3 level” and the like.) I consider competitive partscore hands a source of stochastic variation in the MP result, rather than an opportunity to exercise strategy and wise bidding.
* By the same token, be careful with high-level takeout X’s: CHO GiB is wont to pass for penalties even when it is wrong. In fact, the bots do this even on some low level takeout double auctions: if righty opens 2S and you X, CHO GiB tends to pass with balanced minimums more often than human players. (Could we humans be wrong in this? Perhaps the simulations bear out the robot strategy…)
* Be aware that GiB does not preference normally. So lean towards rebidding your six card suit when competing for a partscore, rather than balancing in your side suit expecting partner to preference.
Playing timed Total Points:
* As Michael stressed, pass. A LOT. In 25 minutes, a good Best Hand total point score is certainly over 5000 and depending on the field over 6000 or 7000 might be necessary to be in the money. You don’t get these scores by playing partials, or even NV games if they take more than a minute. I’m not an expert at this format, but you should definitely be seeing 25-30 boards in 25 minutes, if not more.
* In this format you should lie even more than in matchpoints. If you have a balanced 15 count, partner opens a major, and you have 3 or 4 trumps, I’d say you should jump to game, esp. NV. The chances of slam are slim. (Remember partner is limited to 15!) The time savings of a fast auction are real and valuable.
* If your side clearly has game, and there is a choice of strains, consider letting GiB declare when that is possible. GiB is not a great cardplayer, but it is FAST and time is money.
, teacher = learner
Sam has already given a fantastic answer. There are also some interesting tricks in the cardplay. GIB will often make plays as defender that it knows can’t cost double dummy but do cost in real life. There are a huge number of these and when I play I try to think about those too rather than memorising a list: GIB is creative enough in its misdefences that it often produces new special efforts. A few examples that spring to mind
- Leading out stiff K or Qx off side
- Covering trump pips because if you ran that 7 that you led off the dummy you could pick up its painful four or five card holding.
- Solving two way guesses by leading out the suit or covering the J or T (this happens *all the time*). I’ve even seen it drop Q on the first round from Qx with four trumps outstanding, although I think that bug has been fixed 😉
In terms of this sort of play the easiest advice I can give is “always give GIB a chance to cover”.
I recently saw it bare a K that it was known to hold, offside, because it “knew” that it was going to be squeezed in two tricks’ time. I couldn’t believe it!
, Avid duplicate bridge player.
There are a few different tips:
It is important to know the system that the robots play. You can hover over bids and get explanations of what they meant, what they would mean.
A lot of the robot games are played best hand style where the human player always has at least as many points as any other player. That should influence your play.
Many of the robot games are timed. When that is the case, and when you are playing total points or bingo or some other speed game, you should pass many opening hands. Because you want to get slams and games you want to maximize the value of each hand you play. So a balanced 15 count should be passed if you are playing a timed total points game because you want to wait and get even better hands in the future.