We recently asked Lorand Dali, our software engineer, some questions about GIBBO. He walk us through what’s changed, how it works, and what players should expect at the table.
GIBBO is one of the biggest upgrades GIB has ever had. Many players aren’t fully aware of what changed and what stayed the same. How would you explain it in a simple way? Where should players expect the biggest differences in bidding, card play, and defense?
The biggest improvement is in the basic (free) robot. The main reason is that we enabled Monte Carlo simulation during bidding for the basic robot.
This means that when it’s the robot’s turn to bid, it may consider several possible bids. For each candidate, it makes the bid, fast-forwards to the end of the auction, and evaluates which bid worked better across a sample of possible layouts.
Previously, this was too computationally expensive, so we could only offer it in the advanced GIB. Now the basic GIBBO can do this too.
Another big change is the introduction of “rollouts.” This is a new way to run simulations using neural networks.
We can also now ask questions like: “Given this position, how likely is a player to choose a certain card?” If we observe a specific play, we can estimate the probability of different possible hands.
GIBBO uses a neural network to evaluate bidding outcomes instead of traditional double dummy analysis. For a player who has never touched AI, what does that actually mean at the table?
A simple way to think about a neural network is that it’s a piece of math that can decide what card to play next. In a sense, it acts like a small robot that plays bridge. The interesting part is that the main robot (GIBBO) can use this “small robot” to plan ahead by rolling out the rest of the hand.
A rollout means you simulate playing a card, then let the neural network play the rest of the hand to see what happens. This is an alternative to using a double dummy solver.
There are advantages:
If GIBBO were a human bridge player at your local club, how would you describe their personality at the table?
If GIBBO were human, I think he would be a strong but very single-minded and stubborn player. Occasionally, he would do something out of the blue, and give an explanation that nobody else can relate to.
Can you give a few examples of situations where GIBBO performs noticeably better than the old GIB?
The old basic GIB would always pass when a situation wasn’t covered by an explicit rule. For example, after a 2NT opening, it might pass with 11 points and a long minor. GIBBO will try to find a reasonable bid even when the situation isn’t explicitly defined.
Old GIB also tended to lead passively, especially against notrump. GIBBO is much more aggressive, sometimes even too aggressive (especially against 6NT, but we’ll improve that). GIB also assumed perfect play from opponents. That often led to situations where it didn’t matter what card it played, so it would choose something that looked strange.
No robot is perfect. Where is GIBBO still catching up?
Bidding is still the weakest area, especially in long or complex auctions. If you stretch in competitive auctions, GIBBO will take your bids literally and assume you have the full strength you promised. Because it treats every action seriously, it’s easier to fool than a human. Humans have intuition that something might be wrong and adjust. GIBBO doesn’t.
Defense also still has room for improvement, especially signaling and cooperation.
Declarer play is the strongest area, but in a different way than humans. For example, GIBBO can outperform strong players in some declarer challenges, but still struggles with certain structured problems.
Players sometimes post hands where GIBBO’s decisions look surprising. How should players interpret those moments?
When the robot does something unexpected, I sometimes facepalm. Other times I just smile and move on to the next board. I save the hand and analyze it later. I’m happy when players post examples, as it helps identify patterns in what players find frustrating.
How does player feedback reach you and influence development?
I review reported hands and use them for debugging. Forum discussions are also useful to understand trends and recurring frustrations. Players can also message me directly on BBO (Lorserker).
What’s the best way for players to help improve GIBBO?
The best approach is to share clear examples, ideally using the Handviewer link and specifying whether it was basic or advanced, and whether it was MP or IMP.
If you could take one feature from another bridge AI, what would it be?
Some robots allow configurable systems and conventions, which I like. For defense, I like Bridge Baron’s signaling. From BEN, I would take the sampling, bid simulation, and additional neural network approaches.
What are you most excited to work on next?
I’ve focused mostly on card play so far, since it’s the hardest part. Now it’s time to give more attention to bidding and defensive signaling. I’m excited to work on those next.
Finish this sentence: “Players would be surprised to know that GIBBO…”
“…plays 10,000 cards in its head before choosing the one it actually plays.”
Any final words for players who had a rough session with GIBBO?
First, I apologize. I play with the robots every day, I read the forums, and I understand when you are frustrated. Rest assured we are working hard to optimise GIBBO.
And one tip: try to adapt to the robot. The robot will not adapt to you. To shoot the arrow, you must become the arrow.
That’s GIBBO, through the eyes of the person building it. Some things are already better. Some are still evolving. And behind it all, there’s a lot more going on than it might seem at the table.
If you’ve had a memorable moment with GIBBO, good or bad, we’d love to hear about it. Share it in the comments.
[…] you enjoyed getting to know more about Lorand. We also asked him a few questions about GIBBO. If you’re curious to learn how the robot works and what’s changed, you can read more […]
Hi Lorand,
Yesterday happenned to me something extraordinary, way crazier than any robot blunder I've experienced before.
Holding a 29HCP 2443 hand, I've opened 2C, GIBBO bids 2D, I bid 4NT (28-30HCP) and GIBBO concludes with 7NT, which receives a double from RHGIBBO. Ok, I'm out of an ace, apparently, another crime by GIBBO, nothing new.
GIBBO tables a 4342 hand, two Queens and three Jacks. I'm out of the S Ace, and I play Spades twice from the table towards my hand to make 12 tricks. RHGIBBO (who doubled) plays low twice and I make 7NT doubled. Chilling defense.
This adventure points out not only to faulty software design for the bidding but also to incredible bad defence, pointing to same faulty design.
Cheers,
Paul.
Dear BBO Team,
Whether a robot with perfect memory is "cheating" depends on how "cheating" is defined in the Laws of Bridge. Law 40 B 2(d) of the 2017 Laws of Duplicate Bridge states:” Unless the Regulating Authority provides otherwise, a player is not entitled to any aids to his memory, calculation or technique during the auction period and play.” While humans are generally forbidden from using external aids, robots are programmed to remember every card, giving them a distinct advantage. I believe this is acceptable for learning with free robots, but it creates an uneven playing field in paid tournaments. A robot's built-in memory acts as a permanent "aid" that humans cannot access.
It is the “Unless” part of Law 40 that allows BBO and other platforms to provide viewing capabilities for almost anything (profiles, convention cards, bidding sequences, etc.) during a game. I believe BBO and other providers should allow players to use ‘memory aids’ (apps, AI agents, etc.) while playing in robot tournaments. Such apps already exist, and as they become more common, violations will be nearly impossible to track. Clever users can use these tools easily in ‘Daylong’ tournaments where time is not a factor. This creates an uneven playing field, as some follow the rules while tech-savvy players use aids.
Eventually, online providers may need to provide human players with the same card-counting information known to the robot. Law 40 B 2(d) gives the regulating authority the choice to allow such aids. A law should be equally applicable or relaxed for all types of players. In fact, since you can currently view the previous trick on BBO, why not allow access to the entire history of the hand?
In F2F bridge, physical stamina and memory are part of the challenge. In online bridge, the environment already 'assists' the player by showing the bidding history and the last trick; additionally, there is virtually no time constraint in 'Daylong' robot tournaments. Since tracking external memory apps is almost impossible, the game must eventually change to accommodate this new reality.
Online bridge (‘Digital Bridge’) is evolving into something like a 'take-home exam,' where the emphasis is placed on a player’s bidding and playing techniques rather than their memorization skills, which are strictly required in F2F bridge (‘Physical Bridge’). In my opinion, online bridge is evolving into a unique derivative of the traditional game, driven by the advancement of AI!
Sincerely,
A robot lover
Hi Lorand!
Thank you for the good work! I would like to emphasize our need of getting updated of each major change in the environment of at least these games: Just Play Bridge, Weekly Free Tourney, Free Just Declare, Challenges against Bots. The reason is that my community and probably other ones are evaluating their progress on statistical analysis of their scores in theose games. As the reference moves (hopefully better quality of the robots, no sensible change of the field, different average/variance HCP share of the line), a marker needs to be added as to otherwise explain the changing scores for a cause different than the skills of the player being measured. Thanks a lot in advance!
VJ
Thank you for your thoughtful feedback and suggestions. We’ll pass them along to the team. 🙂
The Basic GIB still keeps bidding crappy suits.
did you get my review? Why didn't you post it?
I have posted 100's of robot complaints for many years now, with specific examples and suggestions that cannot be ignored. But, noooo! I keep increasing my invective for max effect - but Nooo! 🙁 I can write a thick book with all the hands I have analyzed to entertain & inform my friends - including directors, and BBO editorship. I've never had a response, & nothing changes even when the absurd hand evaluations sometimes exceeds 60 silly "Total Points" on a single hand!! The greatest flaw in the ointment is - GIBBO totally ignores crucial lead direct bids. I often take essential ultimate (winning) risks - but the Robot will lead or defend like they are blind - which kills alotto my big scores. I scream for a "REFUIND!" about once a week! ALAS, crickets 🙁 BBO ignores me, even though I have nearly 30,000 Purple-Points and I'm soooo grateful for your platform1! Sincerely. razorsharp 🙂
We understand your frustration, Russell, especially when you’ve spent a long time sharing detailed examples and feedback.
While we may not be able to respond individually to every report, player reports and recurring patterns are reviewed by the team and do help identify areas that need improvement.
For specific hands, the most effective way to report them is through the built-in robot report tool. You can go to your History tab, open the hand, click the 3 bars in the top right corner, then select “Export” → “Send robot report” and include a description of the issue you noticed.
That gives the team the exact hand, auction, and play sequence to review more closely.
Look at my bd1 in EBU comp #26001 Sun 1005. why on earth doesn't gibbo x 4s or bid 5h ???
Hi David, it’s best to share the Handviewer link for the specific hand you mentioned and send it to support@bridgebase.com so our team can review it more closely.
An even better option is to go to your History tab, find the hand where you noticed the unusual GIBBO action, click the 3 bars in the top right corner, then select “Export” → “Send robot report”. You can include a short description of what seemed unusual before clicking OK.
That helps the team review the exact hand, auction, and play sequence in more detail.
I've had a variety of auctions where I bid 4NT which the robot says is quantitative, and then it replies with a key card response
agreed! the description does not match the action in man6 auctions.I have sent in many robot reports…l
Dear Bridge Base Online Team,
I would like to respectfully ask for greater transparency regarding the robot versions used in BBO tournaments and daylong events.
With the recent introduction of GIBBO and the ongoing AI upgrades, robot behavior has clearly changed in many formats. Different tournaments now seem to use different robot configurations, bidding engines, simulation methods, and aggression levels. For serious players, this has a major impact on strategy and expected results.
At the moment, players often do not know:
which exact robot version is being used,
whether the event uses old GIB, GIBBO, or an experimental version,
how aggressive or conservative the robot partnership style is,
whether the robot strictly follows system rules or allows flexible interpretation,
and whether different tournaments use different AI settings.
This matters because robot behavior changes the entire nature of the competition.
For example:
In some tournaments, the robot partner appears to punish almost every small deviation from the system very harshly.
In others, the robots seem much more flexible and adaptive.
Some formats appear extremely aggressive in bidding and leads, while others are more passive.
Without transparency, players cannot properly prepare, adapt, or understand the environment they are competing in.
I believe every tournament should clearly display:
the exact robot version being used,
the AI model or configuration,
the bidding style and aggression profile,
whether experimental features are enabled,
and the date of the last major robot update.
Even a simple label such as:
“Classic GIB”
“Advanced GIBBO”
“Experimental Rollout Version”
“Aggressive Competitive Profile”
would help players enormously.
Transparency builds trust and improves the competitive experience for everyone.
Thank you for continuing to improve the robots and the platform. Many of us appreciate the work being done and simply want clearer information so we can better understand the conditions in each tournament.
Sincerely,
A dedicated BBO player
Thank you for your feedback, RBG. Just to clarify, all GIBs on BBO have now been upgraded to GIBBO.
We’ve actually been identifying the robot type in tournament game descriptions for a while now. Depending on the game, you may also see labels such as Basic GIBBO or Advanced GIBBO (AdvGIBBO) at the table.
We understand the interest in even more transparency around robot behavior and configurations, and our team is continuing to improve how this information is shared with players. 🙂
Out of curiosity, what % of robot reported hands do you guys go over?
How many do you get in on an average day/week/month?
Some recommendations:
Introduce just some basic signallying like (maybe standard for basic and UDCA for advanced). Maybe even suit preference in situations where it's "obvious" attitude and count are irrelevant.
Improve strong 2c bidding. GIBBO seems to always assume 23 HCP. A losing trick count system might be better where 2!c openers show 3 1/2 or less losers. It often likes to jam 6nt thinking there's 33+ combined HCP instead of being patient and listening to the shape the 2c opener has and be ware there might be singletons and voids.
Please make gibbo bid 2d waiting unless it has an opening hand with a good 5+ card suit. Teaching advanced GIBBO Kokish relays over 2c - 2d - 2M would be nice so I can show my 2 suiters before GIBBO starts it's evaulation (I'm aware I might be asking too much).
We don’t currently share exact numbers on robot reports, but reported hands and recurring patterns are reviewed regularly by the team to help identify areas that need improvement. Specific examples and thoughtful suggestions like yours are always valuable.
Great work,Congrats to Lorserker
Very impressive, but how all this come along with GIBBO advertising a certain hand and force for a bid and having in fact an entirely different hand and force ?
How come such a situation is coming up every time GIBBO faces a strong jump bid ? After almost every jump strong bid GIBBO goes to slam with rubbish.
How come that highly distributional hands appear in E-W, when N-S hands are "normal", with 3 or 4 times more incidention than in real face to face bridge ?
How come that I've received no feedback on any of my messages sent after criminal blunders by GIBBO ?
Thanks for sharing these observations, Paul. As Lorand mentioned in the interview, GIBBO tends to take bids very literally, which can sometimes lead to overly aggressive continuations or misunderstandings about partner’s strength and shape.
Regarding unusual distributions, the deals themselves are randomly generated, but we understand that certain patterns can feel frustrating or memorable during play.
While the team may not be able to respond individually to every report, player feedback and reported hands are regularly reviewed to help identify recurring issues and areas that need attention.
Excellent feature, that will increase the momentum towards better quality of the robots. Yet, there is something structural that prevents the robots from reaching valuable bidding skill level: as Asimov made the 3 (lately 4) laws of the robotic mandatory built-in in each positronic robot, the bridge robots should contain dynamic behavior rules that an outstanding bridge player would have listed for the programmer. I had such attempt in my bridge robots, who were performing much better than the ones we see at work, and with much less computing power. So much for my advice 🙂
VJ