Welcome to Project BOLD

Background

As the Founder and Research Director of Cardquant I have extensive experience in conducting original research into the strategy of card games with a very large game space, and teaching my students the insights that result in a powerful, coherent framework. For the last decade I have trained high stakes and ultra-high stakes players in the poker games of Omaha, 5 card Omaha, 6 card Omaha, and Short Deck poker in both cash game and tournament formats. These players have won (in many cases individually) millions of dollars competing in some of the most challenging poker games on the planet.

An essential component of my research has been working at the intersection of computational, analytic, and synthetic approaches to knowledge discovery, dissemination, and strategy. I have provided the domain insights to develop the abstraction for Cardquant’s in-house CFR (Counter Factual Regret) poker solver and designed our in-house poker training web application, Pokermuscle. My lead software developer has a phD in Artificial intelligence, and I personally have degrees in Physics from the University of Oxford and University of Pennsylvania (the latter on a Thouron scholarship).

I conduct research privately on the (much) larger question of the capacity of human hyperintelligence to exceed artificial intelligence in specific domains, a question that mainstream scientific authorities would have you believe is a fait accompli. It was in the course of exploring this area that I came across a series of articles on the progress of robot bridge in July 2023. Having not played bridge for 16 years I had naïvely assumed that the game would largely have been overtaken by robots in a fashion similar to chess, but it rapidly became clear that this was not so. Due to the extraordinarily large game space of bridge, the incomplete information that characterizes the game, and the cooperative elements essential to the bidding, robots were (and are) very far from competing with world class players in the full game. My interest in bridge was reignited, and so began a frenetic period of research that has resulted in the inception of Project BOLD.

Phase 1: Core System Design

From my experience working on machine learning approaches to poker it was clear that, in the specific sub-game of bridge card play detached from the bidding, robots would inevitably achieve a world class level with relatively simple techniques. However, returning to bridge as I was armed with a well-developed information theoretic understanding, it was readily apparent that there was a vast unexplored territory of bidding theory which a mechanistic approach was poorly equipped to traverse. More surprising to me was how little progress had been made in system innovation since I had left the game. Whether the lack of innovation is best explained by the incentive structure of professional bridge, restrictive system regulations by domestic bridge authorities, or by a dearth of creativity, far more energy had been dedicated to refining existing systems rather than trialing new ones.

As part of my early background research I set about reviewing as many systems as I could find, whether available online (the system archives at ecatsbridge.com and bridgewithdan.com were invaluable at this stage), or in book format (including the MOSSO system, coauthored by the irreplaceable Mr. Burn).

Newly reacquainted with the bridge system landscape, I set about applying my own theoretical ideas to establish the foundations of the BOLD system. I recognised at once that the payoff structure of the game was a decisive influence on the power of a bidding system and so elected to build for IMPs scoring- the format used for a number of prestigious international trophies and my own preferred version of the game. During this phase I found the analysis, calculators, and tools at Richard Pavlicek’s site (rpbridge.net) invaluable, and it soon became clear that I would need custom software to proceed with my research at the requisite pace. At this stage I had the Cardquant software team build me a new analytical tool to help with bulk simulation work, and once that was complete the BOLD project truly gained momentum.

Since this is a mere preamble to the project and not a victory lap, I shall not indulge myself by expounding on the details of my research here. Suffice to say that after more than 2,000 hours of research conducted over more than a calendar year I had finally forged a version of the system whose structure remained sufficiently resilient to perturbations that I could move project BOLD to Phase 2.

Phase 2: Partnership Search

Here, dear reader, is where you come in (perhaps). With the core system development work under my belt I am looking to form a partnership to play the BOLD system in high-level competition, in order to achieve the stated mission. I am also looking for strong team mates who are interested in this project, and would be happy to have us on their team playing an unorthodox artificial system. Should there be enough interest, and should I form a strong partnership early on, then I would also be open to teaching team mates the BOLD system to form Team BOLD.

But I get ahead of myself. The remainder of this page is devoted to outlining a few select features of the BOLD system, and helping potentially interested parties to decide whether or not they should apply to join the project.

Whether you take an interest in the project or not, thank you for reading to this point. If you are interested in learning more about the BOLD system, and especially if you are considering applying to join the project, then read on!

BOLD System Overview

The BOLD system has been developed in three stages:

i) The core uncontested system meanings were developed, along with the internal logic of the system. ‘Natural’ systems consistently outperform artificial systems in the myriad of undiscussed situations that arise at the table, where meta-rules are often sufficient to save an experienced partnership from a misunderstanding. A key aspect of the BOLD system that is not evident on the system card fragment shown below is the creation of a family of meta-rules that enable our side to navigate undiscussed situations with aplomb.

ii) The system meanings when opponents intervene over our opening bids were developed.

iii) The system meanings after our opponents open were developed. Unlike the vast majority of systems played in the modern era, these situations have been given as much care and attention as when our side opens. The particular competitive structure that results is one of the most divergent features of the BOLD system.

At this juncture, it is only necessary to share publicly the uncontested features of the BOLD bidding system. What follows is only a quick outline, and not a presentation of the full system:

Opening bids

• 1♣/♦/♥/♠/NT are all artificial, light (typically 10+ HCP), unlimited and forcing. This liberates every bid of 2♣ and above as either a pre-emptive bid, or as an anticipatory bid to handle shapely hands with a major suit that would otherwise risk losing a crucial tempo if they were to be opened at the 1-level.
• 1♣/♦ are double transfer openings showing 4+ cards in the corresponding major. There are several canapé possibilities.
• 1♥ shows either a balanced hand with no 4-card major, or any hand with 54/5+5+ in the minors (not 5440 or 64 in the minors).
• 1♠ shows 8+ HCP and 5+4+ in the majors.
• 1NT shows a hand with an unspecified 6+ card minor, with no side-suit of 5+ cards, and usually no 4-card major.
• As we open so many hands in 1st and 2nd seat, 3rd seat 1-level openings have a higher point count floor, and we usually pass weak balanced hands.
• The 2/3-level opening structure varies according to both position and vulnerability:
  • NV 1/2 our emphasis is on maximizing the raw frequency with which we can take the first action whilst still communicating useful information to our partner. We may open almost every hand with either 4+♠, 5+♥, or 6+♣/♦ . I initially created the 2♣ opening in 2004 and deployed it whilst representing England at youth level in the 2004 Prague European Championships. For theoretical reasons that I will not elaborate on here, I consider it the strongest possible design for an ‘assumed fit’ pre-emptive bid, although I will share that extensive simulation work has confirmed the power of the opening.
    
  •  VUL 1/2 our emphasis is on raising the bidding level quickly whenever we have either a 6-card major, or 5-5 shape with at least one major. This enables our partner to bid vulnerable major suit games aggressively whilst his own hand strength remains ambiguous. Bidding marginal vulnerable major suit games is an essential part of any effective IMP strategy and this structure maximizes the frequency that partner can raise. Structures that are more liberal with shape when vulnerable afford much lower raising frequencies and are prone to offering the opponents a juicy penalty.
    
  • In 3rd seat we use intermediate strength two-level bids with 5+ cards in the suit opened and some side-suit conditions. As we open so light it is very unlikely for third seat to hold only 0-7 high-card points after two passes and so we do not waste bids to describe these rare weak hands.

   

The core uncontested meanings of the BOLD system are displayed on the WBF convention card below.

Strategic Advantages of BOLD

What follows is a brief outline of those strategic advantages of the BOLD system that should be readily apparent to a capable bidding theorist, or indeed any curious bridge player who has succeeded at a competitive level. I have deliberately excluded those strategic advantages whose proper understanding requires familiarity with my original theoretical work so what follows should not be interpreted as an apologetic treatise for the entire system. Our opponents can attempt to figure out counter-strategies for themselves!

1. We open the bidding with an exceptionally high frequency

In 1st seat NV we open around 80% of all starting hands, with 23%RF (Raw Frequency) opened with a pre-emptive bid of 2♣ or higher. VUL we are much more conservative with our pre-emptive bidding, but the light double transfer openings still introduce every hand with 10+ points and a 4-card major to the auction.

2. The 1 and 2-level opening bids consistently communicate useful suit-length information

When opener holds a 4-card major and opens 1♣/♦, he will find partner with at least 4-card support around 34% of the time, and 3-card support another 31% of the time. When he holds a 5-card major and opens 1♣/♦, he will find partner with at least 4-card support around 24% of the time, and 3-card support another 30% of the time. Since the system enables responder to distinguish between 3 and 4-card support on his initial call, we locate most major suit fits on the first round of the auction. This is especially valuable in competition, where many modern methods lose some major suit fits in two common classes of situation. Such methods require the partnership to open 1NT with a 5-card major and risk losing the fit whenever responder lacks the values to invite. They also open 1♣/♦ with most hands with only a 4-card major, often without guaranteeing length in either one or both minor suits.

When opener holds a 6-card minor without a 4-card major he opens 1NT, enabling partner to readily support him, sometimes holding as little as 3-2 in the minor suits. Responder will hold at least 3 cards in both minor suits around 31% of the time, and 3-2 either way in the minors another 14%, and so he is well placed to compete in opener’s minor suit even without knowing which one opener holds. Furthermore, he is relieved from the burden of showing a 4-card major prior to showing his minor suit support, and so can compete in the auction more effectively, whilst retaining the threat of a hidden 4-card major should the opposition choose to compete at the 3-level.  Many modern methods at the very least require responder to search for a 4-4 major fit before showing support and, even when responder lacks a 4-card major, he can support a potentially short minor suit open only rarely. Several systems at least bolster their 1♦ opening to guarantee 5 cards, but even a specific 5-card minor opening is not as well placed to compete as the BOLD 1NT opening showing an unspecified 6-card minor. With 5 diamonds in opener’s hand, he will find 4+ card support opposite (and the consequent 9+ card fit) only 24% of the time.

The useful suit-length information in the 1♥ opening is more subtle, as the restrictions on possible shapes are exploited by responder’s initial calls. With a 5-card major and a 4-card minor responder can guarantee an 8-card fit in at least one of his suits, as he can with 5-4 shape in the minors. With a 6-card minor he is free to pre-empt to the 3-level, usually in anticipation of at least 3-card support, and guaranteed to never find only a singleton opposite.

3. Responder’s initial calls are partitioned according to the contextual utility of information

The majority of responder’s calls fall into one of four classes, each of which has different priorities for the subsequent path of the auction:

i) He has a very weak hand and the partnership is likely to scramble for a fit in a low-level partscore. In most cases he shows this hand type hyper-efficiently by using the lowest available bid.

ii) The partnership has already found a fit, and he wishes to inform opener and communicate some additional useful information.

iii) He has a very strong hand, and wishes to initiate a relay auction to find the best spot for the partnership. In most cases he shows this hand type hyper-efficiently by using the second-lowest available bid. This sequence offers maximal space for the important task of selecting the proper final strain and of course for slam exploration, whilst posing a considerable threat should the opponents attempt to intervene with anything other than unusual suit length. Competent players will no doubt be aware that more slams are available when the point count is split relatively evenly between the two hands, both in terms of the absolute frequency with which such distributions are dealt, and in the cardplay- due to the fluid communication between hands facilitated by the surfeit of entries.

iv) He has an intermediate strength hand, and expects to face a competitive auction with a high frequency. Many of responder’s calls are utilized to express hands with intermediate strength, and are partitioned to communicate crucial suit-length information (often, but not always, in at least one major). If opener holds a 10 count, then he expects his partner to hold 7-11 HCP 46% of the time. With a 13 count, that number increases marginally to 48% of the time. This structure compares very favourably with the partitioning in the vast majority of modern systems. Consider the enduring popularity of the Forcing/Semi-forcing 1NT response to a 1♠ opening, which communicates almost no useful suit information whilst carrying the burden of up to 33% of responding hands!

Partitioning responder’s calls in this fashion enables our side to compete effectively when the balance of the points is roughly even between the two sides, to escape to a low-level partscore when our opponents have the advantage, and to initiate precise relay auctions when we are protected against frivolous intervention by the dual threat of a dominant point count and initial suit-length information.

4. Light, unlimited opening bids make frivolous competition unwise

With plenty of room remaining for our side to scramble a fit at the 1-level, and useful suit-length information presented by every opening bid, our opponents cannot afford to design their counter-strategy around either penalizing our light opening bids or destructively interfering with our auctions. Their best approach is to honestly communicate their hand-type to their partner having started on the back foot.

Precision club systems typically use light limited natural openings to facilitate competitive bidding, which allows responder to accurately assess whether the side belongs in a part-score, game, or slam early in the auction. Yet the cost of bucketing all of the remaining higher HCP hands into the 1♣ opening is well-known, as competent opponents consistently interfere with Strong club auctions, especially when their side is NV. Indeed, many Precision players consider the 1♣ opening the weakest part of their system for exactly this reason.

5. The relay structure has been designed with recurrent patterns and optimized for crucial decisions

Responder’s relay bid after the 1♣/♦ openings makes the relay captain declarer in the most likely final strain. Furthermore, I have designed a set of recurrent patterns that reduces the core shape-relays to a single side of paper, with comparable efficiency to Symmetric Relay. In fact, since most of our relays start at a lower level than typical Symmetric Relay auctions, we usually have more information at a lower step by the time the shape relay auctions terminate. There are several additional complications in the asking bid structures, as I have designed these structures to speciate among different slam types- a labour of love that took many months of research to get right! The memory load is worth the effort for IMPs strategy, since bidding and making a good slam missed at the other table generates such a substantial swing.

If this outline of some advantages of the BOLD system intrigues you, then you may be interested in applying to participate in the project. The next section details how you can do just that.

Should you Apply?

Broadly speaking, there are three criteria that you must meet in order to justify proceeding with an application:

1. You really want to play the BOLD system as written, and are excited to give it a try.
2. You either really want to play the BOLD system with me personally or want to play on the Project BOLD team with me.
3. You are keen to learn new bridge bidding theory from me.

If you bristle at the prospect of taking on the student role for an extended period whilst learning the principles that govern the system, then you certainly should not apply. I am a highly individuated man, with little interest in mass culture, and it takes a person of unusual quality for me to be happy working collaboratively. I have been open about my own evaluation criteria in the filters below, in the full knowledge that those filters will put off a lot of potential applicants. If what you infer about my personality from those filters resonates with you, then it is much more likely that we will be able to form an effective partnership.

Filter 1: Performance Record

You should have a track record of excellence in a competitive domain. Ideally this would be in bridge, but I am open to applications from exceptional high achievers in other domains whose bridge results have been more modest. Whatever the specific details of your background, I am looking for a partner from whom consistent international-level performance at the table is a realistic proposition.

I will look especially kindly on applicants from players who are elite-level defensive players. This is the one area of play which I have not dedicated serious time to so far in the Project, since the human partnership element is so crucial. My own defensive play is presently very far from the standard necessary to be competitive at the top level, and I would be glad to have a reciprocal relationship where I can teach you bidding theory and, in exchange, learn defensive technique.

Filter 2: Aligned personal values

I have witnessed first-hand the degradation of knowledge discovery and dissemination due to the politicisation of academic institutions. Entire academic fields have been corrupted, most notably sociology and psychology, as academics have been forced to adopt false premises in order to sustain their work. The conclusions derived from these false premises are, predictably, used to support the machinations of the incumbent elite. This insight troubled me so greatly that I left institutional academia to pursue my own research independently, at significant personal cost. Naturally I look to associate with people who recognize the severity and extent of this problem.

If you don’t believe in objective morality, and have made your life about making a lot of money and living in comfort, with little regard for the Aristotelian virtues, or any idea of a higher purpose, then please do not apply to Project BOLD. I study and write extensively on philosophical, scientific, and theological matters, and both reason and faith are central to my daily life. It is not possible for me to work closely with someone with a proudly atheistic outlook (anybody struggling with their faith is not excluded). My wife and I live according to a socially traditional model and, whilst I have no interest in imposing my way of life on others, I also have no patience for anyone who seeks to encroach on my own. There are almost no popular left-wing activist causes that I support, and it is likely that any successful applicant to Project BOLD would either be politically right-wing or a studious type with no interest in politics.

Filter 3: Killer instinct

You should be dedicated to winning, even when it hurts the opponent to lose.

This may sound callous, but it is necessary when you are the underdog (as we will be, sooner or later), and is in fact the most moral way to play. If you are the type of player who, when crushing an opponent in a round robin match, releases the gas pedal on the last few boards because you feel sorry for your opponents then you not only let your team mates down, but you let every other team down participating in the competition.

It is essential that we all maintain a high level of physical fitness through regular exercise and personal discipline, as it offers us the best opportunity to compete well for an extended period. Your ‘A’ game may well be inferior to a world class opponent’s ‘A’ game, but if after several days of play you can still play your ‘A’ game, it will be good enough to beat that same opponent’s ‘C’ game. If you would not be comfortable committing to a healthy lifestyle for the duration of the project (I would recommend it long-term!), then you should not apply.

Finally, committing to the preparation and warm-down schedules necessary for peak performance will sometimes mean declining social invitations between matches. Certain parties will no doubt be offended, and label you as ‘intense’, or use some other derogatory term to describe you. If winning trophies matters less to you than rocking the social boat, you aren’t the right fit for this project.

IMPORTANT NOTE: I am completely opposed to any form of cheating, abusive behaviour, or gamesmanship, whether at the bridge table or away from it. None of the above should be interpreted as endorsing such practices.

If you are that rare individual who meets all of the criteria above, then I invite you to apply to participate in Project BOLD!

The Application Process

Phase 1: Complete the BOLD application form

The BOLD application form is designed to be a screening step which filters out low-effort candidates, as well as minimizing wasted time interviewing interested applicants who would not be a good fit for the project. You can find the application form linked here.

Phase 2: Attend a video interview

In the event of a high level of interest in Project BOLD then I will endeavour to video interview every viable applicant personally. The purpose of the video interview is to assess intangibles, such as personal rapport, as well as address any questions that have arisen for either party in the initial search process. If the video interview goes well, and I am happy to explore working together, then I will invite you to participate in a partnership bidding test.

Phase 3: Participate in a partnership bidding test

Should you accept my invitation, then I will share with you a PDF introduction to the system so that you can familiarize yourself with the core principles, and examine the continuations after a 1♦ opening in detail. We will then practice continuations following a 1♦ opening together using the applications available on Cuebids and BridgeBase Online. If both parties consider the partnership to be a promising proposition, then we move to Phase 4.

Phase 4: Prepare for our first tournament

In this phase I will share the rest of the BOLD system with you, and train you in the original theoretical framework that I have developed in order that you can wield the system with finesse. Once we are consistently scoring well against bidding accuracy metrics, we will be ready to compete as a partnership.

The BOLD system is Category 3 for WBF competition, and the EBU is open-minded enough to permit all Category 3 systems in their Level 5 events. Whilst online events in these classfications are rare, they do come up and our fledgling partnership should take advantage of any such opportunity. The real tests begin once we are able to bring BOLD to compete in high-level live events.

I will note that it is extremely unlikely that we will be able to deploy the system in an ACBL event, even at the Open+ chart classification, without extensive consultation with the ACBL, and at least some modifications to the system.

THAT COMPLETES THE PLAN!

If you are excited by the scope of the project and neither the level of commitment involved nor the prospect of working with me has deterred you then congratulations, you are ready to apply! Proceed to the application form by pushing the button below.