How Often Do Bridge Conventions Come Up?

Every newer player runs into the same question: there are hundreds of conventions, so which ones are worth learning first? The usual answer is somebody's opinion. Here is an answer with numbers behind it.

We dealt 2,000,000 random hands by computer and bid each one through its opening bid and the first few calls that followed, using ordinary Standard American methods. A convention is counted when both things happened: the auction actually reached the point where it applies, and somebody held the hand to use it. A convention you never got to use because the opponents opened first doesn't count.

"Per session" assumes a 24-board club game. Convention names in green link to the NVBA handout on that convention.

Competitive bidding is where the action is

Overcalls, weak twos, the limit-raise cuebid, takeout doubles and negative doubles are all about fighting over a partscore, and together they come up roughly thirteen times a session. Nothing else is close.

The limit-raise cuebid deserves a second look, because it's easy to miss for a bid with no glamour attached to it. Whenever your partner opens or overcalls and you hold 3+ card support with 10 or more HCP, the correct call is a cuebid of the opponents' suit — showing a limit raise or better before you've said anything else about your hand. It doesn't matter which of your opponents supplied the suit you're cuebidding, or whether your partner is the opener or the overcaller; the requirement is the same. Skip the cuebid with that hand and you've told partner something too: you don't hold it. The handout covers both auctions in detail.

That has a practical consequence. A pair that handles contested auctions well gets to use that skill on well over half the boards they play. The same pair's slam agreements might come up once or twice in an evening.

Stayman comes up a little more than once every two sessions

This surprises nearly everyone, since Stayman is the first convention most players learn. The explanation isn't Stayman itself. A 1NT opening happens on only about 11% of deals, and partner then needs a four-card major plus about 8 points to use Stayman at all. Multiply those together and you get roughly one deal in forty.

Transfers come up slightly more often, because any five-card major will do and partner's strength doesn't matter.

Rare doesn't mean useless

What the table is genuinely good for is deciding what to learn first, and what to spend your limited partnership discussion time on. If you and your partner have one evening to agree on methods, the top of this list will repay you far sooner than the bottom.

A convention can work on the deals where you never use it

One more caveat, and it cuts against everything a frequency table can measure. Some conventions earn their keep precisely when they don't come up, because agreeing on a meaning for one bid sharpens every neighboring bid at the same time.

Bridge players call this a negative inference: what partner's bid rules out can matter as much as what it shows. Five examples:

• The limit-raise cuebid. Once a cuebid of the opponents' suit is available to show 3+ card support and 10+ HCP, anything else you bid instead — whether partner opened or overcalled — denies that hand. Since this is the entry that shows up most often on this list, it is also the negative inference you'll draw most often.
• Bergen raises. Once 3♣ and 3♦ are the ways to show four-card support, a simple raise of your major denies four trumps. You learn that every time partner raises, not only on the deals the convention gets bid.
• Support doubles. When the double was available and partner didn't use it, partner does not hold three-card support for you. The convention did its work by staying in its box.
• Splinters. A plain raise to game denies a singleton, so partner can rule out a whole class of slam hands without asking a question.
• Michaels. A natural two-level overcall denies the two-suiter, and partner stops guessing about it.

The same reasoning carries into the play, which is where it is easiest to overlook. When responder never bids after an overcall, that silence limits both their pattern and their high-card strength — and you are still holding that information when you make your first discard on defense. A pair with clear agreements collects inferences like these on nearly every board.

The richest inferences of all come from the opening lead. Say partner raised your suit during the auction and then led a different one. The raise was honest; the lead says partner had something better — a sequence, a singleton, an ace-king combination — or holds the ace of your own suit and didn't want to lead it at trick one.

When partner does lead your suit, the card itself carries the message, and what it means depends on the auction. If partner raised you, a middling spot card says partner holds no honor in the suit at all, while a low card usually promises something, typically three to the ten or better. If partner didn't raise you, a fairly high card is usually shortness — a doubleton or a singleton. The same-looking card tells a different story depending on what partner did, or didn't do, a few calls earlier.

In none of that was a single convention involved. Conventions simply add more signposts to an auction, and a defense, that are already full of them.

That is the real answer to a low number in the table. A convention on the rare end of this list may still be working for you on every board you play, silently, through what partner's other bids now promise and deny. The table counts the times you get to bid it. It cannot count the times it told you something by its absence.

How some of these rows relate to each other

Bergen raises and limit raises divide the same raising problem rather than duplicating it: Bergen's 3-level jump covers both the 3-card "mixed" raise and the constructive 4-card raise, while a traditional limit raise reserves the 4-card jump for the 10–12 point invitational hand. A pair that plays Bergen still needs some other way to show a plain invitational raise, so the two rows are a division of labor more than a straight either/or.

Jacoby 2NT and splinter bids are the cleaner overlap: a single hand type — four-card support and game-forcing values — split according to whether responder holds a short suit. Taken together, that game-forcing raise of a major comes up on about 0.4% of deals.

How we counted, and where the numbers are soft

The simulation is deliberately plain, and it is honest about being an estimate. Seven points worth knowing:

• The computer bid every hand the same way: five-card majors, a strong notrump and standard responses, with no partnership agreements beyond that. Change the system and the table changes with it. Acol's weak notrump and four-card majors put a different set of auctions in play, and Precision's strong club takes the strongest hands out of the one-of-a-suit openings altogether.
• Weak twos require a genuine six-card suit, not a token one — the simulation insists on real honor strength in the suit itself, the way a sound weak two does at the table. A looser standard would inflate that row well past what you'd actually see.
• Whichever seat is first, going around the table from an arbitrary starting point, to hold an opening hand is the one who opens. That's a fair model of how an auction actually unfolds — but it also means a convention tied to the opening bid, like weak twos, gets more chances per deal than a convention tied to one specific seat's reaction to a fixed auction, like an overcall.
• Each deal is bid once and the auction stops after the first few calls. Conventions that live later in the auction — fourth suit forcing, new minor forcing, control bids — are therefore floors rather than true rates.
• Only the player directly over the opening bid is considered. Balancing-seat overcalls, doubles and 1NT defenses are real and are not counted here, so overcalls, takeout doubles, the limit-raise cuebid and 1NT defense all run a bit low.
• The 1NT opening is a strict 15–17 balanced hand. Pairs who open 1NT with a 5-4-2-2 shape or a wider range will see more Stayman, transfers and Lebensohl than this.
• The count is of holding the hand, not of the convention changing the result. A takeout double you would have found without any agreement still counts.

What the numbers are not is noisy. The 2,000,000 deals were run as five separate batches of 400,000, and no figure in the table moved by more than a quarter of a percentage point from one batch to the next. Rounding to one decimal place is as fine as the estimate deserves, and every figure is steady at that precision.

For one of the underlying odds this rests on: a hand contains a five-card or longer major about 35% of the time. Our simulated opening bids matched that published frequency closely, which is what gives us confidence in the rest.

Where to go next

• The NVBA lecture list has handouts on most of these conventions.
• Filling out a convention card, section by section.
• The practical Alert guide: which of these need an Alert, and which don't.
• Information for newer players, including where to play and learn.