Killer Bridge Leads

Two tools for one skill. Practice deals you a hand and an auction from a bank of 100,000+ engine-bid deals — you pick your lead, the evidence grades it, and the difficulty follows you. Analyze takes a hand you actually led from — typed, from a BBO screenshot, or from a Handviewer link — and ranks every lead by how often it beats the contract.

Tap a card. Each problem is a deal the engine dealt and bid, chosen because the lead matters, with hundreds of hidden layouts behind it — every one re-bid by the engine to reproduce the auction call for call and solved double-dummy. Tens of millions of engine-bid deals, all told. Your verdict says whether your lead was right and by how much, and lets you browse the layouts. The ladder adapts: five right in a row moves you up, and your rating tracks the climb. Beginner hands are free without limit; the subscription unlocks intermediate to expert.

Enter or import the hand and auction, press Start. Your 13 cards stay fixed; the other 39 are dealt thousands of times, only deals the engine would bid exactly as you heard are kept, and each is solved double-dummy. Every lead is ranked by beat rate and average tricks, with its margin of error. Change a call, edit any seat’s constraints, run more layouts, browse every deal — free, no account. Every practice verdict opens here in one tap.

Other lead simulators deal hands that merely fit a written range for the auction: “12–19 with five spades”. Killer Bridge Leads keeps only layouts a real bidding engine bids exactly the way you heard it — call for call, bids made and bids not made. That is the information a defender actually has at the table, no more and no less, and it’s why these verdicts hold where range-based simulations drift. Every accepted layout is one tap away in the deal browser: check our work.

We tested it. In a 10,000-hand study the sampling reproduced the published structure of expert lead data — any lead works on about one deal in four, our 24% against the published 25%. A bot leading purely from these verified layouts scores 80 on the double-dummy lead-accuracy scale used in cheating investigations (95% CI 77–84): statistically indistinguishable from world-class experts at ~81, and below the 85 that flags a cheat. Every number ships with its margin of error, and a winner is only called clear when the difference is statistically solid.

Yes, and it’s measurable. On about a quarter of deals any lead works; everywhere else the lead matters, and in notrump it decides the hand — among beatable 3NT contracts the best lead wins about 83% of the time against 69% for a random card. The safest-looking card and the winning card differ on one hand in eight. Across verified 3NT pools, a sound textbook lead that ignored the auction ran a median 3 points behind the best lead, and up to 37. That gap is the curriculum.

Two provably fair samplers. The dealer draws the unseen hands exactly uniformly within the ranges the engine states for the auction; verification then keeps only deals the engine re-bids call for call. Practice explores that verified set with symmetric card swaps, so its layouts are uniform over everything the auction allows, with any residual correlation measured and folded into the confidence intervals. If a holding surprises you, that is the engine telling you what the bid really promised.

Because the conditioning differs, deliberately. Books deal from broad written ranges; we keep only deals a disclosed engine bids this exact way. On strong sequences that means a narrower, stronger declarer, so contracts make more often and every lead’s beat rate sits lower — on Bird & Anthias’ Hand 5 (1♠–2♠–4♠) by about a third of a trick, the signature of a stronger declarer pool, not a solver difference. Both are correct answers to different questions. Prefer the book’s framing? Switch verification off and type their ranges.

Double-dummy means both sides play perfectly seeing all 52 cards after the lead — the research standard behind every opening-lead book, with a known bias: it slightly flatters attacking honour leads and undervalues passive or deceptive ones, because declarer never misguesses. Real tables add error on both sides.

Free: the whole analyzer — manual entry, BBO import, auction-driven constraints, the full ranked table, the deal browser, up to 5,000 layouts — and unlimited beginner practice. Earn promotion and you get 30 intermediate-to-expert hands on us. The subscription — $5.99 a month or $49.99 a year, 7-day free trial — unlocks every level, the practice filters and the misses list. Checkout is by Lemon Squeezy, which handles VAT and invoicing worldwide.

Yes, and it’s free: import the hand on lead from a BBO screenshot or a Handviewer link for any finished board. The deal comes straight from BBO, nothing is retyped.

No. It runs in the browser on phone, tablet or computer; the free tier needs no account. On a phone, “Add to Home Screen” installs it like an app.

Hands never leave your device: dealing, bidding and double-dummy solving all run in your browser. The only network calls fetch the practice bank, check the licence and open checkout.

I’m Jorrit Schäfer. I played junior international bridge for Belgium from 2008, including three European Youth Championships, and founded the Imperial College London bridge club, captaining it to the UK university title (the Portland Bowl) in 2013. By profession I’m a data scientist — a PhD in bioinformatics, now a director of advanced analytics in the pharmaceutical industry. Killer Bridge Leads exists for two reasons: opening leads were the part of the game where everyone, me included, was going in mostly blind, with no way to quantify what one lead was worth over another; and there was no dedicated tool for training them. Now there is. You can find me on LinkedIn. The bidding is powered by EPBot / BBA, created by Edward Piwowar and used here with his kind permission; the double-dummy card play comes from DDS, Bo Haglund’s open-source solver. Our thanks to both.