Shanghai Rummy Strategy

Sort for the contract

Most rounds are lost in the first ten seconds, by sorting the hand out of habit. A round asking for two sets and a round asking for two runs want completely different cards, and the ones you would cheerfully throw away in the first are the ones you need in the second.

Look at the contract before you look at your cards. In a set round, group by rank and ignore suit entirely. In a run round, group by suit and hunt for neighbours. In a mixed round, build the run first: a run needs four particular cards in one suit, while a set needs three interchangeable ones, so the run is the harder half and deserves the early commitment.

Keep flexible cards

A card that can only land in one place is worth less than a card that could land in two. This is why middle cards outperform edges.

Aces and kings are the opposite case. They join exactly one run each because there is nothing beyond them, and they cost 15 and 10 points if you are holding them at the end. A low connector is cheap to keep and easy to place. A high edge is expensive and rigid.

The buying decision

In the table version of the game, buying is the single biggest source of both wins and blowouts. Buying gives you the discard plus an extra card off the pile, and you do not discard, so a buy grows your hand by two in a game whose entire object is to run out of cards.

The arithmetic is unforgiving. Two extra cards you cannot place are worth roughly ten to fifty penalty points depending on what they are. So the test is not "is this card useful" but "does this card finish something".

• Buy to complete, not to collect. A card that finishes a meld pays for itself. A card that gives you a promising pair usually does not.
• Buy later rather than earlier. Early in a round you do not yet know which melds will come together, so you cannot tell which cards matter. Late in a round you know exactly what you are missing.
• Count what is left. Your buys are capped per round. Spending all of them by the third turn leaves you with nothing when the round is decided.
• Watch who else wants it. If a player has been visibly collecting hearts, taking a heart away from them can be worth more than the card is worth to you. That is the one good reason to buy something you do not need.

When to spend a joker

A joker is both the most useful card in the deck and the most expensive to be caught holding. Those two facts pull in opposite directions, and the resolution is timing rather than principle.

Early in a round a joker is an option: it can complete whichever meld turns out to be short, and you do not yet know which that will be. Holding it costs nothing while the round is undecided. Once the round starts closing, that flexibility is worth less than the penalty, so commit it and get it out of your hand.

Two limits shape the decision. A meld can never lean on more jokers than natural cards, so a joker is only useful where you already hold most of the meld. And in the long contracts, a run of seven or ten in a single suit, a joker often is not a convenience but the only realistic way to finish.

Reading the discard pile

Every discard is a small piece of information, and over a round they add up to a readable picture.

• What a player takes tells you what they need. Someone lifting a nine of hearts off the pile is building hearts around the nine or collecting nines.
• What a player throws tells you what they have given up on. A run of discarded face cards means nobody at the table is chasing high cards.
• A sudden change of pattern is the loudest signal there is. Players shed their odd cards in the turn or two before they lay down, so a player who suddenly discards two unrelated cards is about to go down.

That last one is the most actionable. When you see it, stop feeding the pile and start thinking about your own exposure.

Discarding to deny

Every card you discard is a card you are offering to the table. When two discards are equally useless to you, the tiebreak should be which one is less useful to everyone else.

This costs nothing and it compounds. If a player has taken two spades from the pile, your spare spade is worth more thrown away later than now. Break up the suits and ranks people are visibly collecting, and prefer to discard into the gaps nobody is working in.

The exception is when someone is one turn from going out. At that point denial is a luxury and shedding your expensive cards is the priority. A joker held back for strategic reasons is 25 points when the round ends without warning.

When to lay down

Laying down the moment you can is the default and usually correct, because until your contract is on the table you cannot get rid of anything except one card per turn. Being down is what turns a good hand into an empty one.

There is one real argument for waiting. The instant your melds are on the table, everyone else who is down gains somewhere to dump their spare cards. If an opponent is clearly a card or two from going out, laying down can hand them exactly the place they needed. Late in a round, holding for a single turn is sometimes worth it. Early in a round it almost never is, because you are giving up several turns of laying off to avoid a risk that has not materialised yet.

Watching the exits

Card counts around the table are public information and most players ignore them. When somebody drops to one or two cards, the round is about to end whether you are ready or not.

That is the moment to stop optimising and start reducing. Play your highest-value dead weight, break up the near-meld you have been nursing, and accept that you will not win this round. The difference between finishing a round on 20 points and on 70 is usually made in these final two turns, not in the twenty before them.

Cutting your losses

The score does not record how close you came. A hand holding two jokers and an ace is 65 points whether it was one card from a perfect meld or complete nonsense.

So when a round has clearly got away from you, change objective. Stop playing to go out and start playing to finish cheaply: shed aces and jokers, keep the low cards, and take a small score instead of gambling on a finish that needs two specific cards. Over ten rounds, the player who never posts a 70 usually beats the player who occasionally posts a zero.

Round by round

The ladder changes character as it climbs, and the right approach changes with it.

• All-set rounds. Sort by rank. Pairs are your currency, and because the deck is doubled, duplicates are welcome rather than wasted.
• All-run rounds. Sort by suit and commit early to two suits rather than collecting fragments of four. Fragments in four suits is the classic losing hand.
• Mixed rounds. The hardest to hold in your head. Build the run first and let the set assemble around whatever is left, since sets are far easier to finish late.
• Long-run rounds. A run of seven or ten is a different problem: you need a deep block of one suit, so take that suit from the discard pile aggressively and plan on a joker or two to bridge the gaps.
• The final round. The contract eats almost the whole hand, so there is nearly nothing spare. Take the discard whenever it touches a suit you are building, and expect the round to be decided by who draws well.

Playing the version here

The game on this site plays the seven-round ladder against three computer opponents, and two of the strategies above do not apply: there is no buying and no Shanghai call. You take the discard only on your own turn.

That makes the discard pile the whole game. Without buying, the only card you can ever take out of turn is none, so every decision reduces to two: is the face-up card better than an unknown one, and which card do I give away. Reading the discards and denying usefully matter more here than in the table game, not less.

The other difference worth planning around is the deal. It grows from eleven cards to fourteen across the last three rounds, so the late contracts leave you almost no spare cards. By round seven you are working with two cards of slack, and a single wasted draw is expensive.

Play a round to try this out, or take today's challenge, where everyone gets the same deal and the only variable is how well you play it. If you need the rules first, start with how to play.