Deal a Gin Rummy hand and look at what you are holding. Most players sort by suit, spot the obvious pairs, and start planning around their highest cards because those are the ones that hurt if they end up as deadwood.
That is backwards. The cards that decide the hand are the ones nobody looks at twice: the fives, sixes and sevens sitting quietly in the middle.
Why a six is worth more than a king
Think about how many different melds a card can belong to.
A king can be part of a set of kings, or a run at the very top of a suit - queen-king, jack-queen-king. It connects downward only. An ace is the same in reverse. These cards have exactly one direction of travel.
A six can be part of a set of sixes, or a run running four-five-six, five-six-seven, or six-seven-eight. It connects in both directions, at multiple offsets. In raw combinatorial terms a middle card belongs to roughly twice as many possible melds as an extreme one.
Now add the deadwood penalty. Face cards count ten. A six counts six. So the middle card is simultaneously more likely to become useful and less costly if it does not.
That is the entire argument, and it leads somewhere counter-intuitive: when choosing between two equally connected cards, keep the middle one and throw the extreme one, even when the extreme one is currently part of a pair.
The early-discard rule that follows
High cards should leave your hand early, while they are still relatively safe to throw. As the hand progresses, opponents' holdings narrow and every discard gets more dangerous - so the time to shed a king is turn two, not turn ten.
This creates a specific rhythm to a well-played hand:
- Turns 1-3: dump unconnected high cards. You are reducing deadwood exposure while the pile is still uninformative.
- Turns 4-8: build. This is where middle cards earn their keep, because the draws you get will connect to them.
- Turns 9+: defend. Throw what is demonstrably safe, and start counting toward a knock.
Most losing hands come from inverting the first and last phases - holding high cards "just in case" early, then being forced to throw them late when they are exactly what the opponent wants.
Pairs are not as valuable as they look
A pair feels like progress. It is one card away from a set, and holding it costs nothing extra.
Except it does. A pair of kings is twenty points of deadwood being carried on the hope of drawing one of two remaining kings - and if an opponent holds one, there is only one. Meanwhile a five-six of the same suit is eleven points chasing four useful cards: the four, the seven, and the other two of each rank for a set.
The general principle: connected sequences beat pairs, because a sequence can extend from either end while a pair can only become a set. Two cards in sequence are typically drawing to roughly twice as many outs as a pair of the same total value.
Counting your outs before committing to a holding is exactly the accounting that the trick-taking games train, where the number of cards that can still hurt you shrinks with every trick played. Anyone who tracks what is left when they play Hearts online is doing the same arithmetic against a different unknown.
What the opponent's discards tell you about their middle
Because good players shed extremes early, a player throwing middle cards in the first few turns is telling you something specific: their hand is concentrated at the top or bottom of the suits, which usually means they are building sets rather than runs.
Conversely, a player who has thrown nothing but face cards for four turns is building runs in the middle, and the cards you most want to avoid feeding them are precisely the ones you want to keep anyway.
This is the two-player version of a much broader skill - inferring a hidden hand from public actions. The fuller-table version of the same engine puts more information on display: when you play Rummy online with several opponents the discard pile becomes a richer record, and in the partnership meld game - play Canasta online - the pile itself becomes a contested object rather than just a signal.
Where luck actually lives
Gin Rummy has a reputation as a luck game among people who have lost at it, and the honest answer is that the deal matters and the draw matters, but neither matters as much as the sequencing.
The reason is that you see a large fraction of the deck over a hand, and you choose from two sources every turn. A player making systematically better keep-or-throw decisions converts a mediocre deal into a knock far more often than the card distribution would suggest.
That places it alongside games where randomness is bounded rather than dominant. When you play Tetris online the piece sequence is a shuffled bag that guarantees one of everything every seven pieces, so the losses come from stack management rather than the draw. Gin Rummy's deck offers a similar guarantee in a looser form: over enough turns, the cards you need pass through.
Contrast that with a game where the randomness genuinely is the engine - play Yahtzee online and no amount of skill changes what the dice show; you only choose where to put the result.
Across the variants
Oklahoma Gin ties the knock threshold to the first upcard, which sharpens everything above - on a low-threshold hand the value of middle cards rises further, because you need very tight deadwood. Straight Gin removes knocking entirely and demands a complete hand, which makes flexibility worth more than safety and turns the middle-card principle into the whole strategy.
From here
For one session, throw every unconnected card above a nine within your first three turns, regardless of how promising it looks. It feels reckless and most players find their deadwood at the end of hands drops sharply.
The rules hub covers melding and knocking, the luck-versus-skill FAQ gives the short version of the section above, and the glossary defines deadwood and the rest. Play on the main game or try today's daily deal, where everyone gets the identical shuffle.
For how the first upcard reshapes a hand before you have played a card, see the companion piece.