What 2,259 real rummy rounds tell us
Original, anonymous data from 13-card Points Rummy games between friends on RummyDen: how rounds end, how often declarations fail and what winning hands look like.
Most rummy advice is passed from player to player: "always drop a bad hand", "jokers win games", "never hold high cards". We wanted to see what actually happens when friends sit down at a table. RummyDen stores the outcome of every finished round so it can show scores and history, so we counted. This page summarises 2,259 rounds of 13-card Points Rummy played by real groups on RummyDen between May and October 2026.
Everything here is aggregate and anonymous: no names, room codes or individual players are reported. The method and its limits are at the end of the page, and we will refresh the figures as more groups play.
The numbers at a glance
| Measure | Result |
|---|---|
| Rounds analysed | 2,259 (3,413 losing seats) |
| Rounds won by a valid declaration | 76.5% |
| Rounds that ended because everyone else dropped | 19.0% |
| Declarations that turned out to be invalid | 8.4% (158 of 1,885) |
| Median length of a round | About 4½ minutes |
| Median points held by a losing player after a valid declaration | 26 |
| Losing hands that hit the 80-point cap | 4.3% |
| Winning hands that used at least one joker | 98.1% |
| Winning hands built as four groups (3 + 3 + 3 + 4) | 94.1% |
How rounds end
Three things can end a round of Points Rummy: someone makes a valid declaration, everyone but one player drops, or a declaration is checked and found invalid (on RummyDen that ends the round: the declarer takes the 80-point penalty and the round goes to the remaining player with the fewest points).
Roughly one round in five never reaches a declaration at all. Most of those are two-player games where one player looks at their opening hand and drops, which ends the round immediately. Our guide to rummy for 2 players breaks down the two-player rounds.
How long does a round take?
We timed 1,780 rounds by the gap between one result and the next at the same table, minus RummyDen's 15-second countdown between rounds (pauses longer than 20 minutes were left out). The median round took about 4½ minutes; half of all rounds finished in 3 to 6½ minutes, and nine in ten in under 8½.
| Table | Rounds timed | Typical round (median) |
|---|---|---|
| 2 players | 1,207 | 4 min |
| 3 players | 316 | 5 min |
| 4 players | 112 | 5 min |
| 5 players | 121 | 6 min |
Rounds that someone won by declaring took a median of just under 5 minutes; rounds that ended because everyone else dropped took under 2. The turn timer matters too: tables on a 30-second timer finished declared rounds in about 4 minutes, while 90-second tables took about 6. If your group wants to fit in ten rounds in an hour, a 30- or 45-second timer is the realistic choice.
About one declaration in twelve is invalid
Of 1,885 declarations, 158 were invalid — 8.4%. That matters because an invalid declaration is the most expensive outcome in Points Rummy: the declarer takes the full 80-point penalty, more than three times the median losing hand. Looking at the arrangements those players submitted, the problems fall into three groups:
| What was missing | Share of invalid declarations |
|---|---|
| No pure sequence at all | 33.5% |
| A pure sequence, but not a second sequence | 28.5% |
| Enough sequences, but one group that was neither a sequence nor a set | 25.3% |
| Not classified by our check (see method) | 12.7% |
Here is what each mistake looks like. These are illustrations, not real players' hands.
The common thread is checking the two sequence rules last, after the hand "looks" finished. Before you place your 14th card in the finish slot, count: one pure sequence? a second sequence of any kind? If you are unsure about a hand, the free rummy hand validator applies the same table checks. Our guide to declaring in rummy walks through the full routine.
How many points do losing players hold?
When someone declares validly, every other player who has not dropped is scored on their unmatched cards, capped at 80. Across 2,457 such losing hands, the median was 26 points; half of all losing hands scored between 13 and 45.
Two things stand out. First, 17.7% of losing hands held fewer than 10 points — these players were one or two cards away from declaring themselves. Second, a full count is rare: only 4.3% of losing hands reached the 80-point cap. High cards cost the most per card, but low cards (2-9) were stranded far more often and made up about 58% of the points losing hands held, as the section on stranded cards below shows. Ace, King, Queen, Jack and 10 count 10 points each; jokers count zero.
What this says about dropping
A first drop costs 20 points and a middle drop 40. Since the median losing hand is 26, a first drop is not automatically cheaper than playing on — you also might win and score zero. But a first drop does protect you from the long tail: about a third of losing hands scored 40 or more. Players seem to know this. 19% of losing seats were first drops, and only 4.4% were middle drops, so most players who leave a round do it before investing a draw.
| Players at the table | Losing seats | First drop | Middle drop |
|---|---|---|---|
| 2 | 1,584 | 16.7% | 5.3% |
| 3 | 771 | 21.1% | 4.5% |
| 4 | 393 | 21.6% | 2.5% |
| 5 | 520 | 20.0% | 1.7% |
| 6 | 145 | 22.1% | 8.3% |
First drops are a little more common at bigger tables, where more opponents means more chances that someone declares before a weak hand recovers. Six-player tables are a small sample, so treat that row with caution.
Which cards get left stranded?
Common advice says high cards are the danger: an unmatched King costs 10 points, a 2 costs only 2. The losing hands in our data show something else. We counted every natural card that losing players were holding when someone declared, and how often it ended up outside a valid group:
Low cards were stranded far more often than high ones — 46% of 2s to 9s against 32% of 10s, picture cards and aces. Losing players were also holding fewer high cards than low ones by the end of a round (1,847 Kings against 2,538 twos). That pattern fits players who already throw away loose high cards and keep cheap low ones "just in case".
Counted in points, low cards still cost losing players more in total. Stranded 2s to 9s added up to 44,627 points (8,819 cards), against 31,953 points (3,195 cards) for stranded 10s, picture cards and aces: a 58/42 split, even though each high card is worth more.
What winning hands look like
We looked at the final arrangement of 1,726 winning declarations.
- Four groups is the norm. 94.1% of winning hands were split 3 + 3 + 3 + 4. Long runs are rare: only about 5% contained a group of five cards or more.
- Sets are common but not required. 74.3% included at least one set; a quarter won with sequences only. 17.4% used a four-card set.
- Two pure sequences happen often. 36.7% of winners had two or more pure sequences, which leaves the jokers free to finish the other groups.
- Jokers are almost universal. 98.1% of winning hands used at least one printed or wild-rank joker, and 63.6% used a printed joker. Two jokers was the most common count.
The joker figure reflects RummyDen's default rules: two decks with printed jokers, plus every card of a randomly chosen wild rank — and on RummyDen a wild-rank card counts as a joker even in its natural place. With that many jokers in play, holding a joker for a sequence or set is usually more valuable than breaking up a natural group to avoid one. Our jokers guide explains how they can and cannot be used.
A typical winning hand
This hand, taken from the data with the wild rank set to 7, shows the usual shape: a pure sequence, a set, a second pure sequence and a four-card impure sequence completed by a printed joker.
What to take from this
- Build the pure sequence first and confirm a second sequence before declaring. Those two checks would have prevented roughly six in ten invalid declarations above.
- Think in four groups. Three groups of three and one of four is the shape almost every winning hand takes.
- Use jokers to finish, not to start. Winners typically used two or three — almost always alongside at least one natural pure sequence.
- Drop early or play it out. Middle drops were rare for a reason: once you have drawn, the cost doubles.
New to the game? Start with how to play or the full rules, then try a round with friends in a private room.
Citing these figures
You are welcome to quote these statistics in articles, videos, forum posts or teaching material. Please credit "RummyDen" and link to this page (rummyden.com/rummy-statistics) so readers can see the method and the date of the data. Spotted something that looks wrong? Tell us through the contact page and we will check it.
How we counted
- Source. Completed rounds of 13-card Points Rummy stored by RummyDen's game server between 18 May and 25 October 2026. Almost all (99%) were played in September and October 2026.
- Rules. Nearly every round used RummyDen's defaults: two 54-card decks, printed and wild-rank jokers, a 20-point first drop, a 40-point middle drop and an 80-point cap. Other tables and house rules can produce different numbers.
- Exclusions. We removed rounds nobody actually played — when every player had left, the turn timer kept dropping them automatically — plus all rounds at tables with practice bots and rounds involving known test accounts.
- Hand analysis. Winning and invalid hands were classified by a simple check that labels each group as a pure sequence, impure sequence, set or neither. It does not replicate every rule the game server applies, which is why 12.7% of invalid declarations are listed as not classified.
- Stranded cards. In losing hands scored on their unmatched cards, we counted every natural card and whether its group was a valid sequence or set. Printed jokers and wild-rank cards are left out because they score nothing. The points figures multiply each rank's stranded cards by its card value.
- When losing scores were taken. Until 25 October 2026 the stored score of a losing hand was taken when the winner's declaration was checked, before the other players' 60-second arrangement window. Players can still regroup in that window, so the final scores shown at the table can be lower than the losing-hand figures here.
- Who played. Groups of friends and family, mostly at two-player tables (70% of rounds). They are casual social games, not tournaments, and the figures describe these tables rather than rummy players in general.
- Privacy. Only aggregate counts are published. RummyDen is free social play with scorekeeping chips that have no cash value; see our privacy policy.