Farkle Strategy

The whole game is one question, asked over and over: roll again, or bank? Here are the numbers behind the answer.

How often you Farkle

Everything starts here. These are exact figures, counted over all possible rolls under the standard rules used on this site (singles, three of a kind and above, straight, three pairs and two triplets):

Those Farkle percentages assume three pairs and a straight both score. If your table does not play them, six dice Farkle 3.1% of the time rather than 2.3%.

The shape of that table is the whole strategy. Six dice almost never Farkle; two dice Farkle nearly half the time. Every die you set aside makes your next roll more dangerous, which is why keeping the maximum number of scoring dice is so often wrong.

The break-even point

Rolling again risks everything accumulated this turn. Call that total T. A single further roll is worth it while the expected gain beats the expected loss, which works out as:

roll again while T < average gain × (1 − Farkle chance) ÷ Farkle chance

This is a deliberately simple model: it values exactly one more roll and ignores both hot dice and the chance to keep rolling after that. Treat it as a floor, not gospel — see the caveat below.

In practice that collapses to a few rules of thumb worth memorising:

• Six or five dice: roll. Nearly always correct.
• Four dice: roll unless you are sitting on 900 or more.
• Three dice: roll under 300, bank over it. This is the decision that comes up most often.
• Two or one die: bank. Almost always — with one important exception.

The hot dice exception

The table above ignores the biggest prize in the game. If you are down to one or two dice and they all score, you get all six back and carry your points forward. That upside is large enough to change the decision when your turn total is modest.

A single die scores one time in three, and the reward is a fresh set of six that will Farkle only 2.3% of the time. With 300 on the table and one die left, rolling is a genuine coin-flip of a decision rather than the obvious blunder the simple model suggests. With 1,500 on the table, bank it and stop being clever.

Keep fewer dice than you can

The most common beginner mistake is sweeping up every scoring die. Suppose you roll 1, 5, 2, 3, 4, 6. Keeping both the 1 and the 5 gives you 150 and leaves four dice. Keeping only the 1 gives you 100 and leaves five.

That 50 points costs you a jump from a 7.7% Farkle chance to 15.7% — you have doubled your risk to gain 50. Early in a turn, take the 100 and roll five.

The same logic makes three 5s (500, three dice left) usually better than breaking them up, and makes a lone 1 out of a bad roll perfectly respectable. The scorekeeper's Select best button follows this rule: when two selections score the same, it keeps the fewer dice.

Getting on the board changes everything

Until you have banked your first 500, points below that threshold are worth nothing. If you are not on the board and your turn total is under 500, you have literally nothing to lose by rolling again — the break-even table does not apply, because banking 450 and banking 0 have exactly the same result.

So: with 450 on the table and two dice left and no score yet, roll. With 450 on the table and two dice left when you are already on the board, bank. Same dice, opposite decision. This is the single biggest adjustment new players fail to make.

Playing the endgame

Once someone crosses 10,000, everyone else gets one final turn. Two consequences:

• If you are behind on your last turn, the break-even table is irrelevant. Banking second place is the same as Farkling. Keep rolling until you pass the leader or bust.
• If you are the one crossing the line, cross it by as much as you can. A 200-point cushion is worth far less than it looks when three players each get a turn to find 2,000.

It also means triggering the end early with a bare 10,000 is often worse than banking 9,800 and taking another turn — you hand every opponent a free shot while adding nothing to your own total.

Reading your opponents' scores

Farkle rewards adjusting to the table. If you are 3,000 behind with the leader near the target, your effective risk tolerance should rise: a cautious 350-point turn does not close a gap that size. If you are ahead and nobody is close, bank early and often and let the Farkle odds grind everyone else down.

The scorekeeper shows everyone's total on every turn for exactly this reason, and the leaderboard tracks each player's Farkle rate over time — which is a surprisingly good measure of whether someone is pushing too hard.