Backgammon Dice Probability Explained

A question needs a board and an event

There is a difference between hoping for a six and asking whether a six can help. A pair of fair dice does not inspect the board. The player does. Before counting, specify exactly what would qualify: at least one particular number, a particular unordered pair, a sum, any double, or a legal hit. These are not interchangeable events. A 6 and 1 can total seven, show a six, and fail to hit a blot if the required destination is closed. Three correct statements can lead to three different board decisions. Write the event in plain language first, then calculate it. That small discipline prevents more probability mistakes than memorizing dozens of percentages.

Why the outcome grid is ordered

Draw six rows for the first die and six columns for the second. Every cell stands for one equally likely ordered result when the dice are independent and fair. The row for 2, column for 5 is different from row 5, column 2. A player may use the same two movement values in either case, but the experiment that produced those values had two arrangements. Double fives occupy only one cell. Therefore the 21 unordered roll types do not have equal probabilities. There are 15 mixed types, each appearing twice in the 36-cell space, and six doubles, each appearing once. Never divide by 21 merely because a roll chart has 21 lines.

Read sums without confusing them with movement

A sum of seven appears in six ordered ways: 1 with 6, 2 with 5, 3 with 4, and their reversals. A sum of twelve appears only as double sixes. The dice sum is arithmetic; movement follows game rules. A seven-pip move in Tavla may require a landing after the first die that is open, then another legal landing for the second die. Under a different blocked configuration, the same sum can be split across two checkers, or partially unplayable. In Long Nard, a single enemy checker makes a point inaccessible, so a tempting sum has no power to cross that destination. The probability table suggests candidate turns, not permission to ignore legal steps.

Understand doubles before counting future turns

Each of the six doubles is one ordered outcome. Together they occupy 6/36, which is one chance in six. Seeing any double is not the same event as seeing 6 and 6: the latter is only 1/36. Doubles are also unusual in how the checker movement is used, since their face value is played four times when legal. That rule changes the possible play, not the probability that the dice show equal numbers. A Long Nard opening double still has to satisfy the game-specific head exception and occupied-point restrictions. A Tavla double cannot move through a closed point simply because it carries four steps.

Keep exact events distinct from approximate predictions

Probability is a fraction before it is a percentage. Eleven qualifying outcomes mean 11/36, or about 30.56 percent. Displaying 31 percent is useful for quick comparison, but it is not a new exact value. Thirty-six equally likely outcomes are the denominator for one roll with two fair dice. For two independent future rolls the underlying sample space contains 36 times 36 ordered pairs of results. There are situations in which the chance changes between turns because a checker is moved, a point is made, or a bar entry becomes possible. When the board changes, redraw the event. Carry forward only the stable probability rules.

Tavla hits are legal-move events

A blot is a single opponent checker on a point. In Tavla Classic or Tavla Modern, you can hit it with a legally arriving checker. But a list of rolls with sufficient total pips is only the beginning. The moving checker must have a legal path using dice steps, intermediate landings cannot be blocked, and any bar checker must re-enter before other moves. A direct shot from a point five pips away needs a five on at least one die and a legal landing. Longer combination shots require enumerating sequences and intermediate conditions. Long Nard has no hits at all, so this calculation should never be presented as a Long Nard tactic.

A bar entry calculation you can verify

For one checker on the Tavla bar, count the enemy home-board points that are closed to entry. Suppose four of the six destinations are closed by at least two enemy checkers. Both dice must show closed values for the turn to fail entry, and there are 4 times 4, or 16, such ordered outcomes. The other 20 outcomes allow at least one entry, so the probability is 20/36. With five closed points, only 11 outcomes allow entry. This calculation gives the chance of at least one checker entering, not necessarily clearing multiple checkers from the bar or completing a subsequent move. It is conditional on the exact six entry points.

Long Nard blocking changes what success means

Suppose a Long Nard checker stands on your 13-point and an opposing checker occupies the intended landing at your 8-point. A roll showing five is mathematically 11/36, but a five does not land there legally. It is a blocked destination. Counting the roll as a successful five-point advance would be a category error. Instead, identify other legal destinations or other checkers that may use that die. The head restriction and rule against a six-point wall trapping all fifteen opposing checkers can matter as well. Probability helps evaluate the available alternatives only after the legality filter has been applied.

Independent rolls are not promises

A player who has missed doubles for five turns may expect the sixth turn to compensate. No such debt exists when the next roll is independent of the previous five. The chance of any double is still 1/6. If you need a six, the chance of seeing one on the next two independent turns is greater than on one turn, but there is still a meaningful chance of missing both. Use complements to calculate a sequence of attempts: first count the chance of failure on each attempt, multiply those failure chances when the rolls are independent, then subtract from one. Do not extend the multiplication blindly to changing board conditions.

The starting pip count is a measurement

The standard Tavla opening has two checkers on the 24-point, five on the 13-point, three on the 8-point and five on the 6-point from the moving player’s perspective. Its pip count is 2 times 24 plus 5 times 13 plus 3 times 8 plus 5 times 6, giving 167. That number records distance remaining, not the chance of winning. Doubles raise the nominal distance available on a roll because the face is used four times. Blocked moves and bearing-off overshoots can waste that nominal movement. The board state, checker distribution and opponent contact still matter far beyond the arithmetic average.

Fair dice, server dice and evidence

Nard Online's owner specification states that online dice are rolled on the server. That describes where the result is generated. It is not by itself a publicly audited proof that every face or combination occurs at its ideal long-run frequency. Fair theoretical dice have a 1/36 chance for each ordered pair; even a fair sample of a few dozen rolls can look lopsided. A valid fairness review would require suitable logs, sample sizes and documented tests, with care not to expose private game data. Do not infer a rigged system from one unlucky evening. Equally, do not claim a certified random-number process without independent evidence.

Probability should improve decisions, not excuse bad ones

Imagine choosing between an easy legal checker advance and a risky play that could expose a Tavla blot. The number 11/36 might describe a relevant direct opposing face, but it is not automatically the chance you will lose the game. The opponent might have other threats, the move might be strategically valuable, or a contact game might justify more risk than a race. In Long Nard, think in terms of reachable points, blocks and timing rather than hits. In both variants, a sound decision combines the correct conditional odds with the rules and the value of the resulting position.

Worked probability notebook

1. A single six

Favorable outcomes: 11 / 36. Probability: 11/36 (about 30.56% ).

The six appears on either die. Count six cells in its row and six in its column, then remove the shared (6,6) cell. Suppose a Tavla blot is exactly six pips away. This 11/36 figure is only the first filter: it presumes a direct shot and says nothing about other legal requirements.

A face appearing is not a guarantee that the destination it suggests is open. Write the target point number before assigning a tactical interpretation.

2. Exactly five and two

Favorable outcomes: 2 / 36. Probability: 2/36 (about 5.56% ).

There are precisely two successful ordered cells: (5,2) and (2,5). The spoken roll name hides that reversal. In a Long Nard opening, a player might use the values for different checkers only when the head rule permits; the exact dice event cannot be enlarged by inventing extra movement orders.

Remember that this is an unordered roll name. The two cells remain separate outcomes in the mathematical model.

3. Double fours

Favorable outcomes: 1 / 36. Probability: 1/36 (about 2.78% ).

Only cell (4,4) qualifies. The four copies of the movement value are a rule of play, not four independent chances to roll that double. If an intermediate landing is blocked, even an exciting double can leave fewer useful actions. Do not convert four movements into a larger probability of seeing the dice.

A named double is a very small event. It becomes strategically large only when four legal movements are possible.

4. Sum of seven

Favorable outcomes: 6 / 36. Probability: 6/36 (about 16.67% ).

The ordered possibilities are (1,6), (2,5), (3,4) and their reversals. This familiar set is exactly one sixth of all rolls. But the player must still find legal intermediate landings. Write a hypothetical Tavla sequence from your 13-point: 13 to 8 using five, then 8 to 6 using two. Either destination can be closed.

Sum probabilities are often useful for puzzles, but they do not prove a continuous checker route is available.

5. The largest sum

Favorable outcomes: 1 / 36. Probability: 1/36 (about 2.78% ).

A sum of twelve requires exactly (6,6). Unlike ordinary mixed rolls, the rules make it four six-pip moves when legal. A player may speak of twenty-four nominal movement pips, but counting that as four independent sixes would multiply the event incorrectly. Treat it as one rare pair with special move usage.

When the sum is extreme, check whether you are accidentally counting a double only once or treating it like an ordinary two-step turn.

6. Ten or more in total

Favorable outcomes: 6 / 36. Probability: 6/36 (about 16.67% ).

The sums ten, eleven and twelve have three, two and one ordered outcomes respectively. Together they produce six favorable cells. The threshold is about a dice total, not how many actual pips you will successfully play. Doubles need separate treatment when translating a total into nominal movement.

This comparison uses the complement, which is usually faster than adding several overlapping success cases.

7. Both dice even

Favorable outcomes: 9 / 36. Probability: 9/36 (about 25.00% ).

Each die must belong to the set {2,4,6}, which has three choices. Three times three equals nine outcomes. This is not a roll of doubles alone: (2,4) qualifies. A condition stated through a set of faces often becomes easier when you count allowed values on each die separately.

The event describes the labels on the dice, not whether the current board offers a useful move with either label.

8. At least a five or six

Favorable outcomes: 20 / 36. Probability: 20/36 (about 55.56% ).

It is quicker to exclude outcomes containing neither five nor six. Then both dice belong to {1,2,3,4}, giving sixteen misses. Subtract them from 36 to obtain twenty successes. This is different from requiring exactly the unordered pair (5,6), which occupies just two outcomes.

Do not confuse an either-or requirement with needing both values at once. The second demand has a smaller favorable set.

9. Sum of eight

Favorable outcomes: 5 / 36. Probability: 5/36 (about 13.89% ).

The five ordered results are (2,6), (3,5), (4,4), (5,3) and (6,2). Notice the double four in the center. Count it once, not twice. A common mistake is to assume every total has an even number of permutations. Symmetric pairs are exceptions.

This is a helpful way to check whether a table of dice totals was weighted correctly before using it for strategy.

10. A one on either die

Favorable outcomes: 11 / 36. Probability: 11/36 (about 30.56% ).

The face one behaves like the face six in this model. There are six cells with a one on die A, six on die B, and one overlap at (1,1). This symmetry is useful for checking a calculation. It does not mean ones and sixes are equally useful on the current board.

The same percentage can describe very different tactical positions; the opponent’s made points change what counts as progress.

11. Any double at all

Favorable outcomes: 6 / 36. Probability: 6/36 (about 16.67% ).

The diagonal of the six-by-six table contains exactly six cells: (1,1) through (6,6). These are disjoint events. Adding their one-cell probabilities gives six out of 36. The next throw does not become less likely to be a double merely because this throw was a double.

A double offers four moves of its value, but the route can be interrupted by legal blocks on any intermediate landing.

12. No double on the next roll

Favorable outcomes: 30 / 36. Probability: 30/36 (about 83.33% ).

Subtract the six diagonal cells from the complete grid. Thirty outcomes show unequal numbers, yielding 5/6. In a Tavla race that may allow two distinct movement values, but it cannot guarantee a strong move: both destinations might still be blocked or strategically poor.

One missed double does not supply credit toward the next throw. Independence means every new attempt begins with the same chances.

13. A double within two throws

Favorable outcomes: 11 / 36. Probability: 11/36 (about 30.56% ).

Avoid adding 1/6 twice. That would count the possibility of doubles on both throws twice. The chance of missing doubles twice is (5/6) squared, or 25/36. Subtract from one and obtain 11/36. This result requires two independent rolls and concerns dice results, not changing board opportunities.

Repeated chances are best counted via the probability of missing every attempt, not by simply adding single-turn percentages.

14. A six across two throws

Favorable outcomes: 671 / 1296. Probability: 671/1296 (about 51.77% ).

A single throw misses six on both dice in 25/36 cases. Two independent throws miss it altogether in (25/36) squared, or 625/1296. Therefore at least one six over four individual dice has probability 671/1296. The denominator expands because this event describes two complete turns rather than one.

In a changing game, only the dice experiment repeats identically. The board-dependent target event may not.

15. Bar entry with two closed points

Favorable outcomes: 32 / 36. Probability: 32/36 (about 88.89% ).

With two closed entry points, only (2 times 2) four outcomes have both dice blocked. Thirty-two remain with at least one open number. This is nearly nine chances out of ten to enter one checker. It says nothing about whether your remaining die can then be played successfully.

With several closed entries, the distinction between at least one re-entry and clearing two bar checkers becomes vital.

16. Bar entry with three closed points

Favorable outcomes: 27 / 36. Probability: 27/36 (about 75.00% ).

Three closed numbers produce nine failed ordered outcomes. Twenty-seven allow at least one entry, exactly three quarters of the grid. Compared with two closed points, the loss is five outcomes, not an equal step in every row. The squared failure count explains the growing effect of a made home board.

These figures concern one checker entering from the Tavla bar. A second checker waiting there creates a different event.

17. Bar entry with four closed points

Favorable outcomes: 20 / 36. Probability: 20/36 (about 55.56% ).

A four-point home board blocks sixteen ordered pairs altogether. Only twenty of the 36 rolls offer an entry for one bar checker. This is still more than half, but the remaining margin is narrower. A single additional made point will eliminate nine more entering outcomes, which is why this stage often feels sharply more restrictive.

This case makes explicit why a mathematically available roll may still not solve a strategically urgent position.

18. Bar entry with five closed points

Favorable outcomes: 11 / 36. Probability: 11/36 (about 30.56% ).

Only one entry number stays open. At least one die must show that number, which occurs in eleven ordered cells. Do not call it one chance in six: the player rolls two dice. Double that open number is included once, and may offer different follow-up possibilities depending on how many checkers wait on the bar.

A fully closed entry board permits no bar re-entry on that turn. It does not mean the next turn must also be closed.

19. Completely closed entry board

Favorable outcomes: 0 / 36. Probability: 0/36 (about 0.00% ).

All six entry numbers are closed, so every pair fails. The chance of entering on this particular turn is zero. This result does not mean the checker is permanently stuck: the opponent may change the home board before the next attempt. Probability is always attached to a specific state and time.

Use the smaller set of favorable outcomes to understand why a specific combination should never be treated as routine.

20. Exactly a total of ten

Favorable outcomes: 3 / 36. Probability: 3/36 (about 8.33% ).

The valid pairs are (4,6), (5,5) and (6,4). Mixed pairs give two results, while the double gives one. That yields three outcomes in total. In a practical bear-off race, a total of ten is not ten guaranteed useful pips, particularly if the checker distribution creates overshoots or unused dice.

The race total is only a distance benchmark. It does not price contact, blocks, or wasted pips during the bear off.

21. At least one face from four to six

Favorable outcomes: 27 / 36. Probability: 27/36 (about 75.00% ).

Failure means both dice show values from {1,2,3}. Nine ordered cells fail and 27 succeed. This broad event is three times as likely as both dice being at least four. Use the complement instead of adding three separate face events, which would double-count rolls containing multiple qualifying faces.

This event checks the usefulness of the complement method when there are many success cases to enumerate.

22. Both dice no higher than three

Favorable outcomes: 9 / 36. Probability: 9/36 (about 25.00% ).

Each die has only three allowed faces: one, two or three. The resulting nine ordered pairs describe a quarter of the whole sample space. Contrast this event with at least one high die: they are mutually exclusive and exhaustive, so their numerators must add to 36.

Read the condition literally before changing the denominator: fair individual dice still have 36 equally weighted ordered outcomes.

Check the arithmetic and the move

Assume two independent fair dice and inspect the condition as written. Check whether you counted the same ordered pair twice or neglected a reversed pair. Only after you confirm the fraction should you examine actual checker choices. For a Tavla position use the moving player’s points from 24 down to 1; for Long Nard name the game explicitly and remember that a single enemy checker closes a landing.

What you can practice next

A sensible next step is to draw the qualifying cells on your own six-by-six grid. Compare the result with the displayed numerator, then alter one feature of the board while keeping the dice experiment unchanged. When a move no longer works, explain precisely which rule prevents it.

Frequently asked questions

1. Why are there 36 possible results?

Each independent six-sided die has six possible faces. Six choices for the first die multiplied by six for the second gives 36 ordered pairs. Each pair has probability 1/36 when both dice are fair.

2. Why are there only 21 named rolls?

Players usually name a roll without caring which die showed which number. Fifteen unequal pairs and six doubles make 21 types. Unequal pairs occur in two orders, while a double has only one order.

3. How often does any double appear?

Six of the 36 ordered cells show equal numbers. That means 6/36, or 1/6, for any double on a single independent roll. Double sixes alone remain 1/36.

4. What is the chance of at least one six?

Eleven ordered outcomes contain a six. Count the six with a six on the first die and the six with a six on the second, then subtract the double-six overlap once.

5. Does a roll totaling seven guarantee a seven-point move?

No. The total equals seven in 6/36 outcomes, but any single checker move must comply with separate dice steps and legal landings. A blocked intermediate point can prevent the intended path.

6. How does the bar-entry probability work?

In Tavla with k closed entry points, both dice fail only when both show one of those k numbers. At least one entry is possible on (36 minus k squared) out of 36 outcomes.

7. Can Long Nard use the same hitting odds?

No. Long Nard has no hitting. A single opposing checker closes a point. You must instead count rolls that permit legal moves around the actual blocks and head restrictions.

8. Will doubles become more likely after several misses?

No. With independent fair rolls, the next turn has the same 1/6 chance of any double regardless of previous throws. The memory of earlier results affects neither die.

9. Are percentages exact when rounded?

The fractions are exact under the fair-dice model. Percentages such as 30.56 percent are rounded displays. Keep the denominator of 36 when comparing small differences.

10. Does server-side rolling prove perfect fairness?

No. It identifies where Nard Online says online rolls are generated. Independent proof of a distribution would require separate evidence and a documented statistical evaluation.

11. Do the odds change between Tavla Classic and Modern?

The dice probabilities do not. Both use ordinary two-die movement. Modern adds a doubling cube to scoring decisions, while Classic does not include it.

12. Where can I practice these calculations?

Draw a six-by-six grid, shade qualifying outcomes, and test each candidate move against the chosen rules. Nard Online also has a live browser version at /play/ for gameplay practice.

Summary

This result applies to dice probabilities under the stated fair independent model. It does not certify Nard Online’s random-number system, establish a winning percentage or prove that a particular engine move is optimal.

Related reading

Dice odds pillar; Rules for Long Nard; Rules for Tavla; Long Nard strategy; Tavla strategy; Game variants; Learn Nard; History; Glossary; Questions; App guide; Online play; Practice in browser.