Backgammon Dice Odds: 36 Rolls and Smarter Decisions

The number on a die is not a prediction

You have waited three turns for a six. The dice land on two and three. It feels as though a six must be getting closer. That feeling is familiar around a nard table, but the dice do not remember the previous throws. On the next independent roll, the chance of seeing at least one six is still 11 out of 36, provided both dice are fair. What has changed is the position, your patience and perhaps the risk that the next roll will be too late.

Probabilities help precisely because they separate those things. A roll belongs to mathematics. A legal move belongs to the rules of a particular game. Choosing a move belongs to strategy. Long Nard has no hitting, and one opposing checker already closes a destination. Tavla Classic permits hitting a lone opposing checker and requires bar re-entry. Tavla Modern adds a doubling cube, which changes scoring decisions but not the physical values shown by the moving dice. Apply the right question to the right game.

The numbers in this guide assume two independent, fair, ordinary six-sided dice. They do not prove the fairness of a physical set, a random-number generator or a specific online match. Nard Online's owner specification says its online dice are rolled on the server. That is a product architecture statement, not a published independent statistical audit. For play, see Nard Online in the browser; for the rules, use Long Nard and Tavla.

Build the 36-outcome table yourself

One die can show 1, 2, 3, 4, 5 or 6. The second can show the same six values. Six choices times six choices gives 36 ordered outcomes. Treat the first die as a row and the second as a column. The cell at row 2, column 5 is (2,5); the cell at row 5, column 2 is (5,2). These are distinct ways the dice can land even though Tavla normally treats them as the same pair of usable move values.

For a roll of 2 and 5, two cells qualify: (2,5) and (5,2). That probability is 2/36, or about 5.56 percent. For double fives, only (5,5) qualifies. That is 1/36, about 2.78 percent. This is why counting the 21 unordered pairs as though they were all equally likely gives incorrect answers. The unordered catalogue has fifteen unequal pairs plus six doubles; it is useful for discussing types of rolls, not as an equally weighted probability space.

If you prefer a physical exercise, draw a six by six grid, place a coin in each cell representing a successful condition and count the coins. A formula becomes much easier to trust when you can identify the actual outcomes it represents. For independent fair dice, the probability of an event is the number of qualifying cells divided by 36. For more complex board shots, that last phrase, qualifying cells, demands care about legality.

Doubles are common as a category, rare as a particular roll

There are six doubles: 1 and 1 through 6 and 6. Together they occupy six of 36 cells, so any double occurs with probability 1/6, about 16.67 percent per turn. A particular double, such as 6 and 6, remains 1/36. Seeing doubles this turn does not make the next turn less likely to produce doubles.

Doubles also affect movement. In both Long Nard and Tavla, the value is played four times, subject to legal moves. A 4 and 4 is four moves of four pips, not a magical sixteen-point jump that ignores intermediate landings. A player might use different checkers, or the same checker repeatedly when the route allows it. Long Nard's head-release restriction and first-turn exceptions still apply; probability does not override them. Read the head rule before treating a tempting double as permission to release several checkers.

Suppose you ask whether any double will appear within three future independent turns. The easiest route is to calculate its opposite. The chance of no double on one turn is 30/36, or 5/6. Three misses in a row occur with probability (5/6) cubed, or 125/216. Subtract that from one: 91/216, approximately 42.13 percent, is the chance of at least one double in three turns. It is not a promise that a useful double will arrive, nor that you will be able to play every movement it offers.

Need a particular face, or several useful faces?

If you need at least one six on two dice, there are six outcomes with a six on the first die and six with a six on the second. The (6,6) outcome appears in both lists, so subtract it once. That leaves 11/36, or 30.56 percent. Exactly one six is 10/36, and two sixes are 1/36. The same calculation works for any single specified value from one through six.

If either five or six would help, the unsuccessful faces are 1 through 4 on each die. Sixteen of 36 outcomes contain neither desired value. Therefore 20/36, about 55.56 percent, show at least one five or six. That is a dice fact, not yet a hit probability. One or both candidate landings might be blocked, your checker might be on the bar, or compulsory move rules might prevent the planned sequence.

A question such as “Can I get a total of seven?” needs a different interpretation. If you mean the two dice add to seven, six ordered outcomes work: (1,6), (2,5), (3,4) and their reversals. That is 6/36. But a legal seven-pip checker move in Tavla can require two distinct legal landings. A blocked intermediate point may make the total unusable on that checker even when the dice sum to seven. Arithmetic is a candidate generator, not the rulebook.

Read the numbered board before counting anything

For the Tavla examples here, your points are numbered 24 down to 1 as your checkers advance. Your home board is points 1 through 6; your outer board is 7 through 12; your opponent's outer board is 13 through 18; and your opponent's home board is 19 through 24. Opposing players number the same physical triangles in reverse. Thus your 1-point is your opponent's 24-point. Always say whose perspective you are using. The central bar is not a twenty-fifth point.

The standard Tavla opening arrangement is two checkers on your 24-point, five on your 13-point, three on your 8-point and five on your 6-point. Its pip count is 167, as we shall calculate below. In Long Nard, all fifteen begin stacked on each player's head, in diagonally opposite corners, and both players move counterclockwise around the physical board toward their own homes. Do not paste a Tavla starting diagram into a Long Nard probability example and call it the same position.

Board numbering is useful even when no diagram exists. “Tavla, my checker on 13, opponent's lone checker on my 8, roll contains five” describes a possible direct hit. “Long Nard, my checker on 13, opponent's lone checker on my 8” describes a blocked destination. Same geometry, different legality. If a pictured position lacks a clear perspective or game label, request those details before trusting a percentage attached to it.

Pip count is distance, not a forecast

A pip count totals the distance your active checkers still need to travel before leaving the board. From the moving player's perspective, a checker on point 13 contributes thirteen pips and a checker on point 6 contributes six. With two on 24, five on 13, three on 8 and five on 6, the standard Tavla opening count is 2 times 24 plus 5 times 13 plus 3 times 8 plus 5 times 6. The total is 167.

Take a smaller teaching position: two checkers on 6, three on 4 and one on 2. Their count is 2 times 6 plus 3 times 4 plus 1 times 2, which equals 26. If all those checkers are in home and a legal 6 and 2 bears off checkers from the 6-point and 2-point, the remaining count falls. But using larger dice on already close checkers can waste potential movement. A raw pip total does not capture whether you will have to pass or overshoot a bearing-off move, or whether a Tavla checker might be hit and sent back to the bar.

For an uncontested race, comparing pip counts gives an initial measure of who has farther to go. It is not a calibrated winning percentage. The number of checkers on each point, their distribution, forced moves and chance of wasted pips all matter. In contact play, the possibility of a hit matters even more. For Long Nard, the absence of hits removes that hazard but a blocked route can still prevent the shortest progress. The Long Nard strategy guide and Tavla strategy guide explain those decisions.

Calculate a direct Tavla hit without exaggerating

Imagine Tavla from your perspective. Your checker is on point 13, and the opponent has one checker on your point 8. The target lies five pips away. Provided it remains a blot, a die showing five permits the direct move 13 to 8. Eleven of 36 rolls include at least one five. Thus 11/36 is the probability of seeing a die value that supports this direct hit, under our simplified position.

That caveat matters. Suppose you also have a checker on the bar. Re-entry has priority, so you cannot assume the five on the dice can simply be used from point 13. Suppose the target is no longer a blot because a second opposing checker was added. The direct hit disappears. Suppose a rule forces a different sequence to use both dice legally. The outcome needs another check. The static calculation is good for understanding candidate shots, but the exact hit probability belongs to the entire position and the set of legal turns.

For a seven-pip target, neither die can show seven. You would need to combine both dice with an intermediate landing that is legal. Three unordered pairs sum to seven: 1 with 6, 2 with 5, and 3 with 4. There are six ordered outcomes in total, but if all relevant intermediate landings are blocked, none produces the desired hit. An attacking move that leaves a checker seven away should not automatically be assigned a 6/36 vulnerability without inspecting those intermediate points.

Why a double contact can still be a poor shot

A player notices that two dice have multiple ways to reach a blot and casually adds the percentages. That often double-counts an outcome which supports more than one route. The safe method is to mark each of the 36 ordered cells once if at least one legal hitting sequence exists. Then divide the count by 36. Different candidate routes may overlap, and legal sequences can interact with forced moves elsewhere on the board.

Consider Tavla with a checker on point 10 and a lone enemy checker on point 7. A three directly reaches the target, subject to the remaining roll and legal-play conditions. If you also own a checker on point 12, a five may offer a second direct route. Do not announce 11/36 plus 11/36 as the hit chance: rolls containing both a three and a five were counted twice. You must take the union of successful outcomes and examine any compulsory-move restrictions. The union of “contains three” and “contains five” is 20/36 before board legality, not 22/36.

In Long Nard the entire calculation changes because checkers cannot be hit. An opponent checker on a point closes it regardless of whether it stands alone or with others. Here probability questions concern arriving at an open landing, passing a head-release bottleneck or finding useful steps while maintaining a lawful block. Moving the same dice mathematics across the two games without moving the rule context is a standard beginner error.

Bar entry in Tavla: an exact conditional model

A Tavla checker that has been hit must re-enter from the bar before that player moves other checkers. From the entering player's perspective, a die value of one enters on point 24, two on 23, and so on until six on point 19. Any entry point occupied by at least two opposing checkers is closed. A point with a single opponent checker is normally hittable and therefore open to entry. These facts make a useful probability table possible.

If the opponent has k of the six entry destinations closed and the remaining destinations are legally open, you fail to enter with one checker when both dice show one of those k closed numbers. That probability is k squared over 36. The chance that at least one die offers entry is (36 minus k squared)/36. The model assumes one checker on the bar and counts the opportunity for an initial entry, not necessarily completion of every later move.

Take the five-point closure: only one die face reaches the sole available destination. Eleven of 36 rolls show that face on at least one die. Now imagine all six entry points closed. No die number works and you cannot use the roll for another checker while the one on the bar remains unable to enter. This arithmetic helps explain why building an opponent's home board is powerful in Tavla. It does not describe Long Nard, which has no hitting or bar entry.

Long Nard blocking is a legal-move problem

Long Nard uses the same dice but not the Tavla bar or blot model. All fifteen checkers start on the head. Ordinarily one checker may leave the head per turn. The first-turn exception is limited by the owner rules to doubles of 6 and 6, 4 and 4, or 3 and 3 when opposing checkers would otherwise block full play. Counting the probability of those doubles does not tell you how often the exception actually applies; its trigger depends on the opponent's position and the opening turn.

One opposing checker blocks an entire landing point in Long Nard. Consider your own checker on your point 17 and a five on one die. Your intended destination is point 12. If one opposing checker occupies point 12, that direct landing is illegal, even though the physical point is not crowded. A different legal checker movement or dice order may be required. Whether a combined roll of two and three can reach point 12 depends on legal intermediate landings and the compulsory-use rules.

There is also the six-point barrier restriction. A chain of six consecutive occupied points cannot lawfully trap all fifteen opposing checkers behind it. You therefore cannot calculate the chance of “completing the perfect wall” merely by listing promising numbers, and then assume any such placement is legal. A real position must be checked for the opponent's pieces as well as your dice. The six-point block guide belongs beside the probability table in any serious study session.

Race estimates require more than dividing by seven

The total on two dice ranges from two to twelve. Its arithmetic mean is seven: each die has mean 3.5. An ordinary non-double pair usually offers the sum of its two numbers as potential forward movement if both values can be played. Doubles are special because the value is played four times. If all movements could be used without blocking or bearing-off waste, the expected nominal pip movement per turn under doubles rules is 49/6, approximately 8.17 pips, not seven.

Here is the calculation. Across all 36 ordered outcomes, adding the two visible faces gives a total of 252 pips, so the ordinary two-dice average is seven. The six doubles contribute 2 times (1+2+3+4+5+6), or 42, to that total. Playing four movements rather than two on doubles adds another 42. Consequently the nominal total becomes 294 over 36 outcomes: 294/36 = 49/6, approximately 8.17 pips. That is a potential movement statistic, not the amount every position can legally use.

Even 49/6 is not a clock that forecasts the finish. A turn can be blocked, an endgame can force a small checker to consume a large die, or the player may not have enough checkers positioned to realize the nominal distance. In Tavla, a hit can add a long journey. In Long Nard, the head rule constrains development. Divide an opening pip count by 8.17 and you do not get an honest predicted number of turns. Use the value to understand the dice distribution, not to claim a fixed game duration.

Bearing off: compare progress with practical waste

In Tavla, all your checkers must be in your home board before you bear off. Imagine one checker each on points 6, 4 and 1. Rolling 6 and 4 can remove the checkers on points 6 and 4 if legal. That is clear use of both values. But suppose all your remaining checkers are on points 2 and 1. A 6 may remove a checker under the ordinary higher-die bearing-off rule when no checker remains on a higher point. It is valid, but the die value exceeds the checker's remaining distance. The surplus does not move a different checker.

A useful practice habit is to record two figures: the pip count before the turn and the amount genuinely eliminated after legal bearing off. Notice how often an apparently large roll fails to produce the largest possible reduction because of checker distribution. Rebalancing a home board earlier may reduce that waste. In a game with contact, however, rushing toward bearing off while leaving a dangerous blot can lose far more than a few pips.

For Long Nard, bearing off also begins only when all fifteen checkers have entered the player's home. Its blocking and movement rules remain in force while a player approaches that stage. A race calculation must not pretend that a checker may leave the board before the game's own condition is satisfied. The beginner guide shows the move sequence before mathematical shortcuts enter the picture.

Fair dice, large samples and ordinary bad runs

When somebody rolls four doubles in a short span, it can feel suspicious. The probability of four specified consecutive turns all being doubles is (1/6) to the fourth power, or 1/1296. That is rare for a particular four-turn sequence, but unusual sequences appear somewhere when very many games are played. Looking at an exciting clip after it happens is not the same experiment as predicting that exact sequence beforehand.

A genuinely useful fairness assessment requires a clearly defined sample, the expected distribution, the observed distribution and an appropriate statistical test. It also requires knowing what events were recorded and whether any filters were applied. Several losses, a single improbable hit or one long stretch without sixes do not establish manipulation. Conversely, saying “dice are server rolled” does not itself establish independent certification. The Nard Online specification describes server-side rolling; publication of any audit claim would require separate supporting evidence.

Offline physical play benefits from ordinary equipment care. Use matched dice where possible, a clean board, a consistent roll procedure and agreement on which dice are valid after landing. Online fair play additionally means respecting the turn clock, not abusing chat and using mute or report functions when needed. The game uses virtual coins with no cash-out or real-money value. Probability education here is about learning a board game, not encouraging stakes or predicting monetary returns.

Board materials matter to the eye, not the probability

A board has 24 narrow triangles, twelve per side, separated by a central bar. Walnut, maple, oak, marble and decorative inlays change how the table looks and feels; they do not change 36 possible ordered dice outcomes. Nard Online offers several cosmetic boards and checker sets, including wood-inspired options and Shusha inlay. Those cosmetics belong to the experience rather than the legal geometry.

For a teaching diagram, choose contrast before decoration. Ivory and burgundy checkers must be distinguishable, the 24 points must remain countable, and the perspective label should stay visible. If a diagram shows a checker on point 13 moving five pips to point 8, number the destination explicitly. Never illustrate a Long Nard hit or put a Tavla checker on an impossible entry point. A beautiful but illegal illustration can teach a wrong move more effectively than a poorly worded paragraph.

If you are buying a physical board for practice, check for a stable opening surface, triangles long enough for fifteen checkers to circulate, smooth checker movement and legible contrasting dice. A board that allows pieces to sit without covering important boundaries makes counting easier. You do not need luxury wood to learn probability. Consistency in how you label points matters much more than the price of the board. This guide does not endorse a retailer or a particular set of equipment.

Five short tests worth trying at the table

Test 1, the double: Before rolling, write down the chance of double threes. The answer is one ordered cell, (3,3), in 36. It is not 1/6; that larger probability concerns any double.

Test 2, the needed four: Imagine you need at least one four to use an open Long Nard landing. Count row four and column four on the grid, removing the duplicate (4,4). The result is 11/36 for seeing a four. Whether you may actually land depends on the board.

Test 3, two helpful numbers: You would accept either a two or a five. The unwanted values are 1, 3, 4 and 6, four faces per die. Sixteen unsuccessful cells leave twenty successes, so the dice probability is 20/36. Then check whether both candidate moves are legal.

Test 4, bar entry: In Tavla, the opponent has closed four of the six entry points. Four squared gives sixteen failed dice outcomes, leaving twenty successes. Your initial chance to enter with one checker is 20/36, about 55.56 percent, assuming the other two entry points are open.

Test 5, pip arithmetic: In Tavla from your perspective, one checker on 24, two on 8 and three on 3 yield 24 plus 16 plus 9, or 49 pips. If someone writes 39, recount the contribution of the 24-point rather than arguing about the dice.

Use these tests as a conversation starter. Ask one person to count the dice grid and another to examine the legal moves. Where the answers differ, the missing information is often a blocked landing, a bar obligation or a confused point perspective. That disagreement is exactly where mathematical literacy becomes practical strategy.

Frequently asked questions

How many dice outcomes are there in backgammon?

There are 36 equally likely ordered outcomes with two fair six-sided dice. When order is ignored, there are 21 different pairs, but they are not all equally likely.

How often do you roll doubles?

Any double has probability 6/36, which is 1/6 or about 16.67 percent. A particular double, such as 2 and 2, has probability 1/36.

What is the chance of rolling at least one six?

It is 11/36, about 30.56 percent. Count six outcomes on each die and subtract the one double-six result counted twice.

Are two and five as likely as double five?

No. Two and five can occur in either order and thus has probability 2/36. Double five has only one ordered outcome, so its probability is 1/36.

What is a pip count in Tavla?

Add each checker's point number from the moving player's perspective. The standard Tavla opening distribution totals 167 pips.

Is 11/36 always the chance of hitting a blot five pips away?

No. It is the chance of rolling at least one five. An actual hit also depends on the legal board position, priorities such as bar entry and forced moves.

Can dice roll a seven?

A die cannot show seven. Two dice can total seven in six ordered outcomes, but using both values as a seven-pip checker move requires legal intermediate landings.

How likely is it to enter from the bar against five closed points?

With one Tavla checker on the bar and exactly one open entry face, the chance that at least one die shows that face is 11/36, about 30.56 percent.

Does Long Nard use bar-entry odds?

No. Long Nard has no hitting and no bar-entry procedure. Its probabilities concern legal travel, head departures, barriers and bearing off.

Do server-rolled dice mean independently certified dice?

No. Nard Online's owner specification identifies server-side dice rolling, but that alone does not substantiate any outside fairness certification.

Is the doubling cube another die?

No. Tavla Modern uses a doubling cube for scoring decisions; it does not produce the checker movement values. Tavla Classic and Long Nard have no doubling cube in Nard Online.

Can probability tell me the exact winner?

No. Dice frequencies describe possible rolls. Legal moves, strategy, timing, contact, board blocks and decisions make each match a changing problem.

Summary

The most useful dice facts fit on a small card: 36 ordered outcomes, 21 unordered roll types, six doubles, 11 outcomes containing a chosen face and two ordered outcomes for any named unequal pair. A pip count measures distance, not destiny. A hit percentage must refer to a specific Tavla position, and a Long Nard blockage must be tested against that game's rules. Learn the grid first, then count only genuinely legal shots. The player who checks assumptions has an advantage over the player who merely quotes percentages.