Hnefatafl

Return to Research Index

References to Norse games abound in the Sagas. Orkneyinga Saga, Greenland Lay of Atli, Hervar's Saga, Fridthjof's Saga, and Gunnlaugs Saga Ormstunga are only a few such. Many problems exist, however, in attempting to use these references to help reconstruct the game. Tafl (table) became hnefatafl (king's table) later in the Viking period as it became necessary to differentiate it from the many other games including chess (skaktafl), and backgammon (kvatrutafl)² coming into play throughout the Norse world Translators often assign an incorrect game when faced with the task of having to translate the old Norse word "tafl". Gordon in his Introduction to Old Norse, for example, translates "tafl" correctly as "table" but goes on to assume that this means the game referred to is similar to the Latin "tables" (backgammon).

Saga copyists in the Middle ages often complicated references by assuming the game to be chess and adding allusions to the rules for that game. Fiske in his book Chess in Iceland attempts to comb the sagas for references to chess. He eliminates many of these references because they obviously refer to a game that is clearly not chess. Most of those rejected references are in fact mistranslations or miscopies of tafl itself. In spite of this we can retain some of the flavour of the game by reading the sagas. A good example is the riddling from Hervar's Saga "who are the maids that fight weaponless around their lord, the brown ever sheltering, and the fair ever attacking him?"³

Debate still rages about which references are to what game.Many interesting papers have been written on the subject some of which are mentioned in the bibliography. Until such time as the difficulties of translation, copy and even writing errors (given that many sagas were written long after the viking era) can be overcome the references themselves will not prove to be helpful in uncovering the rules to this game.

The artifacts associated with hnefatafl provide a rich field for study with six boards having been found to date.

In a farm excavation in Toftanes Eysturoy in the Faroes, half of a 13x13 board was discovered⁴. This oak board is dated to the tenth century and appears to have been a serving platter or plate. This artifact is kept at the Foroya Fornminnissavn in Torshavn as exhibit number 4666/1762.

In 1932 at Ballinderry in Ireland a wooden 7x7 board dating to the late ninth or early tenth century was found⁵. Unlike the others this board has holes for pegs instead of marked squares. It is kept in Dublin at the NMI as artifact 1932:6533.

At Trondheim on the site of the public library, parts of an 11x11 wooden board were discovered⁶. The center is marked as are 6 more squares in a symmetrical pattern This board is dated to the early 1100s and is kept at the Vitenskapsmuseet in Trondheim as artifact N29723/FH414.

Three additional wooden boards exist. The first was found at Wimose on Funen and dated to the Roman iron age⁷. This board has 18 squares and is clearly cut off. In all probability this board was originally 19x19 and was for Alea Evangelii.

The second board⁸ was found in the Gokstad ship burial and dates to the late 9th century and should be in the Universitetets Oldsaksamling in Oslo. This board shows 13 cells across. Several of the cells are marked but without more of the board to place the marks in context it is impossible to tell why it was marked.

The final board was found at Coppergate in York and is dated to the 10th century⁹. I count 16 rows and 3 columns on this board although when the find is discussed in the Bulletin of York Archaeological Trust Interim vol. 6 no 4 18 rows are claimed. Assuming the board to be complete it could have been used to play a 15x15 game in the cells. If the board is incomplete or I am miscounting the number of columns, it could have been used to play Alea Evangelli.

In addition to the six game boards a very large number of playing pieces have been found. Some of the finds are given below to indicate the range of materials used for construction and the approximate ratios of the types of pieces. It is important to note that all but one of these sets of pieces appears to be incomplete. Also of note is that it is very difficult to separate pieces for hnefatafl from those for other games such as backgammon.

• 16 glass - 11 blue, 1 blue/yellow, 4 yellow - Gunnarshaug Norway (c.800 AD)
• One carved bone - Woodperry, Oxfordshire
• Three bone - Boat Grave IX Vendal Parish, Sweden. (10th cent.)
• 2 jet playing - Mote Hill, Warrington
• 24 walrus ivory and a whale bone king - Baldursheimur, Iceland (10th cent.)
• A king of Amber - Roholte, Denmark (10th-11th cent.)
• 2 ivory - Sandnaes Greenland (11th cent)
• 12 antler and tusk - Ile de Groix, France (10th cent)
• 1 ivory - Trondheim (12th cent) (found with a game board)
• 2 ivory - Dublin - (11th cent)
• 1 Ivory king - man like figure - Lund, Sweden (11th cent)
• 27 bone - 1 king, 6 small, 20 large - Birka, Sweden (9th cent)
• 2 bone - Hedeby, Germany
• 2 amber - Schleswig-Holstein, Germany
• 1 Jet - Bawdsey, UK (10th cent)
• 1 Bronze king (man figure) - Eyrarland, Iceland (11th cent)
• 23 Glass (8 black, 15 blue) - Valsgarde, Sweden
• 8 Glass - Birka, Sweden

While all of these sources are valuable and will be used later to attempt the reconstruction none of them is enough to allow us to understand the basic rules of the game. It is fortunate that three written records of variations of this game exist to help fill in the details. There is Linnaeus' diary from July 1732. He went traveling among the Lapps and observed the 9x9 board (Tablut) being played. One translation by Sir J.E. Smith in 1811 explains the rules this way

Tablut is played on a board of eighty-one squares marked as in the diagram. One side consists of the King, who is stationed on the central square, and eight Swedes, who are placed upon the shaded squares. The other side consists of sixteen Russians, who occupy the crosscut squares. All of the pieces have the same move that of the Rook in chess. Play is by alternate moves, and the one player attempts to bring his King to the edge of the board, while the other tries to confine him so that he has no power of moving. In either case the game comes to and end. The King cannot be taken; any other man is taken when two of the opposing men occupy two squares adjacent to it and in the same straight line with it. No other piece than the King can ever play to the central square.¹¹

A simplified version of the diagram that was included in the diary appears below. `' is the king, `' are his Swedes, and `' are the Russians.

At the Welsh National Library a manuscript stored as Peniarth #158 is a book written by Robert ap Ifan in 1587. This 240 page book contains within it a drawing of an eleven by eleven board and a paragraph describing the rules for tawlbrwdd. Murray translates this paragraph as follows:

The above board must be played with a king (brenin) in the centre and twelve men in the places next to him, and twenty-four lie in wait to capture him. These are placed, six in the center of every end of the board and in the six central places. Two players move the pieces, and if one belonging to the king comes between the attackers, he is dead and is thrown out of the play; and if one of the attackers comes between two of the kings men, the same. If the king himself comes between two of the attackers and if you say 'watch your king' before he moves into that place, and he is unable to escape, you catch him. If the other says gwrheill(?) and goes between the two, there is no harm. If the king can go along the line (lacuna here) that side wins the game.¹²

Finally manuscript Oxon 122 held at Corpus Christi College, was written c. 925 AD. It shows the starting positions for a game entitled Alea Evangelii. A long discourse fills the next two folios giving the religious support for these positions. No mention is made of the playing rules but the picture of the board is useful. Correcting some slight copy mistakes to regularize the board produced the following setup. ¹³
 

References to tafl games also exist in the Dimetin Code, and the Laws of Howel Dha and the Book of Rights. These references, however, do not mention the rules, they merely serve to show us how widespread this game was.

In recent times both HJR Murray's A History of Board Games Other Than Chess and A History of Chess and RC Bell's Board and Table Games from Many Civilizations V1&2 give complete versions of the rules and starting positions for all sizes of boards. These modern reconstructions, however, have their own problems. Murray, for example, asserts that "A man can move to a cell between two enemy men without capture"¹⁴ with no suporting documentation Bell often blindly quotes from Murray and his suggested board layout for Tawlbrwdd is wildly at varience with the other board layouts.

Before leaving the topic of what we do know, I would like to point out the flaws inherent with the information presented this far. The manuscripts would at first glance appear to contain all that we need to play these games yet a closer look shows that translations differ in the information they present sometimes quite markedly. For example Frank Lewis¹⁵ draws a different conclusion than Bell in their translations of Robert ap Ifan's manuscript with respet to the king's ability to move between two players. In addition strange variations begin to appear Bell translates Robert Ap Ifan's "tawlbrwdd" as "throw board" and from this assumes that dice were involved in the play.¹⁶ This translation is questioned by many including Frank Lewis.

In addition to the translation problems we face with both Linneaus and Ifan's works, we must accept that these rules were written down hundreds of years after the game was popular. By the late 1500s when Robert ap Ifan wrote chess had long since supplanted hnefatafl. In addition views of the world had changed. In place of the earlier "king as a man" viewpoint of the Norse and Welsh the world had moved to the "divine right of kings". This will have had an effect on the play of this game. How much and in what way is very difficult to ascertain. Finally by using these records we are assuming that both Linnaeus and Robert ap Ifan were good observers and more importantly that they correctly interpreted both what was and wasn't said by the players. We also assume that that they recorded only the facts and not their own opinions. Combined these provide many reasons to hesitate to accept what we find as the whole and complete truth.

To prevent confusion I have included this reference table of the game sizes and names and sources. I am working under the assumption that all of these games belong to the same class of game so that they may all be called "tafl" games.

Name	Fithcheall	Tablut	Tawlbrydd	Hnefatafl	Alea Evangelli
Board Size	7x7	9x9	11x11	13x13	19x19
Source(s)	Drawing 2	Linneaus	Robert ap Ifan, Drawing 3	Drawings 1 & 5	Oxon 122, Drawings 4 & 6

With those restrictions in mind we begin to reconstruct the rules to these games. All three written sources agree that the ratio of attackers to defenders is 2:1 not including the king. They also agree that the king begins the game in the center of the board surrounded by the defenders and that the attackers begin the game around the outside of the board. Both the written rules note that a piece is captured when trapped between two opponents occupying the same line. We can be reasonably sure that these facts are correct. Note that it is not clear if a piece may safely move between two opponents or if he must be attacked. On a practical level it is possible to move safely *temporarily* between two fighters by splitting your attention and defending with a shield against one man and the sword against the other. Culturally norsemen considered themselves to be all equals. Thus it is reasonable that one on one you could not kill one. Engaging his attention then having a friend come up behind him, however, would kill even a good fighter. In addition ap Ifan claims that under some circumstances the king may safely move between two players. Two possible conclusions can be drawn from this. The first that ordinarily the king, and hence all players, may not move between two players. The second that so long as the move is temporary it is safe (the "escape" mentioned). For these reasons I have assumed that a player may safely move between two opponents.

Linneaus tells us that all pieces move as a rook does. This would appear to be supported by ap Ifan who also mentions moving along lines. Unfortunately he cannot confirm Linneaus' unlimited travel along a line. At this point we must turn to the Viking culture to see if any additional information can be gleaned. Many papers have been written linking a society and it's games, so the use of a culture to support a question of a rule appears to be valid. A quick glance at a map of the Norse world shows the extent of their travel. The extensive use of boats allowed them almost unlimited movement at a pace unmatched by other cultures of their time. Thus the culture appears to be adjusted to unlimited rapid travel. These facts would tend to support Linneaus' claim of a rooks unlimited travel. Thus we would appear to be safe using that as a rule.

Linneaus also mentions that only the King may occupy the center square (throne). Of the three archaeological boards that have a center square visible all have it marked in some fashion. In addition the Ockelbo stone shows the center of the board as marked. The only support culturally comes from the Havamaal of the Elder Edda which asks "The guest has arrived. In which seat shall he sit?"¹⁷ Combined these would seem to support the idea that the throne is different from other squares. Based on the translation given above I would have assumed that a player may pass through the throne but not stop on it. ("play *to* the central square").

Two questions then remain. How does either side win the game? Who moves first?

Taking the later question first we have no way of knowing who moves first.For that reason it is listed as one of the variables in the experiments to follow.

The king wins by escaping. This is quite clear from the written record. What exactly counts as an escape is not so clear. Linneaus states that the king must get to the side. Ap Ifan says that he must move along the line but the rest of the sentence is lost from the parchment so we don't know where he must get to. In addition, note that the Ballinderry board and Ockelbo stone both show the corners of the board as marked. Does this mean that the king must play through to a corner? This is another of the experimental questions. A question that came up early in a "to the corner" game was whether the attacker can place one man in each of the four corners thus blocking the king and turning the game into a killing match. It was decided that the corners in this variation were to be treated the same as the throne -- no-one may stop on them. Blocking the corners would then require the attacker to tie up eight men (or usually one third of his strength) making the game more even.

The next problem was to determine the defender's victory conditions. Linneaus is quite clear that you capture (not kill) the king by stopping his ability to move. Thus the attacker must put a man on each of the four sides of the king. Robert ap Ifan brings this into question by mentioning that the king may only move safely between two people if he says gwrheill. Does this mean that if he is caught between two players without saying this that he is dead -- like any other player? Because Linneaus gave such a clear set of rules we played the experimental games with a "4 to capture" rule but as you will see the results bring that into question.

The question of whether the king can assist in the capture of an attacking piece was also raised. I have assumed that he could solely on the basis of culture. During the viking age the king fought just as everyone else did not like the later Middle Ages.

This left only the starting positions to figure out. The 9x9 and 19x19 boards have starting positions and are shown below. Robert ap Ifan gives an incomplete set of starting positions for tawlbrwdd. It is obviously impossible to have "twenty-four lie in wait to capture him. These are placed, six in the center of every end of the board and in the six central places."¹⁸ To begin with, we can not place six in the center of a board that contains an odd number of squares. Also there are not six central places on the four sides of a square. Finally six in the center of every end and six more in the central places would give 30 attackers not twenty-four. If we assume an error in the manuscript and change it to read "twenty-four lie in wait to capture him. These are placed, FIVE in the center of every end of the board and ONE in the central place" we have a scaled up version of the Linnean board for twenty-four attackers. Similarly for the defenders the "twelve men in the places next to him" are not immediately obvious. Below are two possible starting positions for this game.

Similarly the 13x13 board could be laid out as shown below. The advantage to these layouts (especially the second of each pair) is their close correlation to the Linnean layout.
 

The seven by seven board is much more difficult. Two possible options are shown below. Note that the second also explains why the corners of the board are marked.
 

Now that the rules, staring positions and outstanding questions have been understood, we may move into designing the experiments to answer those questions. This method of archaeological reconstruction has some problems which should be mentioned at this time. The players who will be playing these games are unfamiliar with the game and rules. As a result mistakes will be made that may alter the results. Games marked with a `*' in the charts that follow are examples of that. This factor can be alleviated somewhat by increasing the number of games played and keeping the circle of players fairly small so the learning curve does not keep affecting the results. Since the games must be played with volunteers this would drastically increase the amount of time needed to obtain the results. Instead I have asked the players to note when they made a bad mistake that cost them the game and I have marked those games so they can be considered.

To reduce the skew that could be caused by players of unequal skill, wherever possible two games are played one after the other with the same rules. The players change sides between games. In this way a good player will win a game for each side and keep the results balanced. The problem became very apparent as play progressed as some players had a very difficult time adapting to the different style of play this game produces.

The time required to play some of these games becomes prohibitive in our culture where we do not have whole days that can be devoted to drinking and playing (see the Ockelbo stone). I timed some games and noted that when the players were concentrating on the game a round (one move by each player) took less than a minute. With other distractions those round times increased as high as three minutes. When a game can go upwards of one hundred rounds at three minutes a round it takes far to long to play two games back to back.

The data are organized in tables according to the board layout. The first move was recorded as was the question of play to the side or the corner. The side that won is listed as is the number of rounds of play necessary to reach that point. Games marked with a '+' were conceded after that many rounds. Data has been reordered for ease of reading.¹⁹

Tablut (9x9)
Game #	First Move	Side or Corner	Winner	Number of Moves
1	King	Side	King	10
2	King	Side	King	12
3	King	Side	King	29
4	King	Side	King	18
5	King	Side	King	16
6	Attacker	Side	King	17
7	Attacker	Side	King	20
8	Attacker	Side	King	22
9	Attacker	Side	King	8*
10	Attacker	Side	King	27

Alea Evangelii (19x19)
Game #	First Move	Side or Corner	Winner	Number of Moves
1	Attacker	Side	King	19
2	Attacker	Side	Attacker	8*
3	Attacker	Side	King	16
4	Attacker	Side	King	12

Tawlbrwdd (11x11)
Game #	First Move	Side or Corner	Winner	Number of Moves
No Data

Tawlbrwdd (11x11) [second board]
Game #	First Move	Side or Corner	Winner	Number of Moves
1	Attacker	Side	King	32

Hnefatafl (13x13)
Game #	First Move	Side or Corner	Winner	Number of Moves
1	King	Side	King	4
2	King	Side	King	13
3	King	Side	King	45
4	King	Side	King	5
5	King	Side	King	11
6	King	Side	King	4
7	King	Side	King	13
8	King	Side	King	45
9	King	Corner	Attacker	50+
10	King	Corner	Attacker	44
11	King	Corner	King	34*

Hnefatafl (13x13) [second board]
Game #	First Move	Side or Corner	Winner	Number of Moves
1	King	Corner	Attacker	78
2	King	Corner	Attacker	28
3	King	Corner	King	14
4	King	Corner	King	14
5	Attacker	Corner	King	8
6	Attacker	Corner	King	32
7	Attacker	Corner	Attacker	57
8	Attacker	Corner	Attacker	93
9	Attacker	Side	King	5
10	Attacker	Side	King	11
11	Attacker	Side	King	8
12	Attacker	Side	King	10

Fithcheall (7x7)
Game #	First Move	Side or Corner	Winner	Number of Moves
1	King	Side	King	15+
2	King	Side	King	4
3	King	Corner	Attacker	79
4	King	Corner	King	25
5	Attacker	Corner	King	8
6	Attacker	Corner	King	25
7	Attacker	Corner	Attacker	14
8	Attacker	Corner	Attacker	14
9	Attacker	Side	King	4
10	Attacker	Side	King	14

Fithcheall (7x7) [second board]
Game #	First Move	Side or Corner	Winner	Number of Moves
No Data

Although this data is far from complete, enough has been gathered to point to several clear flaws in the basic assumptions. For this reason it is now necessary to analyze the data and redesign the experiments.

Even with the knowledge of the flaws in both our original data and in the method of the experimentation, it quickly becomes clear that the rules Linneaus describes do not provide for a playable game. Hervar's Saga tells us that one should play the king's side as that side wins more often. Yet the data from Tablut indicate that the king wins 100% of the time. When the king moves first he wins the game slightly faster than if the attacker moves first. Regardless, the game is over in an average of 17.9 rounds. The king is simply too powerful on the board and his escape is too easy. This is reflected by the fact that in all of the games where the king needs to get to the side, the king won 30 times to a single victory by the attackers. Even more revealing is that the attacker's victory was the fault of a new player as king who did not yet think in the ways the game requires.

When the game is changed to require the king to play to the corner, we find that the odds are much more even with the attacker winning 9 of 17 games. This too is deceptive. Three of the kings win's were on bad mistakes by the attacker. Removing those games gives 9 of 14 games to the attacker in an average of 40.5 moves. This game is markedly longer and skewed too far in favour of the attackers.

Thus play to the side cannot be the answer to the excessive power of the king. At this point I am willing to drop the exploration of the corner as a valid move. One suggestion has been made that as this game appears to be the results of having a king trapped in his hall by attackers (a common Norse method of killing rulers) we treat the attackers starting positions as the "doors" to the hall. That way the king would have to escape but could only do so through the squares already occupied by the attackers. It is probably worth choosing a board and trying a run of games to see how this plays out but I do not believe at this time that it is the answer. I support my position based on the lack of marked positions on the archaeological boards. It is difficult to remember which 5 squares of the 13 along the outer wall need to be protected and which do not. Note that the throne is different from any other square and it is marked differently. If the "doors" are different why are they not also marked?

It is my belief that this game was changed over time to be more reflective of the game of chess and the modern medieval mindset of the "divine right of the kings". As was mentioned earlier, Robert ap Ifan's rules seem to imply another possible answer to the Linnean rule of stopping the king from moving. If the king is required to say gwrheill and can then proceed safely between two players, this implies that if he does not say that word he cannot proceed in that fashion. It is also worthy of note that the attacker can capture the king with only two men if he first warns the defender by saying "watch your king"²⁰. Taken together these two rules may very well be the remainders of earlier rules that allowed the king to be captured in the same way as any other piece. Combined with playing to the side of the board this may very well be enough to balance the game back to something that is more playable.

To explain the Ballinderry board with it's corner markings we need only assume that the starting positions for Fithchneall are those shown in thesecond layout above. Linneaus himself may have corrupted the game by bringing in the chess capture rule for the king when he heard something similar from the lapps.

The question of who moves first does appear to be answered by the results above. It does not seem to matter appreciably which side moves first under the current rules. For this reason I would suggest that who moves first simply be left as an element of chance much as it is in our current game of chess.

As a result of these experiments it is now possible to conclude that play should conclude when the king has reached the side of the board rather than the corner. In addition it appears that the capture of the king detailed by Linneaus is incorrect. An alternate capture method has been proposed with the king liable to capture by two men instead of four. At this time a new sequence of experiments should be undertaken to determine if this produces a more even game. To further assist this investigation a new translation of Robert ap Ifan's work should be undertaken. This will also help to confirm the assumption that a player may safely move between two opponents.

End Notes

Cited References

Bell, R.C.
1969 Board and Table Games From Many Civilizations I, Oxford University Press, London

Bell, R.C.
1969 Board and Table Games From Many Civilizations II, Oxford University Press, London

Gorden, E.V.
1990 An Introduction to Old Norse, Clarendon Press, Oxford

Graham-Campbell, James
1981 Viking Artefacts A Select Catalogue, British Museum Publications Ltd, London

Interim
1980 Bulletin of York Archaeological Trust Vol 6 No 4, York, UK

Lewis, Frank
1941 Gwerin Ffristial A Thawlbwrdd, Transactions of the Honourable Society of Cymmrodorion, London

Murray, H.J.R.
1913 A History of Chess, Clarendon Press, Oxford

Murray, H.J.R.
1952 A History of Board-Games Other Than Chess, Clarendon Press, Oxford

Robinson, J. Armitage
1969 The Times of Saint Dunstan, Clarendon Press, Oxford

Ward, C.L.
1997 King's Table: Game of the Noble Scandinavians, Tournaments Illuminated Issue 121, Milpitas

Wilson, David M & Rosedahl, Else
1992 From Viking to Crusdaer, Rizzoli International Publications, New York

Bibliography

Fiske, W
1905 Chess in Iceland, Florence

Jones, Gwyn
1961 Gunnlaugs Saga Ormsstunga, Oxford University Press, New York

Magnusson, Magnus & Palsson Hermann
1965 The Vinland Sagas, Penguin Books, Middlesex

McWhite, Eoin
1946 Early Irish Board Games, Eigse: A Journal of Irish Studies Vol V, Dublin

O'Donovan, John
1847 The Book of Rights, Celtic Society, Dublin

Sterckx, Claude
1971 Les Jeux de Damier Cletiques, Societe D'Edition Les Belles Lettres, Paris

Straubhaar, Sandra
1993 Hervor's Saga, Folump Enterprises, Urbana

Return to Research Index