Connect6 235

NOTES

CONNECT6

I-Chen Wu 1, Del-Yen Huang l , and HSlu-Chen Chang l

Hsmchu, Taiwan

ABSTRACT

This note mtroduces the game Connect6, a member of the famtly of the k-m-a-row games,
and mvesttgates some related Issues We analyze the fairness of Connect6 and show that
Connect6 IS potentially fair Then we descnbe other charactensttcs of Connect6, e g , the high
game-tree and state-space complexities Thereafter we present some threat-based wlnnmg
strategies for Connect6 players or programs Fmally, the note descnbes the current
developments of Connect6 and hsts five new challenges

1. INTRODUCTION

Traditionally, the game k-m-a-row IS defined as follows Two players, henceforth represented as Band W,
alternately place one stone, black and white respectively, on one empty squari of an m x n board, B IS
assumed to play first The player who first obtams k consecuttve stones (honzontally, vertically or
diagonally) of his own colour wms the game Recently, Wu and Huang (2005) presented a new family of
k-m-a-row games, Connect(m,n,k,p,q), which are analogous to the tradltlOnal k-m-a-row games, except
that B places q stones Imtially and then both Wand B alternately place p stones subsequently The
addltlOnal parameter q IS a key that slgmficantly mfluences the fairness The games m the family are also
referred to as Connect games For slmpitclty, Connect(k,p,q) denotes the games Connect(ro,ro,k,p,q),
played on mfimte boards

A well-known and popular Connect game IS jive-m-a-row, also called Go-Moku Go-Moku m the free
style (Without any restnctlOn on B) IS Connect(I5,I5,5,I,I) However, fairness has been a major Issue for
Go-Moku The game IS known to favour B over W when played m the free style After each move made
by B, B has one more stone than W, whtle W only has the same number of stones as B after hls 3 move In
order to reduce thiS unfairness, the Japanese ProfesslOnal RenJu AssoclatlOn (1903) Imposed some new
rules to restnct the play ofB for professlOnals For example, B IS forbidden from playmg double three and
double four Expenences of professlOnals mdicate that the game With these restnctlOns, known as RenJu,
sttll favours B Theoretically, B has been proved to wm m the free style by AlliS (1994) and Alhs, Van den
Henk and HuntJens (1996), and under RenJu restrictions by Wagner and VIrag (2001) The RenJu
InternatlOnal Federatton (RIF) (1998) attempted to mcrease the fairness of the game by Imposmg new
rules for the first five moves The RIF (2003) called agam for new openmg rules and listed the
reqUirements for new rules, mdlcatmg that the current rules still need to be Improved for thiS game
However, addmg more rules also mcreases the difficulty of learnmg the game BeSides, the fairness
problem for Go-Moku or RenJu has an Important Side effect ofreducmg the board size Sakata and Ikawa
(1981) mentioned that mcreasmg the board size raised B's advantage Hence, the standard board size was
set to 15 x 15 Indeed, a smaller board lowers the compleXity ofthe game and thus makes It easier to solve
the game

Among the Connect games, Connect(m,n,6,2,l) games for all m and n, referred to as Connect6, are
mterestmg research tOpICS for the nottons of fairness and game compleXity Connect6 games are mtUltlvely

I Department of Computer SCience and InformatIOn Engmeenng, NatIOnal Chiao Tung UnIversity, Hsmchu, Taiwan
EmaIl ICWU@CSIe nctu edu tw, {deryann,hc}@Java cSle nctu edu tw
2 PractIcally, stones are placed on empty mtersectlOns of RenJu or Go boards
3 For breVity we Will use the pronoun he (hiS) where he or she (hiS or her) IS meant
236 ICGA Journal December 2005

fair, m the sense that one player always has one more stone than the other after makmg each move After
Connect6 was presented by Wu and Huang (2005) dunng the II th Advances m Computer Games m
TaipeI, Taiwan, tens of thousands of players, mcludmg many RenJu dan players, played thIS game on a
TaIwan game sIte developed by ThmkNewIdea Inc (2005) So far, these dan players have not been able to
IdentIfy whIch player the game favours In SectIOn 2 we return to thIS tOpIC

As for the board SIzes of the Connect6 games, Wu and Huang (2005) set them to mfimty, 1 e ,
Connect(6,2,1), to maxImIze the game complexItIes However, for playmg games at the Computer
OlympIad we recommend boards wIth 19 x 19 squares or 59 x 59 squares m order to make the game
feasIble to play For Connect(19,19,6,2,1), players can sImply use Go boards to play For
Connect(59,59,6,2,1), players can play on computers or by puttmg 3 x 3 Go boards together Note that 3 x
3 Go boards can be put together to form a board wIth 59 x 59 squares smce placmg two Go boards
together creates one addItIonal Ime between the two boards The Connect(19,19,6,2,1) game IS mitIally
recommended for Computer OlympIad game contests, whIle Connect(59,59,6,2,1) could be used m games
among profeSSIOnals or used m tIe-break games m the future

The remamder of thIS paper IS structured as follows SectIon 2 dIscusses the faIrness Issue of the Connect
games SectIon 3 descnbes the charactenstIcs of the Connect6 games SectIOn 4 descnbes the threat-based
wInnmg strategIes for Connect6 players and programs SectIOn 5 presents some of the current
developments m Connect6 Fmally, SectIOn 6 summanzes thIS paper and lIsts five new challenges

2. FAIRNESS

ThIS sectIOn dIscusses the faIrness Issue of Connect games SubsectIOn 2 I provIdes the defimtIOns of
faIrness SubsectIOn 2 2 uses the strategy-stealIng arguments to show faIrness SubsectIOn 2 3 raIses the
faIrness Issue related to breakaway moves SubsectIOn 2 4 mvestIgates the fairness of Connect6, whIle
SubsectIOn 2 5 mvestIgates the faIrness of other Connect games

2.1 Definitions

In Van den Henk, UiterwIJk, and Van RIJSWIJck (2002) an adequate defimtIOn of faIrness IS gIven It
reads "a game IS consIdered fair If It IS a draw and both players have a roughly equal probabIlIty of
makmg a mIstake" However, practIcally, It IS hard to have a perfect model for ca1culatmg the probabIlIty
of makmg a mIstake, smce some undIscovered strategIes such as makmg breakaway moves, as descnbed
m SubsectIon 2 3, may result m dIfferent probabIlItIes On the contrary, It IS relatIvely easy and possIble to
show when a game IS unfair Below we provIde three dIstmct defimtIOns for unfair games

Definition 1: A game IS definitely unfair, If It has been proved that some player wms the game
For example, smce B wms m Go-Moku (m the free style) as descnbed above, Go-Moku IS defimtely
unfaIr

Definition 2: A game IS monotonically unfair, If It has been proved that one player does not wm the game,
but It has not been proved for the other player
For example, for Connect(k,p,p) or Connect(m,n,k,p,p), W cannot wm, based on the so-called strategy-
stealzng argument, as descnbed III SubsectIOn 22 Thus, Connect(6,1,1), Connect(7,I,I) and
Connect(6,2,2) are all monotomcally unfaIr, smce B has not been proved to wm or tIe m these games
However, smce Connect(8,1,1) has been proved to be a draw by Zetters (1980), It IS not monotomcally
unfaIr

Definition 3: A game IS emplrlcally unfair, If most players, m partIcular profeSSIOnals, have claImed that
the game favours some player
For example, before Go-Moku was solved, Go-Moku was empmcally unfaIr, smce most profeSSIOnals
claImed that B would wm

Definition 4: A game IS consIdered potentzally fair, If It has not yet been shown or claimed to be defimtely
unfaIr, monotomcally unfaIr, or empmcally unfaIr
 Connect6 237

This defimtlOn Indicates that a potentially fair game for the time beIng may not remaIn potentially fair In
the future If a gaIlle remaInS potentially fau any longer, It could have a higher chance to be fair

2.2 Strategy-Stealing Argument

The strategy-stealing argument was first raised by Nash In 1949 (cf Berlekamp, Conway, and Guy, 1982)
and used by other researchers, such as Hales and Jewett (1963), CSlrmaz (1980), and Pluhar (1994), In the
area of combInatonal game theory The argument shows that W does not have a wInmng strategy In
Connect(m,n,k,p,p) Assume by contradictIOn that W has a WInnIng strategy, say S Consider the
followmg B makes the first move at random Then, B Simply follows S (stealmg W's strategy) which
leads to a WIn If the strategy S reqUires B to place on the squares where B placed earlier, B chooses other
squares at random agaIn, Instead Thus, B still remaInS follOWIng the strategy S to WIn the game,
contradictory to the assumptIOn

Based on the same argument, we can also show the follOWIng the higher the value of q In a Connect
game, the higher the chance of B wInnIng Similarly, assume that B has a wmmng strategy S In
Connect(m,n,k,p,q) Then, B Simply follows the strategy S to WIn In Connect(m, n, k,p, q+ 1), as above The
result hInts that the parameter q In Connect games IS a key Significantly InfluenCIng the fairness More fair
and InterestIng Connect games may be found by adjustIng thiS parameter

2.3 Breakaway and Unfairness

ThiS subsectIOn Investigates the fairness Issue related to breakaway moves For a Connect game, a
breakaway move means to place stones far away from the major battle field, where most stones have been
placed W's first move (after B's first move) IS called an mltlal breakaway move, If It IS also a breakaway
move

For Connect(k,p,q) the follOWIng holds IfW makes an Initial breakaway move Without a penalty and B IS
or will be forced to defend Itself agaInst W's move, then the game IS played like Connect(k,p,p) With W
plaYIng first Such games are highly likely to be monotomcally unfair or even defimtely unfair Therefore,
It IS Important to prove that W has a penalty for the Initial breakaway move

2.4 Connect6

Connect6 games are mtUltlvely fair, In the sense that one player always has one more stone than the other
after makIng each move ThiS subsectIOn shows that Connect6 IS, at least, potentially fair for the time
beIng, based on the follOWIng three observatIOns

1 Connect6 has not yet been shown to be defimtely unfair

2 Connect6 has not yet been shown to be monotomcally unfair Most Importantly, Wu and Huang
(2005) proved that B WInS for Connect(6,2,3), hIntIng that W loses If makIng an Initial breakaway
move Due to the penalty, Connect6 cannot be shown to be monotomcally unfair or defimtely unfair
from the argument In Subsection 2 3
3 Connect6 has not yet been claimed to be empmcally unfair Connect6 has been played by tens of
thousands of players IncludIng many RenJu dan players, over an online game system developed by
ThmkNewIdea Inc (2005) So far, none of these dan players have been able to claim which player,
If any, the game favours

The above three POInts conclude that Connect6 IS potentially fau for the time beIng Further eVidence IS
surely needed to determIne the fairness of this game or understand thiS game better

2.5 Connect Games

In addition to the research for Go-Moku and RenJu, many other studies have been undertaken relatIng to
the fairness of k-In-a-row Zetters (1980) proved that W ties when k ~ 8 Many solved Connect games With
p = q = 1 are listed by Van den Henk, UlterwlJk, and Van RIJswIJck (2002)
238 ICGAJoumal December 2005

Smce W does not wm m Connect(m,n,k,p,p) based on the strategy-stealmg argument as descnbed m

SubsectIOn 2 1, many researchers followed an asymmetnc versIOn of rules, called Maker-Breaker, for
slmphclty of combmatonal analysIs In this verSIOn, W IS not allowed to wm In contrast to Maker-
Breaker, the versIOn usmg the nonnal rule (without the restrIctIOn) IS called Maker-Maker Let 8 denote k
- p In the Maker-Breaker verSIOn, Pluhar (2002) proved that W ties under the followmg conditIOn C 8"2>.
80 log2 P + 160 and p "2>. 1000 The condition C IS roughly like 8 = Q(log2 p) This result Imphes that W
stIll ties under the same conditIOn m the Maker-Maker versIOn The result can be easily extended to the
followmg corollary

Corollary 1 For Connect(k,p,q), let k and p satIsfy the conditIOn C (as defined above) For all q, where 1
:0; q :0; p, both Band W tIe 0

Now, mvestlgate unfair Connect games (where either B or W wms) First, B wms when p < Lq/8 2j (48 +
4) For example, If B places 8 2 stones on 8 x 8 squares as a group, then W reqUIres 48 + 4 stones to
defend the group Thus, the above result IS obtamed when B lets Lq/b 2j groups be far away from one
another If q IS not a multiple of 8 2, we can obtam tIghter results For example, for 4 x 4 squares, B can
add 8 addItIOnal stones around the four comers (Wu and Huang, 2005) such that W reqUIres more whIte
stones to defend for each additIOnal black stone If (q mod 8 2) :0; 8 Lq/8 2j, B can put (q mod 8 2) stones
around the comers of the Lq/8 2j groups, and therefore B wms when p < Lq/8 2j (48 + 4) + (q mod 8 2)
Otherwise, B can put 8 Lq/8 2j stones around the comers of the Lq/8 2j groups, and therefore B wms whenp
< Lq/8 2 j (48 + 4) + 8 Lq/82 j Thus, the followmg corollary IS obtamed

Corollary 2 Let 8 = k - P For Connect(k,p, q) games, B WinS when

p < Lq/8 2j (48 + 4) + mm( q mod 8 2, 8 Lq/8 2 j) 0

Moreover, empmcal expenments by Wu and Huang (2005) suggest the followmg conjecture Connect
games WIth 8:0; 3 are likely to be empmcally unfair From the above, an open problem IS whether there are
some more fair and mterestmg Connect games, where (i> 3, (i= O(lOg2P), andp"2>. Lq/8 2j (48+ 4) +
mm(q mod 8 2,8 Lq/8 2j)

3. CHARACTERISTICS

The four mam charactenstlcs of Connect6 are hsted below

1 Connect6 IS potentially fair, based on the argument m SubsectIOn 2 4

2 The rules of Connect6 are qUIte simple to learn In contrast, RenJu mcludes some prohIbited
moves and InternatIOnal RenJu even mcludes an addItIOnal number of openmg rules
3 Connect6 IS a symmetrIc game, If the first move by B IS disregarded
4 Both game-tree and state-space complexities for Connect6 are qUIte high, as descrIbed below

We first consIder the game-tree complexIty of Connect6 ThiS complexity IS much higher than that of Go-
Moku and RenJu, smce placmg on two squares per move mcreases the branchmg factor by about a factor
of half of the board sIze For Connect(l9,19,6,2,J), assume that the average game length IS still 30, the
same as that for Go-Moku (AlliS, 1994) Smce the number of squares on which one stone can be placed IS
about 300, the number of posSibIlities for each move IS about (300*300/2) The game-tree complexity IS
thus approximately (300*300/2)30 "" I 0 140 , which IS slgmficantly hIgher than that of Go-Moku The
complexity IS even higher when usmg a larger board such as Connect(59,59,6,2,1)

Then we consider the state-space complexity The state-space complexIty of Connect( 19,19,6,2,1) IS 10 172 ,
almost the same as that m Go ThiS complexity IS much higher when a larger board, such as
Connect(59,59,6,2,1), IS used
 Connect6 239

4. THREAT-BASED WINNING STRATEGIES

Analogously as in Go-Moku or Renju, threats are the key to winning Connect6 games. This section
describes threat-based winning strategies for Connect6 players and programs. Subsection 4.1 defines
threats for Connect6. Subsection 4.2 describes some winning strategies with threat-based search.

4.1 Threats

Definition 5. For Connect6, assume that one player, say W, cannot connect six. B is said to have t
threats, if and only if W needs to place t stones to prevent B from winning in B's next move. D

++~ •••• III

.
•I
(b)

(c)

Figure I: Threat patterns for Connect6. (a) One threat, (b) two threats and (c) three threats.

Threats in Connect6 are defined in Definition 5. In Figures 1 (a), (b), and (c), B has one, two, and three
threats, respectively. A move is called a single-threat move, if the position has one threat after the move;
and a double-threat move, if the position has two threats. In the case of three threats, B wins because W
requires three stones to defend but only has two stones for a single move. Therefore, a winning strategy is
to have at least three threats.

Lemma 1. In Connect6, consider a single line only. Placing one stone on the line, if it does not connect
six yet, increases the number of threats by at most two.

Proof. Let B place a stone on square s of the line. For each side of s, let W place one stone on the empty
square closest to s. Then, by placing these stones, at most two (one for each side), W must be able to
defend all the new threats created by s. This implies that placing one stone on the line increases the threat
number by at most two. D

Definition 6. In Connect6, consider a single line only. The line includes a dead-l threat for one player,
say B, if B only needs to add (4-1) additional stones to generate one threat. Similarly, the line includes a
live-l threat for B, if B only needs to add (4-1) additional stones to generate two threats. D

11+1++11 (a)

11.11.11 11,111.1.1
(b)
+>.'1 I I I I I•• ~
+~) 1+1++11 11+11++11 111+1+11+11
(c)

--II-I I-I I
(d)

Figure 2: Live-l and dead-l threats for Connect6. (a) live-3, (b) live-2, (c) dead-3, and (d) dead-2 threats.
240 leGA Journal December 2005

Lemma I shows that placing one stone on a line increases the number of threats by at most two. Definition
6 defines the dead-l and live-l threats according to the number of stones that a player must place
subsequently in order to generate one or two more threats. For example, Figure 2 illustrates the cases of
live-3, live-2, dead-3, and dead-2 threats.

Live-3, live-2, dead-3, and dead-2 threats are also important in Connect6, since one more move (two
stones) can transform them into real threats. Among the four threats, live-3, live-2, dead-3 threats are also
called highly potential threats or HP-threats, since one stone can create at least one threat or two stones
can create at least two threats. Among HP-threats, live-3 and dead-3 threats are also called HP3-threats,
which needs only one more stone to have at least one threat. Players usually want to associate real threats
with HP-threats, while attacking. This is a rather useful strategy to adopt when playing Connect6.

4.2 Threat-Based Search Strategies

(a) (b) (c)

Figure 3: Winning sequences in three positions, respectively starting from the moves
(a) at (10,11), (b) at (14,15), (c) at (18,19).

Since Connect6 was presented, many players have leamed how to win by making double-threat moves.
Double-threat moves are quite powerful, since the opponent is forced to reply to both threats. For
example, in Figure 3 (a), after the move at (22, 23), W keeps making double-threat moves to win the
game.

Recently, professionals such as Lee (2005) found some Tsumegos that need to mix single-threat and
double-threat moves together. For example, in Figure 3 (a), the winning sequence of W starting from (10,
II) needs two single-threat moves at (14,15) and (18,19); in Figure 3 (b), the winning sequence from
(14,15) needs one single-threat move at (22,23); in Figure 3 (c), the winning sequence from (18,19) needs
one single-threat move at (22,23). The above three Tsumegos have also been solved by Wu and Chang
(2006).

However, the search trees mixed with single-threat and double-threat moves are usually tremendously
large. In order to reduce the sizes of search trees, Connect6 players and programs should search single-
threat moves only under some conditions. Two heuristics for the conditions are as follows.

1. The opponent, called Defender, has dead-3 threats but no real threats and no live-3 threats, while the
player, called Attacker, has one HP3-threat plus at least two additional HP-threats (without counting
the single real threat). An example is the move at (22, 23) by Attacker, W, in Figure 3 (b). Defender, B,
has only one dead-3 threat and no live-3 threats, while Attacker has one real threat, two HP-3 threats
and two HP-threats.
 Connect6 241

After placmg one stone to defend Attacker's smgle threat, Defender can place the other either to make
a smgle-threat move or to defend one of Attacker's HP-threats For the former, Attacker can place one
stone to reply to Defender's smgle threat, while keepmg placmg the other to make threats m the next
move For the latter, smce Defender has no threats, Attacker can use two HP-threats to make a double-
threat move

2 Defender has no real threats and no HP3-threats, while Attacker has at least two additIonal HP-threats
For example, the move (22,23) m Figure 3 (c)

Smce Defender has no HP3-threats, Defender places one stone to defend Attacker's smgle threat, while
placmg the other to defend one of these HP-threats Thus, Attacker can stIll make threats m the next
move

For Connect6 programs, It IS cntIcal to design good heunstIcs for the conditIOns of makmg smgle-threat
moves while makmg the tree size small

5. CURRENT DEVELOPMENTS

Connect6 has attracted some attentIOn smce ItS first presentatIOn by Wu and Huang (2005) One game
company, ThmkNewldea Inc (2005) has already supported an onlme game system for Connect6, which
has been played by tens of thousands of players In addition, some RenJu dan players m RenJu clubs m
Taiwan also play, study, and teach wmnmg strategies of Connect6 Lee (2005), a RenJu dan player,
mltIated a Connect6 forum, where several Trumegos and 10sekls have been publIshed No players have yet
been able to IdentIfy which player, If any, the game favours

A Connect6 program wntten by Wu and Huang (2005) used an alpha-beta search tree of depth 3 With
threat-space search based on double-threat moves only ThiS program was also connected to the onlme
game system supported by ThmkNewldea Inc (2005) to play agamst human players automatically The
program IS strong enough to wm agamst about 70 per cent of the players A new program wntten by Wu
and Chang (2006) mcorporated smgle-threat moves mto threat-space search as descnbed m SubsectIOn
42 ThiS program was used to solve several pOSItions mcludmg the three m Figure 3

6. SUMMARY AND NEW CHALLENGES

ThiS note mvestlgated several Issues related to the game Connect6 For a qUIck overview we lIst them
below

1 It discussed the fairness Issue of Connect games and showed that Connect6 IS still potentially fair
2 It analyzed other charactenstIcs of Connect6, e g , high game-tree and state-space compleXItIes
3 It presented some threat-based wmnmg strategies for Connect6 players or programs
4 It descnbed the current developments of Connect6

Smce Connect6 IS qUIte new, several new challenges are still open Below we lIst five of such challenges

Present more mterestmg Trumegos or 10sekIs for Connect6

2 Develop new search techmques for deslgnmg Connect6 programs and for solvmg more positIOns
3 Propose fairness models to assist assessmg the fairness of Connect 6 or general Connect games
4 Solve more Connect games, especially m the cases of 5( = k- p) > 3 and 5 = O(lOg2 p)
5 IdentifY some more mterestmg ones among potentially fair Connect games
242 lCGA Journal December 2005

ACKNOWLEDGEMENTS

The authors would lIke to thank the anonymous referees for theIr valuable comments, whIch greatly
Improved thIs note The authors would also lIke to thank Professor Jaap van den Henk for hIs help and
support

7. REFERENCES

AllIs, L V (1994) Searchmg for solutIOns m games and artlficzal mtelligence, Ph D ThesIs, UnIversIty
of LImburg, Maastncht

AllIs, LV, Henk, H J van den, and HuntJens, MPH (1996) Go-Moku Solved by New Search
TechnIques ComputatIOnal Intelligence, Vol 12, pp 7-23

Berlekamp, E R, Conway, J H, and Guy, R K (1982) Wmnmg Ways for your Mathematical Plays,
Vol 2, Games m General AcademIc Press, London

CSlrmaz, L, (1980) On a combmatonal game wIth an applIcatIOn to Go-Moku, Discrete Math, Vol 29,
pp 19-23

Hales, A W , and Jewett, R I (1963) Regulanty and posItIOnal games TransactIOns of the Amencan
MathematIcal SocIety 106, pp 222-229

Henk, H J van den, UlterwlJk, J W H M , and RIJSWIJck, J V (2002) Games solved Now and m the
future Artzjiczal Intelligence, Vol 134, pp 277-311 ISSN 0004-3702

Japanese ProfeSSIOnal RenJu AssocIatIOn, (1903) HIstory of RenJu Rules,

http Ilwww renJusha net/database/oldrule htm

Lee, S W (2005) Josekl and Tsumegos for Connect6 In Connect6 Forum, (m Chmese)
http Ilgroups msn comlconnect61

Pluhar, A (1994) GeneralIzatIOns of the game k-m-a-row, Rutcor Res Rep 15-94

Pluhar, A (2002) The accelerated k-m-a-row game, Theoretical Computer SCience Vol 271, Nos 1-2,
pp 865-875 ISSN 0304-3975

RenJu InternatIOnal FederatIOn (1998) The InternatIOnal Rules of RenJu,

http Ilwww renJu nu/nfrules htm

RenJu InternatIOnal FederatIOn (2003) MOM for the RIF General Assembly,
http Ilwww renJu nu/wc2003/MOM_RIF_030805 htm

Sakata, G and Ikawa, W (1981) Flve-In-A-Row Renju The Ishl Press, Inc, Tokyo, Japan

ThmkNewldea LImIted (2005) CYC game web sIte (m Chmese) http Ilwwwcycgame com

UlterwlJk, J W H M , and Henk, H J van den (2000) The advantage of the InItIatIVe, InformatIOn
SCiences, Vol 122, No 1, pp 43-58
Wagner, J and VIrag, I (2001) Solvmg RenJu, ICGA Journal, Vol 24, No 1, pp 30-34

Wu, I-C and Chang, H -C (2006) Threat-based proof search for Connect6 TechnIcal report, Department
of Computer SCIence and InformatIOn Engmeenng, NatIOnal ChIao Tung UnIVerSIty, Hsmchu, TaIwan
http IIJava cSle nctu edu tw/~lcwu/technIcal-reports/trl-2006 pdf

Wu, I-C and Huang, D -Y (2006) A New FamIly of k-m-a-row Games The 11 th Advances m Computer
Games (ACG 11) Conference, TaIpeI, TaIwan (to be publIshed)

Zetters, T G L (1980) Problem S 10 proposed by R K Guy and J L Selfndge Amer Math Monthly 86
(1979), solutIOn 87(1980), pp 575-576