i
Abuaraanial
5 \
Saarten | tame Pla ine.
6 (Teme GY
-in which we exomme -the problems caps
When we a] -lo plen ohoed in oo woth whuce
thet ogurts Atte plr'eg eda vs.
4
gn ths Chepler be hall Cover We will Covert ate fylleotne—s
@) Alans
& Optus PRCSiON In SAMES
4 ini Paw Algor ithm
G ALPHA ~ BETA PRUNING,
Atames
: eT
«A Chawe Can be do¥jnk ay a geareh
which Apeciries how flee
the Starb
yecl'om _- Mewsieg obuente —s
_. 4, ° Waifial state .
‘a ie Aeb up at— PLAYERB): DoBnes which player has the
wove in a tate s,
— ACTIONS + Return ret af legal Moves,
in a Stoke g,
=> Resuct (5,0) 1 The wiodef tran silvon me del
dofins the pusalh ep Oo move. fran state 4 with
action oo
= “Torminal - “Tost (9; A ovmi na} test ab
a Skate S. which is true when the Gee is over
an. false otbartioise .
— uTiLity (5, p) . A ubbty Prebion also
called an objechive Prclion or payoff Luncheon
clefis fe final numeric value fr a game
that ands in terminal state $ fy a player PVv,
6 mesg,
The inifial stode , ACTions Luelinn ant RESvT
Finehion defne the gene tree — & tyee phate.
He nodes are game lates and the: adges ates.
Moves - Rjuve Bhows park ef the gene tree rf
Tre-tae Te. ork node.
Hay (x)
O
bint Arg dy minimize te objectives oP mare (*).Simpl. Appxeh in Tic zfee - Tee
Jn Aim. algevittn Calculate. all The possible mevey
Arm che curve postin. Gree que of. Noughts 7
Cxosses: the yosull might lok bk.
Sci ble Moves bye fe
~ fx andl Oach rf Hose new ,
Ofkent pleyert
- (Continue Hts expansion unhl 4 Sine pocrdien
| a
fer “Yn plea i Cand
ee
ee
ps Ix \
“TE # Ls we a ct tht
~ \ \ wy rm | sy(,)
a
> THis alpowithes wilt work cod lente
PAdaMiad f oinniet movog af othe Cost ff
2" crovmons Caleul ation o4. Ono wand
et feck, a at the neh lovol O* 2 rox k
£ So on. in bheals
n=@ %)
—— (Cn)!
tak we camo caleulate
lbs and
= 9964lo.
—> “This is mb 40 big
te, ub ib is alan mig pines Novg
Crsses ig Buck a dwpl eee:
Cptinel ‘Decision in Lames
: ip Leouos at tzach
An optimal sohrobery ads lo ovtcomes 2
pod aS og Oboe Bhraters sghon One tS
as go as ong Othoe Ce
on
intalt ole opjonert
Se /
\w2 { oye fame.
, sy nich fain
fame wn WAC qd
SS Considoe & Qoro- SV” (|
oh She plyet is lane once Hy q
of ote ploy,
Je loss3 Static Evoluatin Rakion fine Comple be Row}
Columa } Diggenal open posi hon he SS Conplite. ale Ip
open for 0
Ws ts the Whuclion hou Max (x)
1S wining,
at
Shatin aryaluction is nok ad the
answot tohich plot Is ag dy Win.(Ay
a Migher the Mosult of Pi, te clocune the 0!
-laws sted, agin, “Thad me moiety awl 19% bud
only Ono moves yegew world a fin Oo
p
‘14
ue
—> Tk Lin) 15 en hak another ee io
pick Ake move bth hig hal
ginal a (0% 1m the nox Mave
Lon
noeessany te
a )
wel protect “4
> Te his purpose a Hinimax Ager Hr ae
which Th ovithm 's opponet wil]
® whatever volo “Hue
(hone Hinimax) /
pexb, in
be Aging +o mini mise
algorithm is hey -> MA KIMIZE-
Thus thee orpubot ghoul make. the Mave
Which eve. ik opporsmt Copable of. bing?
te esate damage.
a)u
P the -feso basic. Shot sis,
Ar Minimax Uses ane. +
; :
—» Erle Eypme tree tg “goavehod) ae
1
lea? nodes
Seattth +o & predoffned
—> “vee is
depth and evalua Joh
+ Con pursue Te dyce. by maleing Ayosses
| }
(
as te holy the Oppenest will ploy |
aw Cosh Rackion Can be used 4r dvaluate hors
he opponent is Waly +o peg
> A fhe ovoluatiog Bome Numbot rt
moves Ay we examine total value of
dts Ok +p 24ch plyet
> The qe! is to $d Qa move
which will marimise cthe value f our
mover BOM minimise ‘the valee =p “flea
oppoments moves,
> ~{l algerrt
rr eediure. prooded ruxp
the
bm used fo the
Hinimae A eatech P():
Te Mitra oben.
Prebivn MimMax(a) is
begin
SP Wie deep rough then
jelumn the estimaled sore of this lexP
else
lek Ni, N2, °°: Nm be he Auecessor of Q).
° . 2
4 Nis a Hin nodo then
teturn min {roamnel) | Maninguet
alse
Helurn nachna) - pransiac(e)
and MINIHAY yan Growpl Free for Mini mareAS)
ALPOUN. BETA PRON EN be
=> Minimax seeutch ts exponenlial Uke DES aval
coulda he olipineledd bub can he effec tiaty
enk in half,
= Idea of prunnly -l» obminale. Lowe pols of
ant poveliculane dochnique.
ofee drcee. can ve Wsee
examine ts called Alpe = hele pru ani
=> When applied -b o “shaadored Minimax dree, IPE
Baume move. @S minimax pou! bel
bly rrfhenee
yehurns the
branches + that cane} possibly
pruws aueg
fhe Anat decision.
MAX a4
a As
HIN 5S
/
/
/ b
m/l Nt
AAA
2 re +
Tw A Nodes Atte
in Which ik is MAx's turn to Move
Hin Dedes The Jorminal node Shed
the othet Nodes Aree
Figure! - 4 sjuo- Ply anne sores
Hay mdes”
ak ¥ nodes Ate
He ubib+y yalus for MAX ;wwe Volucs, HAX'S best move
hocause- ib bods -be Hee
ard DIN's
Jaboled woth Aeir minim
at the reek i5 Ay,
shake. with tHe highest minimax value,
best reply is by -bocause TL leads fo Hee Blefe
with lowest minimor — value
On tha Pitancuse
Hee Oh prong allen ar
Geoyme. 16 doccteihe belpid. No neod -t, ovaluate. oo
{ node C.
EISCCESSOVS os
Suppese they hove valus K ard VY. The tha value
A the root 18 5
( n
Uini|Hay rook) 2 MAX | Lin |
Hin (2,%,Y)5 Hin a oS
‘ \
= Hox (a, Hin(2,x,y),2
+ Hox ( 3,2,2°) where 2- Min la ry)er
122
22Alpha - .
a “ pring con be apieh to dros of ary
pth ik 18 oP en pssibl. bo prune entive
svoltecs yothert thon us paves.
Oo ane value of the besk
— valu’) choice we have und 4o
choice pork alow the pth Be HAX.
> pet value ef “the bask (ie, bool
. wet) Choice We hve Burd oe Luc ab
ans Choice point alory ts pot bre MIN).
pun BETA oponinh
Ale aa
(Len highest
wr ot oa
@ baew
[-2 3) BS
AL
A Di?
A rR ivA phe - beta pearch updates Hu Valu Pog arel
pw IF goes Alor @ prenes tha Nena!
Ira fe a ale. 20 Pan sd a. Ylan oP
carrerh Node jg Haron “bh be. worse then
tha Carrerbh yo o- alee.
Pv fro Max Q
Ma) reespectively