A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in...

15
A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park Clyde Kruskal University of Maryland, College Park

Transcript of A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in...

Page 1: A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park.

A Retrograde Approximation Algorithm for One Player Can’t Stop

James Glenn Loyola College in Maryland

Haw-ren Fang University of Maryland, College Park

Clyde Kruskal University of Maryland, College Park

Page 2: A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park.

Can’t Stop Equipment 2 to 4 players 4 6-sided dice Board with

columns 2-12 Colored markers

for each column 3 neutral markers Goal: advance to

top of 3 columns78

910

1112

65

43

2

Page 3: A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park.

Game Play

Roll 4 dice Split into 2 pairs Advance neutral markers in columns

for pair totals Roll again or stop Turn ends if no way to use pair totals

Page 4: A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park.

Example

78

910

1112

65

43

2

Page 5: A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park.

Position Graph

Similar to Backgammon or Yahtzee Bipartite (V1, V2, E) (v2, v1) in E correspond to player

choices (v1, v2) in E correspond to outcomes of

random event

Page 6: A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park.

Can’t Stop Graph

Page 7: A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park.

Comparison to Other Games

Solitaire Yahtzee No cycles

Backgammon General: long cycles Bearing off: only short cycles

Can’t Stop Cycles long but only within one turn components form a DAG

Page 8: A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park.

Can’t Stop Graph

anchor

anchors of other components

Page 9: A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park.

Retrograde Analysis Topologically sort graph Compute position value for each

vertex Start with final states Work back towards start state Player choice: compute max/min over

outgoing edges Random event: weighted average Can’t Stop: retrograde analysis on

components

Page 10: A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park.

Symbolic Analysis

f(B)=f(E) f(C)=min(f(E), f(F)) f(D)=f(A) f(A)=p1f(E)

+p2•min(f(E),f(F))

+p3f(A)A

B C D

E F

p1p2

p3

Page 11: A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park.

Symbolic Analysis for Can’t Stop f(E), f(F) given f(H)=f(F) f(G)=f(A) f(D)=p3f(G)+p4f(H) f(C)=min(f(E),f(F)) f(B)=min(f(D),f(E)) f(A)=1+p1f(B)+p2f(C)=

1+p1min(p3f(A)+p4f(F),f(E))

+p2min(f(E), f(F))A

B C

E F

D

G H

p1 p2

p3 p4

Page 12: A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park.

Numerical Analysis Make a copy of anchor component is now a DAG Guess value of f(A’) f(A) is a function of f(A’) Want fixed point

Function is piecewise linear and continuous

Fast convergence from Newton’s method

A

B C

E F

D

G H

p1 p2

p3 p4

A’

Page 13: A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park.

Retrograde Analysis for Can’t Stop

f(v) = 0 for all final states v For each non-final anchor in reverse

order of topological sort Make a copy of the anchor Topologically sort the anchor’s

component Apply Newton’s method to find value of

anchor; using retrograde analysis on each iteration

Page 14: A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park.

Results

Official game: estimated 3000 years

Dice Size Graph Size Time to Solve Optimal Turns

2 1 225 0.166 sec 1.298

2 2 1,936 0.405 sec 1.347

2 3 9,025 0.601 sec 1.400

3 1 64,372 1.70 sec 1.480

3 2 787,600 5.05 sec 1.722

3 3 4,934,006 23.3 sec 1.890

4 1 20,802,843 5 min 2.187

4 2 289,091,584 59 min 2.454

4 3 2,104,663,011 6 hr 2.700

5 1 7,105,015,062 2.8 days 2.791

Page 15: A Retrograde Approximation Algorithm for One Player Can’t Stop James Glenn Loyola College in Maryland Haw-ren Fang University of Maryland, College Park.

Future Work

Optimizations to Algorithm Better initial estimate Shortcuts when evaluating components

Distributed Algorithm Analysis of 2-player Can’t Stop