隨著觀看線上遊戲直播的人越來越多，觀眾的觀看體驗也受到研究人
員的重視，這個新的研究領域叫做程序化遊戲過程生成。他的研究目標是
利用AI 產生遊戲過程，使用推薦系統推薦給用有不同喜好的觀眾。在本
論文中，以格鬥遊戲為例，我們提出一種方法產生精彩的遊戲過程，透過
結合多樣化遊戲風格和動態難度調整，設計出期望的遊戲過程。我們使用
MCTS 達成不同的遊戲風格，同時為了解決MCTS 強度不足的問題，我
們借助2021 年FightingICE 冠軍AI 的幫助。此外，就我們所知，我們
是第一個使用動作分布來表示不同的遊戲風格，使得可以更方便地結合新
的遊戲風格。最後，我們提出六種指標，分別是動作模式、累計血量差距
的變異數、連續擊中的中段次數、移動距離、來回領先次數和擊中次數來
評斷遊戲過程的精采程度，並探討各指標的重要性與對” 精彩” 的影響程
度。最後，我們邀請了49 位受測者參與我們的問券調查，結果顯示在大
部分的情境中，我們AI 產生的遊戲過程，相較其他AI 產生的遊戲過程，
我們的較為精彩。

With the growing popularity of watching game streaming, the viewer
engagement has been getting attention from researchers, which is a new
field called procedural play generation (PPG). The goal of this applica-
tion is to generate gameplays by AIs and recommend those gameplays to
viewers. We propose a method in that it combines diverse play-styles and
dynamic difficulty adjustment for entertaining gameplay generation using
fighting games as a case study. In this study, we use MCTS to perform
different play-styles. However, since achieving different play-styles with-
out sacrificing the competence of AI is not trivial, we maintain the strength
of MCTS by exploiting the AI of 2021 championship of FightingICE. Fur-
thermore, we use action distribution to represent distinct play-styles for
better mixing of play-styles for further complex design. Finally, we pro-
pose six indicators, which are ”Action Pattern”, ”Variance of Accumulated
HP Difference”, ”Combo Interruption Times”, ”Moving Distance”, ”Lead
Change” and ”Hits”, to access the entertaining level of fighting games and
analyze their relative importance. We also conduct a user study with 49
participants in total, and in most cases they agree with that the gameplays
generated by our proposed AI are more entertaining.

[1] Z. Hilvert-Bruce, J. T. Neill, M. Sjöblom, and J. Hamari, “Social mo-
tivations of live-streaming viewer engagement on twitch,” Computers
in Human Behavior, vol. 84, pp. 58–67, 2018.

[2] R. Thawonmas and T. Harada, “AI for game spectators: Rise of PPG,”
in AAAI Workshops, 2017.

[3] F. Lu, K. Yamamoto, L. H. Nomura, S. Mizuno, Y. Lee, and
R. Thawonmas, “Fighting game artificial intelligence competition
platform,” in 2013 IEEE 2nd Global Conference on Consumer Elec-
tronics (GCCE), pp. 320–323, 2013.

[4] S. Merritt and A. Clauset, “Scoring dynamics across professional team
sports: Tempo, balance and predictability,” EPJ Data Science, vol. 3,
10 2013.

[5] R. Coulom, “Efficient selectivity and backup operators in monte-carlo
tree search,” in Computers and Games, vol. 4630, pp. 72–83, 05 2006.

[6] C. B. Browne, E. Powley, D. Whitehouse, S. M. Lucas, P. I. Cowling,
P. Rohlfshagen, S. Tavener, D. Perez, S. Samothrakis, and S. Colton,
“A survey of monte carlo tree search methods,” IEEE Transactions on
Computational Intelligence and AI in Games, vol. 4, no. 1, pp. 1–43,
2012.

[7] Y. Zhao, I. Borovikov, F. de Mesentier Silva, A. Beirami, J. Rupert,
C. Somers, J. Harder, J. F. Kolen, J. Pinto, R. Pourabolghasem, J. Pes-
trak, H. Chaput, M. Sardari, L. Lin, S. Narravula, N. Aghdaie, and K. A. Zaman, “Winning is not everything: Enhancing game develop-
ment with intelligent agents,” IEEE Transactions on Games, vol. 12,
pp. 199–212, 2020.

[8] D. de Almeida Rocha and J. Cesar Duarte, “Simulating human be-
haviour in games using machine learning,” in 2019 18th Brazil-
ian Symposium on Computer Games and Digital Entertainment
(SBGames), pp. 163–172, 2019.

[9] D. Perez-Liebana, C. Guerrero-Romero, A. Dockhorn, L. Xu, J. Hur-
tado, and D. Jeurissen, “Generating diverse and competitive play-
styles for strategy games,” in 2021 IEEE Conference on Games
(CoG), pp. 1–8, 2021.

[10] A. Dockhorn, J. H. Grueso, D. Jeurissen, L. Xu, and D. P.
Liebana, “Portfolio search and optimization for general strategy
game-playing,” 2021 IEEE Congress on Evolutionary Computation
(CEC), pp. 2085–2092, 2021.

[11] R. Ishii, S. Ito, M. Ishihara, T. Harada, and R. Thawonmas, “Monte-
carlo tree search implementation of fighting game ais having per-
sonas,” in 2018 IEEE Conference on Computational Intelligence and
Games (CIG), pp. 1–8, 2018.

[12] J. T. Kristensen, A. Valdivia, and P. Burelli, “Statistical modelling of
level difficulty in puzzle games,” in 2021 IEEE Conference on Games
(CoG), pp. 1–8, 2021.

[13] A. Sarkar and S. Cooper, “Transforming game difficulty curves using
function composition,” in Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems, CHI ’19, (New York, NY,
USA), p. 1–7, Association for Computing Machinery, 2019.

[14] M. Gonzalez-Duque, R. B. Palm, D. Ha, and S. Risi, “Finding game
levels with the right difficulty in a few trials through intelligent trial-
and-error,” 2020 IEEE Conference on Games (CoG), Aug 2020.

[15] M. Ishihara, S. Ito, R. Ishii, T. Harada, and R. Thawonmas, “Monte-
carlo tree search for implementation of dynamic difficulty adjustment
fighting game ais having believable behaviors,” in 2018 IEEE Confer-
ence on Computational Intelligence and Games (CIG), pp. 1–8, 2018.

[16] M. Shakhova and A. Zagarskikh, “Dynamic difficulty adjustment
with a simplification ability using neuroevolution,” Procedia Com-
puter Science, vol. 156, pp. 395–403, 2019. 8th International Young
Scientists Conference on Computational Science, YSC2019, 24-28
June 2019, Heraklion, Greece.

[17] J. Moon, Y. Choi, T. Park, J. Choi, J.-H. Hong, and K.-J. Kim, “Di-
versifying dynamic difficulty adjustment agent by integrating player
state models into monte-carlo tree search,” Expert Systems with Ap-
plications, vol. 205, p. 117677, 2022.

[18] N. Sorenson and P. Pasquier, “The evolution of fun: Automatic level
design through challenge modeling,” in ICCC, 2010.

[19] S. Ito, M. Ishihara, M. Tamassia, T. Harada, R. Thawonmas, and
F. Zambetta, “Procedural play generation according to play arcs using
monte-carlo tree search,” in 18th International Conference on Intel-
ligent Games and Simulation, GAME-ON 2017, (United Kingdom),
p. 5, EUROSIS, 2017.

[20] R. Ishii, S. Ito, R. Thawonmas, and T. Harada, “A fighting game ai
using highlight cues for generation of entertaining gameplay,” in 2019
IEEE Conference on Games (CoG), pp. 1–6, 2019.

[21] R. Ishii, R. Thawonmas, and T. Harada, “An analysis of highlight-
oriented ai using fighting-game experience,” in 2019 IEEE 8th Global
Conference on Consumer Electronics (GCCE), pp. 551–552, 2019.

[22] R. Ishii, K. Fujimaki, and R. Thawonmas, “Fighting-game game-
play generation using highlight cues,” IEEE Transactions on Games,
pp. 1–1, 2021.

[23] Z. Tang, Y. Zhu, D. Zhao, and S. M. Lucas, “Enhanced rolling horizon
evolution algorithm with opponent model learning,” IEEE Transac-
tions on Games, pp. 1–1, 2020.

[24] D. DCM 筆記, “Logit 模型擬合實戰案例（python）——離散選擇模
型之六,” 2020