以開盤行為知識檢定在台指期市場中隨機漫步假說之合理性

Abstract

開盤氣勢為影響市場當日走勢之重要因素，往往顯示了中長線交易者之企圖心，更包含了所有市場參與者對當日大盤趨勢之期盼，因此投資者如能掌握開盤所產生之行為知識，將有利於當日之投資決策。因此本研究嘗試整合市場輪廓理論及開盤八法之概念為基礎，建構一系列研判台指期市場開盤時控制力量之長線與短線指標，透過倒傳遞類神經網路之學習，分析當日尾盤之漲跌趨勢，發掘市場之潛在知識行為，並提出交易策略進行績效評估，建構一有效的分析模型，達到輔助投資者擬定投資策略，進而檢定隨機漫步假說在台指期市場之合理性。 本研究之實驗結果顯示，整合開盤八法與市場輪廓理論所發展出之指標具有研判當日漲跌趨勢之能力，顯示市場在開盤時存在知識行為，由此推知隨機漫步假說在台指期市場並不成立。而整合開盤八法與市場輪廓理論所建構之指標相較於僅使用開盤八法建構之指標更具有研判當日漲跌趨勢之能力。此外，以市場輪廓理論所建構之指標優於開盤八法所建構之指標。而開盤八法所建構之指標預測能力並無非常顯著，但相較於隨機交易，開盤八法仍能夠小幅提升績效，經過市場輪廓理論輔助下達相輔相成之作用，使預測與獲利能力皆提升。整體而言，透過掌握市場開盤隱含之知識行為，能夠有效提升投資交易之績效。

The market opening provides an excellent opportunity to observe and evaluate the market’s underlying directional conviction. With understanding of market conviction in opening period is helpful for investors to make decision for the day. For this reason, this research is based on market profile and traditional opening patterns to construct indicators applying neural network technically trying to find out the relation between the opening period on Taiwan Index Futures Market and trend of the day. Moreover, this research tests random walk hypothesis on TAIEX Futures market and constructs a model to help investor make decision by finding out the underlying knowledge and behavior in opening period. The results of this research show that the indicators by integrating market profile and tradition opening patterns have ability to forecast the trend of the day, and have better performance than merely constructing by traditional opening patterns. Besides, this result indicates that TAIEX Futures market is not random walk. The performance of indicators constructing by market profile is better than by traditional opening patterns. The ability of forecasting and profitability is not significant but compared to the random trading, traditional opening patterns is still able to slightly improve forecasting and profitability. In conclusion, through understanding the underlying knowledge and behavior in market can effectively improve the accuracy rate of investment transactions and profitability.