本文應用遺傳規劃交易程式來檢驗八個國家的股票指數和外匯巿場的表現，採用移動視窗的方法，測試三組獨立的期間，重新檢驗較早期的研究結果，並繼續延申探討，包括交易報酬與交易行為。實證結果顯示，不論在股票還是外匯巿場，若訓練期間的資料選擇不當，遺傳規劃的獲利表現會不理想。資料形態不但會影響遺傳規劃交易程式的獲利性，同時也決定了程式本身的一些觀察特性。我們另外分析了交易程式的複雜度、演化時間、交易頻率和一致性。交易程式的複雜度和演化時間有正向的相關性，但複雜度和報酬、以及演化時間和報酬之間都只有很弱的關係。這些發現可以讓我們更了解遺傳規劃演化交易策略的過程，有助往後更進一步的研究。

Submitted by Raul do Vale Fonseca (rauldovale@gmail.com) on 2018-09-13T17:06:19Z No. of bitstreams: 1 Dissertacao_Raul_vFinal.pdf: 1127014 bytes, checksum: aabf9f9b4023c4e2ad169118ac8ccf52 (MD5) / Approved for entry into archive by Joana Martorini (joana.martorini@fgv.br) on 2018-09-14T14:39:33Z (GMT) No. of bitstreams: 1 Dissertacao_Raul_vFinal.pdf: 1127014 bytes, checksum: aabf9f9b4023c4e2ad169118ac8ccf52 (MD5) / Approved for entry into archive by Suzane Guimarães (suzane.guimaraes@fgv.br) on 2018-09-14T15:58:08Z (GMT) No. of bitstreams: 1 Dissertacao_Raul_vFinal.pdf: 1127014 bytes, checksum: aabf9f9b4023c4e2ad169118ac8ccf52 (MD5) / Made available in DSpace on 2018-09-14T15:58:08Z (GMT). No. of bitstreams: 1 Dissertacao_Raul_vFinal.pdf: 1127014 bytes, checksum: aabf9f9b4023c4e2ad169118ac8ccf52 (MD5) Previous issue date: 2018-08-20 / Este trabalho constrói estratégias de trading sistemáticas e sintéticas com o objetivo de procurar ferramentas para replicá-las. São testados três modelos de regressão: regressão linear, regressão logística e um tipo de redes neurais artificiais, o multilayer perceptron (MLP). Para comparar a performance das regressões foram usadas três métricas de desvio do valor verdadeiro: diferenças absolutas, diferenças absolutas discretizadas e acurácia. A MLP é mais bem sucedida que as regressões logística e linear ao replicar uma estratégia trend following que usa como parâmetros médias móveis simples. Tentou-se replicar estratégias mean reversion que usam como parâmetros desvios padrão e preços máximos e mínimos num período. Nesses casos não houve clara distinção entre qual regressão foi mais bem sucedida. A acurácia dos modelos ao tentar replicar as estratégias foi maior que o sorteio aleatório em todos os casos. / In this text, systematic trading strategies are manufactured with the goal of finding tools to replicate it. Three regression models are tested: linear regression, logistic regression and one type of artificial neural networks, the multilayer perceptron (MLP). Three measures of the true value deviation were used to compare the performance of the regressions: absolute differences, discretized absolute differences and accuracy. The MLP is best between the three models to replicate one type of trend following strategy that uses simple moving average as parameter. It was attempted to replicate mean reversion strategies that uses standard deviation, maximum and minimum prices of a period as parameters. In this cases, there was no outperforming model to replicate the strategies. The accuracy of the models were better than the random guess in every test.

本研究針對台股期貨與現貨價差、成交量與技術指標融合之期貨交易策略進行獲利分析，以台股期貨與現貨的價差為主體，融合傳統技術指標和量價關係作為進場買賣台股期貨的訊號與指標，採用資料為2001年至2016年加權指數與台指期貨一分鐘資料，經過實證研究後發現，正價差放空與逆價差做多其績效表現優於正價差做多與逆價差放空，這與坊間的使用方法大為不同，另外經過實證結果，我們可以得知，若要以量價關係作為交易策略與指標，長期下來成交量增加做多與成交量減少放空績效較佳，若要以均線作為交易策略與指標，長期下來指數在均線之上做多與指數在均線之下放空績效較佳，也經由實證結果得知，價差策略可以藉由價差濾網與考量除權息因素進行調整，使價差策略績效表現更為突出，另一方面，也實證出價差策略融合成交量形成的新策略，績效表現優於價差策略融合均線形成的策略，本研究最後將價差策略融合成交量形成的新策略，考慮了價差濾網與除權息因素進行改良，並且與大盤績效進行比較，實證結果得知價差策略融合成交量作為的交易策略，績效表現可以擊敗大盤，我們最後將資料區分為兩個時間區間，將價差策略融合成交量的策略進行穩健性檢定，發現在兩個不同時間區間下，策略的績效無明顯差異，因此我們可以說此策略長期下來具有穩定性，這有利於未來進行交易。 / This study focus on profit analysis of futures trading strategy with stock price spread, quantity and technical indicators in Taiwan. With the price spread between the stock index and the futures as main topic, we fusion traditional technical analysis indicators and the relationship of trading volume and price as our signal and indicator to setup a futures trading strategy. Our research data use one-minute data frequency of Taiwan weighted stock index and Taiwan index futures from 2001 to 2016 as analysis period. The empirical result shows that to short sale if bull spread is occurred and to going long if bear spread shows up have better performance than its opposite activity, which is different from the method people use in general. This study also finds that if we attempt to utilize the relationship of trading volume and price as trading strategy and indicator, going long if trading volume increase and to short sale if trading volume decrease will work better in long run period. If we are going to use the moving average as trading strategy and indicators, that we go long for price above the moving average of the stock index and short sale for price below the moving average of the stock will more proper in long run period. Empirical results also demostrate that through spread filter and ex-dividend factor consideration spread strategy can be adjusted accordingly so that spread strategy performance can be more prominent. On the other hand, this study also proves that the performance of new strategy, formed through integration of spread strategy and trading volume strategy, is better than the integration of spread strategy and moving average strategy. Finally, this study integrates the spread strategy and trading volume strategy to formed new strategy, taking into account the improvement of the spread filter and the ex-dividend factor, then compares it with the market performance. The results show that the spread strategy integration with trading volume as a trading strategy and performance indicators can beat the market. We first divide the data into two cycles, then we perfom robustness test to the integration of spread strategy and trading volume strategy. We find out that under both cycles the strategy shows similar result. Thus, we can conclude that this strategy is stabile in long run and would be beneficial in future trading.

This diploma thesis deals with the construction of automated trading strategies, which are built on the basis of market statistics. In the first part are shortly introduced the basic parameters of the E-mini S&P 500. In this chapter is also a detailed description of the process of creating a market statistics. In the second part is described a process of building a number of automated trading strategies. Optimization of these strategies is conducted on historical data. In the second part of this thesis is also shown a discretionary trading system based on market statistics. In order to demonstrate the robustness of automated trading strategies, they are tested on a sample of data outside of the optimization period. The result of the diploma thesis will be an evaluation of the trades and evaluation of the robustness of automated trading strategies.

The aim of the diploma thesis is to test the own trading strategies on the exchange market and evaluate their success and applicability in practice. In the introduction of the diploma thesis, there are described basic parameters and basis for testing, such as tested indicators, tested time period and chosen currency pairs. In the next part of the thesis, selected indicators are compiled into three trading strategies, which are then tested . The final part consists of evaluating the results of all three trading strategies .

As regras definidas pelo Decreto 5.163/2004 trazem incentivos e penalidades aos Distribuidores no processo de apresentação de suas declarações de necessidades de compra de energia ao Ministério de Minas e Energia. Nesse sentido, é importante para os Geradores estabelecer uma metodologia robusta para prever o comportamento dos agentes de distribuição com confiabilidade razoável, de forma a permitir uma preparação adequada para os leilões de que pretendem participar e, adicionalmente, simular os cenários pós-leilões de modo a compreender os efeitos dos preços e volumes contratados no ambiente regulado sobre as condições de contratação no ambiente livre. Este trabalho propõe-se a analisar as referidas regras, apresentando um modelo de otimização utilizando Algoritmos Genéticos que simula o comportamento das distribuidoras, obtendo-se uma importante ferramenta de apoio à definição de estratégias de comercialização de uma empresa geradora. / The rules defined by the Decree 5.163/2004 bring incentives and penalties for Distribution companies to present their power purchase necessity declaration for the Ministry of Mines and Energy. In this sense, it is important for the Generation companies to establish a robust methodology for predicting Distribution companies behavior with enough accountability in order to allow an adequate preparation for the auctions in which those agents intend to participate and, additionally, simulate post auctions scenarios in order to understand the effects of prices and contracted volumes in the regulated environment over the free market contracting conditions. This work is supposed to analyze those rules, presenting an optimization model using Genetic Algorithms, which simulates Distribution companies behavior, getting an important power trading strategy decision support tool for a Generation Company.

This dissertation consists of three essays. The first essay studies the impact of indexing on stock price efficiency. Indexing has experienced substantial growth over the last two decades because it is an effective way of holding a diversified portfolio while minimizing trading costs and taxes. In this paper, we focus on one negative externality of indexing: the effect on efficiency of stock prices. Based on a sample of large and liquid U.S. stocks, we find that greater indexing leads to less efficient stock prices, as indicated by stronger post-earnings-announcement drift, greater deviations of stock prices from the random walk and greater return predictability from lagged order imbalances. We conjecture that reduced incentives for information acquisition and arbitrage induced by indexing are probably the main cause of the degradation in price efficiency, but we find no evidence supporting a direct impact from passive trading or any effect through liquidity. The second essay investigates the effect of price inefficiency on idiosyncratic risk and stock returns. I finds that price inefficiency in individual stocks contributes to expected idiosyncratic volatility. If idiosyncratic risk is priced, greater price inefficiency could be associated with higher expected returns. Consistent with this hypothesis, this paper then finds a positive relation between price inefficiency and future stock returns. This return premium of price inefficiency is not explained by traditional risk factors, illiquidity, or transactions costs. It is also evidently different from the return bias related to Jensen's inequality. This paper thus provides new insights about the determinants of expected stock returns, and new supporting evidence that idiosyncratic risk is priced. The third essay examines whether the upward return bias generated by Jensen's inequality could lead to better performance of equally-weighted (EW) indexes than value-weighted (VW) index when stock prices are not fully efficient. We find that, for a wide range of U.S. stock indexes, EW indexes deliver better four-factor adjusted returns than VW ones do even after deducting transaction costs. Consistent with our hypothesis that the outperformance of EW indexes comes from mispricing, we find that this outperformance concentrates in stocks with greater mispricing, as measured by deviation of stock prices from random walk. Findings in this essay not only imply a potentially winning investment strategy, but also provide new insight into a long-term debate on causes of the outperformance of the EW indexes. / Ph. D.

過去國內文獻大致指出投資人難以依靠券商的投資建議獲利，此與大部份國外文獻的發現相異。本文參考Barber et al. (2001)，建構一個適用於台灣股票市場的研究方法，再以四因子模型做實證。本文以2007年3月至2015年12月，共48987筆卷商個股報告為研究樣本，來探討券商報告的投資建議能否獲利。本文研究結果發現，台灣的券商報告擁有額外的資訊價值，此與Barber et al. (2001)及其他國外文獻大致相同。 本研究依券商的推薦強度建構四個投資組合。發現推薦程度高的投資組合平均月報酬為正，且高於大盤；而推薦程度低的投資組合平均月報酬顯著低於大盤，且擁有顯著的負超額報酬。本文進一步建構買進賣出策略，即買進推薦股票高的投資組合並賣出推薦程度低的投資組合，發現此策略報酬顯著高於零及大盤，且存在顯著的正超額報酬。另外在台股多頭期間，本研究的實證結果更加顯著，推薦程度高的投資組合平均月報酬增加至顯著高於大盤，且超額報酬顯著為正；推薦程度低的投資組合之大盤調整報酬及負超額報酬的顯著程度提高；而買進賣出策略獲得超額報酬的顯著程度也大幅提高。 / Past Taiwanese literatures generally indicated that it is difficult to obtain profit from Taiwanese stock recommendations of brokerage, which is different from most of foreign literatures. Referring to Barber et al. (2001), we improve and build a research methodology applied to Taiwanese stock market, conducting empirical analysis with four-factor model. From March 2007 to December 2015, we use total 48987 brokers’ stock recommendations as sample to investigate whether inventors could earn profit from the broker recommendations. Our empirical results show that Taiwanese broker reports hold additional information, which is consistent with Barber et al. (2001) and most of foreign literatures. According to the strength of recommendation, we construct four portfolios and find that the return of the most favorable portfolio is higher than market, while the return of the least favorable portfolio is significantly smaller than market and holds significantly negative access return. We further construct a long-short strategy, which buys the most favorable portfolios and shorts the least favorable portfolios. The return of this strategy is significantly higher than market, and excess return is significantly positive. During Taiwanese bull market, the significance of our empirical result improves. The significance level of market-adjusted return and access return for both the most favorable and least favorable portfolio is higher. In addition, the significance level of excess return for long-short strategy also greatly improves.

於這篇論文中，我們運White (2000)的Reality Check與Romano and Wolf (2005)的stepwise multiple test檢測交易策略的獲利性以更正資料窺探的偏誤。不同於先前運用資料窺探法則的研究，我們的研究以技術分析及時間序列預測兩者為依歸來建立交易策略，另外我們探討的市場集中在六個主要的亞洲股票市場。大致上，我們發現鮮少證據支持技術交易策略的獲利性；於基礎分析中且考慮交易成本時，只有少數幾個獲利性交易法則出現於兩個興新市場。另外在子樣本期間中，我們發現獲利性策略的表現並不穩定且這幾年來獲利性有逐漸變弱的趨勢。在進階分析中，我們發現沒有任何交易策略表現優越於基本的買進持有策略。 / In this paper, we exam the profitability of trading strategies by using both White’s (2000) Reality Check and Romano and Wolf (2005)s’ stepwise multiple test that correct the data snooping bias. Different from previous studies with the data snooping methodology, our analysis set the universe of forecasts (trading strategies) based on both technical analysis and time series prediction, and the markets which our investigation focuses on are six major Asian stock markets. Overall we find little supportive evidence for the profitability of trading strategies. Our basic analysis shows that there are only few profitable trading strategies detected for two emerging markets while transaction costs are taken into account. Moreover, the performances of the profitable strategies are unstable and the profitability becomes much weaker in the recent years as we find in the sub-periods. In further analysis, we also find that there is no trading strategies in our universe that can outperform the basically buy and hold strategy.

對投資者而言，投資股市的目的就是賺錢，但影響股價因素眾多，我們要如何判斷明天是漲是跌？因此如何建立一個準確的預測模型，一直是財務市場研究的課題之一，然而財務市場一直被認為是一個複雜.充滿不確定性及非線性的動態系統，這也是在建構模型上一個很大的阻礙，本篇研究中使用的EEMD方法則適合解決如金融市場或氣候等此類的非線性問題及有趨勢性的資料上。 在本研究中，我們將EEMD結合ANN建構出兩種不同形式的模型去進行台股個股的預測，也試圖改善ARMA模型使其預測效果較好；此外為了能夠達到分散風險的效果，採用了投資組合的方式，在權重的決定上，我們結合動態與靜態的方式來計算權重；至於在交易策略上，本研究也加入了移動平均線，希望能找到最適合的預測模型，本研究所使用的標的物為曾在該期間被列為注意股票的10檔股票。 另外，我們也分析了影響台股個股價格波動的因素，透過EEMD拆解，我們能夠從中得到具有不同意義的本徵模態函數(IMF)，藉由統計值分析重要的IMF其所代表的意義。例如：影響高頻波動的重要因素為新聞媒體或突發事件，影響中頻的重要因素為法人買賣及季報，而影響低頻的重要因素則為季節循環。 結果顯示，EEMD-ANN Model 1是一個穩健的模型，能夠創造出將近20%的年報酬率，其次為EEMD-ANN Model 2，在搭配移動平均線的策略後，表現與Model 1差不多，但在沒有配合移動平均線策略時，雖報酬率仍為正，但較不穩定，因此從研究結果也可以看到，EEMD-ANN的模型皆表現比ARMA的預測模型好。 / The main purpose of investing is to earn profits for an investor, but there are many factors that can influence stock price. Investments want to know the price will rise or fall tomorrow. Therefore, how to establish an accurate forecasting model is one of the important issue that researched by researchers of financial market. However, the financial market is considered of a complex, uncertainty, and non-linear dynamic systems. These characteristics are obstacles on constructing model. The measure, EEMD, used in this study is suitable to solve questions that are non-linear but have trends such as financial market, climate and so on. In this thesis, we used three models including ARMA model and two types of EEMD-ANN composite models to forecast the stock price. In addition, we tried to improve ARMA model, so a new model was proposed. Through EEMD, the fluctuation of stock price can be decomposed into several IMFs with different economical meanings. Moreover, we adopted portfolio approach to spread risks. We integrate the static weight and the dynamic weight to decide the optimal weights. Also, we added the moving average indicator to our trading strategy. The subject matters in this study are 10 attention stocks. Our results showed that EEMD-ANN Model 1 is a robust model. It is not only the best model but also can produce near 20% of 1-year return ratio. We also find that our EEMD-ANN model have better outcome than those of the traditional ARMA model. Owing to that, the increases of trading performance would be expected via the selected EEMD-ANN model.