BIG NIKKEI [INSERT WITTY TAGLINE] QF206 March2017 Presentation by: Joseph Kang, Noel, Ng Min Min, Li Lin, Paw Jia Xun BIG NIKKEI [INSERT WITTY TAGLINE]
Index Introduction 1. Trading Strategies 2. Performance Metrics 3. 4. Tick-by-Tick Data Analysis 5. Conclusion
Overview: Big Nikkei Futures (Large Contracts) Index Overview Project Objective Index: Big Nikkei Bloomberg Ticker: NK Nikkei 225 Futures is based on The Nikkei Stock Average which is a major Japanese stock index used as a benchmark in various financial instruments To find the optimal holding period by utilizing TD strategies with conservative & non-conservative approaches in conjunction with the 5 different alternative strategies Compare and contrast the P&L generated by the various strategies against a set of performance metrics Project Assumptions Min by min data in separate csv files for different days Per contract commission: 1000 Yen per round trip Trade on Open Maximum 20min holding period 70% in-sample, 30% out-sample Bloomberg Ticker NK Street Name Big Onshore Before July 11 for onshore, July 19 for offshore 8:00 AM to 2:09 PM Default 7:45 AM to 2:09 PM Commission per contract ¥400
2. Trading Strategies
Trading Strategies: TD Countdown & Alternative Strategies Entering into a Long Position TD Countdown Comparing the current close with the low two bars earlier for a potential buy Comparing the current close with the high two bars earlier for a prospective sell Alternative Strategies: Requirements to Enter into a Long Position TD Camouflage The price of the current price bar must be below the close of the previous price bar The close of the current price bar must be above the open of the current price bar The low of the current price bar must be lower than the true low two price bars earlier TD Clop The open of the current price bar must be below the close and open of the previous price bar The market must subsequently trade above both the open and close of the previous bar TD Clopwin The open and close of the current price bar must be contained within the open and close range of the previous price bar The close of the current price bar must be above the close of the prior price bar TD Open The current price bar’s open must be less than the low of the prior price bar It must then trade above that low TD Trap The open of the current price bar Must be within the range of the previous price bar Must then break above the high of that range
Model Upgrade: Inverse Optimiser Inverse Optimiser Description Example: Holding Period: 10 Clop Sell Inverse ? No Yes T-Statistic (In-sample) -.27853 1.1652 T-Statistic (Out-sample) -3.0733 2.1033 Sharpe (In-sample) -0.1783 0.0746 Sharpe (Out-sample) -0.3104 0.2125 Risk-Reward (In-sample) 1.1185 1.5656 Risk-Reward (Out-sample) 0.8140 1.7331 St. Dev. (In-sample) 19,283.8957 St. Dev. (Out-sample) 20,411.7306 P&L (In-sample)1) -83,9000 35,1000 P&L (Out-sample)1) -62,1000 42,5000 Key Feature of Model: Inverse Optimiser With incoming signal, making a “sell” instead of “buy” Comparison between an inversed and non-inversed trade Working on the basis of recycling a “bad” strategy/ signal, Making a profit from a supposed loss P&L are trade-based , inclusive of commissions
3. Index Performance
Performance Metrics: In-Sample T-Statistics for All Trading Strategies In-Sample T-Statistics against Holding Period for All Strategies
Performance Metrics: Breakdown of In-Sample T-Statistics Countdown Buy & Sell Clop Buy & Sell Clopwin Buy & Sell Camouflage Buy & Sell Open Buy & Sell Trap Buy & Sell
Performance Metrics: T-Statistics for Selected Trading Strategies Countdown Buy Countdown Sell Camouflage Buy Trap Sell
Performance Metrics: Reward-Risk Ratio Risk-Reward Ratio for In-Sample Risk-Reward Ratio for Out-Sample
Performance Metrics: Sharpe Ratio Sharpe Ratio for In-Sample Sharpe Ratio for Out-Sample
Performance Metrics (3): Equity Curve & Drawdown In samples Out samples
Performance Metrics (3): Equity Curve & Drawdown In samples Out samples
Performance Metrics (3): Equity Curve & Drawdown In samples Out samples
Performance Metrics (3): Equity Curve & Drawdown In samples Out samples
4. Tick-by-Tick Data Analysis
Preliminary Data Findings In-Sample Ratio X Y Total PnL Trade Count Average PnL St. Dev. Sharpe T-stat RR Ratio 6 10 40000 1852 21.59827214 20.59314048 1.048809052 45.13536075 -1.490390758 1479 21.23826334 -1.413245315 7 -3000 1526 -1.965923984 44.10631711 -0.04457239 -1.741178085 -1.41575381 -10000 1726 -5.793742758 20.65803442 -0.280459537 -11.65173524 -1.441587069 -21000 1576 -13.3248731 43.44745939 -0.30668935 -12.17521957 -1.382809092 9 -50000 1561 -32.03074952 21.93215922 -1.460446699 -57.70150793 -1.435581688 Out-Sample Ratio X Y Total PnL Trade Count Average PnL St. Dev. Sharpe T-stat RR Ratio 6 10 -1096000 821 -1334.957369 32.72022733 -40.79914713 -1169.021943 -1.445965094 Outcome Using optimal X = 6, and Y = 10; While t-stat of in-sample is fairly large = 45.13536075, t-stat of out-sample = -1169.021943 Therefore highlighting the fact that the strategy is not sufficiently robust
5. Conclusion
Several strategies with holding period that generate high t-statistics (>2) for in-sample tests Countdown Buy: 3.75 Countdown Sell: 3.74 Camoflouge Buy: 2.37 Trap Sell: 2.37 Could not replicate the same performance out-sample tests Countdown Buy: - 0.59 Countdown Sell: - 1.27 Camoflouge Buy: - 1.23 Trap Sell: - 1.82 In-sample vs Out-sample equity demonstrated contrasting performances Unlikely that the strategies tested generate alpha