Let me start with old school as long as you are opening only buy (long) positions, you can call yourself an investor. Optimal TAA model is adjusting the weight of each asset class (Stocks, Bonds, Gold and Crude oil) in portfolio on daily base, while Total TAA chooses the asset that has the highest growth potential in the next 24 hours. Even it is possible to use models for ETFs (SPY-TLT- GLD- USO), it is primary developed for futures (ES-ZN-CL-GC). The chat below shows the hypothetical movement of portfolio using TAA models (cost adjusted): 10,86 *Average 3M T-bill rate (1,64%) is used as risk free rate 7,25 *Cost are calculated using Interactive Brokers commissions interactivebrokers.com INSTRUMENTS: ES S&P 500 futures ZN T-notes 10 futures GC Gold futures CL Crude oil futures Front contract (rolled) 1,87 Daily data (21,00 GMT +0:00) from 2 Jan 1993 to 8 Feb 2013 Source: Trading Blox (tradingblox.com) Return: X10 = 900% The chart above shows the model result trading daily intervals on close price, but let s see how would it work if we trade daily interval in other hours: Volume (as % of total daily volume): Time: New York London Data source: Disk Trading http://disktrading.is99.com It is evident that TAA model return are negatively correlated with market volume, because high volume usually causes more volatility and in this situation markets often indicate wrong signals (conversely, in the last year model noted best performance in the most liquid period). If we take a look on table that shows the liquidity expressed as turnover (price
x volume x contract size) of 12 most livid futures on GLOBEX, it is obvious that used instruments are some of most liquid instruments in the world but every market has it`s limits: 100 10 1 12 Most liquid GLOBEX futures: Gold & Crude Oil Turnover per min 1 (GMT -5:00) Tick Contract Daily turnover Time GC CL Time GC CL ZN T-notes 10 138.267.711.582 0:00 3.246.855 469.838 12:00 10.342.218 12.279.459 ES E-min S&P 129.947.087.171 1:00 2.765.459 674.058 13:00 8.948.428 9.936.926 CL Crude oil 55.431.773.025 2:00 3.293.468 824.266 14:00 6.148.824 12.412.097 ZB T-bond 30 52.127.450.564 3:00 4.671.901 1.753.264 15:00 4.098.277 3.711.552 EC EUR 43.957.564.461 4:00 5.790.830 2.496.341 16:00 5.738.273 3.740.625 ND Nasdaq 32.988.632.501 5:00 4.340.213 2.214.713 17:00 806.378 641.041 GC Gold 29.092.942.310 6:00 4.207.161 2.081.645 18:00 5.311.515 3.612.761 ZS Soybeans 14.397.579.548 7:00 4.894.070 2.837.283 19:00 1.651.257 500.955 JY JPY 13.511.313.618 8:00 6.947.599 4.621.640 20:00 4.293.196 828.138 AD AUD 13.150.321.015 9:00 15.612.206 22.099.814 21:00 4.274.615 781.350 NG Natural gas 10.734.118.271 10:00 24.073.217 19.314.037 22:00 2.343.403 701.234 BP GBP 10.125.839.936 11:00 16.610.199 20.205.480 23:00 1.856.782 587.251 Even CL futures (in total) have higher turnover than GC, it is more distributed thought different maturity contracts, so the first (usually most liquid) CL contract makes only 54% of total turnover. The liquidity can be increased with other liquid CL contracts (which can increase costs) and with WBS Crude oil futures (ICE). All in all, it would not be a problem to trade TAA model with 10M, but 50M would be too much. I don t know how would Brent oil or other bond / index futures listed on other exchanges (I have only data from GLOBEX) fit in TAA model, but it is possible that they would have a positive effect. CL volume distribution: CHART BEHIND THE TEXT SHOWS THE TOTAL RETURN OF EQUAL WEIGHT PORTFOLIO WITH STOCKS, BONDS GOLD AND COMMODITIES FUTURES FROM 1970 2012; US MARKET; Source: Bloomberg Well, you can`t make miracle with 4 instrument betting only buy side on daily interval, but since Tactical asset allocation is my doctoral dissertation title, like one of my first models, I have putted it in the presentation. What is my point as long as money supply growth surplus the GDP growth it will bring inflation and this is the main reason why financial markets are growing (+real economy growth). How economy goes thought cycles, some asset class will fall other will grow, but in the long run financial markets are growing. 2008 was the biggest crash from 1931, so it is not logical to expect such a big drawdown in next few decades. So, you want to keep things simple: buy everything and hedge it (options, VIX), put some leverage or find a way to improve this buy and hold (like TAA model). My vision 2013-2020: I believe that gold topped in Sep 2011 and that we will not see higher gold price till 2020, but it can still be a good (uncorrelated) hedge. Commodity prices have more potential, especially if it comes to some conflicts (wars) which are logically to expect in this situation. Western economies are in the falling cycle (opposite from 1980-2000) and we can`t expect high real economic growth, but stocks could grow because they are cheap inflation ate them in the last 10 years. Interest rates will not grow, they will stay low, but I don`t believe they can fall significantly anymore. Like I said, inflation has very important role in financial markets and there must be higher inflation to eat / clean all this debt and this inflation can push stocks and commodity prices up- not real economic growth. My opinion: euro will continue to fall (maybe crash) and dollar will become stronger. 2020 is the next important turning point (1980/82 2000/03 2020 this are most important dates, 2008/10 was just a half of cycle) 1970 1975 1980 1985 1990 1995 2000 2005 2010 1 For example, the turnover per min in 12:00 represent the average turnover in 11:59, 12:00 & 12:01
Since futures are leveraged and you probably would not use all leverage, trading futures leaves majority of portfolio balance in cash, so is clever to invest it in something and stocks can be a good option. But, let s first see how they are liquid - 01 Feb 2013 was an normal day, I went on Reuters site and collected the data from 13 major stock indexes (countries) and calculated a few turnover indicators: TURNOVER PER MIN 2 (Source: Routers) INDEX per min TOP 10 per min Index Currency to $ Trading minutes N stock sum per stock sum per stock FTSE GBP 1,569 510 101 1.080.502.943 10.698.049 366.403.447 36.640.345 20 ETF US USD 1 390 20 106.894.794 5.344.740 92.772.293 9.277.229 S&P 500 USD 1 390 500 306.826.892 613.654 61.043.141 6.104.314 NIKKEI JPY 0,0108 300 225 61.038.156 271.281 21.568.072 2.156.807 IBEX EUR 1,3639 510 33 7.585.534 229.865 6.029.107 602.911 DAX EUR 1,3639 510 30 8.224.544 274.151 5.382.196 538.220 HSI HKD 0,129 330 50 8.702.734 174.055 4.669.577 466.958 CAC EUR 1,3639 510 40 7.782.986 194.575 3.764.548 376.455 Ibovespa BRL 0,5033 420 69 6.357.833 92.143 3.221.972 322.197 AEX EUR 1,3639 510 25 3.478.769 139.151 2.626.655 262.666 S&P TSX CAD 1,004 390 243 9.339.692 38.435 2.212.825 221.283 BELL EUR 1,3639 510 20 1.651.897 82.595 1.583.951 158.395 MXSE MXN 0,0794 390 29 1.221.503 42.121 993.112 99.311 Straits t. SGD 0,8063 480 30 1.248.865 41.629 865.118 86.512 I must admit that Wall street was always my first love and I was very surprised by the domination of FTSE stocks from the aspect of liquidity. But unlike US, UK has 0,5% stamp duty on share transfers which makes stocks trading a mission impossible there is an option to trade them thought CFD (they are excluded from tax law and also have some more advantages) but trading CFDs is much more expensive (CFD brokers commission is usually around 0,1%). For this reasons I have decided to stick to US ETFs and stocks - based on average turnover in the last three months I have created 5 lists of most liquid ETFs and stocks: 5 LISTS CONTENTS AND TURNOVER (per minute in the last 12 months); 3 Source: Yahoo finance ETF 1 ETF 2 STOCKS 1 STOCKS 2 STOCKS 3 SPY 50.274.123 DIA 1.884.278 AAPL 28.962.448 AMZN 2.250.196 FCX 1.603.122 IWM 9.122.370 XLE 2.249.802 GOOG 4.278.250 PFE 2.008.812 ORCL 1.941.283 QQQ 7.135.643 TLT 2.182.156 BAC 4.126.732 AIG 1.698.727 ABT 2.208.093 EEM* 5.126.635 EWZ 2.025.555 MSFT 3.496.623 WFC 2.211.957 CVX 1.742.104 GLD* 4.245.456 FXI* 1.540.366 C 3.315.189 IBM 2.003.812 F 1.284.482 GDX* 1.709.731 IYR 1.219.047 XOM 3.215.868 NFLX 1.234.394 WMT 1.609.277 VXX* 1.155.072 IVV 1.306.655 INTC 2.586.287 QCOM 1.916.914 MRK 1.449.340 EFA 2.497.637 TNA* 1.516.251 JPM 3.241.612 JNJ 2.087.306 HPQ 1.136.644 VWO* 2.149.659 SDS* 568.780 GE 2.370.835 CSCO 1.932.251 GS 1.420.761 XLF 2.437.857 EWJ 452.665 T 2.345.824 PG 1.812.008 VZ 1.522.453 ***The turnover/volume on open and close is usually significantly larger!!! 2 Turnover per min = volume x price / trading minutes 3 * EEM-XLV before 15,4,2003;GLD-XLY before 18,11,2004;GDX-XLU before 23,5,2006;VXX-MDY before 30,1,2009;VWO-LQD before 10,3,2005;FXI-XLK before 12,10,2004;TNA-XLI before 19,11,2008;SDS-XLB before 13,7,2006
Since stocks in US make most volume around open and close, I have made an algorithm that determines the best share of each stock/etf from the list (to be in long or short position) on every open and close of exchange the hypothetical movement of portfolios can be seen from the next chart: Data source: Yahoo finace Cost calculation: since average price of selected instruments was around 50$, cost of spread is 0,0002% per trade (assuming that bid-ask price was 50,00-50,01). Since many stock brokers offer fixed commission (flat) per order, I have assumed 7,5 $ (industry average) per side on 500,000$ value order. All in all, I believe that costs are fairly calculated, especially if we take into account that cost calculation assumed changing the 100% of portfolio on every open/close, even if in some trading points the change of portfolio structure would be insignificant. How it works and possible improvements: Firstly, there is an algorithm/trading strategy that trades all 10 stocks from the list, and this is the main part of this system. After that, second algorithm picks stocks based on performance of first algorithm. Simplest, the first algorithm trades stocks, second manage strategies. Results can be better and more stable if: 1.we pick stocks from list of more than 10 stocks, 2.logicaly create list from 10 stocks that produce good and stable return using first/main algorithm. All in all, I have tried to develop several algorithms for trading stocks (as I said Wall street was my first love) and this is the best one. I tested it with UK/FTSE and the results would be similar (without stamp duty). Well, it assumes market order and this is its disadvantage in the future I will try to make algorithms with limit orders. I didn`t test it in details, but algorithm can be useful even with futures (tested on agricultural), don`t know about currencies/forex.
I hoped that I will finish one intraday (1H/15min) trading system for FX futures, I made a good base but unfortunately, I didn`t have time to finish it jet. I don`t know when it will be done, maybe I need few more days, maybe few months, but it has the potential to be my best trading system. This is why I decided to put one logical and very reliable strategy in the long run VIX-ES. The blue line represents Short-term model which is adjusting the weight of ES and VIX short-term futures on the daily base, while keeping the weight of mid-term VIX futures fixed. Like previous systems, chart shows the results without leverage but it is optimized to be leveraged around x5. Even it produces less return from the mid-term model, it is suitable for the larger amount of portfolio (cash). Oppositely, Mid-term trading model produces more return, but since mid-term VIX contracts that have significant role in this model are less liquid, like the return is more volatile, so it is not proposed to be highly leveraged. WITHOUT MID-TERM VIX CONTRACT Mar 26, 2004 Mar 15, 2006 S&P 500 (ES), SHORT-TERM and MID-TERM VIX future contracts Mar 26, 2004 May 3, 2013 (since the introduction of VIX futures) Daily close data; source: http://cfe.cboe.com/products/historicalvix.aspx From the next table we can see the liquidity of VIX futures which can be traded only during Wall Street session (more/less). If we chose Fixed weight model (only rolling without trading) even 5x leveraged 50M would not be a problem (ES occupies significant proportion), but trading model lowers the potential to not more than 10M. Also, there are many liquid VIX ETFs. VIX FUTURES TURNOVER: Jan May 2013 VIX (VX) futures: Average daily turnover Average turnover per min VIX front contract 884.265.938 2.267.349 VIX - VX2 816.646.823 2.093.966 VIX - VX3 298.905.662 766.425 VIX - VX4 183.312.235 470.031 VIX - VX5 133.258.406 341.688 VIX - VX6 98.392.538 252.289 Mid-term (VX4-6) 414.963.178 1.064.008
1. It is not a question do this models work... the right question is: For how long? If we look, for example, one trading algorithm on S&P 500 1950-now, it was working for 50 years but nothing last forever, people will figure it out sooner or later. The goal of trading system it is to get a trend and the opposite directions (shorting a strategy) would be one of the best strategies for 2008. But my point is that constant aspiration for finding new, original and fresh ways to trade must be an imperative if you want to be successful. 2. Trading is nothing else but allocation between long-short positions of different market instruments...the portfolio management is allocation between trading strategies. I have spent a lot of time and energy trying to make an buy/short algorithm for trading single instruments, one of them you can see upside, stock picking strategy is also mostly based on that kind of algorithm... but I have never found a universal trading system that would work on majority of instruments/markets/periods and I don t believe that it is even possible. Trading only one instrument assumes 3 choices: buy, short or flat, but more instruments provide much more options and possibilities. Secondly, it is much easier to trade strategies nor to trade instruments. 3. More uncorrelated strategies (more diversification) would produce more stable return, and the stability is the key determinant of potential leverage. My head is full of ideas how to improve the existing models like how to try to develop new ones. Models in this presentation are assuming market orders (and in-calculate spreads) but recently I have started to test/play with limit types orders it is not the same thing, but I see it as the logical development direction in which I would like to move in the future. 4. Sky is not the limit, nor is the volume... the limit is the Thank you for viewing my presentation every critic, comments or suggestions are welcomed. For more information you can contact me: Ivo Jelenaca ijelenaca@unipu.hr Mobile: 00385/98-969-5607 PhD student (near end) at Juraj Dobrila University Pula, Croatia