Matlab - IB API

it finally connects to the TWS.

微软广告预测定位

July/2013 trading performance - ES Intraday Trading System

1 contract per trade, Gain 3.7%, Sharpe Ratio 3.3,

Technological Unemployment | 60 minutes

free dowjones 30 download

http://www.masterdatacsv.com/Downloads/Composites/NonSub.htm


https://www.interactivebrokers.com/en/software/demoStandalone.php

Instructions on how to run installed TWS in demo mode:

1. Download IB Trader Workstation(SM) and follow instructions therein.
2. When prompted for username and password

• For Individual Demo
  • User Name: edemo
    Password: demouser
• For Advisor Demo
  • User Name: fdemo
    Password: demouser

login

user: edemo, password: demouser

Intraday trading is more suitable to me

It was upset to me. I haven't found any places allowing me to implement my skills and knowledge of quantitative trading, although I have kept doing quantitative research in these years. Thus I am thinking to start to use quantitative trading strategy to trade my own money seriously. At least, it's better than benefit nothing from my research. But the strategies I've built before are not suitable for me, as they were originally designed for Hedge Funds / CTAs. I need to build some strategies more aggressive to trade such small amount of capital. So I started to work on that.

As I was working on the strategies that are used to handle large size capital. I thought that such high profitable trading strategy might not exist. After sawing a video few month ago, a lady talked about how her team made 41 million USD profits within few years with only 300K USD start-up capital. I then thought that it could be possible.

After doing a research in recent weeks, I found that the market microstructures can be used by traders to make profits have existed in the markets for many many years. Without taking account of market liquidity, the simulation shows that it is possible to generate the remarkable performance with very low drawdown

But the disadvantage of this system is that it can only handle a small amount of capital to trade around 30 - 50 emini contracts per trade.

The Out-Of-Sample Simulation result for DowJones YM $5 future contracts

When God Shuts One Door, He Opens Another

Sometimes our disappointments in life can become God’s appointments, so don’t let closed doors bother you. The things we think of as failures and problems can often end up being blessings in disguise.

The Out-Of-Sample Simulation result for S&P 500 E-mini future contracts

Average holding period: 6 trading days

S&P 500 Out-Of-Sample Simulation results Long-Only Strategey

Model Calibration plays a vital role to form workable trading strategies in reality

Here is video telling the basic idea how to select the model

Option traders use (very) sophisticated heuristics, never the Black–Scholes–Merton formula

Option traders use (very) sophisticated heuristics, never the Black–Scholes–Merton formula

http://talus.maths.usyd.edu.au/u/UG/SM/MATH3075/r/Haug_Taleb_2011.pdf

Inaccuracy of Black-Scholes – Fischer Black and others agreed that the model is flawed

Finally, they agreed the model is flawed. I remember that I had an interview with a European Bank 3 years ago. The interviewer talked B-S model with me. He asked me whether believed B-S model or not. I told him that B-S model is far far away from the real markets and I didn't think that B-S model was right......  Finally, I was rejected after the interview. What a joke.

Anyone's interested in this article. Please visit http://thomsettoptions.com/black-scholes-the-assumptinos-are-simply-wrong-updated/?goback=%2Eanb_1940519_*2_*1_*1_*1_*1_*1

Execution Rules of these trading systems

The executions are based on the signals forecasted on the previous trading day. Meanwhile the systems' parameters are daily self-adapted. Additionally, the slippages set for the simulations are much larger than real market slippages.

Combination of 4 Long/Short Systems 8X Out of Sample Simulation

8X is just a theoretical leverage calculated by the leverage optimal model. I don't think that anyone in reality is gonna use it.

Total Compound Return from Oct/90 - Feb/13: 144,280,539.74%

It means that by using this system, anyone invested 1 dollar in Oct/90. This system could get him/her 1.44 million dollar profits in these years. Although the simulation is totally based on the forecasts and 100% systematic, it's still too good to be true. :-D

Combination of 4 Long/Short Trading System

Total Return 28/Oct/1990 - 08/Feb/2013: 11,378.10%

Statistics for Feb/2010 - Feb/2013

Annual Return: 39.5%

Annual Sharpe Ratio: 1.58 In Average, each system generates over 600+ trades with around 65% winning rate. ( Theoretically the optimal leverage for this system is more than 8X, but here I only use 2X leverage.)

SP500 Long/Short System 1 - Out of sample simulation

Total Return 28/Oct/1990 - 08/Feb/2013: 9,924.18%

Statistics for Feb/2010 - Feb/2013

Annual Return: 48.8%

Annual Sharpe Ratio: 1.81

To design workable a short-selling strategy is much more difficult to me than making a long-buying strategy. But we can see that after implementing the short-selling strategies into the system, both Sharpe Ratio and Compound Returns are improved.

SP500 Long/Short System 2 - Out of sample simulation

Total Return 28/Oct/1990 - 08/Feb/2013: 8,646.38%

Statistics for Feb/2010 - Feb/2013

Annual Return: 35.7%

Annual Sharpe Ratio: 1.50

NASDAQ Long/Short System 2 - Out of sample simulation

NASDAQ Long/Short System 1 - Out of sample simulation

The combination of 4 long-only trend following systems

Total 2X compound Return:	9979.74%
MaxDD:	-22.67%

I notice there is a large drawdown during the period 2009-2010. I've figured the reason causing this drawdown out, and find the way to prevent it happening again in the future. But since out-of-sample simulations are based on the forecasts generated by learning historical data, if the similar things never happened in the history, then the systems based on learning the historical data can't avoid unless we trade less aggressive and use some extra-hedging strategies to balance it. 

So I think that I should accept this drawdown. As drawdowns are nature part of trading, every trader should be able to accept it. And what I want to say is that systematic trading is about dealing with the uncertainty of market movements. Although the similar patterns show on the markets over and over again, the uncertainty always exists. The history teaches us how to deal with the similar situations happening in the future but it can't tell us everything happening in the future.

Also this research reminds me one of Ludwig Wittgenstein's quotes "The world is the totality of facts, not of things."

An example of Simple Kurtosis, Skewness and Mean Asset Selection Filter for Equities

On this chart, you can roughly find what sort of equities you can invest.

Well, I've few more methods to further classify them... :-)

I input 100 stocks into this simple system. After filtering them out and further combining the results with the mean reversion strategy, 5 stocks are selected. They are Stock 3, 28, 55, 60, 67. (Please note: the calculations for Kurtosis, Skewness and Mean have been modified in this example.)

A Robust Stock Picking Method

Why Nikola Tesla was the greatest geek who ever lived?

I found it on the Internet....

James Simons' MIT Speech "Mathematics, Common Sense, and Good Luck: My Life and Careers"

I've listened to this speech for several times. The part I like most in this speech is what he said "don't give up, at least try to not give up". 

http://video.mit.edu/watch/mathematics-common-sense-and-good-luck-my-life-and-careers-9644/

The improvement of execution system has been delayed for 2 weeks

It's time to start to work on it again.

Fuzzy modelling approach may be able to give a better classification result

I'm going to test it...

Market Price Movement and its Patterns

A thought about equity picking model

I used to work on the development of statistical filters and the KPI by blending the stats such as Sortino Ratio, Omega Ratio, Skewness, Kurtosis and so on for selecting and ranking Equities/ETFs/Hedge Funds. Today some thoughts pop into my head few hours after the talks. I ask myself why not implement the statistics of those Equities/ETFs/Hedge Funds and the mean reversion models into the SVM. SVM can handle the selection and filtering things. I use the SVM to do the pattern recognition for trading strategies. Obviously, it can also help me classify and find the patterns within those assets across the investment universe!

Interesting...

Microsoft publishs a new lib for stock charting....

A mean reversion system test (Out-Of-Sample Simulation)

about 80% winning rate

A Noise Or A Pattern?

Do you see the patterns? It's not really noisy, isn't it?

100% quantitative models based systematic trading

The Performances are generated by 100% systematic trading systems, and all the simulations are based on the forecasts. 

The definition of quantitative trading can be found at below link:
http://www.investopedia.com/terms/q/quantitative-trading.asp

Nasdaq Trend-Following System 2 - Out of sample simulation

Total Trades	501
Num of winning trades	296
Num of losing trades	204
% of Winning	59.08%
Average Profit on winning	2.19%
Average Loss on losing	-1.79%

Slippages & Costs have been taken into account

Nasdaq Trend-Following System 1 - Out of sample simulation

Total Trades	498
Num of winning trades	297
Num of losing trades	201
% of Winning	59.64%
Average Profit on winning	2.17%
Average Loss on losing	-1.81%

Slippages & Costs have been taken into account

SP500 Trend-Following System 2 - Out of sample simulation

Total Trades	483
Num of winning trades	296
Num of losing trades	187
% of Winning	61.28%
Average Profit on winning	1.58%
Average Loss on losing	-1.22%

Slippages & Costs have been taken into account.

SP500 Trend-Following System 1 - Out of sample simulation

Total Trades	520
Num of winning trades	318
Num of losing trades	202
% of Winning	61.15%
Average Profit on winning	1.65%
Average Loss on losing	-1.37%

Slippages & Costs are taken into account