So with a few months of research, I’ve been through some of the pain of coming out good strategies be it short term speculative, trend trade, mean reversion, spreads, or just plain time based trading strategies.

These are just some products that are currently on my system:

• US markets Top 10 Volume (so the spreads are 1 cent away, with high volatility/beta though)
• US markets?Top 10 Volatility (must satisfy good spreads)
• Futures markets (Global Index, commodities)
• Correlated US Stocks

Some findings base on my systems:

• Good spreads and volume helped reduce slippages costs, based on Market orders
• Volatility means more trading signals, but does not necessarily means profits
• Futures can be traded like stocks with the same systems, but some trading basics got to be identified and applied.
• eg. Nikkei 225 Futures lacked the “Point” by point spread, instead they are traded with a 5 points per tick, making it hard for Alpha due to the spread.
• On the contrast, the Hang Seng Index futures are great way to trade, because of the tick value.
• Markets are definitely correlated. Systems are up on 20 hours a day, from 8:00am GMT +8 Asian time, to 5:00pm EST US Time.
• By running systems to handle the risk from Hang Seng to Nikkei, and over to Nasdaq and the S&P500 futures, including WTI Crude from Europe to US times, you can see some correlation when one market moves, so is the other, and how the algorithms act upon these information, intra day.
• On the opening bell from 9:30am – 10:00am EST, and on good days (prices are short squeezed onto the Long or Short side), returns can run up to $300-$500, and before 10:00am, it can capture quite a good day’s worth of profits. But so are the risks, when there’s only 10% of the time with such opportunity, and 90% probably losing up to $100-$200. This will be refined and optimized.
• Some of the Algorithm cannot detect what is a random price action and what is an ‘actionable’ trade. So some results are therefore random and something only the system can only manage, but cannot act upon by forecasting or prediction.
• Fortunately, some US Stocks can produce consistent results, even after trading a >5000 shares a day for many many days. o:-)
• For a typical day when all systems are used and all products are added, trades can run up to 500-1000 trades a day, and over a value of USD 2 Million worth of shares traded.
• Adding all up, however the Equity curve are still negative, and points in an inverted 45 degrees, with only about 25-30% of winning ratio and Average/Trade at a negative value

Ahh, some of these stuff are pretty good for record keeping, knowing what works and what not. Thinking back on the days where I had made a few hundreds on an intraday, it spook me to realize how much risk Im actually putting on to produce that kind of returns, that is always inconsistent. Now that the programming of these systems comes at a snap of my fingers, it’s quite easy to “optimize” and fine tune the results.

In my current phase, optimizing the systems for speed and efficiency is a top priority. Tweaking and playing around with the numbers so the magic happens; frequency of signals/trade goes down (reduces brokerage fees) while Profits goes up (cutting useless stuff away).

Also, I have introduce a new strat using pure mean reversion and short trend systems (to counter the mean reverts), and just this system alone is able to produce quite a good run.

So… 3 objectives to meet in 3-6 months time:

1. Trading of more products eg. Futures & Equities, essentially a system able to manage in excess of at least USD 1 Million
2. Average profit per trade > 50 cents, or the Win/Loss ratio > 40% for consistency of at least 100,000 trades over 3 months.
3. Sharpe ratio > 3

Cool, what a post. Next, I’ll be updating my fund performance on my asset allocation base strategies and how they are doing over the past 2 months… (glancing over Municipals Bonds, it was a good bet and a good add to my portfolios since the S&P downgrade in August 2011 =:)

Files under Algorithm Trading, Gambling and statisical theory, Managing Risk | Leave a Comment | No Comments »

Very Volatile markets
In case you wonder what’s up with me these months. It has been a fruitful 6 months, since I started working with Excel, Maths, Models, Historical Data. Take these together and I call it Monte Carlo Simulation.

Initially I started worked with producing Pure Random data, with volatility parameters to simulate the Brownian motion of stock prices. Comparing SGX and NYSE, its a different way of coding it, anyway I was trading the US markets.

http://en.wikipedia.org/wiki/Monte_Carlo_method

So then came along Algorithms, the things that say if A is X and B is Y, then it A/B = 0.02 and bla bla yada. That’s how I came up with my first “Algorithm”, designed to detect market movements and pressure, be it up, down or sideways. By the wy, this was history because it was used as a platform to test out my Monte Carlo (Hail the great 1st Algo!!!)

Great learning though, things like Fat Tails, Probability Distribution Function, Law of Large numbers, Curve fitting, Real life trading, Cut loss simulations and certain Black Swan Scenarios were built in to “Stress” my Algorithm. Then I came up with more and more Algos, Algos that are supposed to detect and manage different stuff, like Breakouts, Opening / Closing, Spreads, Trends, Reversal, and etc…

http://en.wikipedia.org/wiki/Algorithmic_trading

They required more CPU x core speed, and tons of Excel spreads running on my 3 computers, about 30+ of them dealing with different paramters, volatility, different pricing models, etc etc… That took me 2 months.

Computers meets Wall Street

Then it was easy; With random data running all over my Excel, I soon replace them to take in Data from Finance Google. Nevermind HFT running level 2 quotes, trying to Front run Bids/ and Asks at 900 microseconds, (the link from SG to NY is 300 ms on average all major ISP, unless I colo my system to US :> )

http://en.wikipedia.org/wiki/High-frequency_trading

Pulling data from Google Finance was easy. I started coding in Human reactions, or things you would do or behave, automatically. Eventually, I’ll complete the task: Interactive Brokers TWS platform, with my excels connected to its TwsDDE.xls. See my next blog post. It’s magic .

Automation vs Manual

As a Portfolio of strategies, I now allocate my capital of 40% to this Algorithm trading, and 60% to all the other strategies including Asset Allocation based long term trading and other derivatives strats.

Files under Algorithm Trading, Gambling and statisical theory, Portfolio Management | Leave a Comment | No Comments »

I have recently become interested in all sorts of gambling and betting system, especially the opening of Marina Bay Sands and Resort World Sentosa in Singapore.

Understanding that in all games, the House always have some kind of edge (except people to people games), plus the research of Ed Thorp’s card counting and roulette system, I thought I could beat the system.

Yes, I have found a way to simulate how to beat it right here:

http://www.casinator.com/casino-games/play-free-roulette-online.php

Well, I said I found a way to simulate the research in an more effective automated way, on the roulette wheel. And I meant NO! I haven’t found any ways to beat the dealer, based on probability and statistical results.

But if you are interested, set your bets on the roulette table, click auto spin, and let the computer do the spinning, betting and dealing. In no time you realize that there’s no way to beat the dealer, be it the American or Europe 00’s or 0’s system. In the long run, you will soon find yourself out of money =)

Files under Gambling and statisical theory | 1 Comment | 1 Comment »