1) Well, I can't trade real money while I work for an IB. 2) I have an edge, I just don't have a framework outside of work to test theories on 3) I appreciate your concern because your post is full of solid advice. I'm still in the bank. I just don't plan on being there until 2010.
I would take a look at this book, I really enjoyed reading it. The author worked at large firms for 10 years and then went to trade on his own. Very similar to your story. http://www.amazon.com/dp/0470284889...iveASIN=0470284889&adid=1ZAB3NBD6BBG61RB4CCQ& http://blog.traderwerks.com/2009/03/01/quantitative-trading-el-book-review/
I will cheer for you. I also work for a financial company (buy-side) that has been hit hard by the credit crisis, and I feel that my life has been wasted. Plus, since the crisis, the politics has been getting worse, as people realized that there will be fewer positions down the road.
I've decided I simply can't afford to spend too much time writing my own software and that it is cheaper and more cost effective to use other packages. The first goal should be to generate a small, token revenue stream that can cover basic expenses like groceries. So that is the current goal: Can we get a system going that nullifies the cost of groceries? And have it fully automated? I decided on NinjaTrader. It's not perfect, but I think this software can do what I need to do in the short-run, with the added nicety that perfecting the code on NT gives me code re-use options. I'm not a fan of the fact that it depends on Windows, but I'll abstract the core logic in such a fashion that when I move to a higher speed platform, the code will be reusable. Will see if I can generate some better backtest results tonight. Goal is to have something posted by morning. I came home from work and went to bed, got up at 1am, and the plan is to program till around 4am, get about 3 more hours of sleep, and then roll into work.
Ok, so, I read the NinjaTrader docs this morning, and it took me about 3-4 hours to get the long-portfolio side of the equation working. I have to hack something out similarly for the short side, but clearly I'm making some progress. I have to go back through and manually verify the results and fix the hacks that exist in the code. Some parameters I just hardcoded in after having calculated them in R. I'll be back in an hour or two with the short-portfolio side. I'm not really sure yet how to merge long/short results into one strategy with NinjaTrader. I find it somewhat frustrating, but it's still faster and better to do it with NinjaTrader than wasting my time rolling out my own strategy. Here's the first backtest results on a daily time frame, using the same basic strategy: Total Net Profit $2752.00 $1288.00 $1464.00 Gross Profit $4013.00 $2217.00 $1796.00 Gross Loss $-1261.00 $-929.00 $-332.00 Commission $0.00 $0.00 $0.00 Profit Factor 3.18 2.39 5.41 Cumulated Profit 33.90% 15.75% 15.68% Max. Drawdown -14.20% -11.11% -3.47% Sharpe Ratio 0.50 0.37 0.36 Start Date 4/15/2008 End Date 4/15/2009 Total # of Trades 10 5 5 Percent Profitable 70.00% 80.00% 60.00% # of Winning Trades 7 4 3 # of Losing Trades 3 1 2 Average Trade 3.18% 3.25% 3.10% Average Winning Trade 6.75% 6.84% 6.62% Average Losing Trade -5.15% -11.11% -2.17% Ratio avg. Win / avg. Loss 1.31 0.62 3.05 Max. conseq. Winners 3 2 1 Max. conseq. Losers 2 1 1 Largest Winning Trade 11.26% 10.23% 11.26% Largest Losing Trade -11.11% -11.11% -3.47% # of Trades per Day 0.05 0.03 0.03 Avg. Time in Market 21.30 days 17.80 days 24.80 days Avg. Bars in Trade 14.5 12.6 16.4 Profit per Month 4.25% 2.69% 2.23% Max. Time to Recover 92.00 days 64.00 days 96.00 days Average MAE 7.22% 8.27% 6.18% Average MFE 6.37% 5.69% 7.04% Average ETD 3.19% 2.44% 3.94%
You are seriously underestimating how difficult it is to make a living based solely from trading. I too thought that I would just jump into trading and start making $1,000 per day, and I thought that I was being conservative by giving myself at least one year before I started turning a profit. It took me several years just to have a miniscule profitable year, and a couple of more years to become consistently profitable enough to make a living from trading. I tell you this because you should not underestimate the importance of a job that is giving you a source of income. If you are simply trying to generate income to cover groceries, it sounds like you are undercapitalized. In that case, if you are undercapitalized, generating income from trading will not be a turnkey solution that will spit out profits because the markets are constantly changing. The odds of making of living from trading is close to nil if you are undercapitalized. In addition, the best piece of advice someone can give you on ET is throw out backtest results. They are useless. You must forward test your results first on a simulator and then with a small amount of cash for at least 6 months to a year before you start trading with your full account size. Ignore this advice at your peril. If you are being forced out at th IB, I would suggest that you look for another job instead of taking up trading to pay the bills. Give your trading endeavour at least 2-5 years before expecting a profit.
All good advice, but you're assuming I just started yesterday. I didn't. I've been in the business for 5 years. I just want to be independent, and not tied to the institution's desk. There's so much support infrastructure for institutional people that a retail trader has to come up with over again. The real challenge is doing it as a retail trader without the benefits of all of the research, teams of quants, low latency data feeds, etc. Second point to make is: Making grocery money per week is a fair goal for forward testing for 6 months, particularly if I'm trading 100 lots. Do you do $1000 a day in groceries? Who are you? Michael Phelps? I'm shooting for $30 a day, which isn't really that difficult.
I work in a fully automated group, and they have access to much better technology than what I will have access to. I've outlined some goals I want to achieve. When you say you work in a fully automated group , it would help to know in what area does the group specialise , is it an execution desk where orders are being executed for portfolio managers such as equity trades done by automated algorithmic programs , reason I ask is if you are trying to replicate this type of automation or area , I have an idea of the programs you can use on a budget .
I'm in equity derivatives [market-making, not execution for portfolio managers -- self contained, but we do pay out for orderflow to boost our profits], but we have a little bit of small-scale prop stuff going on to complement the main business of spread extraction. I'm trying to leave the group and roll over to stat-arb, or go independent. I'm having a heck of a time getting these retail software packages to do what I want, though.
I've decided to kick NinjaTrader to the curb and just go on programming in R. I'm having a heck of a time getting this program to show me what's going on with any sort of ease, where I suspect I could do better just flat out programming the stuff myself. May be my lack of experience with the software, but I find their multi-instrument model fundamentally confusing. If you want to enter a trade in NT, you have to get the damned thing to place the trade on the next bar update, which means you have to add the first instrument back in as the third instruments. Truly, completely confusing. I need to work with another system that isn't as weird. I'm thinking Python and writing my own code is truly the only way to go. Back to the drawing board. So, so far, I've ruled out e-Signal EFS and NinjaTrader. So my deliverable for Saturday night will be this: Use Python, create a sort of "generator" that will call function objects on each day, and pull out the data that was loaded in via a CSV file. Execute the indicator, and strategy functions, have a position manager, etc. And so on and so forth, and rapidly prototype the system this way. Hopefully Python will let me rapidly prototype something that I want based on what's going on in my head. As far as graphs and such, I guess I'll just export the results as the program executes to a CSV loaded into Excel. Then I'll just let Excel generate the graphics. Epic retail software fail. :-(