Thanks Winston. I just downloaded Tradelink and as I expected I really have no idea what it is. I looked at one of the sample strategies then tried to build it but I don't have any historical data. Is that the issue? If so where can I find some data for free? Sorry for the very open-ended questions, I am blind here.
Since I receive this exact same question many times per month, let me offer a couple of suggestions. I tell all my new people that I would prefer to see a trader come up with a methodology vs. having a programmer come up with a system. Find something that works manually, real time, real money (no "backtesting" or anything, pretty much worthless, especially with today's HFT etc.)..... and then simply automate the process. This can be very simple, from simple Excel DDe links and Active-X to a full blown program. Make sure the whole idea works first, real time.... simple suggestion. And, don't let the naysayers bother you... I, too, understand that most "systems" don't work, but that doesn't mean that you don't have a good idea...just try it out... why waste months/years for the markets to change thus rendering the whole thing moot? Go for it... Let me know how it works out... Don
Hey Don, Thanks for the reply. I am already pretty confident that this strategy can make money. I would still like to backtest it though to see which markets it would perform best in, and how to better size positions, etc...a lot of unanswered things so far. I also don't want a programmer doing this for me. I have taken a lot of time working this thing out and outside of some consultation here and a few friends, I am going to do this alone. For what it is worth, I am also a full-time discretionary trader. I decided that some elements of my current strategy can be automated, and as I was working on that, I came up with some other ideas to implement. Worst case scenario is that this turns out to just be a way to keep me busy on weekends a little bit. But I do think it can work.
Just understand what everyone knows... backtesting can only prove what "won't" work... it can never prove what "will" work. Too many variables involved to make backtesting serious. No way to program in these variables. Something to keep you busy? Sure, why not... probably better than watching a shortened NBA season, LOL. All the best, Don
The reason why back testing only proves what doesn't work is because people don't do it right. They don't have access to good data and they don't understand the differences between executions and simulation. Also people generally look to see if something makes money vs. loses money rather than looking at whether or not the strategy is doing what you intended it to do.
if you're doing automated trading you'll be doing all of * backtesting * backtesting with cross validation * walk forward testing on paper trade account * walk forward testing on live account with small size * live account with regular size this generally makes best use of your capital, and it's not particularly difficult once you get setup.
That's fine. My only point is that we cannot program the factors that make things either work or not work, such as: Prem/DIsc to FV during the transaction time, news items, global economic data, wars (peace, LOL), the myriad fo facts that affect the market every minute of the day. IMO, the market is not in a vacuum, it is affected by everything that goes on. The Euro situation, how would you program that in? (Just as a current relevant factor).... heck, a 5 minute news story could cause dramatic effects... anyway, whatever makes you money... that's what we're in this business for.... Good luck, Don
Yes. i dun give a fxxking sh1t how much time u took to "develop" your whatever strategy, or yada yada. Like I said good luck with your Holy grail. When things are turned automated you have hell lots of sh1t to worry about. let me give u an example. Your strategy is only 5% or the work. Thinking a new strategy, and writing it down on "paper" and into your production systems, is simple. I have develop over 10-20 neural networks, and several algorithms or AI learning. But hey before that: - Now you started learning coding. - You have problems with market feeds - Order system to brokerage reliable ? - Order management. Limit or stop or market ? - How to cancel Orders? How to exit a trade ? Trail stop programmed in, or use your brokers Trial stop ? There are additional 50+ order types, including ghost orders or VWAP/TWAP etc.. - Risk mgmt. How to manage your entire basket or portfolios ? Trade 1 stk at a time ? - Trade extended hours ? How about hedging with Options, Futures ? Pair trade ? And how to write everything I mention on top into computer readable codes ? If I wish you dead, I wouldnt even spend 5 mins typing this. Planning and thinking is easy. Doing is hard. Especially alone. Good Luck .
I agree with you. Backtesting = curve fitting, and doesn't work that well. I program my algos on the go, intra-day. I just wish my AI so strong that it programs by itself and change according to the stocks buy/sell pressure etc
My strategy is only 5% of the work? Sure, but developing a successful strategy is a lot harder than learning how to code or going through all the steps you've mentioned. Why? Because I can find 10000 people who can do that. I can find maybe 3 who can develop a strategy that will work. Some of your other questions are very simple to answer: - Now you started learning coding. - You have problems with market feeds Why would I have problems with feeds? - Order system to brokerage reliable ? Again, this is something that I don't understand yet. - Order management. Limit or stop or market ? Both, it varies depending on what type of trade is being taken. I have all the rules as far as what orders will be used written out. - How to cancel Orders? How to exit a trade ? Trail stop programmed in, or use your brokers Trial stop ? There are additional 50+ order types, including ghost orders or VWAP/TWAP etc.. Orders are cancelled under certain conditions that I have highlighted. Trade exits are all also highlighted depending on certain set ups. Stops, limits, market orders, OCOs are all I will be using. - Risk mgmt. How to manage your entire basket or portfolios ? Trade 1 stk at a time ? Not sure what stk is but this will trade 1-5 markets at any given time. One of the main things I need to figure out is which markets it will be most effective in. As of right now I have one in mind. - Trade extended hours ? How about hedging with Options, Futures ? Pair trade ? This is for futures, not stocks so it will trade around the clock. However, let's say that a pattern is detected in the ES at 2 am. I won't take that. If a pattern is detected in 6E at 2 AM, I will. Thanks for the questions, they are making me think some things over. I hadn't really considered when the strat would NOT trade certain markets.