I would recommend that you dump Metastock. It's system development capabilities are poor and can lead to some bad system decisions. I use TechnifilterPlus, but many people like Wealth-Lab. For stock trading, I have two systems that require a large database in order to have a few trades. I find these kinds of systems do better, by taking just the very best setups out of many, many possibilities. How large a database are you working with? I scan about 3000 symbols every night for the two systems I use and get anywhere from zero to 50 potential trades. A limit order the next day means that of those potential trades, you get, at the most a handful of entries. I have enough datapoints to feel confident that these systems hold up over time. There are a zillion ways to skin this cat, but it sounds like you are pursuing a good method of selecting only the best signals. You can PM me if you'd like to discuss details. ges
It will probably help at this point to get out of the trees and look at the larger picture. Under what conditions are your trades signaling? Do you know? If not, you need to look at how they are signalling vis a vis overall market conditions, which often have a big impact. As one example, if you are backtesting pullbacks, you will find a great number of trades bunched during periods of overall market weakness. The opposite will hold true with breakouts or buying strength, as trades generally don't happen in a complete vacuum. Understanding these issues, and really "getting" what is happening in your system, will help you to tweak it to the point where you feel completely comfortable. If your system has the capabilty to reference an index, say SPY, DIA or QQQ, you might want to include that reference to see how the performance of index impacts your trades. As you are backtesting buying dips, look what happens when SPY is at a low level. Are there more trades? Why? And will that info help you? This is a good exercise to get you thinking of the bigger picture, which you will need in actual trading. And I would say that the converse is true, also. If you have developed a system, no matter how well it backtests, and you don't FULLY understand what's happening, why, and when, you aren't yet ready to trade it. Even with the best system, the market itself will continually try to throw you off. You need to be rock solid in your understanding of what you are doing, to stay with any program.
Thanks Bob111 & kww. Kww, yes, you are correct about the ratios being the same. Sorry, I didn't describe my experience very well. Here's a better explanation. When I backtest a system...if the system generates numerous trades and achieves a 70% win ratio...when I bring that over to actual trading...the win ratio stays pretty close to that 70%. The key is the backtest is performed over different timeframes and on each of those timeframes...many trades are generated. But, when I began to tweak (filter) my systems to increase the win ratio and of course decrease the number of trades...my backtesting might even get above 90% for the win ratio. The problem with this...is my number of trades generated are too low to provide a valid test of the robustness of the system. So, when I bring these systems over to production (real trading)...the win ratio typically drops sometimes 20%, 30%, or more. I might even have to remove the system from production. Hope this helps. My main point to the newbie (traderNik) is the quicker you can get used to the losing trades in the systems you develop...the better a trader of your system you will be. And in turn a better developer of trading systems. Sorry for the long post.
I commented in color at specific places to fast track getting to the solutions you want to use. Basically the content of the color content yields the following: I. A continuous supply of trades. II. Trades that are undertaken only after the trend is established III. Trades that begin after the first distribution in the established trend and are concurrent with the second market accumulation period for the overall trend. IV. The trade duration will be about 2 to 3 days and the average profit per trade will run about 7 to 15%.
InvestorRT at www.linnsoft.com. Fast, easy, very powerful and no programming languages. You can revise and backtest in a fraction of the time of WLD or Tradestation.
I have a few questions about this software: 1) What realtime datafeeds does it support? 2) What kind of charts will it handle. Tick, minute bars, tick bars, candlesticks, etc etc? 3) Does the testing language make use of every tick that comes in, regardless of the bar's time frame you are plotting so that signals can be taken inside a forming bar? nitro
I have a few questions about this software: 1) What realtime datafeeds does it support? A number of data vendors. Check www.linnsoft.com site. When you register the software, you specify which data vendor you want to work with. 2) What kind of charts will it handle. Tick, minute bars, tick bars, candlesticks, etc etc? You can create any type chart with infinite variations. 3) Does the testing language make use of every tick that comes in, regardless of the bar's time frame you are plotting so that signals can be taken inside a forming bar? Correct, so long as you have the data, which you can obtain from a number of sources and format for RT. You can also create scans/signals based on multiple timeframes running at once in realtime, and quote pages displaying many instruments and their levels in regard to various indicators, signals, etc. etc. I would be surprised if there is anything you would want to do in testing or trading that can't be done with RT. Extremely versatile. If you have detailed questions, might want to visit the RT thread at http://groups.yahoo.com/group/linnsoft/messages/, or e-mail them at support@linnsoft.com.
What can I say - a great response. I'll try to keep these comments brief. woodcutter - I had to laugh when I read your post. You can see how long I have been thinking about discretionary trading. Of course you are right - the psych component is taken out of a strictly non-discretionary system. I guess I was just considering the possibility of not having the actual executions automated, and being forced to wake up in the morning and get a signal and pull up the chart and then say 'My God, am I supposed to go long NOW?? Can't I wait for a slight pullback?' The solution would be to forget about the chart and enter the trade. M-Dog - thanks for the detailed replies, I really appreciate your comments. I am actually happy to hear you say that you have profitably traded systems with 60-70% win rates, because my systems are almost all in this range - 66/34 seems really common. I have yet to create one that approached a 90% win rate, with the exception of a couple that trigger 5 times in 20 years. I'm trying to get ready to 'embrace my losses' but I can only hope that I will react in a professional manner. Adequate capitalization seems so crucial here, just in case you happen to run into one of those larger drawdown periods right off the bat. It was really encouraging to read your comment about eventually getting enough systems to get plenty of trades, since that's the way my approach seems to be shaping up, as opposed to having the one killer system. ges - I am in the process of dumping Metastock now. It's too bad because I think that years ago they had a good product but now they aren't interested in making the product better - only in selling data (acquisition by Reuters a few years ago, the old story...I am sure Achelis is not involved any more...founder takes the big buyout and things go downhill). The latest versions are buggy as hell. They couldn't even be bothered to send me the fixes when they came out. Look for my PM. lindq - I will definitely be taking this advice. Until now, I had been trying to develop systems using only price action and the indicators included in my retail s/w package. Of course your idea about correlating signal generation with overall market condition makes sense. I will also be trying to incorporate indicators like VIX and the state of the interest rate markets, and believe it or not, I even have some ideas about developing a system that incorporates fundamental data as part of the entry criteria. I'll try hard to understand why my system is generating trades, not just how. Grob109 - wow.... thanks a lot for this detailed suggestion. I'll definitely check this out. One thing though - I am unclear about what you mean by a 'right to left channel traverse (points 1 and 2)'. Is there any way you could either post a chart with an example or give me a recent example from a particular stock? At any rate, I have an account set up at Multex and I have been doing some screens such as the ones you suggested - I just need to fully understand your entry system. Thanks again to everyone. Any further comments, both +ve and -ve, are welcome.
Do you mean win/loss ratio? As in 66% of the trades were profitable? If so, what is the average gain and average loss? ges