MseTrap

Did you duplicate it exactly? If its still not right, post a data file with the O,L,H,C and the indicator value and I'll take a crack at it. I've a pretty good track record of duplicating indicators from their output.

IMO the Marsland machine learning book linked above is pretty awful. Just skip the book and download the source code from the website. Although it's fairly obfuscated python.

Hi, I'll toss in my $.02 as well. I'm a professional mechanical systems developer and have been on both sides of contract programming. For the most part what works best for you also works best for the programmer. IMO the most important thing is the specification. 9 of 10 times what causes a delay in a project is ambiguity. You need specify what you want done to an exacting detail. When the developer knows exactly what needs to be completed he can give you a very accurate estimate of time/cost. This also prevents "feature creep" ect. As others have mentioned well documented code is very important, make it a requirement of the project. Look for a developer who is passionate about mechanical systems and the markets in general, A developer who is motivated by more than money will go that extra mile to make sure your system works and that your happy. Also a good developer will offer tips and assistance about how to improve your system. just make sure they're aren't trying to squeeze more billing out of you. An signed NDA is a good idea too, It can be done quickly over fax and good developer wont blink at signing one. Be prepared for your mechanical system to not work like you expect!! I've built hundreds of mechanical systems over the years for clients and a majority of the time the system fails to perform like the client expected. This is because of not being totally diligent about your system. Putting the trading logic into software will expose every weakness in the design. If this occurs, get with the developer and have them explain to your satisfaction why you dont see what you expected. The more prep work you do before passing the job to the developer will pay off later. Having the developer give you full access to all the relevant variables of the program as others have suggested is a good way you to tweak the program on your own at your leisure. Good Luck!

Here is what I've used in the past. inputs:DayProfitlimit(500),DayLossLimit(200) variables: canTrade(0), NP( 0 ), OPP( 0 ), PLB4Today( 0 ), ProfToday( 0 ) ; NP = NetProfit ; OPP = OpenPositionProfit ; if date <> date[1] then begin PLB4Today = NP[1] + OPP[1]; CanTrade=1; end; ProfToday = NP + OPP - PLB4Today ; if ProfToday >= DayProfitlimit or ProfToday <= -DayLossLimit then Begin cantrade = 0; end; // In your trades, check 'Cantrade' before trading..

To err is human. This one had me chuckling for the rest of the day. My OSO settings were wrong when I placed the trade and I didn't realize till it was too late. :crap:

Look into 'VTTrader', They offer free software and simulated Forex accounts. Of course its only useful during market hours. I've also looked for something similar but haven't had much luck. You can also download a free trial of "Trademaven", and for 30 days you can get tick replay of the ER2 and YM and practice outside of market hours. The software is crummy but tick replay is nice.

I do this as well when I test additional trading signals during the live hours. What I also do is put the arrow at the particular price the market was when I made the decision and I also will put a small text comment at the arrow telling me what my indications were. what setup signals were in alignment and why I think that trade will work.

Ya, i've developed many strategies for personal use. Optimization is almost an art as much as it is science. You always have to be aware of the failings of tradestations fill logic when backtesting. One the easier ways to help avoid over optimization is only test your inputs on a portion of your back time frame, then later bring in the rest of the data and see how it performs on the new data.

I/we use mostly Tradestation for the signals and logic. some bits are Ruby, and custom dlls when vector language doesnt cut it.

I develop strategies, indicators ect. for a hedgefund and other clients. I just had some questions about how other strat developers deal with certain types of clients and compare notes ect.

Does anyone here program mechanical trading systems for a living or have created any for sale?

I'm not too embarrassed to say that i was stuck in a pattern early on in my trading where I would make good gains in the first 2 hours of trading and then in the next hour give back half my winning as the market would fall into chop. Its strange how a market can start to fall asleep and do more consolidation in a sneaky manner without realizing it, especially if your focused too much on setups. Reading the tape definitely helps cure this. Now, after the 1st 2 hours of the market I either change my strategy to suite the market or just walk away and feel good about it.

All of my indicators are custom. I use an equal combination of oscillating and trend type tools and not the kind that clutter up the price action. Price is always the primary indicator. Also, unlike classic indicators, I write most of my indicators based on identifying reliable patterns and capitalizing on "why" the market is doing what it is.

Despite the tight range, this is one of those fortunate days that the russell was very consistant. You had consistant higher highs and higher lows in the uptrends and lower lows and lower highs of the down trend. Trading from moving averages was ideal for these kinds of days. Also the market didnt suffer a lot a painfull consolidation before changing short term trends. The strategy would have made a profit today. Here are a few things to try as well: 1) Instead of the moving average line, create a 34ema of the highs and 34ema of the lows, use this small channel to help avoid trading into consolidation or ranged cycles, which is the times you want to *not* be using a trendline for setups. 2) Create a MACD of the 5 and 34 ema values to better highlight divergance. This will help when market gets choppy and also provide a stronger signal when that pattern occures near of price of resistance or support.