Building/Testing Tradeplans

[tjnoon 10]

How many trades do you feel are an adequate amount of trades to feel confident that the particular plan you are testing or developing would have a good chance of succeeding? With my own personal testing, I have found that anywhere from 75 to 125 trades seems to hold up. The more the better of course, but when I am looking at my results, it seems my key statistics stabilize somewhere in within that quantity of trades. Any thoughts, opinions, or findings?

[TheNegotiator 10]

75-125 should be a decent number. But it is important to encompass different market conditions and volatility. Bullish/bearish/congested/cyclic conditions could all impact the system on any given trade along with volatility and therefore need to be covered in testing. Otherwise it's like saying you get a hole in one every time you play a golf shot and yet fail to mention the hole was 2ft away each time!

 

TheNegotiator.

[MadMarketScientist 10]

I'm right there with you guys. I think 100 trades typically encompasses almost all scenarios. Obviously any system over time is going to need some tweaking on occasion since it's just the nature of all markets do decide and change their behavior. As long as you make those adjustments based upon what you see over 100 trades and not 5 or 10 trades usually I find that pretty safe.

 

I also like to stress my trading systems by seeing how they would do in the bad times of the day -- times I usually would not trade due to the ranges/volumes/success being tougher. I find if it holds up at least acceptably, like a scratch than I'm onto something good.

 

MMS

[SIUYA 10]

[quote name=MadMarketScientist;110999

I also like to stress my trading systems by seeing how they would do in the bad times of the day -- times I usually would not trade due to the ranges/volumes/success being tougher. I find if it holds up at least acceptably' date=' like a scratch than I'm onto something good.

 

MMS[/quote]

 

for me this is the crucial part - its the opposite of selling a system and only showing the good parts

 

making money is easy - hanging on to it is the hard part.

[TheDude 10]

Given that no 2 market events are under the same circumstance, is there any point in back testing? its always different.

Although, I see where youre coming from - My answer is to make sure the outcomes you test account for the wider background/context. Lets say you test on a x min chart. A set up may work fine in a daily trending market, but will fail on a daily rotating market. What about sentiment etc. Bonds and notes for eg will trade off levels implied by their yields which may not show on a chart. Yet the wise are prepared. Also be aware of other market levels your market is spread against by the big money (who always trade spreads not out rights)

 

Understanding & screen time beats backtesting, but takes years not hours/days.

[teq56 10]

Once upon a time in a statistics class I remember that the more variables your system has the more backtest you need to do. Something about a "T" test or "F" test. Not sure which. The more variables in an equation the larger the data set needs to be in order to be statistically significant. I'm sure google will provide more info than my failing memory. A weaker memory says something like 30 test per variable minimum! 3 indicators or filters means 90 test minimum. Like I said... this was a long time ago. I generally start with 100 before I will even consider looking any harder at an idea.

[tjnoon 10]

All great points. I can't argue with any of them, really. I would say though that backtesting is useful and important to do IF you understand what it tells you and what it won't tell you. If you have an actual tradeplan or are trying to develop one, I think you can learn a lot from a good honest backtest. Remember guys, this is the UTA sponsored forum and so.. , I have to mention that it is a tool that I use to actually manually backtest (yes, click by click, bar by bar) a tradeplan that I am considering throwing real money at. Honestly though, it has made a huge difference for me.

 

I realize it will not be the same as real life trading. But it's a starting point for me. My next step is actually to practice the execution of the plan in real time for a while. Does it continue to hold up? Once I prove it does, I'll begin trading it for real with the smallest position size possible. That's my final test. But it begins with the backtest. What backtesting will do, and I don't think many people think of this or realize this until they experience it first hand, is confirm my understanding of my own tradeplan rules, confirm whether the idea had merit which could indicate that it still might, and, it helps me live with the discomfort of posting losing trades. Even in a backtest, losing trades don't feel good. That is very important because if the tradeplan is a good one, the equity curve will still grow as will my confidence that the winners DO follow the losers. In the end, the most success I have experienced as a trader has come from a good solid manual backtest where I was able to intimately learn the relationship between the wins and losses while also witnessing the equity curve head upwards. That increasing equity curve has within it, all the losing trades and tough sessions, yet the profits continue to grow despite that. Light bulb moments come from THAT experience, at least I can say, it did for me.

 

One of my favorite strategies, a daytrade technique I use to trade the Russell eMini every morning, began as a backtest. Right around the 100th trade of my backtest, my stats started to stablize. It deviated a bit, a couple percentage points up and a couple down, but the overall important stats remained in a good place; a profitable place.

 

I continued backtesting it though and recorded several months worth of trades, over 300 or so, if I remember right. The losses still made me feel uncomfortable, especially the occasional 4th loss in a row, yet my confidence in the system got stronger and stronger as each new stretch of winners followed. THAT in itself, that particular experience, is way worth the price of admission.

 

Now, over 1000 trades later, the stats continue to purr along, like a finely tuned engine and the equity curve consistently posts new profit levels on a steady basis. Even so, this past Monday, for example, was a tough step backwards. Now though, three sessions later, I have hit new equity highs again, winning 6 of the last 7 trades. But then, I knew that would happen because that's what it does. So tomorrow, I have the confidence to show up again, and take the next trade. How cool is that? lol.. But it all started with the backtest. And for me, it was around 100 trades that established the stats that have since remained stable and consistent, and continue to manifest in the ever growing equity curve. So, needless to say, I would encourage anyone who wants to improve their trading to take the backtest process seriously. Afterall, it can't hurt you. And if it could help you like it helped me, then why not? That's my story and I'm sticking with it.. lol..

 

And if you don't want to go through the effort, well, I got a great robot to sell you.. It's called the "Market Destroyer," (designed by Dr. Smith himself) and it is guaranteed to deliver 95% winners. Just set it and forget it.. Trading IS that easy, right? (comes with a free bottle of "CureAll")..

Edited February 10, 2011 by tjnoon