Developing a Trading System Step by Step

[Murray Ruggiero]

Here are the results as produced by TradersStudio for the base moving average set of 21,35,85 , though Friday Jan 27th.

We will add our entry trigger in our next installment.

 

[jem]

I find your work here very interesting. I believe you are coaching us through the development of creating models.

I hope that you will follow this exercise thorught to the creation of something you would eventually trade with say a 50,000 dollars or even a hundred.

What will be very interesting to me is when, a person with your sophistication in system development would say - now it is ready for my money.

I know this question is very premature. But when I see less than 50% winners, I figure that the only people who could trade a system like that are the people trading with other peoples money or money they just do not care about. (which by the way may be the point of trading.)

This is not in any way a knock on the what you are doing here. I think you work here is generous even if does serve multiple purposes.

 

[Murray Ruggiero]

The problem with most trend following systems is that they win about 35%-45% percent of their trades. This is part of the problem with these methods. The only holy grail I know is to trade multiple uncorrelated systems, combine them using money management. This is exacting what TradersStudio was designed for. I would recommend , trading a Bond system and a stock index system combined with a trend following one. In most cases I use counter trend systems in these other markets, these often have higher winning percentages 50%-60%, some intermarket systems can approach 70% winners. When you combine these together ,then the overall winning percentage could be around 50% or higher, which allows you to develop a good overall mix. It also makes you feel better about trading them.

 

[BENG]

Quote from Murray Ruggiero:In terms of Sharpe ratio, I don’t like it as a performance measure because it penalizes sets of parameters that have volatility but positively sloping equity curves. A parabolic upward sloping equity curve would has a relatively low Shape ratio because of volatility.

I think higher volatility without higher return would of course lower the sharpe ratio. For the systems I've simulated so far, I haven't seen a parabolic upward sloping equity curve that have a shape ratio of less than 2. If 2 is a low relative sharpe ratio from your perspective, then please enlight me what do I need do to cheat the mathematics behind the sharpe ratio?

I've been developing futures trading models for multi-markets (stocks, currencies, bonds, and commodities) for a year, I find it extremely difficult to find something that has an edge after slippage. Trend following models would be the default edge for anyone who tries to develop a low profitable model. Intraday trend reversal model is like the hardest thing on earth (at least for me). Let's put it this way, for an intraday model without any slippage, and trade once a day, I need it to make at least 50% per year with a sharpe ratio of 2 (just to give you an idea of my slippage model). It can be done, but I don't know the right approach to do it.

Thanks for running this thread, but I would also like to know is there any other method to start building a model. Should the optimization be run after building the basic generic model, or optimize individual component first, then try to build a model with them?

When will TradersStudio support intraday portfolio backtesting and optimization. This is the reason why I haven't bought the software. I need to build intraday reversal model (feel like committing suicide).

Thanks!

 

[Murray Ruggiero]

Quote from BENG:

I think higher volatility without higher return would of course lower the sharpe ratio. For the systems I've simulated so far, I haven't seen a parabolic upward sloping equity curve that have a shape ratio of less than 2. If 2 is a low relative sharpe ratio from your perspective, then please enlight me what do I need do to cheat the mathematics behind the sharpe ratio?

I've been developing futures trading models for multi-markets (stocks, currencies, bonds, and commodities) for a year, I find it extremely difficult to find something that has an edge after slippage. Trend following models would be the default edge for anyone who tries to develop a low profitable model. Intraday trend reversal model is like the hardest thing on earth (at least for me). Let's put it this way, for an intraday model without any slippage, and trade once a day, I need it to make at least 50% per year with a sharpe ratio of 2 (just to give you an idea of my slippage model). It can be done, but I don't know the right approach to do it.

Thanks for running this thread, but I would also like to know is there any other method to start building a model. Should the optimization be run after building the basic generic model, or optimize individual component first, then try to build a model with them?

When will TradersStudio support intraday portfolio backtesting and optimization. This is the reason why I haven't bought the software. I need to build intraday reversal model (feel like committing suicide).

Thanks!

The sharpe ratio penalizes volatility both positive and negative because we divide by standard deviation, which can be high with a volatile upward path.

I have built many counter trend models which have done well, for example intermarket analysis trading the thirty year bond using UTY , I have discussed this in other threads.

You need to mix trend follow, intermediate trading and short term trading into a basket approach of systems. This is the best way to get high returns with while minimizing risk.


The next upgrade to TradersStudio , which is 2.0 will do intra-day backtesting. We expect real time out this summer. This means you can do your intra-day development when 2.0 is released, and just wait a few months for trading it. We also plan to increase the price after 2.0 is released between $100.00 and $200.00, If you own TradersStudio 1.3.6 it will be a free upgrade to 2.0 Off-Line and a discounted price for the real time upgrade, we plan on charging a discounted price of $399.00 for existing customers. The real time upgrade has a data server for saving historical data on your local machine. It kind of like TradeStation 2000i Global Server.

 

[BENG]

Quote from Murray Ruggiero:

I have built many counter trend models which have done well, for example intermarket analysis trading the thirty year bond using UTY , I have discussed this in other threads.

The next upgrade to TradersStudio , which is 2.0 will do intra-day backtesting. We expect real time out this summer. This means you can do your intra-day development when 2.0 is released, and just wait a few months for trading it. We also plan to increase the price after 2.0 is released between $100.00 and $200.00, If you own TradersStudio 1.3.6 it will be a free upgrade to 2.0 Off-Line and a discounted price for the real time upgrade, we plan on charging a discounted price of $399.00 for existing customers. The real time upgrade has a data server for saving historical data on your local machine. It kind of like TradeStation 2000i Global Server.

Have you built any count trend model that works in a basket of markets after slippage?

What do you use to compute the correlation table? I assume that excel file was prepared by TradersStudio?

In terms of intermarket analysis, how should we find a good basket of markets, It's easy to do analysis based on two markets, but can we logically add more markets or pairs to form a trading system basket?

When will v2.0 offline come out, you only said that the real time version will be out by summer.

Thanks.

 

[Murray Ruggiero]

Quote from BENG:

Have you built any count trend model that works in a basket of markets after slippage?

What do you use to compute the correlation table? I assume that excel file was prepared by TradersStudio?

In terms of intermarket analysis, how should we find a good basket of markets, It's easy to do analysis based on two markets, but can we logically add more markets or pairs to form a trading system basket?

When will v2.0 offline come out, you only said that the real time version will be out by summer.
Thanks.

The workbook I uploaded was produced in TradersStudio 2.0 . We plan on shipping 2.0 offline in in March.

I have built some proprietary counter trend systems that work on baskets. Intermarket systems really only work on a given sector or groups, so it not really basket trading like you think. It helps performance because they trade more frequency and are shorter term trading than the trend following systems.

 

[Murray Ruggiero]

I did not have time to finish the next installment of this tutorial today. I will as soon as I get a chance, I hope by tomorrow night.

I will give you a hint about our next lesson. Sometimes what you think should work does not. What is important to learn is that this is ok. You can learn a lot about why these concepts did not perform as well as expected and based on what you have learned develop concepts that will work.

 

[Murray Ruggiero]

Let's now discuss this next installment. In this lesson you will learn that your best ideas don't always work but you can learn from them. This series is special because we are showing the steps, both good and bad. Most articles show you a cookie cut world where everything works. Our goal here is to give you a journal like report and build this trend following system together.

We modified our triple moving average crossover system so that we added an addition condition for the buy rule. We not only require ShortMA>MedMA and MedMA>LongMA but added the condition that the close is less than the ShortMA, this means we require a pullback before getting into a trade. On the short side we added the condition close is greater than ShortMA. Our original system entered a trade at the market on the next open after the conditions were met. In our new system we added an entry trigger using an opening range breakout based on .30*Average(Truerange,3,0).

This sounds like a good idea, but it does not do very well. It makes a total net profit of $424,981 with an open position profit of $104,014.58. The drawdown does drop, to 111K from 159K. The big problem is the system loses money on the short side, and the long side profits are reduced by about 153K. Our idea made sense; the question is why did it not work.

I have included the rules, a few notes and the spreadsheet of results in the attachment. Please study the trades and analyze, why did this idea make less money. I want to hear your answers, after that I will publish my analysis and tell you where we go from here.

 

[BENG]

Your excel file cannot be opened properly...