One Way to Develop a Winning System | Page 3 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

lets develop a system which is very very simple but effective. 7 replies

How To Develop Winning Automated Trading Systems 15 replies

Change a losing system into a Winning system 68 replies

Develop a System that Fits You, by Dr Van K. Tharp 3 replies

Need Ea for System I am trying to develop 0 replies

Attachments: One Way to Develop a Winning System
Exit Attachments

One Way to Develop a Winning System

  • #41
  • Jul 12, 2007 9:25pm Jul 12, 2007 9:25pm
FXEZ, thanks a lot for sharing your rich experiences here. It helps us traders more than to you. You are generous.

So far I am amazed by the step 4, that's a brilliant aspect to deal with the system development, however, step 4 can be tough in programming tho.

Anyway, bring it on, it is very enlightening to read your writings.
  • #42
  • Jul 13, 2007 3:25pm Jul 13, 2007 3:25pm
Quoting greatserge
yes there are losses and somedays you might not see an opportunity to open a trade. that just puts you a few days behind schedule, thats no big deal.

now if you combine your trading system and my money management, you got yourself a fully functioning system (imo)
I agree with you that money management is a very important part of any trading system. Some say it is the most important part. I think it depends on the kind of system that's being used. My belief is that a unique money management strategy should be added to a system after a very good base of objective rules has been put in place, and customized according to how the data falls. The specific money management strategy used should fit to the system and historical P&L data so that it is best at enhancing system performance without unduly increasing drawdowns or account volatility. I should be adding more on this subject as this discussion progresses.
  • #43
  • Jul 13, 2007 3:28pm Jul 13, 2007 3:28pm
Quoting charvo
FXEZ, thanks a lot for sharing your rich experiences here. It helps us traders more than to you. You are generous.

So far I am amazed by the step 4, that's a brilliant aspect to deal with the system development, however, step 4 can be tough in programming tho.

Anyway, bring it on, it is very enlightening to read your writings.
Thanks for the kudos charvo. Yes step 4 can be a lot of work, particularly in keeping track of all the different combinations of rules and how the system performs both with and without those rules. It takes some time but I've found it to be a very worthwhile and enlightening exercise.
  • #44
  • Jul 14, 2007 8:00pm Jul 14, 2007 8:00pm
Quoting FXEZ
Does MTF stand for multiple time frame system? And if so, are you using the longer time frames as a trend filter and entering on the shorter time frames for timing purposes? I'm curious about this approach because backtesting multiple timeframes isn't that easy to do. But it is something that I want to spend more time looking into.
Ill save you some time.. its all worthless
  • #45
  • Edited 10:55pm Jul 14, 2007 10:45pm | Edited 10:55pm
Quoting FXEZ
I agree with you that money management is a very important part of any trading system. Some say it is the most important part. I think it depends on the kind of system that's being used. My belief is that a unique money management strategy should be added to a system after a very good base of objective rules has been put in place, and customized according to how the data falls. The specific money management strategy used should fit to the system and historical P&L data so that it is best at enhancing system performance without unduly increasing drawdowns or account volatility. I should be adding more on this subject as this discussion progresses.
FXEZ, Awesome thread!!! You are very generous in sharing your experience with everyone. I know it took me a long time to be able to see the richness in other's post's. That was when I stopped thinking that my system was the best and stopped being so close-minded. Anyway's, I have been trying to come up with a good money managment system and have pretty much decided to go with a fixed ratio method to get me out of the gate and then go to a smooth ratio once I get rolling as to cut down on the drastic equity curves involved with fixed ratio. There is a great program out there that is fairly new from what I understand but it's primary purpose is to optimize your money management method. It actually figures out which is best for you based on what you are looking for, example would be if you are looking for net profitability it would find the best setting for that, if a smaller drawdown is what you desire, it will figure out what MM method is best for that. It is called Market System Analyzer. I am in no way affiliated with these people, but I have tried the software and find it very useful. Have you ever dealt with fixed and smooth ratio MM methods?, if so, have you had good reslults with them? Basically a fixed ratio method untilizes a delta. For instance, if I had a starting account of 10k and my delta was 2k, and I was trading 1 lot, once my account balance reached 10,200 I would add another lot. IF my account balance got back down to 10k I would go back to 1 lot. This utilizes equity runs. Now fixed ratio can have some massive equity curves once you get up there in lot size. The example of a delta of 2k is very aggressive. The smooth ratio utilizes a delta but also uses a mulitplier to smooth out the equity curves. There is a great article on it in Active Trader Magazine's website, if I can find It I will post it. I have found nothing else which seems to have better results for a smaller account size than a fixed ratio method.
Thanks again for your great thread and insight.
Pippiper
  • #46
  • Jul 15, 2007 10:28am Jul 15, 2007 10:28am
I just want to make a short comment here:

It seems alot of traders here are going through money management fever

I have read entire books devoted to money management and I think it is absolutely necessary in any trading method or system alike.

But, I want to point out that no mathamathical formula or money management strategy can turn a:

NEGATIVE EXPECTENCY IN A POSITIVE ONE

what that means is, if the system you decide to trade by is going to cause you to lose your money over the long run, then money management can not help you, all it will do is allow you to lose your money slower and hopefully you will pick up on your mistake before it is to late.

Just make sure your entry and exit strategy will give you a positive expectency before you trade it!
  • #47
  • Jul 16, 2007 7:22am Jul 16, 2007 7:22am
Quoting superdezign
I just want to make a short comment here:

It seems alot of traders here are going through money management fever

I have read entire books devoted to money management and I think it is absolutely necessary in any trading method or system alike.

But, I want to point out that no mathamathical formula or money management strategy can turn a:

NEGATIVE EXPECTENCY IN A POSITIVE ONE

what that means is, if the system you decide to trade by is going to cause you to lose your money over the long run, then money management can not help you, all it will do is allow you to lose your money slower and hopefully you will pick up on your mistake before it is to late.

Just make sure your entry and exit strategy will give you a positive expectency before you trade it!
Thanks for the post, you make some good points. This is what I have come to understand. Money Management cannot turn a losing model into a winning model, BUT, poor money management can turn a winning model into a losing model, works both ways!
Pippiper
  • #48
  • Jul 17, 2007 2:04am Jul 17, 2007 2:04am
Quoting superdezign
Ill save you some time.. its all worthless

Appreciate the "heads up" LOL!
  • #49
  • Jul 17, 2007 2:10am Jul 17, 2007 2:10am
Quoting Pippiper
Thanks for the post, you make some good points. This is what I have come to understand. Money Management cannot turn a losing model into a winning model, BUT, poor money management can turn a winning model into a losing model, works both ways!
Thanks for your comments guys. I think this just about sums it up. And I hope to be adding some commentary on money management and a few other topics soon.
  • #50
  • Edited 4:39am Jul 17, 2007 2:16am | Edited 4:39am
Step 5: Evaluate the System:

After we have put our system together piece by piece, testing dozens of alternatives and have taken it apart and found that only the best components remain as a part of our system, it is time to objectively evaluate how well the system actually performs. In addition to evaluating the net profit, max drawdown, the NP/DD ratio and win %, there are a few other things we need to evaluate.

Consistency: The easiest way to evaluate a system’s consistency is to dump the trade history to a spreadsheet like Excel, and chart the profit and loss series. The ideal is to see a pattern of similarly sized profits and similarly sized losses. If your system produces a few really big wins or a few really big losses, this is a red flag that a potential problem could surface. A few really big wins usually means that your system is over-optimized because it relies on too few large wins to make up the bulk of its profits. A few really large losses is a red flag that your system lacks risk control, and that it has the potential to produce an even bigger single loss or string of larger losses that might wipe out or seriously impair the account. While there will be a significant degree of variance in your system results, in general the more consistent the results, the better.

Tradability (Part 1): This is probably the most important thing to evaluate before going live. Probably the easiest way to do this is to code up the system in a format that can be automatically executed on a broker’s platform. For most forex traders this will mean coding it up in an expert advisor for use in Metatrader. While Metatrader’s backtesting capability is suspect, its forward testing capability has proven to be very good, even on demo for most brokers. The exception is on a broker that uses a significantly different price feed on demo vs. live. More on this later.
  • #51
  • Edited 4:39am Jul 17, 2007 2:21am | Edited 4:39am
Profit and Loss Expectations: Initially when we started developing our system, we had some profit and loss expectations about what our ideal system will produce. As system development progresses, those expectations typically shift a bit. This is a good time to reevaluate what it is that you want from your system, and to objectively evaluate whether the system in its current form is capable of meeting your expectations. If your system is capable of meeting your expectations, then there are still some other hurdles that need to be cleared before the system can be implemented on a live account. If your expectations aren’t being met by the current system, either the expectations or the system need to be adjusted.

One of the most important things in trading is to be congruent. At this stage of development it is important to be honest about whether your expectations are being met or not by a given system. Implementing a system that doesn’t meet your minimum profit/drawdown expectations is a disaster in the making because subconsciously you will either attempt to increase results by using too much leverage which almost always dooms a system to failure, or will not be committed to sticking it out when the system goes through a normal drawdown that you aren’t psychologically prepared to handle due to the incongruence of your expectations and the actual performance of your system.

Assuming the system passes all these hurdles during evaluation, it is time to go to work on developing a money management strategy that will enhance the system’s normal positive edge with increased performance while not creating undue problems with over leverage, increased drawdown, lack of congruence, or inconsistency or untradability.
  • #52
  • Edited 3:04pm Jul 17, 2007 3:01am | Edited 3:04pm
Step 6: Pair a money management strategy with our core trading system to enhance overall system performance without that bad aftertaste:

As this step states, our aim is to add a set of money management rules to our core strategy that will outperform the base system without introducing negatives such as drawdown or decreasing overall system expectancy. For any given underlying system there is a money management strategy that will fit that system best based on the performance statistics that you are looking to optimize. I haven’t tried it but apparently there is even a piece of software called Market System Analyzer that purports to do exactly that. Money management can and should be customized to the specific system that is being designed and the strategy used will vary greatly depending on the system’s performance results. For example, the money management strategy that will optimize an 80% win system will be very different than the strategy that will optimize a 30% win system. Fortunately it basically boils down to an understanding of math and an awareness of the pitfalls that can occur. It is important to recognize that when you fit a money management strategy to an underlying system, you can optimize too much, just as it is possible to overly curve fit during the system design process. The specific MM strategy that you employ should be robust and able to hold up during lots of different conditions.

Monte Carlo Simulation: One of the things that I like to do when determining what type of strategy I should use is to dump the trade log into Excel and perform some Monte Carlo simulation to analyze system results. Essentially you want to randomly jumble the trade by trade results to get an idea of what types of extreme strings of profits and losses are possible, as well as how the maximum drawdown can significantly increase or decrease just by jumbling the order of the trade by trade results and performing an analysis. The results are very eye-opening and should be used to generate a more realistic picture of what maximum drawdown and profit and loss streaks may actually look like on future data. One important question to ask is whether your proposed money management strategy holds up under this Monte Carlo simulation, or if it falls apart when profits and losses don't occur in some optimal order. If your proposed money management strategy has problems during this analysis, it should be reevaluated and possibly a less aggressive and more robust approach needs to be used. More on the Monte Carlo method can be found here: http://en.wikipedia.org/wiki/Monte_Carlo_method and you can find more on how to apply this method to trading here: http://www.adaptrade.com/Articles/article-mc.htm

The topic of money management is a large one and there are many good books on this subject. Fortunately, it is one of the few topics where common sense tends to apply in trading and the principles are generally governed by math concepts. What this means is that if a particular money management strategy makes system performance worse in the performance statistics that you're trying to enhance, then don't do it. I’m not going to cover every aspect of money management, but will make a couple of additional points that I think are important in understanding the concept. More to come...
  • #53
  • Jul 17, 2007 3:33pm Jul 17, 2007 3:33pm
Identifying Lot Sizing Drag With the Fixed Dollar Amount of Equity Method: One common and popular theory on lot sizing is that it should be tied directly to changes in equity. The basic principle is that when equity increases, we should increase our trade size, and when equity decreases; we should decrease our trade size. Under the fixed dollar amount of equity method of money management (MM) we will trade 1 lot for every $5,000 in equity. Let’s look at an outlandish example to show how doing this the wrong way can lead to significant drag on the account. Assume a $20,000 account and since we are trading 1 lot for every $5,000 in equity we will trade 4 units or lots. Let’s assume we have a system that wins $100 per lot on each win, and loses $100 per lot on each loss, and has a 50% win rate. Furthermore, there is a series of 10 trades that goes like:

WLLWWLLWWL (W=Win, L=Loss) – note there are 5 wins and 5 losses.

Using what we know about our MM strategy our P&L results look like:

Starting Equity: 20,000 – trade 1 lot per $5,000 equity
Inserted Code
  [font=&quot]P&L     Equity      Nxt Lots     Trade#[/font]
  [font=&quot]+400     20,400        4         1[/font]
  [font=&quot]-400     20,000        4         2[/font]
  [font=&quot]-400     19,600        3         3[/font]
  [font=&quot]+300     19,900        3         4[/font]
  [font=&quot]+300     20,200        4         5[/font]
  [font=&quot]-400     19,800        3         6[/font]
  [font=&quot]-300     19,500        3         7[/font]
  [font=&quot]+300     19,800        3         8[/font]
  [font=&quot]+300     20,100        4         9[/font]
  [font=&quot]-400     19,700        3         10[/font]
Compare this result to using even lot sizing and we have a difference of –300 created by a money management strategy that is sub-optimal or that introduces DRAG. Unfortunately it isn’t possible to create a MM strategy that has a net positive boost on overall results without affecting drawdown and other statistics (due to increasing leverage), but it is possible to create a negative drag on equity by using a sub-optimal strategy. Drag can become a major negative factor in system results when the impact over hundreds of trades is analyzed.

I’m going to admit that I made this discussion so far very elementary. The basic idea is that when you decrease lot size after losses, you need bigger wins to come back when you base your lot sizing on open equity levels. But is there a method to size trades that eliminates drag from the equation?

More on generic position sizing strategies can be found here: http://www.adaptrade.com/Articles/article-ps.htm
  • #54
  • Edited 4:57pm Jul 17, 2007 3:42pm | Edited 4:57pm
Drag-Free Position Sizing: My own preference is to think in terms of equity plateaus. In the example if reaching $20,000 were a plateau, you would NEVER reduce trade size once you have increased your size at that level. In actual practice, there are times when you need to have the flexibility to pull the plug for various “life” reasons that may have nothing to do with system performance. But when done on a systematic basis, in my view the practice of reducing lot sizing after losses should be avoided in order to avoid the negative drag that this introduces. This drag can be quite significant. Is it any wonder that some FX dealer platforms have a built-in default % of equity sizing option for new orders?! Just a note: using the concept of equity plateaus is actually a slight martingale strategy. You increase drawdown risk by keeping lot size constant during a drawdown, and betting that the system will recover from its drawdown normally. If the account is on its way to blowout, using equity plateaus will hasten the decline, so it tends to be a strategy that can increase overall risk on the assumption that many small recoveries are more likely to occur than a total account blowout with a given method.

I’ve attached an analysis in Excel comparing the base Fixed Ratio position sizing method with the Equity Plateau method of Fixed Ratio position sizing to provide an example of the difference between the two and what it translates to in terms of net profit and drawdowns within a series of trades. Hopefully it will become obvious why I favor the Equity Plateau method over the base method. As an added bonus I included the effect of using fractional lot sizing using the enhanced Equity Plateau method, and the results are almost double profits over the Basic method, with only a 50% increase in drawdown!
Attached File(s)
File Type: zip FRPS_EquityPlateau.zip   35 KB | 766 downloads
  • #55
  • Jul 18, 2007 3:49pm Jul 18, 2007 3:49pm
Step 7 Forward Testing - the Reality Check: Now that we have evaluated our system to ensure that it meets our profit and loss expectations, meets the criteria of consistency, and can be coded for our broker’s platform, we should delve a bit deeper to ensure that our system is in fact tradable in its current form. We also have added in a set of money management rules that should enhance our system’s performance. We need to get all of this coded up and start testing to ensure that it is operating as intended.

I’m assuming at this point that any system you’re evaluating has been sufficiently backtested and isn’t over-optimized. I suggest demoing systems for at least 2-3 weeks and thoroughly evaluating each trade decision made by the system before making any decision to go live. You will undoubtedly find bugs in your code as you forward test and observe your system under forward testing conditions. When you correct a bug in the code, or make any change to the code, whether it be adding a new parameter as an input, or making a slight change to the internal system or order handling logic, the forward testing clock should be reset. There is no need to rush into going live, and a full and thorough examination is prudent before making that final decision. Systems that haven’t been backtested at all should be demoed for several months to see the whole gamut of possible up and down cycles, and to particularly understand what type of drawdowns are possible. Essentially when you make the decision to demo a system, you are committed to seeing if it performs the same way in live trading as it does in testing. Analyzing trade history and comparing backtests with forward tests is critical. It is important here to analyze system logic to ensure that the system is actually executing how it should be according to the system’s initial design and intent.

The other main difference between live and demo trading on Metatrader brokers is in slippage and fills. There is no way to know exactly how your system will perform on the live server until you try it out because this will mostly depend on your broker. Typically demo and live server performance should be very similar. But the best bet is to start very small when you decide to go live, and evaluate the slippage and fills / execution differences on the live server vs. the demo server, which means you should continue running your forward tests even after you go live.

You will undoubtedly need to add in position management code to deal with issues of real-time fills. How will you deal with orders that are not filled at the specified price? Will you continue attempting to fill the order at the market price, or will you only execute an order near where the original signal was given? Not all FX brokers are equally good at providing quick fills, so this becomes an exercise in understanding how your broker fills orders. How will your system respond during times of widened spreads (such as during fundamental news reports)? The other thing to keep in mind is that system performance may vary SIGNIFICANTLY from broker to broker depending on spreads, fills and the prices provided by your broker.
  • #56
  • Edited 4:58pm Jul 18, 2007 4:06pm | Edited 4:58pm
I received some questions via PM and email that I’d like to answer here.

Question: What is expectancy?

Answer: Expectancy is a mathematical formula that can be stated as follows:

Expectancy = Average Win * Win% - (Average Loss * Loss%)

Average Win is the sum of all the wins divided by the total number of wins and Average Loss is the sum of all the losses divided by the total number of losses. Loss% is 1-Win%.

Positive expectancy values indicate systems that are profitable, while systems with negative expectancies are unprofitable. As a result all things being equal, systems with higher expectancies are better than systems with lower expectancies.
  • #57
  • Jul 18, 2007 4:36pm Jul 18, 2007 4:36pm
Question: Could you please tell me how to properly backtest my trading strategy?

Answer: The answer to this question really boils down to what you’re trying to accomplish and the type of strategy you are attempting to develop. If you are developing an end of day strategy, your backtesting requirements are going to be less strenuous than if you are developing a system that requires high frequency data (data less than 1 hour in length).

There are really two basic ways to backtest: manually and mechanically. If you have no other resources available such as backtesting software, or ability to program your strategy, manual backtesting might be your only option. The key with this type of backtesting is to keep it as objective as possible. The human mind tends to focus on what it is attempting to find, and to exclude what it isn’t trying to find. So the mind may tend to find successful outcomes and ignore unsuccessful outcomes. Your job as a manual backtester is to be objective and find all outcomes in the data, good or bad and to report the results accordingly.

Most of you are probably familiar with Metatrader software. I don’t suggest relying on its backtesting results. It has many flaws and limitations. But it can give a decent idea of what you can expect with a given rule set. But I wouldn’t suggest using it on data less than 1 hour in length. The reason is the data granularity Metatrader uses to simulate test results. The software is free, highly accessible and there is a lot of information out there on using it for backtesting. If you are intent on using it for backtesting, do yourself a favor and educate yourself on the best practices for Metatrader backtesting before getting too involved in the project. There is a way to increase modeling quality but I would still strongly suggest not relying on backtest results on data less than an hour in length.

If I were evaluating potential backtesting software the first questions I would ask is whether that software is capable of handling bid/ask data, and if it has accurate measures for position leverage and margin utilization on open positions. In other words, will it accurately simulate a margin call? I don’t have any suggestions for good backtesting software that is commercially available. I didn’t see any I liked, so I wrote my own backtester and collect my own bid/ask data for testing.
  • #58
  • Jul 18, 2007 4:52pm Jul 18, 2007 4:52pm
Question: What should I base my system on?

Answer: There are a few fundamental concepts that I think any good system should take into account. The concepts that I have found to be most valuable are support and resistance, trend, breakouts, and volatility.

Support and resistance can be defined as previous levels where price has turned. Typically that means previous highs and lows. The most common way of analyzing support and resistance is through drawing channels or trendlines.

Trend is an interesting concept. The more clearly you can define “trend” the better you will get at capturing profits from your definition of trend. There are trends on all timeframes. How small a timeframe that works for your definition of trend really depends a great degree on execution costs. If you can define trend sooner, you have an opportunity to capture that trend sooner. Most indicators were developed as a way to identify and capture trend. Most of them do a poor job of this. I don’t favor using lots of indicators to determine trend. I favor using support and resistance in combination with volatility to define trend.

Volatility is a great concept that can be defined in a number of ways. It can be defined as momentum or rate of change. I think it is most useful to define it in terms of support and resistance. When price moves through a support or resistance level by a given volatility figure, you have what may be defined as a breakout or trend signal.

The combinations are really endless, but I favor sticking closely to the basic concepts of trend, breakouts, volatility, and support and resistance because they have proven to be robust concepts that can be objectively defined and tested.
  • #59
  • Jul 26, 2007 12:15am Jul 26, 2007 12:15am
Hi, nice work and I've been following and waiting to learn more :-)

Is there a step on adjusting the system further after going live? or some statistics analysis on the system performance?

Trade well!
  • #60
  • Aug 24, 2007 2:11pm Aug 24, 2007 2:11pm
system is just a system....
without discipline... even with a good system you can not make money

regards,
"Live is for living not for working"
0 traders viewing now