You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
Extremely simple strategies with > 100% annual return
- Thread starter jcl
- Start date
Quote from jcl:
Selling tools that promise a dream is a sort of contradiction: if you have such a tool, and it works, you obviously don't need to sell it. I have no intention to sell tools. Nowhere on the Zorro website is anyone asking you for money.
Does anybody know how we can try the LowPass filters in Matlab or R? I am curious and would like to try and compare them... Thanks!
Quote from jcl:
Selling tools that promise a dream is a sort of contradiction: if you have such a tool, and it works, you obviously don't need to sell it. I have no intention to sell tools. Nowhere on the Zorro website is anyone asking you for money.
Notice the very key word *almost* which keeps things perfectly legal. Okay, may be you are not directly selling them, but there are other ways to monetize once you have a big following (see google or facebook for examples). I might sound a little sarcastic, but I am quite serious and I have always considered this as a possibility for my own future plans. I figured, I will try my best to improve my algorithm and run it for a few years. If it proves to be statistically insignificant, then I can always figure out how to polish and distribute my tools. Afterall, I have spent a huge amount of time writing everything from scratch as well. And it may be that in my hands, I just couldn't find that algorithm -- it could be that others might find it with these tools. However, I also believe that it is probably harder to find an algorithm using existing tools for several reasons. First to make things polished and user-friendly enough, it takes a lot of simplification and coding effort such that usually only about 10% of what you would like gets pushed out as a release. Second, that 10% is likely those that are the simplest to implement, and the simplest to specify by the end user. Third, that 10% must be the most convincing set of tools in order to receive widespread adoption which leads to the fourth: simplicity in itself is good. However, if it is obvious (anybody can think of it), then it will likely be exploited and averaged out in the market. Fifth, if it is obvious and simple, but hard to implement, then again introducing a canned package will basically give that edge an expiration date, as more and more people exploit that pattern. Therefore, I believe it is more likely to find something if you implement things yourself. Just my opinion.
I think Zorro works without admin rights. On the "What's New" page you can find an email address for a download link to the current version, which is a stable beta: http://zorro-trader.comQuote from mizhael:
Interesting... where are the free software and tools? If they are available without the requirement for system admin rights... then I will be able to use them on my PC... (I don't have admin rights:=() I want to give them a try... Thanks
Quote from jcl:
Sure, here's the performance sheet:
The equity curves and other details of both simple strategies can be found in the Zorro workshops 4 and 5:Code:BackTest Workshop5_1 EUR/USD - performance report Test period 04.05.2008-30.12.2011 Lookback period 541 bars (130 days) Assumed spread 2.6 PIP (roll -0.01/-0.01) Assumed slippage 2.0 sec Contracts per lot 7800 Gross win (loss) $10116 (-$5090) Average profit $1375/year, $115/month, $5/day Max drawdown -$1563 Max down time 71 weeks from Sep 2008 Largest margin $400 Trade volume $1395980 = $381958 per year Capital required $1963 Number of trades 64 = 17 per year Percent winning 19% Avg trade return $79 (+$843 -$98) Max win (loss) $1658 (-$190) Avg trade time 43 bars (10 days) Max trade time 447 bars (107 days) Time in market 47% Max open trades 4 Max loss streak 10 Annual return 70% Profit factor 1.99 Sharpe ratio 0.63 Kelly criterion 0.56 Ulcer index 17.48% Prediction error 72% Fac Win/Loss Profit EUR/USD:L 0.3 3/28 909 EUR/USD:S 1.7 9/24 4116
http://zorro-trader.com/en/tutorial_trade.htm
http://zorro-trader.com/en/tutorial_fisher.htm
This system is a joke. I do not think your LPF can generate systems that are better than a simple MA crossover. Why don't you provide backtest results for 10 - 15 years? There is free FX data all over the place.
Quote from jcl:
Selling tools that promise a dream is a sort of contradiction: if you have such a tool, and it works, you obviously don't need to sell it. I have no intention to sell tools. Nowhere on the Zorro website is anyone asking you for money.
Time is a factor.
Why develop a free tool following instead of trading?
Yes, we're confronted with the same considerations. The existing trade platforms have not very good tools for developing strategies. It starts with their script languages, which are mostly clumsy to use and slow as hell. We found that we had to write the tool first before we could start developing strategies. But we tried to write it already in a polished and user-friendly way, because when you do that at the end, it never gets done. We also tried to make the system as open as possible so that you can implement anything and are not hampered by some system limitations.Quote from ssrrkk:
Notice the very key word *almost* which keeps things perfectly legal. Okay, may be you are not directly selling them, but there are other ways to monetize once you have a big following (see google or facebook for examples). I might sound a little sarcastic, but I am quite serious and I have always considered this as a possibility for my own future plans. I figured, I will try my best to improve my algorithm and run it for a few years. If it proves to be statistically insignificant, then I can always figure out how to polish and distribute my tools. Afterall, I have spent a huge amount of time writing everything from scratch as well. And it may be that in my hands, I just couldn't find that algorithm -- it could be that others might find it with these tools. However, I also believe that it is probably harder to find an algorithm using existing tools for several reasons. First to make things polished and user-friendly enough, it takes a lot of simplification and coding effort such that usually only about 10% of what you would like gets pushed out as a release. Second, that 10% is likely those that are the simplest to implement, and the simplest to specify by the end user. Third, that 10% must be the most convincing set of tools in order to receive widespread adoption which leads to the fourth: simplicity in itself is good. However, if it is obvious (anybody can think of it), then it will likely be exploited and averaged out in the market. Fifth, if it is obvious and simple, but hard to implement, then again introducing a canned package will basically give that edge an expiration date, as more and more people exploit that pattern. Therefore, I believe it is more likely to find something if you implement things yourself. Just my opinion.
i dont understand the hype around sharpe ratio........if you truely could replicate 0.72 sharpe over the long haul (10 year backtest year on year profitable), then of course it would be more than tradeable.
People can't even manage to find one postive expectancy system in their life time, that is how hard this game is.
Why not use cents per share or another gauge to measure your system. Sharpe ratio has many flaws.
Also there is no point posting a backtested system on here unless you have done the following:
1. 10 years of data with every year upon year being profitable.
2. limit and market orders in backtest should not be considered until it exceeds price on both entry and exit otherwise your wasting your time and the results you are producing are going to be fake.
3. same parameters over the whole course of the 10 year data with no curve fitting to suit years, one set of parameters only.
4. Make sure commissions are in.
Do you realise that bookmakers, casino's only have a small edge themselves?, infact Lescor who is a consistantly profitable trader here on ET states himself that his edge is something like only 2 cents per share.
Do the above then post results please.
regards,
People can't even manage to find one postive expectancy system in their life time, that is how hard this game is.
Why not use cents per share or another gauge to measure your system. Sharpe ratio has many flaws.
Also there is no point posting a backtested system on here unless you have done the following:
1. 10 years of data with every year upon year being profitable.
2. limit and market orders in backtest should not be considered until it exceeds price on both entry and exit otherwise your wasting your time and the results you are producing are going to be fake.
3. same parameters over the whole course of the 10 year data with no curve fitting to suit years, one set of parameters only.
4. Make sure commissions are in.
Do you realise that bookmakers, casino's only have a small edge themselves?, infact Lescor who is a consistantly profitable trader here on ET states himself that his edge is something like only 2 cents per share.
Do the above then post results please.
regards,
Quote from ssrrkk:
two words: curve fitting -- if in fact all you can do is curve fit, then your sharpe ratio and profit factor and risk reward will be out of this world by creating a following compared to trading.
In this particular case it appears that a good system cannot be produced using as a bait the low pass filter concept. If that were true, most electrical engineers would be billioner traders. Only promises were offered but no actual good results.