Yeah that movie was hilarious. As for your question, I understand what you are saying now. Some of the settings I use change depending on how far back the sample goes, and what timeframe is being used. Here are what I used to dismiss:
Return/DD Ratio < 2.5
Number of Trades – this depends on the strategy timeframe used, and how far you’re looking back to sample. If it’s a 4hr strategy, I want at least 1 trade a week on average. If it’s a 5min strategy I want at least 3 trades a week. You just have to do the math accordingly. Example: 5 year test period, 4hr strategy would be “Number of Trades < 250" would be dismissed.
% stagnation > 15% this one is real important. This will eliminate strategies that performed real well during a very small period of time. I want strategies that steadily increase constantly.
Net Profit – this depends on lot size and time period. I started with .01 lots, 5 years I want about “Net Profit < $500". But you'll have to tinker with this as each strategy and pair will have different ranges.
Net Profit (OOS) – this depends on how long your OOS range is. Mine is usually a year and for .01 lots for 1 year I want at least a $50 gain for that year, which is half the average of $100 gain I require for IS periods.
Hope that helps.
The Net Profit is in dollars, not %. If it’s .01 lots, I want about $100 Net Profit in sample a year, on average. So if testing for a 5 year period, I only want strategies that have a Net Profit in sample of $500 or more total. So the dismiss setting is “Net Profit < $500".
As for stagnation, what is your question on that one?
Portfolio is drawing down after an explosive +42% month in June (FxBook). So this drawdown seems healthy overall, as long as it’s reasonable it will be part of the expected “measured drawdown”. Drawdown periods are the most important aspect to watch in strategy development in my view, so the coming weeks for these strategies will be important in how well I developed StrategyQuant strategies and implemented them in real time live trading.
Overall drawdown is now just over 10%, and during testing I always tried to keep drawdown below 15%, so there’s still some room. But also keep in mind that I was using .01 lots only during testing, and in live trading I bumped the strategies up to .02 lots, and as it climbs higher I will continue to increase the lot size to compound profits over time. Regardless, I’d like for overall drawdown to remain under 15%, but must be under 30%, otherwise I must re-evaluate position sizing, or possibly re-evaluate the strategies.
The key for me at this point is not necessarily to make as much money as possible, but that my strategies unfold in real time live trading similarly to how they performed during testing. So far so go, but only 3 months in so we have to let this run further.
Hi guys ,
im struggling and i need your advise .. can you tell me what king of sl and tp you search for ?
is sl and pl required ?
or you use only indicator exits ?
fixed pips , atr ?
which one of these makes it more robust . as i am finding good strategies but failing the robust test always .
btw i am not using aaditional data as it makes the search so slow
I am also in DD on my automated portfolio.
The Public one in my link was started in mid-March near a peak of a rally. Its private counter-part which has been running for over a year is still positive for this year because of a really nice rally in January and February.
20% DD is normal for my portfolio and its a very slow account as 80% of it is strategies I made in EA Wizard that trade only D1 and have only about 20 signals per year(a bunch of those occurred Jan/Feb with the huge Dollar rally). The other 20% is SQ H1 strategies that trade more frequently with low risk to reduce stagnation. I don’t expect the public link to look “normal” for atleast a year or 2. Right now the sample is far too small and is not reflective of what a higher sample looks like. Nothing new in the market.
Gentmat, I would recommend some type of stop loss. Depending on the timeframe used, your stop parameters will need the be adjusted accordingly. Try doing a Simple Optimization and rounding your numbers off. Then, find the most common used numbers in all the optimization results and use those for your walk forward tests. If you’re having trouble with the Monte Carlo analysis, make sure your position sizing and account size is appropriate. if you’re trading standard lots on a $5,000 account, it will fail in drawdown percentages. When generations my strategies, set SQ to only produce results with less than 15% drawdown.
See if that helps.
Your portfolio looks good Todd, mine is still going down the last days, yours up I see. Good work! Any insight on the strategies? Mine are mainly breakout (e.g. buy stop at high / sell stop at low) ones, those don´t work well in the current market as there are no clear trends especially on EURUSD and breakouts are ever reversing. Are yours mainly working with market orders or also buy stop / sell stop breakout style? Just a little direction would be great to know:)
Also: how many trades do your strategies have in the backtest? I am always looking for at least 1000 trades in from 2001 to 2015, so that the statistical edge is not to small. What´s your experience there and how many years back do you optimize if I may ask?
Oh and most important: which timeframe are your best working ones? I use M30 / H1 only right now as anything above seems to produce to few trades for me and anything below doesn´t find enough stable strategies.
You must be logged in to reply to this topic.