A Little Survey - On Expectations
I'm trying to Gage the type of performance I should shoot for. Just a simple Yes or No answer here would do. Have any programmer here managed to create an EA with the follow results or better? If your answer is Yes, can you specify the PF, MD and Average #Trades per year. Thanks Guys.
Average:
2% or Greater Profit Factor.
20% or Less Maximum Draw-down.
200 or More Trades per Year.
No
Backtest or live/forward results?
Backtest results -> sure, its no problem to get those kinds of results.
But that is the problem with backtesting. Think of it like this...let's say you happen to you watch the World Series Poker Tour on TV one evening and you happened to write down what every player's cards were, how much they put into the pot, etc. And then you say to yourself "can I create an automated poker player code that would have beat everyone in the WSOP tournament last night?". So you set about backtesting until you have a strategy with parameters that just seem to magically know when to hold'em and when to fold'em...based on the poker hands data you collected while watching that one tournament.
Your win rates are high, your drawdown is low, and you hands-in are high.
Now you attempt to use that same (over) optimized strategy in the next WSOP tournament...and you watch it fail miserably. How could this be!? All your stats were so high from the backtesting. Ah but your backtesting basically amounted to you just creating a system of holding and folding that coincidentally fit neatly into the specific card and hand history pattern that developed over time during just that particular evening of poker play...you hadn't actually developed a good strategy at all, merely one that was optimized to coincidentally do well within the backdrop of a specific timeseries...a non-reoccuring timeseries at that. Which is why the same strategy failed (and will continue to do so) when you attempted to use it on a different timeseries.
So you have a question to answer...how to know when backtester is legitimizing your strategy versus when backtester is simply over-optimizing your parameter set such that by pure random chance the resultant strategy/parameter set opened and closed trades for a net profit in a way that will never be reproduced in future timeseries?
Back-Tested results as well as forward Tested results are welcomed. I don't believe allot of people have forward tested results for over a year. But i could be wrong.
Phillip, I understand your point. However, World Series Of Poker and Everyday Poker are two different monsters. The strategies that work in High Stakes Poker or Grand Event's Poker are very different than say On-line Low Stakes Poker. Your example illustrates nicely what wouldn't work but it does Not say what will. Example, if I play like Phil Harvey then I will Not fail.
I don't believe a Professional Poker player can clearly describe in detail what goes through his head in every situation. At least not enough to Teach someone else how to play exactly like him. This type of confusion is what I want to first take out of trading. I want to be the type of trader who when it's time to pull the trigger I don't hesitate. In my experience, winning in Risk ventures is not about System, Expectation, or Money Management. It's about Emotions, the emotions to stick with the System, satisfied with the Expectation and patience with the Money Management.
A professional poker player Never know if he's going to win the up coming tournament. All he knows is if he employs his system, he should come up on top more times then if he plays with his gut feelings. 10 people playing a game where they win 2 dollars for heads and lose 1 dollar for tails. 8 out of 10 will lose their money on this good game because they don't know optimum bet. And if you told them what the optimum bet is 6 out of 10 would still lose because of greed.
I'm trying to Gage the type of performance I should shoot for. Just a simple Yes or No answer here would do. Have any programmer here managed to create an EA with the follow results or better? If your answer is Yes, can you specify the PF, MD and Average #Trades per year. Thanks Guys.
Average:
2% or Greater Profit Factor.
20% or Less Maximum Draw-down.
200 or More Trades per Year.
not yet but im working on it !!
2% or Greater Profit Factor.
Do you really only want a 2% profit factor? So profit factor would be 1.02 or greater? Or did you mean a profit factor of "2"?
Sorry if I'm confusing, what I'm looking for is 2.0. The best I've been able to do is 1.30 on systems that places decent amount of orders. Of-course I hit up to 10.0 percent but if it only placed 100 orders in 10 years, i just disregard it. My thinking is that alot of people with curve-fitting systems would have achieved 2.0. But if that's hard to hit in back-tests, how much harder is it dealing with real money? I'm not limiting myself to any pre-defined #, I just want to be realistic. Also, tho I didn't specify, I'm not interested in results that trade in one direction for limited time. Like Shorting the EUR/USD this year (I can create a system that don't lose doing that). What I prefer are systems with results over 10 years going in both directions. If you can curve-fit over 10 years and obtain 5% returns, I definitely wanna hear from you :). There could be a patten there.
So why 2%? Well the 2-books (lol) along with on-line resources I've read usually suggest not risking more than 2% of un-allocated capital. In my bj experience your optimum bet coincide with your profit-factor (expectation in bj). I think the equation for pf, expectation and roi are the same but i could be wrong. I can always look it up, thats for sure. I realize no EA can operate like a Black Box for a Long Time. There's a point when the creator may need to Re-View the program. I define that point as when the EA exceeds the Theoretical Maximum Draw-downs.
You need to read this: Minimizing your risk of ruin
Thanks Phillip. Very nice article indeed, I'm adding this one to my Trading Tool Box. It takes me back to the good old days when I was trying to figure out the meaning of the bell shape curve. Questions like 2-Standard Deviation vs 3, which is better. What Risk of Ruin should I accept, 10%, 1% or 1/1000. Imho, if someone is new to Risk appetite they definitely need to cross this bridge. No one can answer the question of which ROR is better for the trader. The trader starting with the 10% ROR may become a millionaire while his counter part starting with 1% ROR could go broke. Talk about bad luck, something which happens to 1 in 100 people just happened to you on your first try.
Fixed fractional position sizing in theory is suppose to guarantee against losing all your money (ROR). No matter how much you lose, you could always break what's remaining down into fraction and bet that. In practical terms, it doesn't work as well because of 2 reasons. 1-table minimum or minimum account size is going to force you to over-bet. And 2-If you could keep your money management- who has time for returns trading 2% of their remaining 10 dollars. Bottom line is these numbers are mostly good for comparative values or to figure out what has a better chance. My advice is if someone don't have allot of money and time to keep adjusting during the downward sparrow, don't even think about playing the game.
Personally, I like the Kelly Criterion. Prefer the ROR of 1% to begin with and will stop playing/trading at 13% or higher. For the good old Bell shape curve, anything withing 3-Standard Deviation -/+ wouldn't work my nerves. 4/sd or more then we have a problem and need to re-evaluate the variables coming up with these numbers.
Attached is an example that meets your stated requirements in the op, but I wouldn't put money into it no-way no-how. Over-optimization is a beautiful thing if you like good looking backtest results.
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
I'm trying to Gage the type of performance I should shoot for. Just a simple Yes or No answer here would do. Have any programmer here managed to create an EA with the follow results or better? If your answer is Yes, can you specify the PF, MD and Average #Trades per year. Thanks Guys.
Average:
2% or Greater Profit Factor.
20% or Less Maximum Draw-down.
200 or More Trades per Year.