nonlinear, Glad to see you got a successfull strategy out of the tips I had given you. Mine is optimized on both net gain and on profit factor. However, probably would have been smarter to optimize on max draw down instead. I plan for this week to leave it running, just tell it to not actually place the orders at the exchange. To see if things start to pick back up. I think the past 2 weeks has just been unlucky. These types of draw downs have been seen in the system before. However they showed up much more frequently in 2005 and before than recently. So, the draw downs are not proof the system has broken, only that conditions may be returning to more of a 2005 and before type environment. Which the system was still profitable back then, but with wilder turns.
He was able to do that because, well, he probably knows how to trade. So he took the basic frame work that you mentioned in your other thread and designed a bot around it that would mimic the habits of a good trader (OK, here's a hint for free - a good intra-day trader won't allow a position to move against by over $1,000 multiple times.). JJ
If you include the number 4 in my previous post, plus the number of times I said it in Frost's first Live Bot thread, double plus the times I said it in his original thread, the count is actually more like 20 or 30 times. You think it's sinking in yet? JJ
I'd throw the net gain away as an optimization criteria and make profit factor a primary factor. Then, out of the top strategies, I'd pick the one with the lowest drawdown.
Or you could just have the bot randomly flip a coin before entering a trade. Heads = Long Tails = Short Coin Flip Algorithm
add a class to your bot which calls jimmyjam every trade and asks for his opinion. collect of course.
Frost, here is a little twist around your original strategy that makes the drawdowns lower. Notations: S = strength of the shorter term trend (negative or positive depending on the trend direction) L = strength of the longer term trend (negative or positive depending on the trend direction) T<sub>s</sub> = threshold for the strength of the shorter term trend (fixed positive number) T<sub>L</sub> = threshold for the strength of the longer term trend (fixed positive number) T<sub>e</sub> = exit threshold (fixed positive number) DELTA = S - L (negative or positive depending on the direction of the divergence between the two trends) Original strategy (always either long or short): Long entry: when L > T<sub>L</sub> and S < -T<sub>s</sub> Short entry: when L < -T<sub>L</sub> and S > T<sub>s</sub> Original strategy with a twist (can be either in the market or flat): Long entry: when L > T<sub>L</sub> and S < -T<sub>s</sub> Exit long: when DELTA > T<sub>e</sub> Short entry: when L < -T<sub>L</sub> and S > T<sub>s</sub> Exit short: when DELTA < -T<sub>e</sub> Essentially, the "twist" is to allow the strategy to take profit, instead of waiting to reverse from short to long and vice versa. Hope the pseudo code makes sense. See if this works for you in your back test.