Hi, for couple of days I've been playing around with simple EA that will enter both long and short using RSI trend detection criteria and entry depending on %Williams'. Entry setup is based on: Trend detection: T1. RSI ranging between 10 - 50 downtrend T2. RSI ranging between 50 - 90 uptrend Entry point: E1. %R peak > -20 when T1 - SELL E2. %R peak < -80 when T2 - BUY Self explanatory picture is attached Attachment 1797639 TP and SL are fixed SL = 250*Point TP = 350*Point I use mainly 15M chart and its only ~150-200 transactions in a year on a pair - not much but using all possible pairs can do a lot. I even tried to look for trading hours but not much out of it Attachment 1797662 It has been proven in many back-tests that it can be profitable in majors Attachment 1797655 but there is one BUT: Initially i tried with default 14 period RSI and %W. Then i realized i got better results when 13 period is used, then i realized that best IndPeriod varies over time. Like this year 13 is the best but last year or quarter it was 16. It was supposed to be simple EA but then i decided to play also with: break even when in profit, lot optimization, risk management, trading hours and it became quite complex robot Here is where i got stuck and i think i could use some help. 1. Do you think we could somewhat optimize robot "on-line" to align IndPeriod for future? Like close loop after 3 months optimization and readjusting for following month? 2. Do you think of some criteria how to filter bad setup and increase profit? 3. Any other comments or inputs? EA is attached with good settings for this year USD/JPY, go play around with it if you want and let me know your thoughts Cheers! Attached Thumbnails Attached Images (click to enlarge) Attached Images Attached Image Attached Files Attached Files RSI_%R.pdf 187 KB RSI_%R_rev.8.ex4 17 KB Source.