Zachary Standley 

Is back-testing really necessary

Backtest for quantitative trading. Forward test for discretionary trading.

I haven't backtested the strategy I've been using for almost 2 years now. It's outperforming my algo strats by a longs shot. Only you will know if you need to backtest. In general, most strategies should be backtested. I would presume most traders curve fit their backtests until the results are to their liking. My backtests consist of running the strategy through 20 instruments and then picking the top 8 to work with.

For the average algo/backtested trader i would suggest the majority of systems are nothing much beyond Discretionary trading systems converted to maths , clearly this isnt an easily accomplished task and is where most fail . The average discretionary trader really is a rules based trader in the main and 'if" they could test and quantify their trades mathematically it becomes exponentially easier to identify areas that are letting you down . Backtesting done on robust contexted mathematical systems will 100% improve your trading faster than any other process . AND if it doesnt i suggest your conversion of thoughts into maths is where you are failing , not the actual backtest process . Logical progressive code writing is not easy and i think many make the mistake of giving up quickly after getting no tangible results in the short term . I can attest that after taking the algo route that solid and regularly reliable results took years not months . Even a failed backtest that most would bin immediately can be extremely revealing . Just analyzing equity curve against market price action can reveal weak points in literally seconds . what type of market conditions are the Drawdowns occurring in and is there a solid correlation between the 2 ? if there is you have an area to isolate and improve .

If you go down this road be patient , work hard on converting all the rules of 'discretionary " systems into binary maths decision trees . Some areas of the flow chart of questions you ask price and time need more than 2 outputs with further questioning to get that binary result . Contrary to a lot of popular opinion robust algos will be complex . too many think simple is best and dont challenge their intellect enough to find the answers to complex decision trees . Keep the complex as simple as you can sure but not at the expense of robust binary decisions . There are no shortcuts going down this path . Its a challenging task to turn thoughts into maths and developing the skillset/tools to achieve this but its the shortest path to building robust binary systems /// if , or , then , else , once , while , do , for are all aspects that can and need to be included in robust systems . CONTEXT adds substance .. Good luck to all