How do you know when successfully backtested strategy will succeed in real trading?

As a novice, I would think using statistical tools such as Kolmogorov–Smirnov test to measure the returns between the expected results from back or walkforward test against the returns from real/paper trading. If the resulting distributions does not match, then something is probably not right.

At least this is the method I uses. Would love to hear from the experts here if I am being naive

 

Just my own opinion - backtesting can be a fine screening tool to identify potentially worthwhile strategies. But over a couple decades of doing this I have seldom seen backtested results correlate sufficiently to live trading results (for my own satisfaction at least). In trading you are taking calculated risk based upon historical information. You can identify better than average risk/reward opportunities but you cannot predict future market behavior. YMMV, just my own 2 cents.

 

IMHO (after a few brewskis):
Backtesting is similar to farm equipment. Without greasing and routine maintenance, you will surely end up broken down in the field! Even with, it there is risk.
AKA, once the "bugs are worked out" only takes you so far! You must insure you address changing markets.
Not doing Backtesting, makes sense only if it is NOT your money! (AKA-for lawyers and similar leaches)

 

There are also backtesting software that would actually allow you to forward test with various possible scenarios. These would be really helpful but one thing to be cautious about is that you REALLY have to make sure the forward testing scenarios really do reflect the market irl otherwise the results won't be very reliable.

I have said it before, one thing that backtesting or forward testing absolutely does NOT account for is execution quality. Depending on how you trade, that can be VERY important but that unfortunately is something that you would only be able to find out when you trade live.

 

A well conducted back testing strategy gives the confidence to trader or analysts to move forward and assures them that the strategy is fundamentally sound and is likely to give profits when implemented in real trading. A well established back test which gives sub optimal results urges the traders either to change or reject the strategy.

 

A lot of good posts on this thread, so this thought stands on others' shoulders:
On any given day, a strategy that has worked in the past still faces an uncertain future -- whether that past performance is in live trading or empirical back-testing, the strategy's ability to maintain "success" going forward depends on its performance going forward, and must always be treated as qualified success, "subject to check" and all that. Whether from backtesting or from live-trades, prior success is "indicative but not dispositive" w.r.t. success going forward.

 

This is a good way to go about it.

 

IMO: It is wise to think about your expectations and insure any confidence you build in your strategy or backtest is appropriate. If your strategy is developed/refined by backtesting, you may be subjected to getting what you asked for instead of what you desired! A trading strategy/plan should be one that exploits an edge. If you do not understand your edge and how your strategy is exploiting it, you may be "chasing your tail" (like betting on the dude that tossed a coin and hit head 80 times in a row! --)

A well designed backtest will allow one to observe how the strategy would have performed in the past. This does not, however infer future performance will be the same, as things change, and it is possible your strategy is not as "robust" as you may think.

 

You can never know that because the drivers of profits can disappear anytime.

 

I understand points 1 and 2. I fail to understand point 3. Why should longer time-frame work better than shorter time-frame in backtesting? Shouldn't shorter time-frame work better since you have more data for testing when you go into intra-day?