Quote from dom993:
That's what ultimately lead me to developing & trading automated systems ... I always lost money discretionary trading, at some point I realized I wasn't confident in my trading plan because I had not backtested it, and forward test on sim was not enough to convince me.
But it took me several years of R&D and over a dozen non-viable trading systems before starting to get some backtesting results not immediately invalidated by forward testing or live trading.
My lastest system works a lot better than any prior one ... I believe the main reasons for that to be:
1) no trade management artifact - as strange as it seems ... but too often, the "natural" outcome of a pattern is interfered by trade management
2) using many patterns at once (currently 24)
3) having a large sample-set for a least a few patterns :
- 2 are ~3000 instances in-sample (6 years)
- 1 is 2000+ instances
- 2 are 1000+ instances
- 6 are 500+ instances
4) having a trade frequency large enough that the law of large numbers can play-out in a reasonable time horizon (the system generates on average 6 trading signals per day)
That's interesting phenomenon, cause now as you are profitable you are confident about your automated system, right?
Would you be as confident trading it manually? Exactly the same rules I mean, just manual execution.

