If candle stick analysis was really that great and predictable...

If candle stick analysis was really that great and predictable then don't you think Algo's would already have closed the spread? I can pretty much guarantee you'll find no statistical significance here, you're sample size is far too small to begin with. For example there would be a small probability you could flip a coin 20 times and get heads 80% of the time, however after 1000's of coin flips you would be left with something closer to 50%. I think you'll find the same applies here and results would eventually flatten out after 1000s of outcomes- the law of large numbers. You'll also find there's a difference between your back testing results and your "real world" results, where in some cases it would be impossible to get filled at the exact entry point you've anticipated.

Not that it proves anything, however lets even see you post up the exact entry and exit points, with volumes to support your percentage per trade.