Baluda wrote:This is where we disagree. We are definitely not looking to 'beautify' data.
I'd say that's
precisely what the original future-looking TMA does. It makes the TMA plot "pretty" so it looks prescient when looking at historic charts. I consider it analogous to a zig-zag indicator that marks the exact top and bottom of each swing, but only in hindsight. You cannot buy and sell at the zig-zag's swing maxima/minima. An inexperienced person who looks at the zig-zag might think he's found the Grail but we know that's not true. The zig-zag may be useful for certain historical analysis -- how often does this market swing, what's the average swing size, etc -- but it's useless for actual realtime trading signals.
The TMA actually *IS* useful for realtime trading, but by definition that means it's only using the "past data" part of the triangle -- in realtime the "future data" doesn't exist yet. I honestly don't understand why anybody would want to destroy that useful actual-trading information and plot a "beautified" full future-looking triangular MA. I can't see any benefit to prettifying the actual realtime signal into an unattainable perfect TMA curve. (Can you? I'm honestly puzzled why anybody would do this.) But I can see VERY strong benefits to seeing the ACTUAL TMA (or whatever you call it) values, as you would have seen them in realtime. That allows you to scan back and see how the indicator performed in realtime, giving you an idea how your indicator / system / etc would have performed. NNBob mentions this benefit for daily/weekly/monthly pivots, and a non-prettified "TMA" would gain the same benefit.
Another issue is your wish to backtest, visual or otherwise. It is my strong belief that the importance of (visual) backtesting is highly overrated. Remember the past can never be used to expect/predict the future.
Here I must
vehemently disagree. If the past gives us no prediction ability for the future, then the entire activity of trading is impossible. The only way you can trade (other than techniques like pure arbitrage which do not attempt to predict future price movement) is to say "I've seen this work in the past, and I hope it will continue to work in the future." Whether that is trendlines, pivots, sunspots, chicken entrails, whatever -- you observe a pattern and you expect it to repeat in the future. Otherwise you're just making random guesses. Backtesting is simply a formalized process for observing and measuring past behavior.
Even live "forward testing" is simply backtesting
in realtime. If you forward test it and it works well, you say "OK, it seems to be working, I'll try it with real money now." If the past had no predictive power, the forward test you did
in the past would have no predictive power for live real-money trading.
I have over 10 years of experience building and trading backtested systems. At one point I was trading over US$10MM of client money, using backtested systems I developed, and until the 9/11 attacks changed the whole market psychology, I was one of the top-rated CTAs in the business. Backtesting WORKS, if done properly. You can never predict the future with 100% accuracy, but if you work at it and use appropriate techniques, you can predict the future with enough accuracy to make a consistent profit.
