That is true indeed. For example I have coded a relative volume VWAP. It analyzes the intraday volume over the last year and then calculates a volume-weighted average price by using that forecasted volume distribution rather than real volume. And I have not posted it, because you need to wait a little after applying it to a chart. However, it is only slow, while the historical data is being treated, in real-time those backward calculations are not necessary, as the results are sitting in an array and do not need to be recalculated with every incoming tick.
The NinjaTrader forum VWAP however calculates back to the beginning of the session with every incoming tick, so this will definitely kill your CPU.
The approach is like running a backtest prior to trading, The problem is that you should have a training and a forward testing period for any backtest, and only trade, if the forecast parameters calculated by using the training period have well performed during the forward testing period. I do not think that the indicator can do that, so it will not be a substitute for proper backtesting.
Last edited by Fat Tails; June 22nd, 2011 at 09:10 AM.
It seems like it would make sense to find these patterns while NOT trading. Then later not recalculate that pattern for each tic. It seems like once you find some averaged pattern that yields a profit, you would just code that pattern and have your script point that out.
Most of the calculations are not necessary when loading historical data. For this reason, I do not run these calculations when loading the data, they are only run in real-time. Therefore, the indicator loads quickly. In real-time, the code only runs on the bar close (CalculateOnBarClose=true). If bars are closing every second, the indicator may have issues, but for most traders this should not be an issue.
I would also agree with your point that PROPER backtesting would require some forward testing as well. However, I do believe the indicator is performing a simple backtest in the sense that it identifies the current market pattern, looks at historical data to find what the market has done after that pattern occurred historically, then provides statistics similar to a backtest.
I know a lot of ya'll believe in back testing, but it just don't hold water with me. I have seen a gazillion strategies that all back tested well but failed in real time trading.
What I do put some stock in is gathering empirical data and drilling down to find commonality. For a year I kept an excel spreadsheet on every trade I could categorize. At the end of the year I ran the stats and found that certain formations did indeed have a better percentage of success than others. I still trade off of what i learned from that data. The percentages will fluctuate depending on the mood of the market, but the stats have held up.
Granted, it is quicker and easier to just run a back test . But then again I have heard a lot of guys wives and girlfriends complaining that being quick is not a good thang...
I'm just a simple man trading a simple plan.
My daddy always said, "Every day above ground is a good day!"
I know this is an old thread, i just found it while searched more information about this indicator. I guess many of the questions here could been answered if there was a trial for this indicator. I wonder why some of their products have trials but this one not...