|
PureBytes Links
Trading Reference Links
|
> Does anyone have any experience with amy of Lee Gettes systems?
I experimented with his TT30 system "with Market Mapping." The basic
TT30 did very well during TS4 backtesting of the same period that
Gettess used, but totally fell apart out of sample. Even during his
in-sample periods, the system averaged only about $50-150 per trade,
and I don't think he included any slippage in his calculations.
Conclusion: totally curve-fit.
Furthermore, a friend tried the TT30 system on TS2k. Again, it
looked great during backtesting during Gettess' test period using TS4-
style testing. Then my friend turned on TS2k's "tick level
resolution" feature, where it **accurately** simulates backtesting
instead of doing the charitable "guesses" TS usually does. Result:
during Gettess' training period, the system would have actually lost
money in realtime trading. Conclusion: either Gettess never traded
this thing, or he KNEW it was just a fraud generated by TS's (or,
rather, System Writer's) faulty backtest algorithm.
Then I looked at the "Market Mapping" variant. This was a breakout
method that used "pattern recognition" to tailor its behavior
depending on whether the day was an inside/outside/up/down day,
whether the open was above/below/gapup/gapdown from yesterday's
close, etc. Gettess provided a table of ONE HUNDRED AND TWENTY-EIGHT
different parameters for each of the different conditions.
Conclusion: massively curve-fit to the Nth degree.
If this is typical of his work, I don't think I'll waste any more
time on Gettess systems.
Gary
|