|
PureBytes Links
Trading Reference Links
|
Jonathan Stewart Dempster wrote:
>
>
> BRITISH POUND-Daily 07/01/88 - 07/17/98
>
> Performance Summary: All Trades
>
> Total net profit $ 149712.50 Open position P/L $ 0.00
> Gross profit $ 189600.00 Gross loss $ -39887.50
>
> Total # of trades 251 Percent profitable 83%
> Number winning trades 208 Number losing trades 43
>
> Largest winning trade $ 6987.50 Largest losing trade $ -1662.50
> Average winning trade $ 911.54 Average losing trade $ -927.62
> Ratio avg win/avg loss 0.98 Avg trade(win & loss)$ 596.46
>
> Max consec. winners 23 Max consec. losers 3
> Avg # bars in winners 1 Avg # bars in losers 1
>
> Max intraday drawdown $ -5787.50
> Profit factor 4.75 Max # contracts held 2
> Account size required $ 9787.50 Return on account 1530%
>
I'm certainly no TS expert but, I am concerned by the avg # bars in
winners/losers = 1. This means most times if not always the trade is
exited on the entry day. What are the type of entry/exit rules - and do
they need intraday data to function or TEST? With end of day data TS
obviously cannot know the sequence of prices between open and close.
If your rule said buy at open and sell at close then your tests are
accurate. (However, do note that Bridge endofday data includes the
previous night sesssion as the open, as far as i know, as a Bridge
subscriber - at least the daily downloaded data includes Globex.)
If you have a limit order to enter with an exit at the close that would
be ok too, to test with endofday data. But where you can get into
trouble is: if you start combining entries, stops, and profit targets
which really require knowing how the prices got from open to close, to
know how the trade would have gone.
When TS is not supplied with sufficient info, it attempts to make some
assumptions about the intrabar movement. Obviously this is just a
guess. If your trade lasts a number of bars, this probably actually
improves the approximation that is your testing. However, if you test a
situation which really requires intraday data, but only give TS endofday
data, the results can be unreliable; in the extreme case, they may be
meaningless.
In reality TS should probably warn when its approximations are more
than a certain % of the trades entire move. But currrently it doesnt'.
It makes its best guess with the data supplied but if you test a day
trading system with endofday data, the test probably does not mean
much.
You will have to think about your trading rules to determine whether
you need intraday data or not. Anytime you find your system exiting on
the same bar as entry more than once in a while, you need to consider
whether you need data of the next lower time span.
Conrad Bowers
|