|
PureBytes Links
Trading Reference Links
|
Hello David,
When you're dealing with this type of intraday trading frequency, you
may want to do another comparison: Offset your five minute bars. In
other words, make the close of your first 5 minute bar at the close of
two minutes, and then close your bars every 5 minutes after that.
There are other ways to do this, but what comes to mind is to force
your 5 minute system to run on 1 minute bars and write additional
code to keep track of 5 minute intervals on one
minute charts. Adjust your indicators accordingly.
But as a control, I would test your system on the close of every 5th
one minute bar to see if there is any distortion. Then, implement the
offset. Try another offset.
You may be surprised by what you find. I think this is part of the reason
why there is a difference between cash and futures: simply the offset
of the time intervals since the exchanges' time stamps are not
perfectly synchronized.
You may also find this performance distortion when applying the same
system to the same market, but using two different datafeeds.
Saturday, April 05, 2003, 11:14:57 PM, you wrote:
DC> I realize this has been discussed here before, but I have been realizing
DC> lately that testing systems on the cash SP is kind of a waste of time. In
DC> the most jarring example of this phenomenon that I have yet encountered, I
DC> was trying of late to build a system based on the NYSE Tick index. I was
DC> using the cash index because my futures data was not properly aligned with
DC> my $Tick data in terms of time stamps, but I figured, if I can get good
DC> results, the difference in the futures shouldn't be all that bad.
DC> I started with the simple premise of buying whenever the NYSE $Tick closed
DC> above a certain level and selling whenever it closed below a certain level.
DC> An optimization run and subsequent OOS runs suggested that 200/-200 was a
DC> good place to start. Much to my surprise just this basic system returned
DC> pretty stellar results on SPX. It traded a lot, but still had a respectable
DC> avg. trade size of $250(+/-) or so. And the equity curve was quite linear.
DC> I proceeded to refine it by adding stops and trend filters and so forth
DC> until I had it earning $2.5 million over a 4 year period and looking as
DC> though I drew the equity curve with a ruler. The DD was so small it wasn't
DC> even worth mentioning.
DC> Well this was just too damn good to be true.
DC> So I did a run through on futures. Here are the results, cash vs. futures,
DC> 5 minute charts, 3 month period between Dec 17/02 and March 16/03:
DC> CASH SPX:
DC> Total Net Profit $122,467.50 Open position P/L$995.00
DC> Gross Profit $192,030.00 Gross Loss($69,562.50)
DC> Total # of trades428 Percent profitable54.44%
DC> Number winning trades233 Number losing trades195
DC> Largest winning trade$8,432.50 Largest losing trade($2,512.50)
DC> Average winning trade$824.16 Average losing trade($356.73)
DC> Ratio avg win/avg loss2.31 Avg trade (win & loss)$286.14
DC> Max consec. Winners7 Max consec. losers7
DC> Avg # bars in winners14 Avg # bars in losers7
DC> Max intraday drawdown($4,937.50)
DC> Profit Factor2.76 Max # contracts held1
DC> Account size required$4,937.50 Return on account2480.35%
DC> S&P FUTURES
DC> Performance Summary: All Trades
DC> Total Net Profit $2,187.50 Open position P/L$1,312.50
DC> Gross Profit $138,000.00 Gross Loss($135,812.50)
DC> Total # of trades428 Percent profitable36.45%
DC> Number winning trades156 Number losing trades272
DC> Largest winning trade$8,000.00 Largest losing trade($3,062.50)
DC> Average winning trade$884.62 Average losing trade($499.31)
DC> Ratio avg win/avg loss1.77 Avg trade (win & loss)$5.11
DC> Max consec. Winners7 Max consec. losers14
DC> Avg # bars in winners18 Avg # bars in losers7
DC> Max intraday drawdown($15,187.50)
DC> Profit Factor1.02 Max # contracts held1
DC> Account size required$15,187.50 Return on account14.40%
DC> I realize this system trades a whole hell of a lot, and that these
DC> differences aren't as pronounced on a system that trades less frequently.
DC> But I just wanted to illustrate the degree to which SPX can fool you.
DC> David
DC> ---
|