Flatline....

[Glenn]

ok, the patient is dead....


attached is an equity curve of a home-grown trading strat tested on data as far back as 1990.

the inputs (all two of them) are dynamic, and so should therefore respond to current prevailing market conditions..

it is counter-trend..


performance has been rather stellar (50pts per trade, 80% win rate), until we hit 2004, whereupon it has basically flatlined.

this...
is...
annoying....

looks like 60,000 points is providing overhead resistance..


anyway, the point of my posting was..

has anyone come up with a decent mechanical strat that has performed well on the Dow during 2004? all my trend-following ones have got a similar pattern, as have my counter trend ones.


it cant be due to current market volatility can it, as the inputs are based on recent volatility levels......?

Some questions FC.

1. You backtested it from 1990 until when ?
2. Did you forward-test from the end of the backtest period ? If so, for how long and what was the outcome of the forward test ?
3. Have you used any Optimisation in obtaining your operational parameters?
4. What is the calculated Expectancy of the system ?
5. Have you considered that the Dow has been in a pretty narrow range for 2004 when compared to previous years ? The performance of your model may be mirroring this fact and may therefore change for the better along with the market if its behaviour reverts to norm.
6."it cant be due to current market volatility can it,......" Depends upon what you mean by current. ? Volatility is measured over a period of time, so it depends which period you select.


More questions to follow depending upon your response.

Glenn

 

[TheBramble]

the majority of traders trade breakouts, and go long on the break of a high. i on the other hand usually (circumstances permitting) go short..

Which is fine and dandy if you're able to confirm it's the end of a passive accumulation phase leading into a joe public buying climax. Normally the volume will give a clue. As will the previously 'placid' movement of the stock. But just shorting on every break of a High would I imagine be a 50/50 proposition at best.

Is this the system for which you're showing us the equity curve? A reverse BO regardless of volume and recent activity?

 

[FetteredChinos]

Some questions FC.

1. You backtested it from 1990 until when ?
2. Did you forward-test from the end of the backtest period ? If so, for how long and what was the outcome of the forward test ?
3. Have you used any Optimisation in obtaining your operational parameters?
4. What is the calculated Expectancy of the system ?
5. Have you considered that the Dow has been in a pretty narrow range for 2004 when compared to previous years ? The performance of your model may be mirroring this fact and may therefore change for the better along with the market if its behaviour reverts to norm.
6."it cant be due to current market volatility can it,......" Depends upon what you mean by current. ? Volatility is measured over a period of time, so it depends which period you select.


More questions to follow depending upon your response.

Glenn

1) 1990 - early 2003.. been trading a derivative of it since later 2003..
2) fwd test has been pretty much the flatline you see on the graph. hence the need to derivatives of it.....
3) no optimisation at all. literally entry and exit are based on 2 volatility factors (1 weekly, the other daily). no dodgy fudge-factors in there at all. and indeed, running the test for any % of volatility returns a similar shaped equity curve...
4) expectancy? well the Profit factor is 2.7 . not sure of the calc for expectancy (i suppose i could dig it up on here somewhere)
5) agreed. but since this is counter-trend and volalitly based, it shouldnt matter what the range is. having said that, it did perform best when the monthly volatility was high in 1998-2002.. perhaps that might explain things
6) the volatility is based on the past 4/5 weeks data.. thats pretty current?

FC

 

[FetteredChinos]

Which is fine and dandy if you're able to confirm it's the end of a passive accumulation phase leading into a joe public buying climax. Normally the volume will give a clue. As will the previously 'placid' movement of the stock. But just shorting on every break of a High would I imagine be a 50/50 proposition at best.

Is this the system for which you're showing us the equity curve? A reverse BO regardless of volume and recent activity?

no entirely a BO..but no volume at all.

forecast highs/lows based on recent volatility (past 4/5 weeks only)...

with regard to fading every single breakout :- closing trades intraday has a negative expectancy on pretty much every instrument i have tested on, including the oboe, but those breakouts have a tendency to reverse over the course of the next couple of days.. usually a failed re-test of the high/low..

FC

 

[Glenn]

1) 1990 - early 2003.. been trading a derivative of it since later 2003..
2) fwd test has been pretty much the flatline you see on the graph. hence the need to derivatives of it.....
3) no optimisation at all. literally entry and exit are based on 2 volatility factors (1 weekly, the other daily). no dodgy fudge-factors in there at all. and indeed, running the test for any % of volatility returns a similar shaped equity curve...
4) expectancy? well the Profit factor is 2.7 . not sure of the calc for expectancy (i suppose i could dig it up on here somewhere)
5) agreed. but since this is counter-trend and volalitly based, it shouldnt matter what the range is. having said that, it did perform best when the monthly volatility was high in 1998-2002.. perhaps that might explain things
6) the volatility is based on the past 4/5 weeks data.. thats pretty current?

FC

Stating the obvious I know, but your forward test has highlighted that the system does not work in all conditions.
You suggest that tweaking your volty variables makes no difference. Therefore there is a variable missing (or more than one). Something longer term needed it seems.

I.r.o. 'current' volty I was thinking more in terms of 6-12 months rather than 4-5 weeks. You will not pick up the year-long consolidation of the Dow with such a short timescale.

Expectancy must be greater than zero, otherwise the system will fail you no matter what the win rate !!!

Glenn