Example Trading System

Hi,
I`m a little bit lost in my own mass of testresults... I did some tests with correlation of the entryprice against yesterdays close, high/low gave less meaningful results...uuh i don`t use it at the time, i think there was some little edge if entry is not to far in the opposite side of YDclose, not too exciting effects i remember. Give me 2-3 days, i check it again and i `ll try not only the entryprice but the middlepoint of the OR. Some correlation to YD close you find in the GAP rules i mentioned on the little website...

Michael
 
I seem to remember you mentioning that you used TradeStation as your backtesting environment. Is that the case?

As I understand it, TradeStation allows you to test only one market at a time. I think other platforms allow the testing of strategies on entire portfolios. Will look into it.


Quote from mechtrader41:

Hi,
I`m a little bit lost in my own mass of testresults... I did some tests with correlation of the entryprice against yesterdays close, high/low gave less meaningful results...uuh i don`t use it at the time, i think there was some little edge if entry is not to far in the opposite side of YDclose, not too exciting effects i remember. Give me 2-3 days, i check it again and i `ll try not only the entryprice but the middlepoint of the OR. Some correlation to YD close you find in the GAP rules i mentioned on the little website...

Michael
More...
 
Yes, I use good old TS 2000i and IB as Broker/Datafeed and it`s right you can not test/optimize portfolios but that`s not an issue for me. Good with TS is, you find very much discussion and files to use/modify, makes things easier. I´m not very good in programm languages.

Wealthlab could be nice, is able to autotrade with Interactive Brokers and can do portfolios. To autotrade TS you need 3rd party tools like hyperorder, but no problem either.

Michael
 
Quote from mechtrader41:

AM,
i did it, here the nice results about the effects of the position of the midpoint of the OR against yesterday`s close. I added it at the end of the htm file, you find it here:
http://trader.gmxhome.de/breakout/breakouttest.htm
More...

Great. Interesting to see some effects vs the OR's position. My feeling is that this information should only be used to add to bet sizes rather than filter trades out.

To further improve our understanding of the Context, I think spending time on estimating volatilty would be useful. Have you got any ideas here?

In terms of entries, have you found any differences between a simple breakthrough of the OR versus an x minute bar close outside the OR?

In terms of exits, perhaps there's room for some diversification. If one was trading 4 lots, one could: close 1 lot on a profit taking level; 1 lot MOC; 1 lot on a Trailing stop; 1 lot on a timed exit. Might reduce profitability per trade but improve robustness. Theorectically, if the above was useful, there would be some optimising to do between the various options.

Any thoughts?
AM
 
I`m just a little bit short of time, will go in detail later...
The volatility stuff is the ATR Filter, it`s some derivate from the NR4/NR7 stuff T.Crabel first mentioned and is a lot used by Linda Raschke. Volatility is somewhat cyclic and much more predictable than price. The condition of high Vola 3 days ago and following consolidation/low vola is a high probability of a higher vola day again>>>mentioned on my site.

The exit stuff i tried a lot earlier, but found no real edge, not in scaling out nor in pyramiding, but i`ll try again with new ideas...

In general, scaling out works only if you have an entry with a high probability of some profit >>>High %Win, any entry with a lower Prob. is bad for scaling out, cause you cut your profits and have double loss if stopped out. If this makes any sense is depended from the version/filters of BO system and the %win.

If you look at the study with the stopsize, you see that a large stop is serious, on some days we see large retracements, sometimes faking out the other side of OR but ending positive or with relatively low loss. That`s a reason a trailingstop is not working, i just use a width trailingstop very late in the day

An other idea is adding, if you`re right you have more profit, so may be wait for the very high probability trades, wait for a little retracement and add to your position....

But as already mentioned i found no edge in both, but i keep on working....

I`m busy a few days now...

Michael
 
It's great to see you guys taking the orignal system and finding real ways to improve the performance. Keep up the good work!

It's interesting to see that Mechtrader41 has found that Monday's are not good days for breakouts. I remember back in 2002, when we had very volatile markets, that Mondays were the best days for breakouts. If only the markets would stay consistent, it'd make our jobs so much easier!
 
Quote from sidinuk:

It's great to see you guys taking the orignal system and finding real ways to improve the performance. Keep up the good work!

It's interesting to see that Mechtrader41 has found that Monday's are not good days for breakouts. I remember back in 2002, when we had very volatile markets, that Mondays were the best days for breakouts. If only the markets would stay consistent, it'd make our jobs so much easier!
More...

Hey there.

Good to see you're still here! Thanks once again for providing the catalyst for this discussion.

Regarding non-stationary markets, it would be interesting to compile a quarterly analysis of the various factors to see exactly how they change over time.

Your observation that Monday's were good in 2002 strengthens my hunch that all days of the week should be traded. Quarterly seasonal analysis + bet size adjustments = improved performance? How close is that to curve fitting? I guess it depends 'how' you curve fit. On that note, an interesting article from Curtis Faith on the subject.

http://www.tradingblox.com/tradingblox/optimization_paradox.htm

Cheers,
AM
 
Hi sidinuk,
I remember a study from Larry Williams about day of week ranges, historically over 20-25 years or so, most markets had the lowest range on monday, the time from ~ 2000 -2002/3 was a very unique high volatility environment, may be not seen ever in Length/strength and may be we see nothing similar for many years. Under more normal conditions the monday thing seems to have more statistical relevance over time. More to saisonals, i found it better to standby the last 3 days to expiration/thirdFriday.

greetings

Michael
 
Saisonal stuff....there is a difference in moneyflow/activity in dayofweeks in bull or bearmarkets. In something like a bullmarket (or not a ugly bearmarket) you have a high probability for example for a very simple trade. If monday closed negative, buy tuesdays open and exit MOC. With some filters like not overbought osc. you get a % win ~ 75% or so, works not in bearmarkets. Same with endofmonth effect, with some filters buy later than the 29th if close < close[1] and exit partly at 1-2 atr(10), or after 15 days, you get a ~90% win. So the activity on mondays, in more quiet environment and best in bullmarkets the traders come from weekend and are not to fast in beginning the workingweek, everything more or less ok, so they really start working tuesday. Not so in bearmarkets/highVola, they have fear and a lot to think about the ugly positions they held over weekend and made some thougths, and can`t await monday to correct some possible mistake.

So you have to analyse what kind of marketenvironment is at the present, and what means that for the behavior of the marketpartitiants in cases of speed of movements(change length of OR?), day of week, more latedayreversals ? It`s somewhat easy and very complex if you look in the details....

Michael
 
You must log in or register to reply here.
Back
Top