GSB Forums

Truisms and other filters

admin - 18-5-2017 at 07:27 PM

Need you ideas for Truism's
This lead to the dramatic improvement of all metrics on 20.4
A Truism's is something that is generally true, but not always true.
IE eat healthy and you live longer.
In trading it could be only trade with the trend.
Don't enter in the last 30 minutes of the day on a day trading system.

Carl - 19-5-2017 at 12:09 AM

something like these?

truerange > avgtruerange(5) (increasing volatility)
close > average(close,200) (trend)
closeD > openD (upday)
best time window for entering the market
max number of entries in a day


gmoney - 19-5-2017 at 08:31 AM

One to add would be to only daytrade when the VIX > 18 (long term median). That can be too restrictive, so using VIX > 15, 12 (or some similar measure).

Bar Patterns

andrew.gibbs777 - 19-5-2017 at 03:33 PM

Gaps

Open of day below Low or Above High of previous daily bar (by a certain amount from 0 to XX (Offset)):

Traditional Gaps

OpenD(0) OpenD(0)>HighD(1) +Average(Range of Daily,XX)*XX (Generally a signal to go short)

Overnight Gaps

OpenD(0)<>CloseD(1) +/-Offset

Ranges (Intraday or Daily)

Previous Bar is a Narrow Range or Doji

Example

Range of Daily < Average(Range of Daily, X) *0.XX

Previous Bar is a Large Range bar

Example

Range of Daily>Average(Range of Daily, X)*1.XX

Directional (Intraday or Daily)

Close<(>)Close[1] +/-Offset
Close < Low[1] +/-Offset
Close > High[1] +/-Offset
High(Low) > High(Low)[1] +/-Offset
High(Low) < High(Low)[1] +/-Offset
Close > (<)Open +/-Offset
Highest(OHL or C,Length)
Lowest(OHL or C,Length)
Indicators do a good job when analysing direction of more than just the previous bar.
ADX(Length)> (<) XX
Consecutive Bars. Example: Close < Close[1] and close[1] < close[2] and close[2] < close[3]


Seasonal (At Risk of Over Optimisation...WFO will highlight if the case)

Trading Days in Month 1-23

Week of Month 1-5 (5th week will always be a short week of up to 3 days)

Day of Week (Mondays stand out as being different by way of following a weekend when compared to other days of week)

Economic News Dates (Non Farm payroll)

Options Expiry Week (or day)

Month

Public Holidays

rws - 29-6-2017 at 10:09 AM

good ideas!


Quote: Originally posted by andrew.gibbs777  
Gaps

Open of day below Low or Above High of previous daily bar (by a certain amount from 0 to XX (Offset)):

Traditional Gaps

OpenD(0) OpenD(0)>HighD(1) +Average(Range of Daily,XX)*XX (Generally a signal to go short)

Overnight Gaps

OpenD(0)<>CloseD(1) +/-Offset

Ranges (Intraday or Daily)

Previous Bar is a Narrow Range or Doji

Example

Range of Daily < Average(Range of Daily, X) *0.XX

Previous Bar is a Large Range bar

Example

Range of Daily>Average(Range of Daily, X)*1.XX

Directional (Intraday or Daily)

Close<(>)Close[1] +/-Offset
Close < Low[1] +/-Offset
Close > High[1] +/-Offset
High(Low) > High(Low)[1] +/-Offset
High(Low) < High(Low)[1] +/-Offset
Close > (<)Open +/-Offset
Highest(OHL or C,Length)
Lowest(OHL or C,Length)
Indicators do a good job when analysing direction of more than just the previous bar.
ADX(Length)> (<) XX
Consecutive Bars. Example: Close < Close[1] and close[1] < close[2] and close[2] < close[3]


Seasonal (At Risk of Over Optimisation...WFO will highlight if the case)

Trading Days in Month 1-23

Week of Month 1-5 (5th week will always be a short week of up to 3 days)

Day of Week (Mondays stand out as being different by way of following a weekend when compared to other days of week)

Economic News Dates (Non Farm payroll)

Options Expiry Week (or day)

Month

Public Holidays

boosted - 15-7-2017 at 02:35 PM

I would like to offer adding the exact opposite of the existing "truism" filter CloseLessPrevCloseD. That "truism" could be used for back-tests that involve only Shorts. Since you are not back-testing for any Longs the opposite of the above "truism" might yield good results.

I have no data to back this idea up, but it sounds like a logical approach if you are just looking for Shorts in a back-test.

admin - 16-7-2017 at 08:09 PM

Quote: Originally posted by boosted  
I would like to offer adding the exact opposite of the existing "truism" filter CloseLessPrevCloseD. That "truism" could be used for back-tests that involve only Shorts. Since you are not back-testing for any Longs the opposite of the above "truism" might yield good results.

I have no data to back this idea up, but it sounds like a logical approach if you are just looking for Shorts in a back-test.


The opposite Truism is done for longs and shorts. If you switch it around results on indices are very very bad.

Truism enhancement is on the to do list, but too many other things ahead of it. Lots of little things, stock support (to be finished), time of day, entries max today,intraday dd, wf multi-pass and wf individual parameters in isolation, removal of duplicate algebra when using indic1*w1*indicator2*w2 are my short term priorities

boothy - 2-6-2018 at 07:33 PM

Market Internals

What about adding market internals for secondary filters? I have heard this can work well on index futures to reduce drawdowns etc. Have not tested myself though.

eg. UVOL, DVOL, TICK, VIX, Stocks above/below 50sma etc.

admin - 3-6-2018 at 03:09 PM

Quote: Originally posted by boothy  
Market Internals

What about adding market internals for secondary filters? I have heard this can work well on index futures to reduce drawdowns etc. Have not tested myself though.

eg. UVOL, DVOL, TICK, VIX, Stocks above/below 50sma etc.

You can do this now but as primary filters. See under documentation, custom indicators / functions
http://trademaid.info/forum/viewthread.php?tid=17#pid1731