rws
Member
Posts: 114
Registered: 12-6-2017
Member Is Offline
Mood: No Mood
|
|
Flat period in ES between 2002-2007
I have tested many different settings and always seem to find
a flat period between 2002 and 2007.
If I show X axis with time I see this happening for thousands of
sytems. I tested with 40/60% and 80/10/10 and in both cases
I see very flat 2002-2007 in sample performance while SP500
went almost up 2X
It is good that GSB does well OOS but it is not unimpossible that a period like 2002-2007 would happen again OOS. It is good to see a system working
in all kinds of markets even in sample.
Did someone found a setting that did perfom well in sample 2002-2007?
|
|
|
rws
Member
Posts: 114
Registered: 12-6-2017
Member Is Offline
Mood: No Mood
|
|
I also tested with 15 minutes ES data and in that case the flat period in sample tends to be from 2003 to 2007. Also in this case for the majority
of the found systems (>95%).
I wonder if the pearson definition is based on #trades or time X-ax.
A high pearson is of less value with long flat periods I would think.
I also have the impression that the longer and flatter the in sample
low activity, the more chance I also see this OOS.
I wonder if somebody found settings that don't have these long periods
of inactivity at least in sample.
I still see great OOS after 4 years of inactivity IS but you would not like
to have that kind of inactivity OOS.
I hope there will be fitness criteria that allow inactivity and custom formulas.
|
|
|
andrew.gibbs777
Junior Member
Posts: 17
Registered: 11-5-2017
Member Is Offline
Mood: No Mood
|
|
Try using IDX or SPX as secondary data...
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
This characteristic is much less visible in TS and GSB when you use x axis on trade number, not bars. It doesnt change the reality.
Its not such a bad thing, as the volatility is so low in the period, you are likely to loose money if the system is active during this period.
Better to not trade in bad periods, than loose money. Dont think Ive every seen an ES system in the 2005 2006 period that does well.
Best case it makes a little money.
|
|
|
rws
Member
Posts: 114
Registered: 12-6-2017
Member Is Offline
Mood: No Mood
|
|
When choosing geneticAlgorithm for secondary filter indicator
the long flat periods are gone.
This was very obvious when testing 27.8 which only allow geneticAlgorithm
Why is genetic algorithm adviced for things like forex and previous
day for ES or stock indices?
Quote: Originally posted by admin  | This characteristic is much less visible in TS and GSB when you use x axis on trade number, not bars. It doesnt change the reality.
Its not such a bad thing, as the volatility is so low in the period, you are likely to loose money if the system is active during this period.
Better to not trade in bad periods, than loose money. Dont think Ive every seen an ES system in the 2005 2006 period that does well.
Best case it makes a little money. |
|
|
|
rws
Member
Posts: 114
Registered: 12-6-2017
Member Is Offline
Mood: No Mood
|
|
Oke I did some testing myself.
It appears that geneticAlgorithm needs much more time to find systems
than previous day on ES.
| Quote: Originally posted by rws | When choosing geneticAlgorithm for secondary filter indicator
the long flat periods are gone.
This was very obvious when testing 27.8 which only allow geneticAlgorithm
Why is genetic algorithm adviced for things like forex and previous
day for ES or stock indices?
| Quote: Originally posted by admin | This characteristic is much less visible in TS and GSB when you use x axis on trade number, not bars. It doesnt change the reality.
Its not such a bad thing, as the volatility is so low in the period, you are likely to loose money if the system is active during this period.
Better to not trade in bad periods, than loose money. Dont think Ive every seen an ES system in the 2005 2006 period that does well.
Best case it makes a little money. |
|
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
correct because on ES, the closeMinusClosed filter is so effective. Our goal is to have genetic AND closeMinusCloseD filter as secondary filters +
some others. Its discussed else where on the forum here and at other areas too.
http://www.trademaid.info/forum/viewthread.php?tid=14
|
|
|
![[*]](images/davis/default_icon.gif){kind=icon}
