- ago
Hi,

I have an issue with my rule based strategy which uses a position sizing of 20% (so 5 trades simultaneously):

When I run it with a setting of last 10 years, it shows me 3 shares I should hold currently:
F, DOW, IBM
and one signal for buying a fourth:
ORCL

However, when I run it for the last 20 years, I get no signal and according to the positions I should hold the following two shares at the moment:
C, DOW

Why doesn't the ORCL signal appear?

I also tried it out with several date ranges:
Until 22.8.: Signals from 20.8. show up
Until 23.8.: Signals from 23.8. show up
Until 24.8.: No Signals show up

So, probably on the 22.8. it shows me the signals from 20.8., because it was the last trading day, right?
But on the 23.8. it shows no signals at all, because the last trading day did not result in signals. Do I understand it correclty?

However, the problem with the 10 years - 20 years differences bothers me, because I cannot understand why the strategy works differently depending on the backtesting period.

Thanks for your help!
Werner
0
669
9 Replies

Reply

Bookmark

Sort
Cone8
 ( 28.25% )
- ago
#1
QUOTE:
why the strategy works differently depending on the backtesting period.

We can help probably only by reviewing the code/rules. There are certainly lots of ways that a strategy can made to be give different results - too many to start guessing. And, if it's a bug, we'll need to reproduce it to fix it.

Let's see some code, the symbols in your DataSet, and Strategy Settings.
0
- ago
#2
Could be:
1. Not enough funds to take all trades due to the 100% equity sizing (20% * 5).
2. Using an unstable indicator (seed period).
etc.

Please follow @Cone's advice.
0
- ago
#3
But I am using an additional 10% margin.
0
- ago
#4
Buy at market open
Accumdist less than SMA(Close/18) 1 bar ago
SMA(Close/3) less than SMA(Close/13)
Accumdist less than Accumdist 3 bars ago
Transaction Weight RSI(Close/19) lowest value
RSI(Close/100) less than 100000000,00
Sell after 4 bars

0
- ago
#5
Dataset = S&P100
Starting Cap = 100000
Position Size = 20%
Margin Factor = 1,10
0
- ago
#6
Are you using Wealth-Data's S&P 100 DataSet? Its composition isn't fixed, with constituents entering and exiting the index at dates when they were inducted or expelled. The changing index composition could also serve as an explanation for what you're experiencing.
0
- ago
#7
Yes, Wealth-Datas S&P100
0
Cone8
 ( 28.25% )
- ago
#8
Can't explain it or help without the code, but margin factor 1.1 isn't necessarily sufficient to reduce NSF positions to zero. How many NSF Positions are reported in the Metrics Report > Positions?
0
- ago
#9
My point is that since the Wealth-Data portfolio is dynamic, the variable index composition could be affected by this. For example, wernerhh's Transaction weight's lowest value could change depending on the period as stocks get included to or expelled from the index.
0

Reply

Bookmark

Sort