qldfrog weekly Skate inspired system

[StockyGuy]

Much respect to the ? for tellin' it like it is over long time frame - whether good, bad or ugly. Trading journal type threads often end abruptly once some real drawdown is experienced.

 

[qldfrog]

Much respect to the ? for tellin' it like it is over long time frame - whether, good or ugly. Trading journal type threads often end abruptly once some real drawdown is experienced.

With weekly systems, it can take a while for protection to trigger, GTFO to start sell all etc.
Xnt fell 2pc last week
Standard small cap ETFs fell 4pc, and my systems went "just" double.not a surprise. Drawdowns do happen, these are quite aggressive systems and a dd of 20pc is not unexpected.
Obviously, I'd prefer this not to happen, but this is what it is.
I expect to go into losses on weeks like this, the losses earlier this year were an issue, i fixed that,
I just hope this will end soon or that my systems will go cash if this carries on.they should.

 

[qldfrog]

a so so week..a bit disappointing to be honest but spring reloaded and system back in:
the daily:
the intra day volatility experiment: 1d trade only +0.5% or 100$ does not hold on weekend to remove the general trend effect
my normal dailies:
MOA: back fully invested -$0.6k or -1.2% bad
DL guppy :back fully invested flat -0.1% had to pay the brokerage fees
volatility ASX +$0.6k or +1.25%
volatility US +$0.65k or +4.2%

weekly:
QF sectors: flat+0.04%
QFDuc: -$0.8k or -1%
overall systems: 0 flat mehhhh

the discretionary gained $3k which was quite niceand lifted the spirit for the weekend

 

[qldfrog]

another week...And a smiling frog
System ok but ...

xnt up 0.58% this week
my dailies:
intra day volatility experiment: +0.57% or $90 paying the cappuccinos
MOA: good +$1.1k or +2.18%, fully invested
DLGuppy disappointing: -$0.3k or -0.68%, fully invested
volatility asx: +$0.7k or +1.56%, fully invested
volatility US: +$0.6k or +3.55%, fully invested

weekly:
QFsec mehhh +$0.4k or +0.53%, fully invested
QFDuc bad -$0.8k or -1.06% ,86% invested

so basically back to fully invested
overall for the week:+1.7k or +0.55% so yes, just matching the xnt;
the discretionary did a killing(>10k) with gold and silver plus a few other play so going toward retrieving some of these abysmal losses

 

[qldfrog]

last week up, this week down;
I am back in the scheme: winning around lunchtime and down 1% from that at the end of the session; all that on flat or rising XAO during that time..
Why?? Unsure ...same code, same specs mostly,
As a reference , xnt managed +0.73%.

This week the dailies were NOT GOOD and the weekly system good
intra day volatility experiment:+$180 or +1.2%
MOA -$1.8k or -3.5%
DLGuppy -$2.4k or -5.8%
volatility ASX +0.7k or +1.5%
volatility US+0.7k or +4.06%

weekly
QFSec +$0.6k or 0.77%..as last week matching xnt : no more not less
QFDuc good +$2.3k or 3.1%

overall:+$300 for the week, fully invested in all
a week to forget with an overall loss [System plus discretionary] of $2.3k;
Of interest:
Lost $6.5k just today..mostly due to a red only night on the NYSE.
A first for me where every single owned share/etf went down out of my 13 different ones..
volatility, PM, green economy, Russia,China,Blockchain, bear ETF rural..
ahhh well

 

[qldfrog]

a warm welcome back to Mr @Skate who just presented a a few weekly system backtests;
sadly I have no platinum subscription so I might have some bias in my backtests but it has been a while since i have been able to compare some backtesting
For info these are the results for my weekly:
QFDuc v23 active since 28/08
calendar year:23/10/2020 to now:
and for the last 2 years: 23/10/2019 to now

These QFDuc results compare very favorably to both ADX and Platinum backtests so that is a bit of a relief as these results are not tweaked...
But I am sure the Ulcer index of Mr Skate is much more serene than mine..

 

[qldfrog]

now for the QF sector bias weekly:
one calendar year:


and last 2 calendar years:

this version is live since 17/08 and indeed results have recovered with live trading for that weekly system

 

[qldfrog]

I have to say QFDuc has been disappointing:
backtests are stunning but I did not see much better real life results since end of august [when that version became live], very volatile and portfolio down roughly 10% since whereas xnt is basically flat

 

[Skate]

a warm welcome back to Mr @Skate who just presented a a few weekly system backtests;
sadly I have no platinum subscription so I might have some bias in my backtests but it has been a while since i have been able to compare some backtesting

These QFDuc results compare very favorably to both ADX and Platinum backtests so that is a bit of a relief as these results are not tweaked...
But I am sure the Ulcer index of Mr Skate is much more serene than mine..

I post with the best intentions
The backtest results of your "QFDuc v23" raise a "Red Flag" for me.

The old adage, "When something seems too good to be true, it probably is" - meaning those figures are excellent but not believable.

Skate.

 

[qldfrog]

I post with the best intentions
The backtest results of your "QFDuc v23" raise a "Red Flag" for me.

The old adage, "When something seems too good to be true, it probably is" - meaning those figures are excellent but not believable.

Skate.

true but I could not find anything like looking forward ..nor many time when explore results differ from week to week
These periods were not in my development testing so not exactly custom fitted
But volatility is intense:
these are actual results per week

27/08/2021​	3.1997%
3/09/2021​	0.2974%
10/09/2021​	0.6691%
17/09/2021​	10.3125%
24/09/2021​	-11.3284%
1/10/2021​	-8.6293%
8/10/2021​	-1.0223%
15/10/2021​	-1.0558%
22/10/2021​	3.0996%

Please also note on a different subject: system QFSec results above start at 80k not 100k so please look at % for comparison

 

[peter2]

I agree with @Skate. There's got to be an error in your system with regards to back testing. The results are unbelievably good.
450 trades pa. is 9 trades pw. This is doable if your pos size is very small but if your initial positions are small you can't make huge individual profits (max win of 32K ?). If your pos size is large then the system would be unable to do 450 trades pa. Is the back test investing money that is unavailable or already used?

You mention that the live performance of QFDuc has been disappointing since going live. You wouldn't have any disappointing periods if QFDuc was as good as indicated by the back-test (fast recovery factor, small max DD). This would indicate that there is an error.

It appears that both back test reports contain the same error. I'm unable to assist you further as I don't use Amibroker.

 

[qldfrog]

I agree with @Skate. There's got to be an error in your system with regards to back testing. The results are unbelievably good.
450 trades pa. is 9 trades pw. This is doable if your pos size is very small but if your initial positions are small you can't make huge individual profits (max win of 32K ?). If your pos size is large then the system would be unable to do 450 trades pa. Is the back test investing money that is unavailable or already used?

You mention that the live performance of QFDuc has been disappointing since going live. You wouldn't have any disappointing periods if QFDuc was as good as indicated by the back-test (fast recovery factor, small max DD). This would indicate that there is an error.

It appears that both back test reports contain the same error. I'm unable to assist you further as I don't use Amibroker.

I will try to deepen the research further , and hope to be able to do that during the week
And will share this here or in my thread

 

[qldfrog]

too good to be true BT on QFDuc v23;
the analysis;
Context
that system uses 20 positions, started with equity of 85k on 28/08 when v23 was switched on, and was 55% invested so we will experience a border effect/legacy effect for a dozen or so earlier packets
in real world, we experienced: 70 buy and sell, BT on the same period 61 so same same
out of the 70, we had 4 earlier investment sales
so each week on average we got 8 to 9 sell and matching buy

real trades did lose 10k on these new trades triggered by v23

BT got a win of 24k obviously, that's quite a difference
Backtest equity curve:

so same shape but obviously reduced gain, increased losses;
Are we dealing with the same tickers?
seems so but we have sometimes big differences in the exit prices between backtest and real:
one example is telling:
real trade:

YOJ

13/09/2021​

30/08/2021​

18297​

$0.23

$10.00

$4,218.31

$0.195​

$ 10.00

$ 3,557.92

-$ 660.40

-15.66%​

and Backtest:

YOJ

Long

3/09/2021​

0.23​

17/09/2021​

0.22​

-4.35%​

-205.05​

-4.82%​

Same dates BT using different time but these are the same trades;
I lose $660 vs BT $205
so i sold YOJ at open at 19.5c on the13/09
based on my data on the 13/09 (week of the 17/09) the open was indeed at 19.5c not 22c
worse:
YOJ did not reach 22c at open during any following day of the month!!!!
WTF??

Where is amibroker getting that better value? do I have a code issue with a sell price computation?
That is the next check in the next post

 

[qldfrog]

yes code error, I had some code trying to enforce the reality that I can not sell at open at a price lower/higher than around 3% of last close (obviously more for smaller cents shares by there is a variable limit imposed by my trader;
obviously code not working;
so i just set sellprice=open
BT profits do halve=> +10k
I do not expect any different trade

 

[qldfrog]

the new equity curve:

ok now back to do we have the same trades between BT and real:
I have 7 differences out of the 60+ trades:
see graphically below: in yellow the executed ones:

I missed:

yeap 12k gone: why? I just missed the entries as they went past my buy orders and replaced by lower in ranking packets the next week
, or some ripple effects from the 55% already invested at the start vs a pure empty backtest start run
So
in conclusion:
1) we had a code error,
2) once fixed, BT matches the real trade but for: human errors, and issue with limits on open order
Note I forgot to mention the realm: fully paid ordinary stocks + ETF

 

[qldfrog]

Now if i run the fixed code with the year BT:

and 2 y BT, wo historical data



So all very good for me but the issue is then:
how can I actually get these open buy and sell price;
AND not miss all stratospheric rises and get a few cents extra cost and less win on most trades which end up killing my returns
I was already worried about this as it was a simple observation done daily on all my systems with rejected trades or missed trades

 

[qldfrog]

Time for a rest

 

[qldfrog]

So now the hard data is here,yes,my attempt at simulating the missed opportunities due to open order restrictions backfired for backtesting.
Now fixed.the system is "solid" but slippage is killing it.
I always dismissed slippage, was wrong .my initial systems were trading less volatile shares and my orders were executed.
not anymore and i pay dearly
So options are:
1)
Ensure orders are first in the queue..but much is out of my control,
2) get more traded tickers. So stricter volume ema etc
3) being a weekly system,let run the order longer during the week..but then i will probably just catch extra dogs of the selection

 

[Skate]

So now the hard data is here,yes,my attempt at simulating the missed opportunities due to open order restrictions backfired for backtesting.
Now fixed.the system is "solid" but slippage is killing it.
I always dismissed slippage, was wrong .my initial systems were trading less volatile shares and my orders were executed.
not anymore and i pay dearly
So options are:
1)
Ensure orders are first in the queue..but much is out of my control,
2) get more traded tickers. So stricter volume ema etc
3) being a weekly system,let run the order longer during the week..but then i will probably just catch extra dogs of the selection

Adaption is the name of the game
Different times require different methods. "The Platinum Strategy" hopefully goes somewhat to addressing these concerns. Backtesting & paper trading over the last 6 months have resulted in profitability but as of today, that's going to be tested.

Skate.

 

[qldfrog]

To illustrate the differences between live and BT;
of the 37 trades requested at open this morning:
1 was missed:
I had an order for a parcel 8840 PNN at $0.475
PNN opened at 48c to close at 56s
so I missed $700 just on that trade today ;
The BT will capture the win.
Not that it was a bad day : my systems went up1.8% today, but yes we will always have some losses along the way