No changes for today
Printable View
No changes for today
No changes for today.
Because we are stuck in a barely moving market, I will postpone daily updates until something effectively changes in my strategy.
Hopefully next week I will post a new paper that holds my improved strategy using 6 models.
PdP
pdp,
looking forward to reading about your latest work on your strategies.
have a nice weekend,
lisa
Last week I finished my renewed back-testing. I have used the warmest weekend of the year here in Belgium to write a summarisation. I post it here. As with my previous papers, I look forward to any feedback. I am still very eager to learn and improve my trading.
PdP
[QUOTE=pdp-brugge;23421]Last week I finished my renewed back-testing. I have used the warmest weekend of the year here in Belgium to write a summarisation. I post it here. As with my previous papers, I look forward to any feedback. I am still very eager to learn and improve my trading.
PdP[/QUOTE]
Hi Pdp,
Fantastic work !
I will follow along with keen interest !
Will you also show the ATR levels on the table ?
Trev
[QUOTE=pdp-brugge;23421]Last week I finished my renewed back-testing. I have used the warmest weekend of the year here in Belgium to write a summarisation. I post it here. As with my previous papers, I look forward to any feedback. I am still very eager to learn and improve my trading.
PdP[/QUOTE]
Very interesting work.
Did you find out whet the strategy has not worked well in the last few months?
Pascal
[ATTACH=CONFIG]15543[/ATTACH]
Pascal,
Good Question!
Until the beginning of March, I had a long series of large winning trades, my strategy was in "full-force" mode (Kelly Criterion +/-25% and NoFear at 0). See the trades marked in green.
From that point on the market went for more then 4 months nowhere. My strategy had a sequence of losing trades. The first bunch of those trades where made in "full-force" and let to a serious draw down. See the trades marked in red.
I have noted that the strategy does not cope well in a sideways environment. Perhaps (and this is a feeling) is following MF-based EOD signals not optimal for brief short trades. The largest negative trades where made around March 7, April 11 and May 5. This where brief periods where the market changed direction every few days. Nearly all of the losing trades between March 7 and April 17 where short trades.
[QUOTE=pdp-brugge;23421]Last week I finished my renewed back-testing. I have used the warmest weekend of the year here in Belgium to write a summarisation. I post it here. As with my previous papers, I look forward to any feedback. I am still very eager to learn and improve my trading.
PdP[/QUOTE]
PdP,
sorry to be a knucklehead, but i cannot find any XLX_Trades file on the website. could you or someone point me in the right direction?
thanks for sharing your work!
lisa
Lisa, you clock on the link shown by the arrow.
pascal
[ATTACH=CONFIG]15545[/ATTACH]
exellent, and thanks, pascal! told you i was a knucklehead.
lisa
No changes in signals or trades.
No changes in signals or trades.
[QUOTE=pdp-brugge;23421]Last week I finished my renewed back-testing. I have used the warmest weekend of the year here in Belgium to write a summarisation. I post it here. As with my previous papers, I look forward to any feedback. I am still very eager to learn and improve my trading.
PdP[/QUOTE]
Hi PdP,
A question please on you latest excellent paper.
If you are stopped out of a position on a stop loss signal but the model remains on a buy signal would you re-enter on the following day ?
Thanks in advance.
Trev
Trev,
Yes, if the signal is still valid, I open a new trade on the day after the stop-loss occurred. On several occasions the stop-loss occurs at times that the MFs are changing. If a stop-loss happens, most of the times it is beneficial because larger loss is prevented.
PdP
[QUOTE=pdp-brugge;23442]Trev,
Yes, if the signal is still valid, I open a new trade on the day after the stop-loss occurred. On several occasions the stop-loss occurs at times that the MFs are changing. If a stop-loss happens, most of the times it is beneficial because larger loss is prevented.
PdP[/QUOTE]
PdP,
Thanks for that.
If the original buy signal was 24 days ago would you still re-purchase after a stop loss if the instrument is still on a buy signal ?
Trev
Yes. I would follow the system. The most important thing that I have learned since I started trading is never to be emotional. Very hard to master but very necessarily. Ignoring a valid signal just after a stop-loss is emotional. In my back tests I found several series of "small" stop-losses followed by a huge win. You never know if you take a signal what the result is going to be. Because of that uncertainty you need stop-loss setting. That is the mean reason why I did all the work to calculate the ideal stop-loss settings. The distance between the initial start of trade and the initial stop-loss level is the risk I take per trade. By optimizing that risk level it is possible to achieve in the end very nice returns.
[QUOTE=pdp-brugge;23444]Yes. I would follow the system. The most important thing that I have learned since I started trading is never to be emotional. Very hard to master but very necessarily. Ignoring a valid signal just after a stop-loss is emotional. In my back tests I found several series of "small" stop-losses followed by a huge win. You never know if you take a signal what the result is going to be. Because of that uncertainty you need stop-loss setting. That is the mean reason why I did all the work to calculate the ideal stop-loss settings. The distance between the initial start of trade and the initial stop-loss level is the risk I take per trade. By optimizing that risk level it is possible to achieve in the end very nice returns.[/QUOTE]
Just to be certain, in your paper you state that you take the trade, following a signal change, the following day at the open. Do I take it that you also take ALL trades after a stop loss if the original signal is still valid.
(After 22 years full time trading I fully understand your comments about emotional trading) :O(
Trev
Yep, I take all the trades after a stop-loss and if the signal is still valid.
Of course: the filters keep on playing.
Sometimes the filters force me to close a trade because the odds or not in favor to continue a trade once the filters states the signal is no longer valid.
An example: let's state that at a certain time a signal came to buy XLF because the signal in the XLX-files of the EV-website changes to "Bought. The filter for a XLF Bought signal is that Combo-MF may not be in Cash. If Combo-MF is at that time not in Cash, I will take the trade. In during that trade the Combo-MF goes to Cash, then I will close that XLF trade.
[QUOTE=pdp-brugge;23446]Yep, I take all the trades after a stop-loss and if the signal is still valid.
Of course: the filters keep on playing.
Sometimes the filters force me to close a trade because the odds or not in favor to continue a trade once the filters states the signal is no longer valid.
An example: let's state that at a certain time a signal came to buy XLF because the signal in the XLX-files of the EV-website changes to "Bought. The filter for a XLF Bought signal is that Combo-MF may not be in Cash. If Combo-MF is at that time not in Cash, I will take the trade. In during that trade the Combo-MF goes to Cash, then I will close that XLF trade.[/QUOTE]
Understood thanks :O)
Trev
Due to a technical issue with my Reuters Datalink, I am not able to prepare my daily routines.
Once this issue is solved, I will post the daily update.
No changes in signals or trades.
No changes
[QUOTE=pdp-brugge;23474]No changes[/QUOTE]
Hi Pdp,
Regarding stop losses.
You state you use the leveraged TQQQ,UYM,FAS,UXI,TECL and XLY as your preferred trading instruments.
Do I understand correctly that the stop losses are calculated based on the basic XLX's metrics in your daily tables and this is the trigger for a sale of the relavant leveraged instrument ?
(ie the ATR(3) figures relate to the basic XLX's and not the leveraged instrument).
Thanks,
Trev
Trev,
The XLX signals determine when to take a trade and the direction of the trade. The incidental filters limit some of the trade signals.
The stop-loss strategy is calculated based on historical executed trades following the XLX signals but executed with the leveraged instruments.
The ATR(3) I use (and which is displayed in the daily table) is that of the leveraged instrument.
PdP
[QUOTE=pdp-brugge;23476]Trev,
The XLX signals determine when to take a trade and the direction of the trade. The incidental filters limit some of the trade signals.
The stop-loss strategy is calculated based on historical executed trades following the XLX signals but executed with the leveraged instruments.
The ATR(3) I use (and which is displayed in the daily table) is that of the leveraged instrument.
PdP[/QUOTE]
Thanks.
Trev
pdp,
first, wow! this is really a lot of work you have undertaken, and i appreciate you efforts not only because of the time you put into this, but also because you inspired me to look at the daily signals more in depth. i put the daily XLX signals into a spreadsheet and recorded the trades from january of 2010 to the end of july 2012 and found the results amazing. this exercise made me realize i have been an idiot for not following them!
i have some questions about your findings though. why did you decide not to use XLE, XLV, XLY, and GDX as the results seemed quite good? why did you decide to use XLB as the results seemed not so good? maybe i have done something incorrectly and will double check. also, i used QQQ in place of XLK as the holdings look similar and found the results improved quite a bit over XLK. i realize that would be somewhat redundant with the Combo signal, but are they a little redundant already? just curious.
as far as atr3, kelly, and no fear, i'll have to take you word for those. one thing i don't understand is the no fear. how do you know what the drawdown is currently to apply the no fear? is that the drawdown from when the most recent signal is given? sorry, not being a math person i don't really know how to ask the question.
i think i'm starting to drive my boyfriend crazy with all of my excel questions, but i was wondering what software you use for testing. it might be over my head, but at least that would give him some peace since he would not be able to answer questions about that program.
so again, amazing job and thanks!
lisa
Lisa,
I decided not to use the XLE, XLV, XLU, and GDX signals because, at first sight, they where not looking as promising as the others. This first look I based on the raw signals. To me, the XLB, XLI, XLF, XLK and XLY where the best 5. I wanted also to limit the number of models. I did back tests with different number of models. In theory I have 11 models (Combo-MF, GDX and the nine S&P500 sub-sectors). My back tests showed me that using a maximum of 6 models was financially the best.
I agree that the trades from the Combo-MF signals and from the XLK signals are redundant at some point. I found enough differences though (in the result of the trades). I use QQQ (or QLD or TQQQ) for trading the Combo-MF signals. I use TECL as instrument for the XLK signals. I think there is a sufficient enough difference between the Nasdaq-100 companies and the S&P500 Tech-companies. I must confess that I did not thoroughly had a look at the exact composition of the 2 etfs.
How did I do the back tests? I created a database using a proprietary program I created. I have been a software developer for more then 30 years, so this was doable for me. This application enable me to generate thousands of runs with each time a slight different parameter (ATR, Stop-loss, ...). I guess doing this with plain Excel is not realizable.
The Nofear was calculated from the draw downs. I calculate the running equity after each trade. The size of the draw downs determined the NoFear factor.
I hope this makes it a bit more clear.
Do not hesitate to post more questions. This enables me to think again and it gives me a fresh new approach to some items.
PdP
[QUOTE=pdp-brugge;23483]Lisa,
I decided not to use the XLE, XLV, XLU, and GDX signals because, at first sight, they where not looking as promising as the others. This first look I based on the raw signals. To me, the XLB, XLI, XLF, XLK and XLY where the best 5. I wanted also to limit the number of models. I did back tests with different number of models. In theory I have 11 models (Combo-MF, GDX and the nine S&P500 sub-sectors). My back tests showed me that using a maximum of 6 models was financially the best.
I agree that the trades from the Combo-MF signals and from the XLK signals are redundant at some point. I found enough differences though (in the result of the trades). I use QQQ (or QLD or TQQQ) for trading the Combo-MF signals. I use TECL as instrument for the XLK signals. I think there is a sufficient enough difference between the Nasdaq-100 companies and the S&P500 Tech-companies. I must confess that I did not thoroughly had a look at the exact composition of the 2 etfs.
How did I do the back tests? I created a database using a proprietary program I created. I have been a software developer for more then 30 years, so this was doable for me. This application enable me to generate thousands of runs with each time a slight different parameter (ATR, Stop-loss, ...). I guess doing this with plain Excel is not realizable.
The Nofear was calculated from the draw downs. I calculate the running equity after each trade. The size of the draw downs determined the NoFear factor.
I hope this makes it a bit more clear.
Do not hesitate to post more questions. This enables me to think again and it gives me a fresh new approach to some items.
PdP[/QUOTE]
PdP,
You stated you will be reveiwing the sytem quarterly and so I presume you will change the mix of XLX's if warrented ?
Trev
I am going to close my UXI trade today at the open. I started this trade at the open of August 3. At the close last Friday the return of this trade was 4.37%. I will close this trade because the 50MA of XLI went above the 200MA.
My back tests have shown that trades following a "Bought" signal are not beneficial when the 50MA of XLI is above the 200MA and the 20DMF is in cash. Since the beginning of 2010 there have been 11 XLI trades on a "Bought" signal. 6 of them where positive. 4 out of the 5 negative trades where when the 20DMF was in cash and the 50MA of the XLI was above the 200MA. The balance between the positive and negative trades was in favor of the negative ones. That is the reason I skip the XLI signal "Bought" when the 20DMF is in cash and the 50MA is above the 200MA. We had a cross of the 50MA and the 200MA last Friday. On August 2, when the XLI signal "Bought" was given, the 50MA was below the 200MA, so I made the trade. Now that the 50MA is again above the 200MA, I will close that trade.
pdp,
instead of treating the data as EOD, i used the opening price of the day of the signal and am now redoing my excel files. so far, i have still found XLE to be better than XLB but will let you know when i am finished to compare.
sigh.
lisa
Lisa,
Not sure if I am confusing your terminology, but shouldn't you be using "day after the signal" than "day of the signal," meaning if signal was generated at close of 8/13, the earliest you could act would be open of 8/14. Otherwise you are biasing your results by looking ahead.
Harry
harry,
that is exactly my mistake. i essentially jumped the gun and am now correcting my errors. unfortunately, the returns are now worse.
lisa
Lisa,
While you're doing your work again, maybe calculate also trading fees and taxes. I reduce each trade with 0.50% for fees and taxes. Over 2½ years more then 300 trades makes a lot of fees and taxes.
PdP
[QUOTE=pdp-brugge;23490]I am going to close my UXI trade today at the open. I started this trade at the open of August 3. At the close last Friday the return of this trade was 4.37%. I will close this trade because the 50MA of XLI went above the 200MA.
My back tests have shown that trades following a "Bought" signal are not beneficial when the 50MA of XLI is above the 200MA and the 20DMF is in cash. Since the beginning of 2010 there have been 11 XLI trades on a "Bought" signal. 6 of them where positive. 4 out of the 5 negative trades where when the 20DMF was in cash and the 50MA of the XLI was above the 200MA. The balance between the positive and negative trades was in favor of the negative ones. That is the reason I skip the XLI signal "Bought" when the 20DMF is in cash and the 50MA is above the 200MA. We had a cross of the 50MA and the 200MA last Friday. On August 2, when the XLI signal "Bought" was given, the 50MA was below the 200MA, so I made the trade. Now that the 50MA is again above the 200MA, I will close that trade.[/QUOTE]
PdP,
Something I don't quite understand here.
I calculated the highest stop price which occured on 20th Aug at $50.56 - (.7003*1.8) = 49.30
The price on the 23rd of August dropped through this level intraday. UXI should have been stopped out then should it not ?
Thanks in advance.
Trev
Trev,
Do be honest: for now I am using mental stops. I calculate the (mental) stops each day based on the stop-loss levels of the base model XLI (and his ATR). The base model XLI did not have a stop-loss last week.
I have just done a simulation and you are right: the UXI did have a intraday stop-loss on the 22th @ 49.6792.
The main reason for the mental stops is just psychological.
In fact, the XLK model had also a mental stop-loss last Friday. Because I treated it like a mental stop, it did not fire.
I know that one day I should use real stop loss orders. For now, I hesitate that because someone once told me that stop loss orders are easy to spot by trained traders (HFTs ?) and they are used to trigger bull/bear traps. That's the main reason I do not use real stop loss orders for now.
PdP
No changes in signals nor trades.
[QUOTE=manucastle;23489]PdP,
You stated you will be reveiwing the sytem quarterly and so I presume you will change the mix of XLX's if warrented ?
Trev[/QUOTE]
PdP,
Did you have a chance to think about an answer to the above question please ?
Also, as with many systems we have seen here do you think there is a chance that the excellent results you have achieved in your tests are a product of data mining ? Often when a model's equity curve declines after testing this can be caused by over optimization.
I really appreciate your hard work but understand these difficult questions need to be faced.
Thanks in advance.
Trev
Trev,
I do intent to review all the criteria I use on a regular base. The next time I will redo the complete study will probably be end of September. Now that my back test environment is setup (the programming is done) all it takes is changing the parameters and let the system run a couple of thousands simulations. When I say redo, I mean also evaluating if other XLX models become more interesting then the ones I have been using.
Of course I use data-mining techniques (I have been a software-developer of analytical accounting software in the past). I have been careful not to over optimize the system. Because my window of data is rather small (January 2010 up until now) the statistical evidence is also rather small. I hope over time this will improve simply because my database will grow. Because my equity curve has already had a serious decline, I have been quit honest. The equity curve has been going up again the last 4 weeks.
PdP
[QUOTE=pdp-brugge;23510]Trev,
I do intent to review all the criteria I use on a regular base. The next time I will redo the complete study will probably be end of September. Now that my back test environment is setup (the programming is done) all it takes is changing the parameters and let the system run a couple of thousands simulations. When I say redo, I mean also evaluating if other XLX models become more interesting then the ones I have been using.
Of course I use data-mining techniques (I have been a software-developer of analytical accounting software in the past). I have been careful not to over optimize the system. Because my window of data is rather small (January 2010 up until now) the statistical evidence is also rather small. I hope over time this will improve simply because my database will grow. Because my equity curve has already had a serious decline, I have been quit honest. The equity curve has been going up again the last 4 weeks.
PdP[/QUOTE]
Thanks very much for your quick reply.
It will be fascinating to follow along with its development.
If you are away from your desk or on holiday will we still recieve the tables ?
Trev
that's my intention...