Looking thru the SPY System historical trade data I took a look
at the distribution of trades by gross P/L. No big surprises, but a
couple of thoughts that reinforce following the system signals and
removing discretion.
- the top 10 trades (259 total) accounted for 25% of the
system's gross P/L
- that's 4% of the system trades accounting for 25% of the gross
P/L
- that's $153K out of $604K gross P/L (single entry system)
You never know when a "top 10" trade is being executed. If you
cherry pick trades or take profits early you may be significantly
compromising the potential P/L of the system. This is one case of
being a lemming being a good thing. Good trading everyone!
That reminds me of what mutual fund companies say about
trying to time the market, e.g. "Missing just a few of the
best days in the market can have a dramatic negative effect on
overall portfolio results". What they conveniently fail to
mention is the dramatic positive impact of missing a few of the
worst days. My money is on Matt's system. It has beaten
"buy and hold" by a country mile. Knock me down and call me a
lemming!
Very interesting work tumbler! Taking the conversation a
different way might also be slightly fun as you were discussing how
cherry picking trades or deviating from the system in general could
have a profound effect on someones overall results; what if we
could optimize leverage to put the trades more in our favor?
I am imagining that you were using the single entry system for
your analysis but what if we then took the top 10 trades and
reviewed if they ever used Multi entry. If so how many
entries did it take? Can we utilize the first trade or 2 with
the SPY but then use subsequent signals to trade leveraged vehicles
like the at the money calls or even futures. In general I am
a subscriber that more leverage equals more problems BUT is there a
way we can put the odds in our favor and then sprinkle in some
leverage to try and outpace gains. I might spend my weekend
looking over those trades and matching them with the Multi
entry.
FYI, I was/am managing the system for Matt while he is away and
serve as a daily back-up but I was not integral in the development
so my comments when discussing the system, unless I am calling out
a new signal, should be taken more as if I was just a member.
On the subject of additional leverage for the SPY system.
I have been playing around with manually backtesting (my
computer skills are limited in terms of computerized backtesting)
using SPY for the first buy. If there is a 2nd signal sell
all the SPY and go into SSO. Then sell it all and go into
UPRO for the 3rd signal. For the 4th I keep the UPRO and buy
another 1/3 to add to the position to equal 4x leverage.
As I write this it seems like crazy risk by the 3rd and 4th leg
of the trade but the drawdowns are not too bad. Almost all
the historic trade series end up positive. The only ones that
don't are the ones that don't have a positive last leg to them.
Those do have bigger loses, but again not immensely
bigger. And there are so few trade series in the history that
don't end on a positive trade
I think the key to this working is the fact that the system
looks to end a long trade on an oversold bounce. So even when
a trade is going poorly, all but a few end up closing in the plus
column.
In order to compare it to the $100,000 back test one would trade
$75k on the first 3 legs and use that last 25k for the final UPRO
buy in the eventuality of a 4th leg.
Unfortunately I don't have any nice spreadsheets to post since
all my work is penciled out longhand. Also I used SPY
data for all the legs of a trade and just multiplied the results by
2, 3 or 4 for the appropriate leverage of any given leg.
On trades from 5/2010 to 5/2011 (that I did use SSO and UPRO
data for this) I got a return of 67% for this technique vs 30% for
the basic multi entry system over the past 12 months.
After I found this I went back to 1995 and have worked forward
to 1999 so far.
Basic SPY System "Mo"
Leverage technique
1995 8%
17%
1996 15%
41%
1997 27%
55%
1998 22%
47%
1999 20%
53%
These seem to be fairly out sized returns. I hope I
haven't made some fundamental error in logic (or math) that
proves these all wrong.
I hope my english is good enough that you see my point
I disagree that a more leveraged version (2x and 3x funds,
options....) brings a better result. Fact is, that when you take
UPRO instead of SPY you are willing to increase your account size
(profits, but also risks) by the factor 3, and when you are able to
take another 1/3 for the fourth entry, your account size has
increased from US$ 100'000 to 400'000. Why not put all the 400'000
into the first entry (single system) and your profits would be 4
times higher. Remember, the single entry system brings the highest
profit, but with the highest drawdowns.
In my opinion the only possibility to increase the profit (next
to increasing the account size of course) is finding out in advance
how chances are that we will get a second, third or fourth entry. I
don't know if this is possible. Maybe Matt gets a first entry
signal, but the system also tells him that we are rather early and
there is a good chance that we will get more entries. In this case
we should not put much money into the first entry. Or
maybe the system tells Matt that the market is very stretched and
the first entry will probably be the last. Then we should put
everything into this first entry. We just don't know if there are
"better" first entry calls than others. Maybe Matt can tell
something if chances are always the same for getting second, third
or fourth entries.
Another thing is the risk factor. We should not not forget that
risks are high. Of course the system did not have a huge drawdown
in backtest. This is logical, beacuse Matt would have changed the
system parameters. It doesn't mean the risks are limited because
there were only small drawdowns.
Just wanted to mention that in case someone is basically willing
to play the system with US$ 50'000, increased the size with
borrowed money to 100'000 because of the profitability of the
system and plays now UPRO (3x) instead of SPY because it's so easy
earned money...... Maybe we'll all find out that it's not so easy
after a fourth entry....
That reminds me of Hedge Funds and banks a few years ago. Greed
and risk models that relied on the past.....
zwyss - I suspect we will never know if there are "better"
first entry calls than others. In any case, you can estimate
the chances you mention by looking at the trade history.
There were 251 total trades. 145 were 1-entry trades.
68 were 2-entry trades. 24 were 3-entry trades. 14 were
4-entry trades.
Whenever there was a 1st entry signal:
It was a 1-entry trade 58% of the time (145/251).
It was a 2-entry trade 27%...(68/251)
It was a 3-entry trade 9%...
It was a 4-entry trade 6%...
Whenever there was a 1st entry signal AND a 2nd entry signal:
It was a 2-entry trade 64% of the time (68/(68+24+14)).
It was a 3-entry trade 23%...
It was a 4-entry trade 13%...
Whenever there was a 1st entry signal AND a 2nd entry signal
AND a 3rd entry signal:
It was a 3-entry trade 63% of the time (24/(24+14)).
It was a 4-entry trade 37% of the time (14/(24+14)).
So no matter which entry signal you are talking about (1st,
2nd, or 3rd), it was followed by another entry signal about
35% to 40% of the time.
Well, Matt is the only person who can give an answer to the
question if there are stronger and weaker entry signals or if all
signals are about equal. This system was once compared to a rubber
band which bounces back if stretched too much. Maybe the entries
were set when the rubber band is stretched 70%, 80%, 90% and 95%.
That would mean that there is always the same chance of getting
next entries because we just don't know if the market turns at 65%,
78%, 92 % or maybe the band explodes which would not be
good.......
zwyss, You make very valid points that one needs to
take under consideration when deciding whether or not to deviate
from the basics of the system.
Perhaps Matt can shed some light on the point you raise re:
large drawdowns having been inadvertently "filtered"
out of the historic data during system design.
along the lines of what I have pointed out -- one must be
prepared to take the losses, and also to sit through the
drawdowns. And my experience is that generally we
over-estimate our psychological ability to do
this. That said, I think
the discussion of how to allocate the entries -- within the
parameters of your chosen total capital -- is very valuable.
tumbler, and everyone -- thanks so much for sharing all this
stuff. I just can't post much lately and can't do much in the
evenings, but want to really thank you.
tj6p -- your post, too, thanks very much. Also, I want to
say, this is much the way peter Campbell trades his systems.
As the system scales in, the added positions are larger. If
the scale-ins work at all, which the obviously do, it really makes
sense.
Posted by burkmere on 13th of May 2011 at 12:12 pm
The SPY System seems to be about probability. It's high
probability trading. What I especially am impressed about is
that even with the drawdowns, particuarly if you use the
multi-entry system, the greatest loss out of approx. 260 trades is
1.3%!! So, it helps out psychologically, too, to know that
you can scale in at lower prices and out of the losses (less than
4% of the trades in the multi-entry) the highest is 1.3% and I
believe the average is less than half a percent.
Very interesting and a concept I agree with. Cherry
picking systems, thereby turning them into discretionary trading,
is almost always counter-productive.
Now, a question. Since the second buy signal in the
multi-entry scheme has a higher success rate than the only entry in
the single entry system, I wonder how it would work using options.
There are fewer trades, but would having a higher success
rate coupled with leveraging compensate for that?
Tom, it's an honor to have the same thoughts as you.
Unfortunately, it's doesn't look so good the other way
around. But maybe we can both figure out how to infuse some
option leverage into the SPY system while controlling risk.
And Tumbler, thanks for all your great ideas.
You just made my weekend! Now we just need to put Tumbler
back to work crunching numbers while you and I sit back and reap
the benefits of the system and his additional analysis!
Follow the system
Posted by tumbler on 13th of May 2011 at 12:06 pm
Looking thru the SPY System historical trade data I took a look at the distribution of trades by gross P/L. No big surprises, but a couple of thoughts that reinforce following the system signals and removing discretion.
- the top 10 trades (259 total) accounted for 25% of the system's gross P/L
- that's 4% of the system trades accounting for 25% of the gross P/L
- that's $153K out of $604K gross P/L (single entry system)
You never know when a "top 10" trade is being executed. If you cherry pick trades or take profits early you may be significantly compromising the potential P/L of the system. This is one case of being a lemming being a good thing. Good trading everyone!
That reminds me of what
Posted by yojimbo on 13th of May 2011 at 04:37 pm
That reminds me of what mutual fund companies say about trying to time the market, e.g. "Missing just a few of the best days in the market can have a dramatic negative effect on overall portfolio results". What they conveniently fail to mention is the dramatic positive impact of missing a few of the worst days. My money is on Matt's system. It has beaten "buy and hold" by a country mile. Knock me down and call me a lemming!
Very interesting work tumbler! Taking
Posted by tom on 13th of May 2011 at 12:22 pm
Very interesting work tumbler! Taking the conversation a different way might also be slightly fun as you were discussing how cherry picking trades or deviating from the system in general could have a profound effect on someones overall results; what if we could optimize leverage to put the trades more in our favor?
I am imagining that you were using the single entry system for your analysis but what if we then took the top 10 trades and reviewed if they ever used Multi entry. If so how many entries did it take? Can we utilize the first trade or 2 with the SPY but then use subsequent signals to trade leveraged vehicles like the at the money calls or even futures. In general I am a subscriber that more leverage equals more problems BUT is there a way we can put the odds in our favor and then sprinkle in some leverage to try and outpace gains. I might spend my weekend looking over those trades and matching them with the Multi entry.
FYI, I was/am managing the system for Matt while he is away and serve as a daily back-up but I was not integral in the development so my comments when discussing the system, unless I am calling out a new signal, should be taken more as if I was just a member.
"mo leverage"
Posted by tj6p on 13th of May 2011 at 05:04 pm
Tom,
On the subject of additional leverage for the SPY system. I have been playing around with manually backtesting (my computer skills are limited in terms of computerized backtesting) using SPY for the first buy. If there is a 2nd signal sell all the SPY and go into SSO. Then sell it all and go into UPRO for the 3rd signal. For the 4th I keep the UPRO and buy another 1/3 to add to the position to equal 4x leverage.
As I write this it seems like crazy risk by the 3rd and 4th leg of the trade but the drawdowns are not too bad. Almost all the historic trade series end up positive. The only ones that don't are the ones that don't have a positive last leg to them. Those do have bigger loses, but again not immensely bigger. And there are so few trade series in the history that don't end on a positive trade
I think the key to this working is the fact that the system looks to end a long trade on an oversold bounce. So even when a trade is going poorly, all but a few end up closing in the plus column.
In order to compare it to the $100,000 back test one would trade $75k on the first 3 legs and use that last 25k for the final UPRO buy in the eventuality of a 4th leg.
Unfortunately I don't have any nice spreadsheets to post since all my work is penciled out longhand. Also I used SPY data for all the legs of a trade and just multiplied the results by 2, 3 or 4 for the appropriate leverage of any given leg.
On trades from 5/2010 to 5/2011 (that I did use SSO and UPRO data for this) I got a return of 67% for this technique vs 30% for the basic multi entry system over the past 12 months.
After I found this I went back to 1995 and have worked forward to 1999 so far.
Basic SPY System "Mo" Leverage technique
1995 8% 17%
1996 15% 41%
1997 27% 55%
1998 22% 47%
1999 20% 53%
These seem to be fairly out sized returns. I hope I haven't made some fundamental error in logic (or math) that proves these all wrong.
"mo leverage"
Posted by mkom_prod on 14th of May 2011 at 09:49 am
Leverage
Posted by zwyss on 14th of May 2011 at 05:11 am
I hope my english is good enough that you see my point
I disagree that a more leveraged version (2x and 3x funds, options....) brings a better result. Fact is, that when you take UPRO instead of SPY you are willing to increase your account size (profits, but also risks) by the factor 3, and when you are able to take another 1/3 for the fourth entry, your account size has increased from US$ 100'000 to 400'000. Why not put all the 400'000 into the first entry (single system) and your profits would be 4 times higher. Remember, the single entry system brings the highest profit, but with the highest drawdowns.
In my opinion the only possibility to increase the profit (next to increasing the account size of course) is finding out in advance how chances are that we will get a second, third or fourth entry. I don't know if this is possible. Maybe Matt gets a first entry signal, but the system also tells him that we are rather early and there is a good chance that we will get more entries. In this case we should not put much money into the first entry. Or maybe the system tells Matt that the market is very stretched and the first entry will probably be the last. Then we should put everything into this first entry. We just don't know if there are "better" first entry calls than others. Maybe Matt can tell something if chances are always the same for getting second, third or fourth entries.
Another thing is the risk factor. We should not not forget that risks are high. Of course the system did not have a huge drawdown in backtest. This is logical, beacuse Matt would have changed the system parameters. It doesn't mean the risks are limited because there were only small drawdowns.
Just wanted to mention that in case someone is basically willing to play the system with US$ 50'000, increased the size with borrowed money to 100'000 because of the profitability of the system and plays now UPRO (3x) instead of SPY because it's so easy earned money...... Maybe we'll all find out that it's not so easy after a fourth entry....
That reminds me of Hedge Funds and banks a few years ago. Greed and risk models that relied on the past.....
zwyss - I suspect we will
Posted by yojimbo on 14th of May 2011 at 11:41 am
zwyss - I suspect we will never know if there are "better" first entry calls than others. In any case, you can estimate the chances you mention by looking at the trade history. There were 251 total trades. 145 were 1-entry trades. 68 were 2-entry trades. 24 were 3-entry trades. 14 were 4-entry trades.
Whenever there was a 1st entry signal:
It was a 1-entry trade 58% of the time (145/251).
It was a 2-entry trade 27%...(68/251)
It was a 3-entry trade 9%...
It was a 4-entry trade 6%...
Whenever there was a 1st entry signal AND a 2nd entry signal:
It was a 2-entry trade 64% of the time (68/(68+24+14)).
It was a 3-entry trade 23%...
It was a 4-entry trade 13%...
Whenever there was a 1st entry signal AND a 2nd entry signal AND a 3rd entry signal:
It was a 3-entry trade 63% of the time (24/(24+14)).
It was a 4-entry trade 37% of the time (14/(24+14)).
So no matter which entry signal you are talking about (1st, 2nd, or 3rd), it was followed by another entry signal about 35% to 40% of the time.
nice statistics, thanks.
Posted by Michael on 14th of May 2011 at 03:29 pm
nice statistics, thanks.
yojimbo, nice statistics Well, Matt is
Posted by zwyss on 14th of May 2011 at 12:53 pm
yojimbo, nice statistics
Well, Matt is the only person who can give an answer to the question if there are stronger and weaker entry signals or if all signals are about equal. This system was once compared to a rubber band which bounces back if stretched too much. Maybe the entries were set when the rubber band is stretched 70%, 80%, 90% and 95%. That would mean that there is always the same chance of getting next entries because we just don't know if the market turns at 65%, 78%, 92 % or maybe the band explodes which would not be good.......
My trade count may be
Posted by yojimbo on 14th of May 2011 at 01:42 pm
My trade count may be slightly off as I was getting bleary-eyed trying to distinguish blue from blue-gray from mauve. Close enough for jazz.
zwyss, You make very
Posted by tj6p on 14th of May 2011 at 09:30 am
zwyss, You make very valid points that one needs to take under consideration when deciding whether or not to deviate from the basics of the system.
Perhaps Matt can shed some light on the point you raise re: large drawdowns having been inadvertently "filtered" out of the historic data during system design.
Thanks for the thoughtful input.
along the lines of what
Posted by Michael on 14th of May 2011 at 08:34 am
along the lines of what I have pointed out -- one must be prepared to take the losses, and also to sit through the drawdowns. And my experience is that generally we over-estimate our psychological ability to do this. That said, I think the discussion of how to allocate the entries -- within the parameters of your chosen total capital -- is very valuable.
tumbler, and everyone -- thanks
Posted by Michael on 13th of May 2011 at 06:16 pm
tumbler, and everyone -- thanks so much for sharing all this stuff. I just can't post much lately and can't do much in the evenings, but want to really thank you.
tj6p -- your post, too, thanks very much. Also, I want to say, this is much the way peter Campbell trades his systems. As the system scales in, the added positions are larger. If the scale-ins work at all, which the obviously do, it really makes sense.
ditto what Michael said....especially the
Posted by marketguy on 13th of May 2011 at 08:27 pm
ditto what Michael said....especially the option breakdown (absolutely awesome to see)....
thanks all!
MG
Great work. Thanks for that.
Posted by lessarda on 13th of May 2011 at 12:17 pm
Probability...
Posted by burkmere on 13th of May 2011 at 12:12 pm
The SPY System seems to be about probability. It's high probability trading. What I especially am impressed about is that even with the drawdowns, particuarly if you use the multi-entry system, the greatest loss out of approx. 260 trades is 1.3%!! So, it helps out psychologically, too, to know that you can scale in at lower prices and out of the losses (less than 4% of the trades in the multi-entry) the highest is 1.3% and I believe the average is less than half a percent.
Very interesting and a concept
Posted by algyros on 13th of May 2011 at 12:12 pm
Very interesting and a concept I agree with. Cherry picking systems, thereby turning them into discretionary trading, is almost always counter-productive.
Now, a question. Since the second buy signal in the multi-entry scheme has a higher success rate than the only entry in the single entry system, I wonder how it would work using options. There are fewer trades, but would having a higher success rate coupled with leveraging compensate for that?
I think we were basically
Posted by tom on 13th of May 2011 at 12:25 pm
I think we were basically having the same thoughts. Unfortunately for you I don't know if that makes you more or less intelligent! lol
Tom, it's an honor to
Posted by algyros on 13th of May 2011 at 01:42 pm
Tom, it's an honor to have the same thoughts as you. Unfortunately, it's doesn't look so good the other way around. But maybe we can both figure out how to infuse some option leverage into the SPY system while controlling risk. And Tumbler, thanks for all your great ideas.
You just made my weekend!
Posted by tom on 13th of May 2011 at 03:36 pm
You just made my weekend! Now we just need to put Tumbler back to work crunching numbers while you and I sit back and reap the benefits of the system and his additional analysis!
is matt out for a few more weeks?
Posted by onoffon on 13th of May 2011 at 03:50 pm
No no no, Matt was
Posted by tom on 13th of May 2011 at 04:08 pm
No no no, Matt was technically back today and catching back-up.
Multi-entry with options
Posted by tumbler on 13th of May 2011 at 12:18 pm
algyros - very interesting...hadn't thought about that. Will take a look at that over the weekend.