Registered User
Joined: 6/6/2005
Posts: 1,157
|
Hi everyone.
Today Mammon brought up the thread "Some Interesting Numbers" from a couple years ago. At that time we did a test to see which stocks performed better, the strongest Russel 1000 stocks from the prior month, or the weakest Russell 1000 stocks from the prior month.
We went back 10 years and determined that buying the strongest Russel 1000 stocks from the prior month was a decent momentum strategy that beat the index performance.
The original thread is here:
http://forums.worden.com/Default.aspx?g=posts&t=26798
I thought it would be interesting to forward test this idea real time to see if we can come up with any refinesments/new ideas and to see how tradable the idea really is.
The idea is to simply buy the top 5 stocks on the 1st day of the month and sell them on the last day of the month. No entry signal. No exit signal. No stop. Just 5 equal weight positions. I will be trading it live to see how it feels.
I will be taking small positions and making notes on what might work as far as optimal stops, etc. I will also be looking at the weakest 5 stocks, though I won't be trading them.
Here are the stocks and their June 2009 performance. All comments are welcome. Good luck all.
David John Hall
|
|
Registered User
Joined: 1/30/2009
Posts: 267
|
Thank you DJH. This should be very interesting to watch.
|
|
Registered User
Joined: 1/28/2005
Posts: 6,049
|
I think if your going to trade high momentum stocks.
Some type of filter should be applied to stop trading
in a bear market.
(long-term moving average on the SP-500 for example)
You probably would have had your head handed to you
in 2008. (if your not using stops)
Maybe a basket of inverse ETF's could be traded when
the 500 is under the average.
Thanks
diceman
|
|
Registered User
Joined: 6/6/2005
Posts: 1,157
|
Very good point diceman - I think I'll go back and look at 2008 and post the results for that year here. I'll include inverse ETFs alongside the Russel - should be very interesting. And I'll think about stops. Bigblock mentioned offsetting the risk on the other side with options. David John Hall
|
|
Registered User
Joined: 10/7/2004
Posts: 816
|
DJH,
Are you going to use the same criteria (top and bottom 5 - best monthly change %) as the other thread - or are you adding any addition criteria - based on your success since the original analysis?
Regards
Bob Mc
|
|
Registered User
Joined: 6/6/2005
Posts: 1,157
|
Hi Bob,
I have used the exact same criteria as the original thread, using only % gained (top 5) or lost (bottom 5)during the prior month. I must say I am very eager to get home and do 2008 tonight -- I'll post the results here.
David John Hall
|
|
Registered User
Joined: 12/31/2005
Posts: 2,499
|
Cool idea! Do you have backtesting results? Granted the membership bias would deminish the vaulue of the backtest.
|
|
Registered User
Joined: 12/31/2005
Posts: 2,499
|
Pairs Strategy:
A related idea, reducing volatility and risk, pair trading the top 5 best as longs with the bottom 5 worst as shorts. This would provide a hedge against the market direction.
|
|
Registered User
Joined: 12/31/2005
Posts: 2,499
|
QUOTE (jas0501) Pairs Strategy:
A related idea, reducing volatility and risk, pair trading the top 5 best as longs with the bottom 5 worst as shorts. This would provide a hedge against the market direction.
Oh for want of edit feature on posts....
It should read
This would provide a hedge against the market direction change.
|
|
Registered User
Joined: 6/6/2005
Posts: 1,157
|
That's an excellent idea, jas. As I'll be going forward a year and recording the results of the longs and shorts it will be easy to determine how well it would work for me and/or any changes we could make.
The only backtesting done against this idea was in the orginal thread where we manually went back about 10 years. I was really happy with the results but have never traded it live until now.
David John Hall
|
|
Registered User
Joined: 7/1/2008
Posts: 889
|
I don't think it would provide a hedge against market change...
from my experience when the market changes direction the market leaders get crushed the most and the laggards the least.
this could be happening now, in fact, in the markets
|
|
Registered User
Joined: 2/5/2006
Posts: 1,148
|
QUOTE (jas0501) QUOTE (jas0501) Pairs Strategy:
A related idea, reducing volatility and risk, pair trading the top 5 best as longs with the bottom 5 worst as shorts. This would provide a hedge against the market direction.
Oh for want of edit feature on posts....
It should read
This would provide a hedge against the market direction change.
if your a good market timer, and utilize good a money management strategy, you don't need to hedge.
|
|
Registered User
Joined: 12/31/2005
Posts: 2,499
|
QUOTE (funnymony) QUOTE (jas0501) QUOTE (jas0501) Pairs Strategy:
A related idea, reducing volatility and risk, pair trading the top 5 best as longs with the bottom 5 worst as shorts. This would provide a hedge against the market direction.
Oh for want of edit feature on posts....
It should read
This would provide a hedge against the market direction change.
if your a good market timer, and utilize good a money management strategy, you don't need to hedge.
Sure, but since this strategy could buy and hold for a month the hedge is worth investigating as I am assuming market timing is not part of this exploration. Granted a hedged approach often does not show as good a return month to month. But the goal of the hedge is to avoid drastic drawdowns. The paired strategy should reduce the volatility as the expense of some profit.
That is the assumption. If I get a chance I will run some backscans as get a rough performance profile.
As I stated earlier backtest of membership watchlists, as in the Russell 1000, suffer from the fact the the current membership is not what the membership was 5 years ago. So when backtesting for July 2004 one isn't really testing the bottom 5% of the Russell1000 as of July 2004. One is testing the bottom 5% of the current membership, as of 7/04. Membership bias, a form of suvivorship bias.
I expect this membership bias to amplify the long profits and reduce the short profits, as missing members from 5 years ago are typically delisted or defunct with a few merges mixed in. Thus the best shorts as of 7/04 may not be part of the backtest.
Oh for historically accurrate prices and symbols, data for all symbols trading any given day. But that data is expensive to get.
|
|
Registered User
Joined: 6/6/2005
Posts: 1,157
|
Hi everyone,
Here is the breakdown for 2008. First of all I think we all deserve a round of applause for even trading through any of this. Truly, we can say "we were there" when they talk about the crash of 2008. Amazing. That any of us survived at all is a wonder to me.
As you can see, the numbers are pretty horrible. On the far right I have the cumulative and average numbers for the year, plus a couple trading simulations. At first glance you might look at the bottom 5 and think that it beat the top 5, but then we have to remember the distribution of the numbers. When we take them month by month with a simulated equity amount we see the top 5 does fare better.
Next, I took the bottom 5 short and made a KILLING...until December wiped me out completely.
Then I looked at the Russell 1000's performance for the year and saw that it declined 38.15%. Our trading the top 5 would have resulted in a loss of 33.47% excluding transaction costs.
I have to say that there is nothing thrilling about that.
One thing that I am constantly amazed by is the amount of times the top5 for any month had similar gains of around 30 - 33%. Over and over again, 30% popped up. Just an interesting quirk.
Anyway, I'm open to any and all interpretations of this data. i just thought it would be interesting to look at.
David John Hall
|
|
Registered User
Joined: 12/31/2005
Posts: 2,499
|
David cool numbers. Instead of posting an image you might try ]cutting and pasting from excel. Just add borders in excel before copying. See some garbage numbers below as as example.
Also there seems to be a problem with the list line of Bottom 5 - Sort. The -$567,735.63 loss on a starting balance of $469,552.25 looks wrong.
I think with these above number corrected the performance of the "paired strategy" would be resonable.
| $18,301.00 |
$0.00 |
0.000% |
$18,301.00 |
| $12,800.00 |
$0.00 |
0.000% |
$12,800.00 |
| $22,000.00 |
$0.00 |
0.000% |
$22,000.00 |
| $12,800.00 |
$0.00 |
0.000% |
$12,800.00 |
| $18,000.00 |
$0.00 |
0.000% |
$18,000.00 |
| $6,400.00 |
$0.00 |
0.000% |
$6,400.00 |
| $86,444.80 |
$0.00 |
0.000% |
$86,444.80 |
| $68,480.00 |
$0.00 |
0.000% |
$68,480.00 |
| $65,000.00 |
$0.00 |
0.000% |
$65,000.00 |
| $19,324.85 |
$0.00 |
0.000% |
$19,324.85 |
Before copying doing Menu: format=>column=>autofit selection will resize as well
| $18,301.00 |
$0.00 |
0.000% |
$18,301.00 |
| $12,800.00 |
$0.00 |
0.000% |
$12,800.00 |
| $22,000.00 |
$0.00 |
0.000% |
$22,000.00 |
| $12,800.00 |
$0.00 |
0.000% |
$12,800.00 |
| $18,000.00 |
$0.00 |
0.000% |
$18,000.00 |
| $6,400.00 |
$0.00 |
0.000% |
$6,400.00 |
| $86,444.80 |
$0.00 |
0.000% |
$86,444.80 |
| $68,480.00 |
$0.00 |
0.000% |
$68,480.00 |
| $65,000.00 |
$0.00 |
0.000% |
$65,000.00 |
| $19,324.85 |
$0.00 |
0.000% |
$19,324.85 |
|
|
Registered User
Joined: 6/6/2005
Posts: 1,157
|
Hey jas, nice trick! Never knew that was a possibility.
Looking at that last number, you're right. Let's just assume it's any number less than zero and a lot less than the $463,000.00 you would have had at the beginning of the month. :)
Looking at the numbers I also see that the 1st number in the top 5 should be .1653 not .01653 -- what can I say -- rushing before work. I will revisit this tonight.
I think I'm also going to do the 1st half of 2009 tonight.
David John Hall
|
|
Registered User
Joined: 12/31/2005
Posts: 2,499
|
Is it
Buy on open first day of month
sell on close last day of month
If so Dec's PLD would be
Buy 3.91
Sell 13.89
Perf (13.89-3.91)/3.91 = 2.55 not 4.49
|
|
Registered User
Joined: 6/6/2005
Posts: 1,157
|
Hi jas,
I create a "c" key scan of the month and just log what the report shows me. Sometimes the first trading day is the 3rd of the month, etc. I'll have to revisit that one and see if anything went wacky...
|
|
Registered User
Joined: 3/29/2005
Posts: 56
|
QUOTE (davidjohnhall) Hi everyone, Here is the breakdown for 2008.
As you can see, the numbers are pretty horrible. I looked at the Russell 1000's performance for the year and saw that it declined 38.15%. Our trading the top 5 would have resulted in a loss of 33.47% excluding transaction costs.
I have to say that there is nothing thrilling about that.
Anyway, I'm open to any and all interpretations of this data. i just thought it would be interesting to look at.
Interesting (and depressing!)
1) What if you add a criterion that the Russell has to be up over the previous month in order to go long on the top 5, and it has to be down in order to short the bottom 5?
2) Instead of a straight monthly ROC, use 4/10 * [ROC for the most recent week] + 3/10 * ROC[-1] + 2/10 * ROC[-2] + 1/10 * ROC[-3] ? You could use other weights, maybe based on Fibonacci ratios. But divide each by a certain amount so the sum equals 100 percent.
|
|
Registered User
Joined: 3/29/2005
Posts: 56
|
Never mind my first suggestion [I wish we could edit our posts!], if the system tracks the index performance that closely this should not make much difference.
|
|
Registered User
Joined: 6/6/2005
Posts: 1,157
|
Hi Vvrd,
I liked your suggestion. Are you sure it has no merit? So you're saying, use the ROC from the prior month to weight the current month's position.
I agree, the numbers are depressing, I'd like to do 2009 to see how it compares to the index when stocks were flying as they were in March-May.
Another test i did but abandonned on the original thread was to look at stocks that had nearly zero ROC the prior month. They turned out to be pretty haphazard because they could take off in either direction and pretty much cancelled each other out.
For the most part I enjoy excersises such as this because it gets me thinking.
I also find it interesting how the same names come up again and agian.
David John Hall
|
|
Registered User
Joined: 1/28/2005
Posts: 6,049
|
It should be remembered this is a 100% long strategy
with no stops. So I see nothing wrong with it performing
slightly better than the indexes in one of the great bear markets.
(combined with the larger gains during the bull market)
I still like the idea of using a long-term MAV as a filter
not to do this in a bear market.
(probably the simplest solution)
Another possibility would be to put a bunch of 1 to 1
inverse etf's in a watchlist with the Russell stocks.
(DOG,SH,MYY,PSQ,RWM,SBB,EFZ,EUM)
(of course this is a pure momentum play. You would
skip the 5 weakest short hedge as the etf's would
probably be a the bottom in a bull or you could keep the
short just only short stocks and not the inverse etf's)
The idea would be they would move to the top in a bear
market and hedge your positions.
(not sure I would trust the 2X, 3X inverse as they may move to
the top in minor corrections)
However they may require more time. Like a 1 quarter sort
or 6 month sort.
(over a 1 month period there may still be stocks that beat the
etf's)
Thanks
diceman
|
|
Registered User
Joined: 12/31/2005
Posts: 2,499
|
to backtest this with Stockfinder ...
I'm still trying to wrap my head around this. It seems like a multi-step process:
Pass 1
1. compute the monthly performance for each symbol
2. compute the ranking top-x and bottom-x
Pass 2
3. execute the trades
a. exit if not still in the top/bottom x
Complicating factor: ties in the ranking cause more than x to qualify. What to use as a tie-breaker?
Two passes are required since Stockfinder processes for each symbol from firstbar to lastbar.
Somes choices:
1. output the monthly performance numbers, compute the ranknings and read them back in.
2. output the performance numbers an do the processing in Excel
3. trick Stockfinder into doing 2 passes and manage the performance history as static values.
Time will tell....
when I get a chance to attempt this...
|
|
Registered User
Joined: 6/6/2005
Posts: 1,157
|
Hi jas,
Sounds like once you get it set up it willo be a lot easier than the manual way I've been doing it which takes about 30 min per year! lol
When it comes to a tie breaker I usually use something very simple like alphabetical or price to keep it as simple as possible.
I'm very interested to see your results. Please keep us posted!
David John Hall
|
|
Registered User
Joined: 3/29/2005
Posts: 56
|
QUOTE (davidjohnhall) Hi Vvrd,
I liked your suggestion. Are you sure it has no merit?
Not without testing it, but since the top 5 results is just slightly better than the index itself, it probably won't change things much.
QUOTESo you're saying, use the ROC from the prior month to weight the current month's position.
Not exactly, I was saying don't go long unless the ROC of the Russell or other underlying index is also up in the previous month, and don't go short on the bottm X unless the index ROC is negative. BUT, [1] if you wait until the ROC is above zero you might miss most of the reversal;
[2] you say the bottom X don't decline as much in a down period as the top X, so you might want to short the previous month's top X when the index is in decline;
[3] What I really had in mind was a weighted segmented ROC to get in early (which leaves you prey to whipsaws); weekly ROC[-3] is weighted the least, 1/10; the next most recent week had a weight of 2/10, week[-2] is mult by 3/10, and the most recent week's ROC has a full weight of 4/10 (the sum of the weights should equal 1). Or you could just use an EMA of the ROC.
I made a graph showing my formula, with a standard 4-week ROC overlaid, for the past several years of the Russell 3000 but I can't seem to attach it to this message, since I don't have a webpage to upload it to.
QUOTEAnother test i did but abandoned on the original thread was to look at stocks that had nearly zero ROC the prior month. They turned out to be pretty haphazard because they could take off in either direction and pretty much cancelled each other out.
What if you looked at stocks that were 2 standard deviations above their recent bottom, or outside ATR? Or that broke their Bollinger band/envelope and then moved above it?
QUOTEI also find it interesting how the same names come up again and again.
Like alchemists searching for a way to transmute 1s and 0s into gold...
|
|
Registered User
Joined: 12/31/2005
Posts: 2,499
|
QUOTE (davidjohnhall) Hi jas,
Sounds like once you get it set up it willo be a lot easier than the manual way I've been doing it which takes about 30 min per year! lol
When it comes to a tie breaker I usually use something very simple like alphabetical or price to keep it as simple as possible.
I'm very interested to see your results. Please keep us posted!
David John Hall
Yes but it should read
Sounds like IFyou get it set up it will be a lot easier than the manual way.
As you can see it is a big if.
|
|
Registered User
Joined: 1/28/2005
Posts: 6,049
|
Is the problem that they don't have a monthly buy/sell criteria?
If it helps simply the test process.
Selecting candidates by a 21 day percent change and
holding for 21 days would probably give a reasonable approximation
of the results.
Thanks
diceman
|
|
Registered User
Joined: 3/29/2005
Posts: 56
|
Is there a way to include hard drive charts in TC forum posts?
|
|
Registered User
Joined: 6/6/2005
Posts: 1,157
|
Hi jas,
Yeah, I wasn't assuming you'd put it together, so no worries there. That big IF could be attached to each one of my posts as well.
David John Hall
|
|
Registered User
Joined: 6/6/2005
Posts: 1,157
|
Like alchemists searching for a way to transmute 1s and 0s into gold...[/QUOTE]
Nah, not like alchemists, like speculators. I've always been an eccentric speculator. I suppose that's what appeals to me. It's interesting to me how some patterns in the market (Edwards and Magee, most standard indicators, etc.) are acceptable, but when you start to move toward the fringe you're just asking for it. LOL There's quite a lot in the market that no one understands, and it's pure joy for me to delve deeper into it.
P.S. There's a thread someonwehere on the message board about putting charts into posts. I'm not 100% sure what you mean by hard drive charts though...
David John Hall
|
|
Registered User
Joined: 3/29/2005
Posts: 56
|
QUOTEQUOTELike alchemists searching for a way to transmute 1s and 0s into gold...
Nah, not like alchemists, like speculators.
You'll have to excuse the metaphor, I've been reading Neal Stephenson's [u]Baroque Cycle[/u]
QUOTEP.S. There's a thread someonwehere on the message board about putting charts into posts. I'm not 100% sure what you mean by hard drive charts though...
Charts that reside on my computer, not on a website where I can provide a url.
|
|
Registered User
Joined: 12/31/2005
Posts: 2,499
|
QUOTE (Vyrd) Is there a way to include hard drive charts in TC forum posts?
see
Worden Discussion Forum » Market Discussion » Stock and Market Talk » Inserting A Chart into A Worden post
|
|
Registered User
Joined: 3/29/2005
Posts: 56
|
The bmp method didn't work, maybe I will get a photbucket account.
|
|
Registered User
Joined: 8/15/2006
Posts: 132
|
Just out of curiosity has anyone looked at the stocks that are added and removed form the Russell and the performance within a certain period? If no one has looked at this I will when I return home this fall but it would seem to me there has to be a profitable window in that change. Not to derail this thread, just a thought that came to me when reading it.
Scott
|
|
Registered User
Joined: 6/6/2005
Posts: 1,157
|
Hey mmscottyb,
I read a study once that showed an edge for stocks removed/added to the S&P500 but completely forgot about it. I don't mind this derailment at all as I think we have all established that the momentum play is too depressing to warrant much more energy. LOL
David John Hall
|
|
|
Guest-1 |