Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Hi Jack,
Yes, I'm currently working on another update to fix the problem of foreign (to the US) quotes only going back 200 periods. Yahoo! is a great source for quotes, but they do some funky things.
US stocks work for all dates, but non-US (e.g. Canadian) stocks don't. I'll post another note here when everything is updated.
Regarding Cdn RSP, the Fundamental Analyzer does not officially work with non-US stocks (at least that's what you'll read in the user's guide). However it does actually work for many Canadian stocks if you preface the ticker symbol with "CA:" (no quotes, e.g. CA:NT or CA:ABX). This is because the data are retrieved from MSN rather than Yahoo!
Since MSN uses the CA: (rather than .TO) designation, I decided not to officially support non-US stocks in this release. However AI 3.0 SP1's Fundamental Analyzer will use Yahoo! as its data source and so will support all non-US stocks.
I hope that clarifies things a bit. Let me know if you have any other questions.
Hey ROC, The answer is...
A very long time!
Here's what I'd do if I was in that situation:
Calculate the Cost to Operate each vehicle annually. I'm assuming only fuel costs are relevant here and the price of gas is $2 a gallon (you can adjust the fuel price if you like, but you get the idea).
Car A: 10,000 mi / 1 year * 1 gallon / 25 mi * $2 / 1 gallon = $800 per year.
Car B: 10,000 mi / 1 year * 1 gallon / 50 mi * $2 / 1 gallon = $400 per year.
The difference is $400 a year in Car B's favour.
The difference in initial cost is $7000.
So we basically have $7000 at work for us at 8% the first year if we choose Car A. So after one year we would have $7,560.
After one year with Car B, we would have $432 (i.e. $400 at 8%). I'm simplifying things by assuming that the entire cost of fuel will be paid at the start of each year (if you want to be stricter, you can calculate the interest you'd receive on the savings EACH time you fill up Car A, but again you get the idea).
So the problem can be viewed as... when does the total value of investing $400 a year at 8% overtake the total value of a lump sum investment of $7,000 at an annual rate of 8%?
After 50 years, $400 invested annually at 8% would be worth less than $267,700 while a $7,000 lump sum at 8% would be worth $328,311.29
By that time both cars would be sitting at the bottom of a scrap yard. If you went with the more strict calculation then the results would be even better for Car A.
Therefore if the ONLY thing you were concerned about was fuel cost then Car A is the way to go. However, as you're well aware, there are other factors that might cause Car B to come out ahead (such as reliability, resale value, maintenance costs, insurance, etc., etc., etc.).
This is a good example of why Albert Einstein purportedly said that compounding is the 8th wonder of the world. That Einstein really knew his stuff
Service Pack 3b for Automatic Investor 2.0 is now available.
It fixes the problem with retrieving data from Yahoo!
You can get your Service Pack from here--> http://www.automaticinvestor.com/upgradecenter.html (look under the heading Automatic Investor 2.0 Service Pack 3b).
Note that support for AI 2.0 will be withdrawn in a few months, so you should upgrade to AI 3.0 as soon as possible (AI 3.0 is far superior anyways, so it's worthwhile to upgrade even while AI 2.0 is currently being supported).
NOTE: You do NOT require this Service Pack if you're currently using AI 3.0.
If there are any questions, feel free to ask!
Hi Don,
Got your PM. Yes, that will be coming in one of the future updates. I'm also planning to redo the entire reporting functionality.
Here's an example of how easy it is to test stocks with Automatic Investor 3.0.
First, I use Yahoo!'s stock screener to screen for stocks that meet my criteria. In this case I'm looking for:
1) Stocks in any industry.
2) Share price <= $10
3) Average share volume >= 1 million per day.
4) Market Cap. >= 1 Billion.
5) Beta >= 2
6) Annual Revenues >= 250 million.
This gives me a list of 30 stocks.
The next thing I do is select these stocks right off the Web page, then copy and paste them into a text file.
My text file looks like this...
------- Start of Text File -------
ATML Atmel Corp 7.19 7.87M 3.42B 4.17 $1.33B Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
BRCD Brocade Communications Systems Inc 6.64 10.38M 1.72B 3.314 $547.20M Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
CPN Calpine Corp 4.66 12.52M 1.94B 2.243 $8.92B Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
CHTR Charter Communications Inc 4.70 6.43M 1.39B 2.608 $4.82B Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
CIEN Ciena Corp 5.00 12.67M 2.37B 2.629 $279.08M Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
CBB Cincinnati Bell Inc 4.21 1.56M 1.03B 2.579 $1.56B Quote, Chart, News, Profile, Reports, Research, Msgs, Insider, Financials, Analyst Ratings
CPWR Compuware Corp 7.51 2.41M 2.89B 2.912 $1.26B Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
CNXT Conexant Systems Inc 6.07 11.29M 2.79B 3.185 $633.11M Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
CCK Crown Holdings Inc 8.87 1.12M 1.46B 2.009 $6.63B Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials
DYN Dynegy Inc 3.78 4.81M 1.42B 2.431 $5.79B Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
GTW Gateway Inc 6.13 4.02M 1.99B 2.624 $3.40B Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
GMST Gemstar-TV Guide International Inc 6.46 3.51M 2.70B 3.417 $878.65M Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
JDSU JDS Uniphase Corp 4.28 45.09M 6.41B 3.001 $626.30M Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
LSI LSI Logic Corp 9.42 3.70M 3.60B 2.426 $1.69B Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
LU Lucent Technologies Inc 4.28 72.46M 18.23B 2.87 $8.65B Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
MXO Maxtor Corp 8.02 4.85M 1.98B 2.262 $4.09B Quote, Chart, News, Profile, Reports, Research, Msgs, Insider, Financials, Analyst Ratings
NT Nortel Networks Corp 5.87 52.81M 24.50B 3.858 $9.81B Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
ONNN ON Semiconductor Corp 7.32 3.61M 1.85B 3.474 $1.07B Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials
PMTC Parametric Technology Corp 5.13 1.43M 1.37B 2.328 $656.78M Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
Q Qwest Communications International Inc (New)
RFMD RF Micro Devices Inc 8.49 10.76M 1.58B 3.159 $626.29M Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
RRI Reliant Resources Inc 8.05 2.03M 2.38B 2.965 $11.00B Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
SLR Solectron Corp (Delware) 5.61 6.72M 4.71B 3.189 $11.57B Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
PCS Sprint PCS Group 9.54 13.33M 9.89B 2.933 $12.69B Quote, Chart, News, Profile, Reports, Research, Msgs, Insider, Analyst Ratings
SUNW Sun Microsystems Inc 4.55 49.40M 14.96B 2.646 $11.20B Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
TIBX TIBCO Software Inc 8.88 3.50M 1.76B 3.031 $274.96M Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
TCP Telesp Celular Participacoes SA 8.94 1.12M 4.19B 2.863 $1.77B Quote, Chart, News, Profile, Reports, Research, Msgs, Financials
TLAB Tellabs Inc 9.00 4.65M 3.74B 2.542 $980.40M Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
AES The AES Corp 8.37 2.25M 5.26B 2.088 $8.41B Quote, Chart, News, Profile, Reports, Research, SEC, Msgs, Insider, Financials, Analyst Ratings
UCOMA UnitedGlobalCom Inc
------- End of Text File -------
Ugly! I know. However the text file is not meant for me to read, so I don't particularly care what it looks like. The text file's sole purpose is to give me an easy way to get the ticker symbols into AI 3.0. Note that each symbol is on its own line followed by a space and then a bunch of other information.
The AI 3.0 import function will retrieve the symbol from the line and ignore everything else on that line.
Next I import the symbols into an AI 3.0 Research Portfolio. This is done with just 2 mouse clicks. The following screen shot shows the ticker symbols imported into a research portfolio on the Historical Analyzer screen.
I then select the desired Model, the Start and End Dates, the Original Investment amount, the Commission and then click the Run Analysis button.
After the analysis completes the results show up in the Report field. Note that AI 3.0 returned $1,967,887.19 compared to Buy and Hold's $635,761.52 over the 10 year period.
Specific details for each stock analyzed can be viewed by scrolling down the report.
I can also make notes directly on the report and then save it to a file for later viewing or printing.
The entire process took me a little less than 5 minutes (including the time it took to process all of the ticker symbols), in fact I spent more time writing this description than actually analyzing the stocks.
I'm also able to save the ticker symbols as a new Research Portfolio so that if I want to run another historical analysis (or use the Fundamental Analyzer or the Asset Allocator) I don't have to go through the import process again. I just load the Research Portfolio and I'm ready to go.
As you can see, using a stock screening tool and AI 3.0, you can quickly and easily determine which stocks and funds are the best places in which to invest your money.
Hi Jack,
Yes, Yahoo! recently changed their quote format again and this has caused problems with the AI 2.0 historical analyzer and the optimizer. It's also caused Screen-O-Matic to break.
However AI 3.0 uses the new format so if you use AI 3.0, it will work correctly.
Patches for AI 2.0 and SOM will be available in the next few weeks. I'll post a note here when they're ready.
Hi uptime,
For complete details, see http://www.automaticinvestor.com/upgradecenter.html
It explains what you need to know. If you have any questions, please send them to the AI support desk (email address is on the web page listed above).
Automatic Investor 3.0 is now available.
Free trials are waiting here --> http://www.automaticinvestor.com
Automatic Investor 3.0 is now available.
http://www.automaticinvestor.com
I just finished writing a new book called the Pragmatic Investor. It has absolutely nothing to do with AIM, but I think it will be useful to many people and I think it's one of my better efforts.
If you're interested take a look here --> http://www.pragmaticinvestor.com
Hi Matt,
The trades were executed using the closing price for that day, so your point is valid about not being able to necessarily fill at that price the next trading day.
However there are two reasons why I don't think it matters for these tests.
1) Not all fills would have been at a worse price.
Since stocks go up, over time, more than they go down (at least the good ones), I'd expect that if we did this long enough, the fills would actually be better on average, on the sell side, rather than worse. They'd be worse on the buy side, on average.
So the results should actually be somewhat better if there were more sales and somewhat worse if there were more purchases.
But with an increasing market over time, the sales should outnumber the purchases. With longer update periods the same concept would apply, but with less frequency.
2) The difference in closing prices and opening prices are not large enough to account for the differences in results between daily, weekly and monthly updates.
Hi Steve,
I found a couple of posts including http://www.investorshub.com/boards/read_msg.asp?message_id=283024 and http://www.investorshub.com/boards/replies.asp?msg=283188
They're over two years old, I didn't realize how much time had passed since I ran those tests. I don't see any figures, but perhaps they're somewhere on the SI board. I'll look for them there when I have a chance.
Hi Steve,
"I believe Mark H. (aptus) of Automatic Investor has done thousands of these simulations, and determined that Daily Checkups most often perform better than weekly, monthly or quarterly."
Yes, that's correct. The vast majority of AIM-friendly stocks performed better with daily updates (most non-AIM-friendly, i.e. low volatility stocks, also did better with daily updates, however I didn't run as many tests on those).
Also, Weekly usually beat Monthly, but I didn't test quarterly updates. Tests were done using closing prices.
There were some situations where monthly updates were better, but these were confined to situations where prices fell quickly over short periods of time (in these cases daily updates usually ran out of cash before monthly updates, but in a real portfolio this can be mitigated using some of the cash management techniques discussed on this board).
For long-term outlooks, shorter update periods significantly outperformed longer update periods. This makes sense as the shorter update periods are able to capture more of the volatility that longer update periods miss.
BTW, if you own AI 2.0 (or the trial version) you can do your own tests using the multi-security historical analyzer. Set up a list of ticker symbols and then run the analyzer three times -- once using daily updates, once for weekly and once for monthly. Save each run result and compare for yourself.
Hi TF,
"That is why you do not recommend it for individual stocks I guess"
I think you may be confusing two separate topics. One is portfolio optimization (or diversification or asset allocation) using MPT. That's when I don't recommend individual stocks.
The other is parameter optimization. For example, finding the best Buy and Sell Resistance settings for a stock with particular characteristics. That's what I was talking about when I mentioned optimizing for number of shares or PC.
When we optimize parameters (or as AI calls it, build models), we're trying to find a set of parameters that provide the best returns for a stock with a particular set of characteristics.
One important characteristic is the volatility. We don't try to predict when a stock will move up or down, rather our thinking is that this stock, based on its history, will move up and/or down with a certain volatility at some point in the future.
When it does we react to it by buying or selling as appropriate. So unlike TA, we're not predicting the when.
An example might clarify. Let's say I'm a longbow hunter and I'm waiting for a deer. However I only have a 20 foot gap between two cliffs. Based on historical viewing, I know a deer will run across this gap at some point in the future. When it does, I need to shoot immediately or I'll miss my opportunity because the deer is in the gap for only a second or two.
If I have a longbow and I'm waiting for a deer to appear I need to predict when that deer will appear. The reason is that I can't pull my bowstring back and hold it there indefinitely because my arms will tire.
So I need to make a predication on when the deer will appear and then pull back my bow string and ready my arrow. If I'm right and the deer does appear, I'm ready to shoot.
If I'm wrong and the deer doesn't appear, my arms will tire and I'll have to relax my bow string.
If my arrow isn't ready when the deer appears I'll miss my shot because it will take me too long to pull back the bow string and shoot. The deer will be out of the gap before I can do that.
In essence I need to predict when the deer will appear and I need to be right within a very small tolerance.
However if I'm armed with a crossbow, rather than a longbow, the situation changes.
I don't need to predict when the deer will appear anymore. Rather I set my crossbow and wait. When the deer appears, I react by firing. My arms don't get tired holding back the bow string.
And that is essentially the difference between AIM and TA. AIM reacts when the price moves. TA predicts when the price will move. AIM is ready and pulls the trigger when the correct event occurs. TA tries to predict when to pull the trigger. If the event doesn't occur when predicted, TA fails.
Back to Model building...
When I build a model I'm not trying to predict when an event will occur. I'm assuming the event will occur at some undetermined time and I'm waiting to pull the trigger. If the event doesn't occur for 2 months, that's okay. I'll wait until it does. Whenever it does, I'm ready to react.
So I use the historical volatility in a stock as a guide to setting my parameters. Stocks that have certain volatility characteristics will work better with a certain model while others with different volatility characteristics will work better with different models.
Lichello did the same thing. Except he tried to find a model that would work well with all stocks. Given that he didn't have a computer and did all his work with a pencil, he did very well. The models he found work reasonably well for many stocks.
However there are other models that work much better for stocks with different characteristics. Thus we try to place stocks with similar characteristics into groups.
Then we build models for those groups, test them and if they're robust enough, use them.
We are actually predicting that a stock's characteristics will be similar to what it was in the past, so even with AIM we are predicting something (if we want to wring out the best performance from our stocks).
But this usually works because a stock that has been quite volatile in the past tends to continue that way until something changes its behaviour.
So it's much easier to predict volatility than when a price will move up or down.
"With AIM you do not care which way it goes.... you REACT to which way it goes."
That's true, but with AIM you do care that it is volatile. You can react only if your chosen stock has that particular characteristic. If it's not volatile enough, you will not react.
"So you are optimizing for something that probably will not happen."
Not at all. You are optimizing for something that has a very high probability of happening. However you don't know when it will happen. And that's the difference.
If you were optimizing based on when the event would happen, you'd be trying to predict the market, which you can't do consistently.
One last example...
This is real, so if you want to take me up on it just let me know
Following is a list of events. Let's say I have $100,000 that I'm willing to bet in certain circumstances. If the event occurs, I win your $100K, if it doesn't you win my $100K.
Here are the events...
1) IBM's closing price will be higher in the next trading day? I don't know, so I'll decline the bet.
2) IBM's closing price will be lower in the next trading day? I have no idea, so again I'll decline.
3) Over the next year there will be at least one time when IBM's price will close lower than the preceeding trading day's closing price? That's very probable. In fact it's so probable that I will take that bet.
4) Over the next year there will be at least one time when IBM's price will close higher than the preceeding trading day's closing price? Same answer as above. I'll take that bet.
And that's the difference between predicting price movement TA style and predicting volatility AIM style.
And that's also why you stand a much better chance at building robust models for AIM than for using historical price movement to predict future price movement.
Hi TF,
Although we are, in the end, trying to maximize the portfolio value, sometimes AIM will give us the best portfolio value in the short term at the expense of long term performance.
This is similar to short term traders who try to follow ST trends and make more money relative to an investor with a long-term view, but only in the short-term.
In the long-run, the investor with the long-term view usually ends up doing better. Actively Managed mutual funds have this problem. There is a huge incentive for fund managers to make their performance look good in the short-term (e.g. every year), but that penalizes them in the long run (however they usually don't care because they're compensated on short-term performance).
So AIM might go mostly to cash when the stock is down (if we're optimizaing for Portfolio Value) and therefore have a greater portfolio value over the optimizing period than B&H. However B&H might have significantly more shares (I've seen this happen many times in my historical tests).
But when the share price rises again, B&H's portfolio value would outpace AIM's. If we took a longer term view and optimized for number of shares, then we might hold more shares than B&H when the price was down. Our portfolio value might not be as large as if we optimized for Portfolio Value at that particular moment, but when the share price inevitably rose again, we'd be better off.
According to Tom, a similar line of thinking can be applied to maximizing PC. I think the concept is a sound one.
However I've not done enough backtesting to know when (and under what conditions) optimizing for number of shares or PC is better. But Tom seems to have done quite a bit of work on his study and his conclusion is that optimizing for PC is worthwhile.
I hope that once AI 3.0 is released others will try these types of things and add to the body of information. Until now, it's been onerous (and took a long time) to optimize anything, let alone optimize for specific things.
With AI 3.0 I hope to change that. Now people can take 10 minutes and optimize in ways they couldn't before. This will lead to new ideas and perhaps a better way to do things.
I think if people are given the tools to easily do difficult things, they'll come up with improvements.
Hi Tom,
You convinced me that optimizing for PC is a good thing. So I've added it to the AI 3.0 Genetic Optimizer.
After running a few tests, the results do appear better than optimizing for Number of Shares. The Buys and Sells appear to be more efficient too (at least after running thorougly exhaustive, 100% scientific and statistically sound tests of 5 runs
Here's a screenshot of the modified Genetic Optimizer...
Hi again LC,
"So cash should have a negative return due to inflation. Inflation is making the future value of the cash worth less in present value dollars."
Yes, but cash is also earning interest. So if the interest it earns is equal to or exceeds inflation, then purchasing power is preserved. Or am I missing something in your statement?
Hi LC,
Thanks for the clarification. That's an interesting concept. Rather than choose your own risk level, let the IW choose it for you.
It incorporates the IW at the MPT level.
I like the idea!
Hi TF,
"Is there a site where I can see the correlation of these ETF's too each other?"
I'm not aware of one. However keep in mind that even if you do find a site (perhaps someone can jump in with a URL), if you're AIMing the ETFs you need to use the correlations for the AIM results, NOT the underlying ETF results.
BTW, for the most part your investment portfolio is exactly what I envision people should hold if they want to make a reasonable return with moderate risk.
"But compared to the two mutual funds I owned (Fidelity Magellan and Stein Roe Young Investors)I did much better."
I'm with you on this one. I am generally against actively managed mutual funds. See http://www.automaticinvestor.com/articles/mutualfunds.html for the reasons.
Hi LC,
"To me that includes the cash"
Yes, you are right. Cash is important. Take a look at the chart below...
Once you find the best portfolio on the Efficient Frontier, you then add cash to give you your desired risk level.
"...and what you think the risk and return of that investment is..."
When using index funds you can generally use historical data to do this.
"For cash I would use a negative return of what ever you think the inflation rate is. I think I would then compare the cash percentage to what the idiot wave is telling you to start up with. I would then adjust the the MPT over all risk until the two cash percentages are close. That will give you the starting percentages. You would then let aim take over."
I'm not sure I follow this line of thinking, pehaps an example would clarify.
Hi TF,
"I know buying the ETF's last year worked very well for me."
Buying almost anything last year would have worked out well because most things (even individual equities) went up.
But just because someone does well in a rising market doesn't mean he or she should ignore sound asset allocation theory because there will come a time when parts of the market will go down even while other parts are going up.
When that happens a wise asset allocation policy will allow for efficiently buying low and selling high (via rebalancing).
The fact that you bought ETFs probably means that you are already somewhat diversified. It sounds like you're not using any specific strategy to diversify other than purchasing your ETFs. Which for most people is just fine.
My feeling, however, is that if you look at your portfolio as a whole and view the interactions between the individual ETFs as an important piece of the puzzle, rather than simply viewing each ETF as a separate, discrete component that doesn't have anything to do with the other components, you can diversify in a better way. That's the premise of MPT.
Whether you do this or not, it sounds like your investment strategy is better than most people's anyway, it's just that I think you could do a little better with MPT and your ETFs.
Hi TF,
"I want a portfolio with a 100% return and a 0% standard deviation.... how do I choose that?"
That portfolio is not attainable. Portfolios above the Efficient Frontier are unattainable. Portfolios below the EF are inefficient and portfolios on the EF are the ones you want to choose.
The reason I state that MPT should not be used with individual equities is exactly because past results do not ensure future performance.
It's the same with index funds and sector ETFs, however it is much easier to get reasonable Expected Returns, SDs and covariances with index funds than with individual equities.
If you use a good enough sample of historical data you should still be able to take advantage of MPT's ability to diversify more efficiently than by just using an ad-hoc method.
One of the nice things about MPT is that there's substantial slack allowed. Things aren't measured to decimal place accuracy. Even being 10% off on some of the allocations can still give you benefits.
When I say, "simply choose," I'm not implying that what happened in the past is guaranteed to happen in the future, rather I'm saying this is the best tool available to diversify. It's not perfect, but it beats anything else out there. So you "simply choose" the return you're after and let MPT tell you how to allocate your assets.
You might not get that return, you might get more, you might get less, but over the long-run you should do better than if you simply took a wild guess at your asset allocations, ignoring correlation and SDs.
"I agree that diversifying is desireable but I wonder if you can expect any particular results."
I know diversifying is a good thing and I believe you can expect better results using MPT with index funds and sector ETFs than by diversifying by the seat-of-your-pants.
Hi Tom,
"If we know we're going to use a risk modifier like AIM, how do we know where along the MPT curve we should be selecting our portfolio?"
One of the reasons I like the MPT concept is that it is very flexible. We don't tell the MPT alogorithm how we're managing our equities.
We can manage some using the Buy and Hold method, others using AIM and still others using a short term trading program.
The only thing MPT needs are the returns, standard deviations (SD) and covariances (or correlation coefficients).
How we actually get those values (whether we used AIM or not), doesn't matter to MPT.
So, we choose our portfolio from the Efficient Frontier the same regardless of the underlying equity management method.
If we're looking for a 10% return we simply choose the portfolio that gives us the 10% return and MPT will automatically give us the portfolio with the lowest SD (it might be 8%, but it might not be).
If we're willing to accept a maximum of 8% SD, then we select that portfolio and MPT will give use the highest expected return for that 8% SD.
There will, most likely, be differences in the Efficient Frontier curve if we choose to manage our equities with AIM (rather than using B&H) because the very fact we use AIM will change the Expected Return and SD values (hopefully lower SD and/or increase Return). It will also change the Covariances.
But the MPT curve will be shifted appropriately without you having to "tell" it you used AIM. Then if you want a 10% return, you'll get a portfolio allocation that gives you a lower SD than if you selected the 10% return from an unshifted MPT curve (the unshifted curve might result from managing your equities with the B&H strategy).
The advantage to using AIM is that we can invest in riskier equities than we normally would. The reason can be seen as follows.
If we had an equity that has an expected return of 12% with a SD of 12% (using the B&H strategy) and we manage it with AIM, the actual return might be 13% with a SD of 9%.
If we have other such securities in our portfolio and AIM improves their returns and/or lowers their risks in a similar fashion, then when we feed these equities into the MPT algorithm the EF curve will be shifted up and to the left.
This directly translates into us being able to choose a risk level with which we're comfortable but the resulting portfolio might contain more of the riskier equities than we'd normally choose (had we not managed them with AIM).
After saying all that I need to add my usual caveat, that I forgot to add in my last response. MPT should not be used to allocate individual stocks (I've explained the reasoning for this in the past so I'll refrain from repeating it here).
However it can be used fairly successfully on index funds and sector ETFs if you use a long enough historical data period to obtain the expected returns, SDs and covariances.
Hi Tom,
That is interesting stuff. Since Conrad brought up the number of shares optimization I've been interested in seeing how well it works. But it looks like you got there first and then some
So do you think that optimizing for maximum PC is better than optimizing for maximum shares in the general case, or just in some specific cases?
I wish I had more time to run some tests (the Genetic Optimizer in AI 3.0 makes optimization almost fun), but with AI 3.0 scheduled for release next month plus two other projects I'm working on, these tests will have to wait (but it sounds like if I wait long enough you'll have discovered all the answers
Don Carlson has sent me some interesting notes that I'd also like to analyze, but again there's that time thing (I need to either retire or hire some AI elves at the Automatic Investor Factory!)
If you do update your study, please post the results. It sounds like I might need to add a few more options to the Genetic Optimizer (i.e. Optimize for maximum PC, etc.)
Hi Tom,
When using AIM with MPT it's important to separate the two. I do this separation by defining the AIM view as the Micro view and the MPT view as the Macro view.
The Micro and Macro views are independent. So MPT does not know about AIM's cash reserves and AIM portfolios do not know about MPT.
Therefore it's not actually AIM's cash reserve, as such, that is shifting the Efficient Frontier (EF). Rather it's AIM's management of the portfolio.
For example, if I have a stock, stock A, and I find that, over some period, its expected return (E) is 10% and its Standard Deviation (SD) is 8%, and I combine it with another stock, stock B, with its own E and SD, then I can use MPT to show me an EF for those two stocks.
I'd then choose the portfolio, on that EF, that best suits my Reward/Risk needs.
If I now use AIM to manage stock A, I might find that the AIM results of stock A are not E = 12% and SD = 6%. In essence AIM has used its cash reserve and algorithm to increase the expected return and decrease the SD.
So when I use MPT to calculate the EF for my AIM managed stock A and my non-AIM manged stock B, I'll find that the EF has been shifted upwards (i.e. higher E) and to the left (i.e. lower SD).
If I then use AIM to manage stock B and it gives better results (i.e. lower SD and higher E) than non-AIM managed stock B results, then when I use MPT to calculate the EF for AIM-managed stock A and AIM-managed stock B, I'll find the EF has shifted up and to the left again.
However MPT didn't actually use AIM's cash directly to achieve this shift. Rather AIM used the cash to minimize the risk (and/or increase E) at the Micro-level and MPT then used the AIM E and SD values to create the better EF.
Now since MPT is independent of the underlying securities, we can view it at the Macro level without needing to know whether the underlying securities are AIM-managed or not (all we care about are the E and SD values for each security in our portfolios).
Therefore when we add risk-free cash to the MPT model, we get something like the following...
What happens is we first determine our optimal portfolio from the EF and then add cash to the mix. This moves our portfolio along the straight line (rather than along the EF curve).
So when we add cash, we reduce risk (possibly to a lower value than we could get without cash), however we also reduce the expected return. Note that as we move along the straight line to the left, there are portfolios on that line that have a better return for the same SD, compared to the portfolios on the EF (so adding a little bit of cash can actually increase your returns and reduce your risk, or a least hold it constant).
But if we add too much cash, we then start to reduce our returns.
On the other side of the coin, if we use margin (i.e. negative cash), we increase risk, but also increase E (again, we move off the EF curve and follow the straight line). Note again that using a little bit of margin can increase our returns for the same level of risk compared to portfolios on the EF. But if we use too much margin, our risk level goes up.
The beauty of MPT is that you can use it with any investment method (not just AIM). However when using it with AIM, you're generally able to get better EFs because AIM serves to inherently reduce risk.
Hi Pete,
There's an article at the Motley Fool describing 5 stocks under $10. It's here --> http://www.fool.com/news/commentary/2004/commentary040227RAM.htm
Here's an article from Yahoo! on RSS...
NEW YORK - E-mail is crippled, concussed by an irrepressible spam stream. Web surfing can be equally confounding, a wobbly wade through bursts of pop-ups and loudmouthed video ads.
And that may explain the excitement these days over a somewhat crude but nifty software tool that automatically delivers updated information to your computer directly from your favorite Web sites.
Enthusiasts see these Web feeds as sketching the outline of the next Net revolution.
The technology behind them is called RSS and I rely on it daily to consult The New York Times, the BBC, CNET News, Slashdot and a few dozen other Web sites that employ RSS to make the very latest news stories or bits of commentary available for the plucking.
Aided by software on my computer that goes out and retrieves my feeds, I swiftly sort through headlines and summaries. By clicking on included hyperlinks, I can visit originating sites for more detail.
"For an average Internet user who regularly visits about 50 Web sites, rather than have to go visit those 50 sites wouldn't it be cool if those sites could somehow visit you? And not only that, but if they could also tell you when they've changed?" said Greg Reinacker, head of NewsGator, which sells an add-on for Microsoft's Outlook e-mail client that offers one leading way to read feeds.
Hundreds of thousands of Web feeds are available, spurred by the popularity of Web logs, which account for their bulk. One site that has been sorting feeds since 2001, Syndicat8.com, added 7,326 in January — its biggest monthly jump — to its collection of more than 53,000 information streams.
Some of that upsurge was election year fever as Democratic presidential candidates led by Howard Dean (news - web sites) daily turned on the RSS spigot to "broadcast" to supporters.
But Web feeds are no Howard-come-lately. Info generators of all kinds — big media, government and non-profits alike — are embracing them.
Disney leverages the technology to deliver video clips for ESPN.com and ABCNews.com. Apple's iTunes generates a feed to alert subscribers to its latest sounds.
Anyone who builds a Web site can incorporate Web feeds. If it lives on the Web, it can be brought to your desktop — or to your wireless device, for that matter.
Human Rights Watch keeps activists current with feeds sorted by region. The U.S. Geological Survey (news - web sites)'s feeds let seismologists immediately know where the world is shaking.
The U.S. Product Safety Commission just began providing recall notices via RSS. General Motors offers feeds on topics including safety and automotive tech. And a growing number of companies use feeds to disseminate info internally.
"If you're not reading it in RSS you're wasting your time," declaimed Microsoft's blogging evangelist, Robert Scoble, who says he subscribes to nearly 1,300 feeds.
RSS has been called the TiVo (news - web sites) of the Web, the first "killer app" of the anticipated automation of social and commercial transactions online using the Web's second-generation XML (extensible markup language) standard.
Alas, you'll not find the tools for handling RSS in your Microsoft Windows operating system. Not yet, anyway.
You've got to go out and get them, just like you had to download Netscape or one of its competitors in 1994 when you wanted a Web browser.
But the writing is on the wall. And it's not graffiti; the feeds are spam-free — though advertising may be pumped through some eventually.
Yahoo and Google recently embraced Web feeds, and Microsoft is expected to incorporate tools for managing them in its next-generation operating system, code-named Longhorn.
Yahoo's new search engine trolls through RSS feeds in addition to Web pages. And a five-person company called Feedster.com is trying to build a business around customizing searches of 500,000 feeds — and then delivering you the search results in a single feed.
RSS feeds vary in length and capability. Depending on how a Web site decides to serve them up and how a given aggregator wants to organize them, feeds can be simple, spare text or bold and multimedia-flashy.
And that's what makes them both exciting and frustrating.
First, it's not simple for the non-techie to configure RSS. If they're obvious on a Web page, the feeds generally are offered as orange buttons that read "XML" or "RSS." There's no uniformity to feeds, though the best include a good headline and a succinct summary. You can choose to have feeds delivered to your desktop or gathered by a Web-based service.
"It can be really hard to get people to look at it. I tried to get my father, who is a news junkie, to look at it and he wouldn't," conceded blogging guru Dave Winer, who created the Web-based aggregator Radio Userland.
Programmers who've developed rival versions of RSS since its 1999 invention — primarily by Winer and folks at Netscape — can't agree on what RSS is supposed to stand for. Winer's preference is Really Simple Syndication (RDF Site Summary and Rich Site Summary are the other options).
At least it's nothing like the fiasco of 1997 known as "push technology" and incarnate in PointCast, which wrote its death warrant by clogging hard drives and crashing operating systems as it delivered updated information to subscribers.
RSS is more pull than push. Your aggregator retrieves the updated material from the feed-offering Web site at set time intervals.
For an introduction, My.Yahoo.com offers a dumbed-down beta version. Web-based aggregators including FastBuzz.com and Bloglines.com are popular because there's no software to download — and they're free.
FeedDemon, a downloadable cross between an e-mail client and a Web browser, is feature-packed and costs $30. NetNewsWire for the Mac, also a download, costs $40.
If only the RSS prophets would stop squabbling.
Winer is among those who consider the standard complete; others insist it must become more versatile if it's to be an engine of the next-generation Internet — a smarter, two-way street rather than just a blind delivery vehicle.
Anil Dash, vice president of business development for Six Apart, whose Movable Type is among the Web's leading blogging products, says RSS is broken. He promotes a more robust and flexible alternative called Atom that got a big boost when Blogger.com, Google's blogging service, began supporting it in January.
As with most technologies, the market will settle these scores. But first, the market itself has to develop.
Major content providers want to ensure that any feeds they offer drive traffic back to their Web sites.
"The benefit to us is we're distributing our headlines and the users come back to the site," said Catherine Levene, vice president for business development at New York Times Digital, which has been quietly offering feeds for two years.
Many RSS-watchers predict Web feeds will eventually morph into ad-delivery vehicles because it can be expensive to run a Web site that serves up hundreds of thousands of feeds daily, draining bandwidth.
Nevertheless, boosters like Jeremy Zawodny, a software engineer at Yahoo who promoted RSS feeds there, are convinced that 2004 will be the year the technology goes mainstream.
"Remember when you first starting seeing URLs appear on billboards and at the end of movie trailers?" Zawodny wrote in his blog in December. "It's going to be like that. One day we're just going to look around and realize that RSS is popping up all over the place. And a couple years later, we'll all wonder how we ever got along without it."
For Registered Automatic Investor 2.0 Users Only!
If you're a registered Automatic Investor 2.0 user, here's how you can upgrade to Automatic Investor 3.0 --> http://www.automaticinvestor.com/upgradeorder.html
Screen shots of the redesigned Historical Analyzer.
Here's the screen you use to analyze single securities...
When the analysis completes you can view the Trade recommendations...
and view the Price Chart (with buy and sell indicators)...
Or view the Portfolio Chart...
You can also analyze multiple securities at once using the Multiple Security Historical Analyzer...
It ties in seamlessly with the Research Portfolios you've already created and allows you to save the results in a report that can be viewed in WordPad or Microsoft Word.
Hello again Tom,
"I guess I'll have to wait for AI 3.0 to come out to try this test!"
AI 3.0 is currently being Beta Tested. I had stopped accepting Beta testers because I find any more than 5 takes too much of my time (and I seem to have less and less each day
Also I've found that 5 good people find all the major bugs and most of the minor ones AND they come up with excellent suggestions and ideas.
However if you'd like a Beta copy I will be very happy to send you one. You don't have to Beta test it (of course if you'd like to then that would be great) but you can use it to run whatever tests you'd like.
I'm finding the Fundamental Analyzer and the Genetic Optimizer to be the most useful tools.
Let me know and I'll send you an email with the super secret Beta URL
Hi Tom,
I'll look for your note. It's nice to know I'm not breaking new ground and I can start from an established base. That's a much easier way to go.
Hi Conrad,
If you design software correctly, things like this are easy to do. By separting the GUI from the behind the scenes functionality (and creating a different class for each piece of the behind the scenes functionality) you're able to change the GUI and/or the functionality quite easily.
The Genetic Optimizer (GO) uses the AIM engine class and that class was already storing number of shares as part of its function (since it was needed to calculate Portfolio Value).
The GO makes its decisions based on a Fitness function.
Sooooooo, all I had to do was use a different Fitness function (one that maximized number of shares rather than Portfolio Value) to make this work.
I could do the same thing for any number of attributes or AIM parameters (I just need to write a different Fitness function for whatever I want to maximize and pass it to the Genetic Optimizer) -- but I think I'll leave things as they are for now (at least until AI 3.0 hits the streets).
I've only done some cursory tests on optimizing for number of shares and I don't expect to do much more in the near future. However once AI 3.0 is released, perhaps others will post their thoughts.
Eventually I will get back to the Number of Shares idea though.
Hi Conrad,
Your optimize for number of shares idea was so intriguing that I couldn't wait until after releasing AI 3.0 to try it.
So I did and found the results very good. That led directly to me adding this functionality to the Genetic Optimizer. So now AI 3.0 users can optimize by Portfolio Value or by Number of Shares.
Thanks again for the idea. Now here's a screen shot of the enhancement...
Thanks for the ideas Tom.
With the Genetic Optimizer it is easy to optimize for different things (by simply changing the Fitness function), so we could conceivably give the user a choice of what to optimize for.
Once AI 3.0 is out the door I'll play around with this idea.
Hi Conrad,
"I have to keep doing my optimizations manually!"
Ouch! That must have stopped being fun a long time ago
To answer your questions...
1) The Buy and Sell Resistance settings are equivalent to Tom Veale's split SAFE. AIMers do not use Resistance AND SAFE, because they are the same thing.
2) That is a very good idea. I've often thought about what would happen if the PC was adjusted after a Sale, but it didn't get out of the pure thought phase.
I think that you are correct, a SPCF genetically optimized would probably not be 0. That is something that I will have to look into at some point in the future. Thank you for the idea.
3) You can infer that the target of the optimization is to maximize the Portfolio value by looking at the ranking in the Genetic Optimizer display. And in fact it is just that. However I had not thought of maximizing the number of shares. But this is another good idea (you're really on tonight For a good stock, the probability is strongly in favour of it rising over the long-term, so having more shares would be a long-term benefit.
If we use the B&H share quantity as the benchmark, then that can give us the "return" so to speak.
Here's another idea. If we use AIM to maximize the number of shares and then modify the algorithm to never go below the initial B&H number (once it gets above it of course, while it's initially below we would AIM normally), we would always beat B&H as soon as AIM had more shares. This would somewhat loosely mimic LD-AIM in that although we would have a heap of shares, they could be counted as virtual shares since they wouldn't be sold.
We couldn't start out with 0% cash, but with a sufficiently volatile stock even 20% cash might do the trick.
I must say that you certainly have some good ideas. How is Vortex coming along? The last I heard you had put it on the back burner because of time constraints.
I hope you dust it off and continue developing it as I think it has some very good concepts. One thing that I'll be working on after AI 3.0 is released is to move AI over to the .NET framework.
As part of this effort I'm building a general framework into which you can plug in various engines. One engine will be AIM (of course), another will be HUSKY and another might be LD-AIM. There's also a short-term trading concept I've developed that will be pluggable into the framework.
The GUI will be configured using XML configuration files. So a typical plugin will consist of an engine, a database and one or more XML config files.
All the generic functionality (checkpoints, asset allocation, fundamental analysis, charts and such) will be completely reusable.
Therefore if you're interested, I'd be willing to explore a Vortex plugin at that time. However this is at least a year away, but it's something to think about.
Hi Tom,
Yes, I like the chart feature. As they say, a picture is worth a thousand words
And one more screen shot...
This is a screen shot of the redesigned Portfolio Manager...
Note that you can double-click on an entry in the list to move to that portfolio.
You can also click on the charts to expand them (and see more detail).
Hi Keith,
No, I didn't notice the split (I own about 10 stocks total, so I usually follow them very closely and have some watch lists set up for others I might be interested in).
As I mentioned previously I purchased PLNR a few days ago. I think it will make an excellent AIM stock and the fundamentals are top notch.
I usually run the report for my stocks and my watch lists every day or two (since it doesn't take too long). As prices change the scores also change, so I like to see how my stocks are doing in that respect. However you can certainly run it once a quarter if you'd like.
I've posted a preview of some additional AI 3.0 screen shots here --> http://www.investorshub.com/boards/read_msg.asp?message_id=2429690
It covers the Fundamental Analyzer, the MPT Asset Allocator, the Genetic Parameter Optimizer and the RSS News Reader.
AI 3.0 marks a huge improvement over AI 1.0 which was released in August 2000. I think it's now the complete tool for AIM investors, but it's not done yet.
I have a rather long list of enhancements that were suggested by AI users (as well as some I personally want to add). So much to do, so little time
Automatic Investor 3.0 Screen shots Preview part IV
Everyone likes to keep up to date with the News. But sometimes it's a hassle going to all your favorite News sites looking for interesting articles.
That's where RSS comes in. Do a Google search on "RSS" to learn more. It's fast becoming a major phenomenon on the Web and if you haven't heard about it, you will soon.
Most of the large sites (including Yahoo!, the Motley Fool, Reuters and such) have RSS News Feeds. You can subscribe to any News Feed of interest and the news from that feed will be delivered right to your desktop.
And you can do all this from within Automatic Investor 3.0.
Here's what the AI news reader looks like...
Once you've started reading News this way, you won't want to go back to the old way again.
That's the end of the AI 3.0 preview tour. If you're a current AI 2.0 user, you'll receive an email telling you when the AI 3.0 upgrade is available.
If you're not a current AI user, please check the AI website over the coming weeks for details.
As usual, questions are always welcome.
Automatic Investor 3.0 Screen shots Preview part III
AI contains two built-in Parameter Optimizers: Brute-Force and Genetic.
Once parameters are optimized they can be saved as a Model for later reuse. Models can also be exported and imported so you can share AI models with other AI users.
The Brute-Force Optimizer is used when you want to find the absolute best parameters. However it can take a very long time depending on how many parameters you're trying to optimize simultaneously.
The Genetic Optimizer (based on Biology's Natural Selection) does the job much more quickly. Although it doesn't guarantee the absolute best parameters, it comes reasonably close. And you can optimize all of the important parameters at once.
There's also an added bonus in that you can select any of the top 100 Models.
Here's the Brute-Force Optimizer screen...
It functions the same way as in AI 2.0.
Here's the Genetic Optimizer screen...
Simply click the Optimize button to start the Optimization process.
When it completes, you'll see...
Note that the best portfolios are listed in descending order. You can double-click on the desired portfolio to view the details...
This is where you can also save your parameters as a Model if desired.
A price chart (with Buy and Sell indicators) is also available on the fourth Tab...
The AI and Buy and Hold portfolio values are also displayed at the top of the chart.
Finally we'll turn our attention to the RSS News Reader...