Where'd all the PCA users go? http://boards.fool.com/Message.asp?mid=16441200It would appear that the oldest and the newest software products are the most popular.Best regards, Tom
Hey Tom, I feel like the Lone Ranger with no Tonto. Yes I am the lone PCA guy.John (whitelake@SI)
Hi All,With this talk about which SW, I think it would be prudent to put the AIM algorithm first. It does not really matter which SW is used, or if SW is used at all. First and foremost the algorithm should be understood. In anything that requires or concerns money we should never do something because "The SW Said So". With a possible few exceptions, we all need from time to time a little bit of help in understanding just what AIM is and how it works. That is the big advantage of these boards, we can deepen our understanding and help each other along.Having that of my chest, please carry on and discuss whatever SW package you like :-)Best,Rien.(And keep in mind that SW is NEVER error free!)
Hi AIMers,I am new to the AIM community but have been an AIM user for many years. I started out with a simple Lichello AIM but modified the way of calculating buys and sells as well as the use of the Portfolio Control.At first I used my Apple Classic and an Excel 1.5 spreadsheet. Later I moved up the ladder and at last I used a Mac Centris 650 and Excel 4.0.(Old stuff but very functional). About 1,5 years ago I have moved sideways to a Pentium III platform. I must say Programming is not my strong point so I got to where I am with sweat, cursing, tears and sleepless nights.... but I understood everything did so I have a good feel for the underlying concepts of my AIM.With my Mac programs and equations I got a friend so far to make us a DOS Program called VORTEX. This has now matured into a Windows 95/98/02+program. VORTEX is an AIM that is strongly tailored to the way I think an AIM program should work but I have little experience as to what others AIM-users think a program should do. Via the website of Tom Veale I have come into contact with this website and I plan to learn what other AIMers think about various ways of using AIM concepts. Obviously I am also interested to know if my AIM software is of interest to other AIMers. I like to get a discussion going on the fundamental aspect of the Portfolio Control, and how various people use it in a different way than Lichello did, and how it is used to generate a buy/sell advise.I know many AIMers refer often back to Lichello and say: Mr. L says this. Mr L says that. For example: Let's start an AIM Portfolio worth "y" at y =1000=pc1. Then (pc1-y)=0. Fine! This states that no equity is to be acquired. So far so good. The price drops to 500 after a plane hits a skyscraper: pc1-y=500. This gives food for thought. In a neutral strategy equity @ 500 is acquired(I ignore the Lichello SAFE). y =1000 again and pc2=pc1+250=1250...and this is rather peculiar in a Lichello AIM. The test at the same stock rate at which equity is acquired gives a buy-signal!!!! The test pc2-y=250 and the advice generator advises to buy another 250 worth of stock...I am from Missouri....(Now I am confused. I thought I came from Holland !!!). So I do what the AIM instructs me to do, and this happens: dy=250...y= 1250...pc3=pc2+125= 1375--->Testing Testing----->pc3-y=125. Again a buy -advise!I know that with the use of the SAFE the buys are moderated but that does not eliminate the problem that each time I test for (pc-y) there is a buy-advice generated until the advice is zero! Every time I read about an AIMer adding 50% of the equity that was purchased to the pc(n) to get a new pc(n+1) I feel my throat being squeezed shut, just as happened when I read The Crucible of Henry Miller. How is it possible that the Lichello Flaw still prevails after about 30 years since the Lichello creation? Lichello never mentioned this flaw in his book(2nd or 3rd ed). In practice, if you do the Test only once in a while, after you noticed that prices had changed, you would never notice this flaw. You only notice it if you do the test at the same stock price as before. The Flaw is a serious one: A curious investor from Missouri would test for advise and will find that he is forced to keep buying equity without the stock prices having moved one bit. He would be wasting his money on provisions until he found an honest broker!Of course, some AIMers know this flaw. But do all AIMers know it? What do they do about it? How do they accept this? Simply by not doing the Advice Test if they know the stock has not moved yet? That would not be a satisfactory solution.I have analysed this problem over the years and concluded that although the Lichello ideas were novel they were certainly not the gospel for AIMers. I have created an AIM alternative and I call it the "Vortex Method". The Lichello Flaw is removed from this method and no matter at what moment I look at the portfolio and do the Vortex Test (pc-y), I get a zero buy/sell advice if the stock price has not changed. This is, I believe, among others, a strong feature of the program VORTEX.I am seeking feedback from AIMers that are interested in finding a solution to the Lichello Flaw. In the first instance we can discuss the problem to determine how serious it is for you. Secondly, interested AIMers can download VORTEX from the website www.vortexcw.nl and discuss the issue on this platform and give me feedback on VORTEX. In case you wish to contact me personally please use email@example.comRegards to all,Conrad Winkelman
Hello Vortex,The funny thing is, I always viewed what you refer to as a "flaw" as a feature. My view is that AIM will recommend you purchase equity when the price goes down. Now there are two ways to do this.1. Purchase all the equity you'll need at the lower price once, or2. time your purchases so that you don't purchase all the equity at once.AIM's mantra is to never purchase all equity at once (now I'm not 100% in agreement with this since backtesting results indicate that for many stocks starting with 100% equity is actually a good thing. However this is the AIM way of thinking, and after the initial investment, I do agree with it) and that's why you split your initial investment into two parts.Therefore it makes complete sense to time your purchases, after a stock price decline, so that buys are spread out. This follows the same logic as the initial investment (i.e. we'll hold off on investing everything in equity until we see what it does tomorrow -- or next week, or next month). Lichello refers to this as "pumping the brakes."When you add SAFE and minimum buys/sells to the mix, the reality is that you don't go on infinitely purchasing smaller and smaller amounts of a stock whose price hasn't moved. What you're really doing is waiting to see what the price will do before purchasing more shares. If it stays the same, you purchase a little more and wait again. Because the bias of the stock market over the long term is upwards (and AIM is meant to be a long-term strategy), it makes sense to purchase a finite amount of equity at the same price. AIM goes one step further and tries to not do it all at once.Of course this mechanism doesn't kick in until the equity value falls for the first time. So no matter how much we buy, we're still better off than the buy and hold position."Of course, some AIMers know this flaw. But do all AIMers know it? What do they do about it? How do they accept this? Simply by not doing the Advice Test if they know the stock has not moved yet? That would not be a satisfactory solution."Furthermore, most people have a fixed update interval (i.e. daily, weekly, monthly, quarterly, etc.) so they don't care about whether or not the equity value has changed. They simply update based on a predetermined time-period."I look at the portfolio and do the Vortex Test (pc-y), I get a zero buy/sell advice if the stock price has not changed." Having said that, I would be most interested in seeing your algorithm that gives zero buy/sell advice for non-changing prices.In the end theories are only as good as their results. The way to determine which method is better is to backtest a variety of stocks with varying characteristics over multiple time-periods, using both methods, and see which one wins. I haven't back tested your Vortex method, however I've run thousands of backtests using AIM and am quite comfortable with the algorithm (although I never use BTB)."Lichello never mentioned this flaw in his book(2nd or 3rd ed)."Lichello did mention this, however he didn't view it as a flaw. On page 57, of the 3rd edition, he states, "Sometimes when a declining market suddenly shrieks to a stop, AIM will pump its brakes to avoid going into a spin. Instead of curtailing our investments entirely, it will call for sharply decreasing investments for a month or two -- and then nothing -- just to pick up a few cheap shares in case the market turns around and heads upward again."In fact he says, "Of all the features we incorporated into AIM, this automatic check was the hardest to design and is the one of which we are most proud." So it appears that Lichello knew about the "flaw" you describe, but felt it not only wasn't a flaw, but was a huge benefit.At the present time, I tend to agree.regards,Mark.www.automaticinvestor.com
Conrad: I am seeking feedback from AIMers that are interested in finding a solution to the Lichello Flaw. In the first instance we can discuss the problem to determine how serious it is for you.OK, you call it the Lichello Flaw, others call it a feature. But I think Mark/aptus only points out your problem when he describes the 'flaw' as 'pumping the brakes'. The solution is easy with todays spreadsheets, but would complicate Lichello's pen and paper method a lot. That is why Lichello had to live with it and put the best face upon it by remarking that AIM continued to fish a bit at the bottom. At least, that is the impression I get from Mark. Your point is, that AIM continues fishing at exactly the same price (or even a price a bit higher!).Again, the solution is easy, if you feel the need. (Ignoring the flaw, and even welcoming its effects is a valid solution too, IMO.) The only thing you have to change is the buy advice. The buy advice no longer needs to reference the PC before the buy, but the PC after the buy. This points directly to the trouble Lichello would have faced: the PC after the buy is influenced by the buy, so the buy advice would need info from the buy advice, and the formula has to refer to its own outcome, to be solved by iteration. Today's spreadsheets take this in their stride, and a formula likeBuyAdvice = (PC+(BuyAdvice/2))-(ShareValue*(1+Safe))is no problem. In Excel 2000, I had to activate the iteration function to make this work, otherwise Excel refused to calculate the circular reference.Perhaps smarter minds than I will be able to remove the iteration with the help of calculus or other black magic. And of course, the pencil method still works too:1. Calculate the buy advice2. Calculate the PC after this buy.3. Calculate the second buy advice.4. Repeat steps 2 and 3 until you no longer get a buy advice.5. Add all buy advices and buy.In step 3, the minimum buy amount is no longer relevant (that condition was already met in step 1), but you could still use it. When you don't, you have to repeat steps 2 and 3 more times, until the advice drops below 1 share, or you get tired of adding ever smaller amounts of shares, whichever comes first.Regards,Karel
Hi Karel and Conrad,Concerning the "flaw". If you perceive it as a flaw there is an easy solution. I have posted it before, but it goes like this:1. Buy the stock according to AIM BTB advice.2. Add the number of stocks bought to the account but do not update the PC.3. Calculate the stock price were the next AIM BTB buy advice will be (with the unmodified PC).4. Increment the PC with half the buy value, as usual.5. Simulate the buy price you found in step 3 and check how many shares (or amount) you will need to buy at this price.Now you have a buy price, a volume and updated AIM parameters.No circular references needed.I have actually started doing this, as in the high volatility tech stocks the AIM buys would not cover enough price span before I ran out of cash.Best,Rien.BTW that makes three dutchies at this board :-)
Yes, that is a nice approximation, and should suffice in most cases. But easy? You only remove the circular reference by a two stage procedure (not quite an iteration, but close enough).But why avoid the circular reference? It works very well, and can be included easily in any standard AIM spreadsheet: it is a little addition to the formula in the cell calculating the buy advice. I don't think things get much easier than that. The spreadsheet then does the iteration, not you!Regards,Karel
Hi Karel,But why avoid the circular reference? It works very wellAre you by any chance a programmer? (I am)I have noticed that no two programmers hold the same view of what is easy and simple. :-) :-)Best,Rien.
HI CW, The "Flaw" is only a flaw if it causes a problem in the real world. In Mr. Lichello's book he refers to "pumping the brakes" when describing this event. He considers the first purchase to be the initial purchase at a certain price. With minor price changes, normally SAFE acts to inhibit the further purchase of shares at that same price. When a major drop in price happens, then SAFE is no longer able to contain AIM's enthusiasm in purchasing. Once the first purchase is made, AIM can actually purchase shares at a price HIGHER than the one just made! Again, Mr. L. says this is just AIM completing the purchase in a designated price range. I know that many of the new AIM users that come from short term trading methods attempt to soup up AIM to make it more active. This is because they are used to lots of activity, not because their account will necessarily benefit from it. Mr. Lichello assumed that a month would go by between trades. If the value had fallen enough to trigger a buy in December and now, after a month, the price was still down, then Mr. L thought it was prudent to complete the purchase at the same price. Again, Looking only at the Buy/Sell Advice column and not the Market Order column one could start feeling that there's a mathematical flaw in the system. If one worked only off the Advice column, AIM would snowball into uncontrolled buying on each cycle. Since this is a "new" thread (or should I say "newly active") on TMF, there's bound to be lots of questions. On Silicon Investor and also here on TMF I ask that historical examples of reasonable length be used to show the value of changes made to the AIM algorithm. Also changes made should be tested with both bull and bear markets represented. That way we can see what the changes are. I came up with a measure which I call "Return On Capital At Risk" (ROCAR) as a measure of performance. Some changes will increase "Total Return." Some changes will increase ROCAR. The changes that prove to be most valuable are the ones that improve Total Return with ROCAR improving or at least not declining. The idea is that if we're gaining in total return but ROCAR is declining, then we're just making the program more risky with the change. Best regards, Tom
Hallo Rien,No, I wouldn't call myself a programmer, but I do know something about programming. So you may be right about the 'easy' part.:-)Karel
Hi Karel, Rien & others,In reference to the Lichello Flaw: Whether or not it is a problem is obviously a personal matter. With regards to the fact that AIM is supposed to be an automatic investment manager I would not tolerate it and I have not allowed it in my program VORTEX. I am still interested however how others dealt with it. Karel wrote: Perhaps smarter minds than I will be able to remove the iteration with the help of calculus or other black magic. And of course, the pencil method still works too:1. Calculate the buy advice2. Calculate the PC after this buy.3. Calculate the second buy advice.4. Repeat steps 2 and 3 until you no longer get a buy advice.5. Add all buy advices and buy.In step 3, the minimum buy amount is no longer relevant (that condition was already met in step 1), but you could still use it. When you don't, you have to repeat steps 2 and 3 more times, until the advice drops below 1 share, or you get tired of adding ever smaller amounts of shares, whichever comes first.Regards,KarelOK Karel, you have hit upon the "essence" of the solution that I have implemented in VORTEX. Bravo!I have used White Magic though!The details of my solution are being prepared for making them available on my website but I will briefly summarise it here. They are already presented in my Dutch booklet "The Vortex Method". The idea you proposed is in fact an excellent answer to the problem but it would make the AIM more aggressive than intended. The uncorrected Accumulated Buy is about double the Lichello Buy Advice. I have a solution for that too! The VORTEX Solution (apart from the process of analysing and synthesizing) is in effect very simple.1 As Karel suggested I accumulate the respective buy-advices:Buy-Order = SUM( B1 + B2 + B3 +.....Bn)=SumB n----> oo (infinity)Note that B1 is the Lichello Buy Advise without the SAFE,and the Sum (Bi's) is the Residue Buy, and that I call "The Flaw"This results in Buy-Order = b*B1 where "b" is the unknown to be found.2 For each Bi I calculate each new pci with factor "f" instead of 0,5.The value of B2 etc. is dependant on the factor "f" obviously. The factor "f" may be freely chosen and dictates in fact how aggressive thebuy-strategy will be...you can already "see" that if f=0 the PC will not be updated at all and then pc2-pc1. Perfect! It will be clear later!The reasoning dictates that "b" will be a function of "f". This too islogical as well as desirable. The Final Buy-Order b*B1 will be "tempered" with the proper choice of "f"(Applying the Brakes).3 In the synthesis the unknown "b" is eliminated by substitution.Hence Sum (Bi's) is only a function of "f"4 The expression "SUMB" reduces to B1*( 1 + F1 + F2 +F3+....Fn) n---->ooand this infinite series function reduces to a simple function of "f"if "f" is limited from 0 to 1 then SumB= F(f)*B1.5 The end-result is then, for the initial condition pc1=y1 and after a portfolio value change from y1--->y2B1=(pc1-y2) is the "Lichello Buy" without the SAFE.______________________________________________________ Equity Value Update: y3 = Y2 + F(f)*(pc1-y2) = y2 + BuyPc Value Update: pc2 = pc1 + f*Buy_______________________________________________________And this is even simpler than the Lichello construction because after every buy or every sell the new pc2 = y3 and this gives the Test (pc2-y3)= identically zero, and the Buy-signal = zeroThe same formula can apply for buys as well as sells, but with a separate factor "f" for the sells the functionality of the AIM is increased so that for selling a different aggressivity can be used. This produces great flexibility for biasing the portfolio management towards a number of different strategies: For example: aggressive buying + aggressive selling--->The Tiger AIMaggressive buying + zero selling---------->Growth Mode AIMZero buying + aggressive selling---------->Liquidation Mode AIMZero buying + zero selling--------------->Holding Mode (NO-AIM)Considering that with the choice of the factors f-buy and f-sellthe investor can "Apply the Brakes" of "Step on the Accelerator". The investor can choose any level of stock trading activity he desires.This means for example that the idea of a separate SAFE can be eliminated.Furthermore, the investor can add equity at any time as he pleases, or take away equity at any time he pleases, and the AIM will automatically adjust via the function pc = yafter every buy or sell. In essence the VORTEX solution for the AIM is to RESET the system to initial conditions pc=y1 just like it was a Virgin AIM that is as simple as Lichello wanted it....just infinitely more functional.
Hi Conrad,Since I know that you have commercial interests in the Vortex method, let me point you to the TMF rules for posting to these boards. In short, do not disclose anything you want to keep the "rights" to.Best,Rien.
Hi Karel,You wrote:Again, the solution is easy, if you feel the need. (Ignoring the flaw, and even welcoming its effects is a valid solution too, IMO.) The only thing you have to change is the buy advice. The buy advice no longer needs to reference the PC before the buy, but the PC after the buy. This points directly to the trouble Lichello would have faced: the PC after the buy is influenced by the buy, so the buy advice would need info from the buy advice, and the formula has to refer to its own outcome, to be solved by iteration. Today's spreadsheets take this in their stride, and a formula likeBuyAdvice = (PC+(BuyAdvice/2))-(ShareValue*(1+Safe))_____________________________________________________________________<This is an interesting approach but I do not immediately understandhow you got it, but it needs no intteration if you see that it has a unique solution:<<If this is soved for one variable we get:<<BA=PC+BA/2-SV*(1+S)-----> BA-BA/2=PC-SV*(1+S)---->BA/2=PC-SV*(1+S)<<BA=2*(PC-SV*(1+S))---->Now if S=0-----> BA=2*(pc-SV) EXACTLY OK !!!!<<This states that the Buy Deficit in the Lichello Case is 100%. If <one buys twice the Lichello Buy-Advice EVERY Lichello AIMer would <scream Bloody Murder as the buys are far too agressive. The inclusion <of the SAFE would not change that much. My solution I presented in my <previous reply was a lot more work...especially as I had to understand <the fundamentals of what I was trying to achieve....I did it about 3 <years ago. But by using the factor "f" instead of 0,5 I came up with a <solution that provided flexibility as well. I just wonder how your <solution would work with "f" instrad of 0,5. Let"s see:<BA-BA*f=PC-SV*(1+S)---->BA*(1-f)=PC-SV*(1+S)<<BA= 1/(1-f)*(PC-SV*(1+S)---->RIGHT ON!!! <This is the same solution as I got! If you now calculate the new PC <then:< PC2=PC1 + f*BA< and this is the identical solution I have derived at via presetting<the PC-corrector with the f*Buy and then afterwards calculating that <the Buy = F(f)*(PC-SV) for which F(f)= 1/(1-f)<< This of course raises the question if with the use of the SAFE "S" <included the PC2=VS1+Buy so that for the next test PC-VS2=0<At the moment I missed a full night sleep, so I will check that out later____________________________________________________________________Karel, your solution is much shorter than the one I used. It's amazing how this solution is so simple to get to!Thanks,Conrad
Hi Mark and/or Aptus,I have just reported back to Rien and Karel on their comments to my FLAW-Argument. Without meaning to be critical in a negative way I sense strongly that many AIMers try to explain away or justify the Lichello Flaw/Feature. Is is not satisfactory to do so unless it can be demonstrated that it is an intentional Feature! I do agree that if one desires to live with it that it is an acceptable approach, but to try to explain it away as if it was so intended is like buying a car and then to discover it is missing a wheel and accepting it as if the car maker intended it like hat.To design a system to derive at an automatic means for calculating an Buy-Order means that it should do so in a single step. The Lichello Buy-Order from the Test (pc-y) is fine. Even to automatically apply a SAFE I can accept..this is only done because the Buy=(PC-y) appeared too aggressive for Lichello. This way you can use any SAFE that is to your liking. No problem! I do reject an Order Algorithm that leaves a residue. The fact that Lichello AIMers accept the Flaw as Applying the Brakes is not satisfactory. It is easy to apply brakes without having a residue dangling on the Advice Generator. Having said that I would like to refer you to the solution I presented to Karel and Rien. You will find it in this Discussion Group. If you can not find it let me know and I will find out the exact reference.Incidentially, Karel (or Quarel???) presented an idea for a solution that he used on his Excel spraed sheet and his soltion is identical to mine, although I think Karel proposed it as an answer to my question and not as means for actually using it..I do not know.In the VORTEX program I have inplemented the solution. Here it is:Buy Order: Buy = (pc1-y2)*F(f) y2 is the equity value after dropping from y1New Equity y3 = y2 + BuyPc-corrector: Pc2 = PC1 + f*Buy The factor "f" is an replacement for the Lichello 0,5 !!!Processing now the requirement that after the addition of the Buy the pc should be equal to y3 gives as Buy Signal pc2 = y3 Thus pc2-y3 = 0 exactlyThis results in the fact the Vortex Buy Order Modifier on the Lichello Buy Order(neglecting SAFE) is: F(f) = 1/(1-f) whith "f" being freely assignable number less than 1( for f > 1 you geta meaningless negative Buy-Order).The f-factor alows you to "Apply the Brakes" as hard as you want! Even to the point that you inhibit buys and sells completely!!! The LiChello SAFE can be disposed of.I hope you like my solution. Also it is the solution Karel found but he may not have proceded to the significant conclsion that it is the answer in a practical sense that states that for al buys and sells the pc-corrector always gives pc2=y3 identicallyresulting in a zero buy-signal if the stock price does not change.QEDI notice that you are on the website of Automatic Investor. Are you the designer of that SW? In essence then we may be competitors. However I do not like that name. The Vortex Solution is already incorporated in VORTEX. If you like it we could discuss a joint venture. I have not yet commercialised VORTEX seriously. In Holland some early versions are sold but I do not yet have an active market. On that score you are way ahead of me. Besides, VORTEX has a Dutch Help File.At this very time I am preparing to expose VORTEX to the market again and I would be interested in producing an English version. Also, the programmer needs to implement the Graphics package yet, and I need tofine tune the interface(Windows) as investors may want to see some features we do not yet show. I need to find a SW marketeer that can put VORTEX on the market for me in the USA. Feel free to try-out VORTEX.Downloading from: www.vortexcw.nl---> Vortex Method----Stock Market.I have also developed a program for Investment Clubs with few or many members. Its called Income & Expense. It is a program that couples with VORTEX and for a club it provides personalised statements for allocated incomes and expenses in proportion of the participation. The novelty is that the participants can step in/out at will any time and the allocations automatically adapt to these canges.I look forward to your reply. I can be reached at firstname.lastname@example.orgPS: Tom Veale referred me to several people in regards to my question to him on software marketing in the USA. I am sure yur name was on the list but you got to me first!Regards,Conrad Winkelman
Hello Conrad,"I sense strongly that many AIMers try to explain away or justify the Lichello Flaw/Feature. Is is not satisfactory to do so unless it can be demonstrated that it is an intentional Feature!"Perhaps some AIMers do try to justify things about AIM (Lichello left a number of things unsaid, so naturally contradictions and problems rear their ugly heads once in awhile), however my point was that Lichello knew about this "flaw" and seemed to think it was a "feature."As I mentioned in one of my previous posts, Lichello wrote the following on page 57 of the 3rd edition of his book:"Sometimes when a declining market suddenly shrieks to a stop, AIM will pump its brakes to avoid going into a spin. Instead of curtailing our investments entirely, it will call for sharply decreasing investments for a month or two -- and then nothing -- just to pick up a few cheap shares in case the market turns around and heads upward again.Of all the features we incorporated into AIM, this automatic check was the hardest to design and is the one of which we are most proud." Now I suppose you can take this to mean that he wouldn't/couldn't solve the "flaw" you brought up and decided to put a positive spin on it, however I'd just as soon take him at his word and believe he deliberately incorporated the "flaw/feature" into his algorithm. And the fact is that I like it (for reasons I've already given in my previous post).Having said that, I really don't disagree with your solution as I don't believe it will make that much of a difference in real-world results and it's mathematically more elegant. However I prefer to use the original algorithm that Lichello created."I notice that you are on the website of Automatic Investor. Are you the designer of that SW? "Yes, I'm the guy responsible for Automatic Investor. In fact I'm working on AI 2.0 which is scheduled to be released in March. There are also a number of other AIM software vendors, so there is competition in the AIM software space. However I believe this is a good thing as it serves to move AIM software ahead faster than if there were only 1 or 2 vendors.So welcome to the club. At this time I'm not sure how we could work together (being competitors and all ;0), but I'll send you an email next week to discuss further.Regards,Mark.www.automaticinvestor.com
Hi Rien,I know what you mean but I am talking about making a simple thing simpler. I doubt in any case that it is an effective policy to keep simple things secret.Thanks a lot for the warning.Conrad
Hello Mark,OK, I accept that certain facts can be interpreted in various ways. My query flaw/feature has been more than satisfactorily dealt with. I certainly appreciated the suggestion of Rien that led to the identical solution that I had arrived at. No need for me to press the point any farther. I am pleased that you called my solution elegant. That coming from a SW expert! I took some time to roam over your website. I am impressed. It is a very pleasing interface and especially informative on the discussion sections. Of course on all the things that are said abut AI I can find my self agreeing with it. As I can see it now A.I. appears to be a well-developed program...Ages ahead of VORTEX, especially what the marketing is concerned. Incidentally, I was today searching the TMF-site for discussions on artificial intelligence but found none. I used AI as a search key and only later I realized that it is also stands for your Automatic Investor!I also found out that you are located in Canada. I lived in Canada for 21 years, 20 of them in Vancouver! I may be travelling to Canada soon so id I do I would try to visit your office.Regards for now,Conrad Winkelman
Hello Conrad,Yes, I'm in Vancouver (Maple Ridge to be specific). I saw the West Coast Energy building on your site and figured you had to spend at least some time here.It's interesting that you mention Artificial Intelligence because that was the area I concentrated on while I was in university (SFU). Back then I was interested in Natural Language Processing, but since then I've added Machine Learning to the list and find all aspects of AI fascinating (well except for the Spielberg movie -- now that was one of the worst films I've ever seen. Some people actually got up and left the theatre half way through. And the ending? Well I won't even start a commentary on that ;-).Anyway, once I get AI 2.0 out the door in March I plan on adding some AI features to a future update. I've mapped out an expert system design and will be implementing a genetic algorithm for the optimizer (and if somebody makes a breakthrough in NLP, you might just be speaking to Automatic Investor 25.0 -- Artificial Intelligence Edition (AIAI???)).I hope things go well for you with your VORTEX software, and keep sharing your ideas. That's how we all learn.Regards,Mark.www.automaticinvestor.com
Conrad: This is an interesting approach but I do not immediately understand how you got it, but it needs no intteration if you see that it has a unique solution:Indeed it has! Very simple too, I could have worked that out myself, if I wasn't so impressed by the iteration. Thank you! Finding the formula was really easy, and in my message I already pointed out how I arrived at it. I started with the standard Lichello formula for the buy advice:BuyAdvice = PC-(ShareValue*(1+Safe))You require that the buy advice is neutral after updating, so I argued: then we can't use the 'old' PC, but we need to use the new PC, which equals the old PC + half the buy advice, so:BuyAdvice = (PC+(BuyAdvice/2))-(ShareValue*(1+Safe))That is all! Of course, the formula refers to its own result. (In the spreadheet, BuyAdvice is the cell that receives the result from the formula to the right of the equals sign.) Now I happened to be translating the part of an Excel book on circular references, so I didn't look twice to see that the formula was really very simple to solve. Which you did, with the added twist of making the divisor for the addition to PC a variable, and doing away with the Safe.In another message you say you don't know whether I meant my message just as answer, or that I was a convert :-) Well, I don't know. I would like to do some backtests first, first with Safe and then with your further tweaks.Thank you for putting me on this trail. There is a lot to sniff out here!Karel
Hoi Mark,I find it fascinating how people who at some point in time tend to have various "connectors" in their past. On the point of "artificial" intelligence I could start up a separate discussion line on the TMF. My main line on interest in that AINT direction is the contention that the adjective "artificial" is redundant. I did not see the Spielberg movie you mentioned, but if our "connectors" run deep then I suspect that your aversion was that the popular representation of intelligent machines is so far off of what intelligent machines are actually like!Well, let's leave this for the AINT-Line on TMF. On the idea of AI + AI (for investing) as you are supposing it may be very interesting to see what AIMing will be like then...I would like to present your idea as AI*e^at to emphasize its exponential potential.... but then, what would people do?Regards,Conrad
Hi Karel, Mentioned on the SI thread some time ago was also another "tweak" that I don't think has been mentioned here. For the "frequent flyers" in the AIM group who like to check up on AIM very often, someone had built a "decay" into the Buy/Sell Advice and Market Order columns. The "decay" feature prohibited purchasing extra shares without a significant discount after the initial AIM buy. Then, as time passes, the discount is reduced, eventually all the way back to standard AIM. So, on Day One we make an AIM Buy. On Day Two it would require a further price discount of say 30% for AIM to trigger a Market Order. Day Three 29%, Day 4 28%, etc. At the end of 30 days, we would be back to standard AIM. So, if we'd seen a good discount immediately after making our AIM buy, this process would allow us to make additional purchases. At the end of the 30 day decay period if the price was exactly the same as our last buy and a Market Order was still being generated at that price, we could execute it. This solution gives our Equity Warehouse Purchasing Dept. some discipline and brings it back to a once/month buy if further discount isn't seen. It also allows us to buy more frequently should deep discounts be available on a very short term basis. I'm sorry I can't give proper credit for this idea as I can't recall who first proposed it. Best regards, Tom
In May 2000, the first AIM convention was held in Las Vegas but I could not attend, instead Tom posted my report at http://www.aim-users.com/aimbrief.htmFig 3 shows how to calculate the next buy/sell prices and Fig. 5 illustrate whatbuy/sell and vealies does to the next target prices.
Hello Tom,Well, I may not be a 'frequent flyer', but I do like to get to the airport every day to see the action!I think I have found the message (by axp) on SI: http://www.siliconinvestor.com/stocktalk/msg.gsp?msgid=15463320It looks like a nice idea and it addresses the 'get the action, but don't trade too often' issue that is somewhere on my backtest list! This is a useful addition, thank you!Best regards,Karel
Hi K, Thanks for digging that one up! Yes it was AXP who mentioned the Decay idea. I really like it. It does open up a new realm of testing possibilities. What's the proper decay % and what is the time frame for the decay. Well, I know there's plenty of folks out there who will be testing these things! :-)Best regards, Tom
Newbie question about the AI software: when doing historical testing, does the software account for splits and/or distributions? These events would seem to make historical prices meaningless unless it can do it on a split adjusted basis. Thanks!The pup
Hello Rien,Going back to your suggestion for circumventing the "flaw", from a curiosity point of view, I tried to visualise your 5-step procedure but somehow I fail to "get it". I mean, it does not appear to solve the basic problem I addressed.In the first two steps(igoring the SAFE for now) the equity value returns to its orginal value y1...(at the start y1=pc1 also)y3=y1+(pc1-y2)=y1 pc2=pc1 as you suggest ----->zero buy signal. OK!Your next step is "fuzzy" to me. You calculate "next prices" using pc1=pc2, OK some prive differentials result..I assume you use the SAFE idea for that. But now you raise the pc as usual pc3=pc2 +1/2(next buy advice) and as I see it there is again a buy-signal as pc3-y3 is not zero! The buy-advice is again 1/2 of the next calculated buy, and his is exactly the problem I addressed.I think I see what is happening: After you have bought the fist lot and did not update the pc you calcuated a "next buy" signal and substituted that for the (pc-y) criterion. Then after updating the pc "as usual" you simply ignore the new buy-signal that results and stick to the calculated "next buy" you calculated. I would agree that this will stop you from buying in on the new buy-signal but the methods appears to be more of a "fudging-approach" rather than anything else. One might as wel simply use the original method and ignore the residue Buy-signal and only buy if the stock price has dropped and then implement the buy as advised, with again ignoring the residue buy-signal.After having reviewed the many discussions I have concluded, as I have said before, that using the pc-corrector pc=y eacht time a buy or a sell is activated appears to be the most effective solution. With selecting an appropriate "aggression" for the buy and sell seperately we can tune the AIM-behaviour as we see fit. With the addition of an appropriate Resistance for the Buy and Sell activity we can "time" the action so that a sell-signal on a trend-reversal is not actuated too soon(creating a delay-factor on rising prices).I think that with this I have reached the end of my quest for a simple and convenient Advice Generator that gives a zero buy-sell signal after every buy and sell. It would appear unnecessary for me to dwell on this issue much longer.I hope the discussions have provided more insight into the possibilities for AIMers. At the least I have come to recognize the merit of the 1/2buy pc-corrector on the bottom of a Dip...the Residue Buy at that point produces a delay for the selling off of the equity. The equity value has to rise at least to above the previous 1/2Buy amount in order to produce a sell signal. I agree that this is a positive featue of pc2=pc1+1/2Buy pc-corrector.Is there more to say?Conrad
Correction on previous posting:I wrote to Rien:In the first two steps(igoring the SAFE for now) the equity value returns to its orginal value y1...(at the start y1=pc1 also)y3=y1+(pc1-y2)=y1 pc2=pc1 as you suggest ----->zero buy signal. OK!<?I>Oviously this is incorrect for y3(the updated equity value. The correct expression for y3 is:y3=y2+(pc1-y2)=y1 with pc2=pc1...Conrad
Hi Conrad,I would agree that this will stop you from buying in on the new buy-signal but the methods appears to be more of a "fudging-approach" rather than anything else. One might as wel simply use the original method and ignore the residue Buy-signal and only buy if the stock price has dropped and then implement the buy as advised, with again ignoring the residue buy-signal.I should maybe clarify that I had this idea not because of the "flaw" but because I found that the buy's were to closely spaced. I wanted a bit more "price difference" between buy points. And that is exactly what it does.Best,Rien
Rien,OK, I "get it". There sure are more roads to Rome than the few I knew!One of these days we all meet in Rome en talk about our journey some more.The discussions of the last few weeks have given me a chance to understand how various serious AIMers create interesting ways of making their AIM more flexible and making it perform better than at first.Regards,Conrad
Best Of |
Favorites & Replies |
Start a New Board |
My Fool |
BATS data provided in real-time. NYSE, NASDAQ and NYSEMKT data delayed 15 minutes.
Real-Time prices provided by BATS. Market data provided by Interactive Data.
Company fundamental data provided by Mornin