From extropians-request@extropy.org Sun Oct 24 02:46:17 1993 Return-Path: Received: from usc.edu by chaph.usc.edu (4.1/SMI-4.1+ucs-3.0) id AA14151; Sun, 24 Oct 93 02:46:16 PDT Errors-To: Extropians-Request@gnu.ai.mit.edu Received: from news.panix.com by usc.edu (4.1/SMI-3.0DEV3-USC+3.1) id AA27603; Sun, 24 Oct 93 02:46:12 PDT Errors-To: Extropians-Request@gnu.ai.mit.edu Received: by news.panix.com id AA26052 (5.65c/IDA-1.4.4 for more@usc.edu); Sun, 24 Oct 1993 05:40:32 -0400 Date: Sun, 24 Oct 1993 05:40:32 -0400 Message-Id: <199310240940.AA26052@news.panix.com> To: Extropians@extropy.org From: Extropians@extropy.org Subject: Extropians Digest X-Extropian-Date: October 24, 373 P.N.O. [09:40:07 UTC] Reply-To: extropians@extropy.org Errors-To: Extropians-Request@gnu.ai.mit.edu Status: RO Extropians Digest Sun, 24 Oct 93 Volume 93 : Issue 296 Today's Topics: [1 msgs] Beating the Stock Market? [7 msgs] Beating the Stock Market? [1 msgs] Genetic Programming and stocks: the "data" question [2 msgs] Help! [1 msgs] OUTREACH: Magazines [2 msgs] QUERY: off shore accounts [2 msgs] scams [2 msgs] Administrivia: No admin msg. Approximate Size: 53645 bytes. ---------------------------------------------------------------------- Date: Sat, 23 Oct 1993 07:52:15 -0400 (EDT) From: Harry Shapiro Subject: Help! If exclude is not working, it is because you are sending commands from an address that is different from the address we are using to send you mail. Send me that address. /hawk a conscious being, Christina M Grimes wrote: > > It's not as if I don't enjoy the discussion on the list, but I seriously > a) don't have the time to read/respond and b) don't have the space in my > quota on the mail server; it keeps filling up, and I have to wade > through dozens of messages to find ones from friends (which I often > miss). I've tried to ::exclude all three times, but it hasn't worked. > I'm not sure why. I'll give it one more shot, I guess, but there seems > to be something up with the program. > > Thanks a lot, > Ratha > -- Harry S. Hawk habs@extropy.org Electronic Communications Officer, Extropy Institute Inc. The Extropians Mailing List, Since 1991 ------------------------------ Date: Sat, 23 Oct 93 5:33:30 PDT From: szabo@netcom.com (Nick Szabo) Subject: Beating the Stock Market? I made some doodles this evening towards AIT theories of time series compression and arbitrage. It would be interesting to see if the new Kolmogorov complexity book has come up with something similar, and I'd love to hear comments, critique, & elaborations of this stuff. We start with the "brilliant penny" scenario, where we want to find a model that predicts a random binary sequence x. If we flip 2^|x| pennies |x| times each, chances are one of the sequences will match, the "brilliant penny" that magically "predicts" the original sequence. To model a random series like this requires fully |x| bits: H(x)=|x|. For a sequence x where odd bits are random, while even bits are 1, H(x) = |x|/2 + Htm("even bits are 1"). Now let's create two random temporal sequences x and y. The ground rule for doing trading here is that we can read any prices at time T H(x) = |x| = 8 y = 11000010 => H(y) = |y| = 8 = |x| By themselves both x and y are random, but we notice that for even bits in y, y(t) = x(t-1) xor 1. So for sufficiently large |x|, H(x,y) < H(x) + H(y). Specifically, H(x,y) = H(x) + H(y)/2 + Htm(f(x,y)) = 3|x|/2 + Htm(f(x,y)) Where f(x,y) = "if (t even) then y(t) = x(t-1) xor 1". >From f(x,y) we construct the trading strategy if (t even) and x(t-1)==1 buy y(t) if (t odd) and x(t-1)==0 sell y(t) and gain a profit of 4 bits on the sequence. ($.50!) So we have an arbitrage model "f" which reduces the mutual complexity of sequences x and y which, by themselves, look random. Now we said the sequences "looked random", but there might be other good Turing machines that compress the time series, either involving arbitrage between x and y, or individually. Our goal is to find, not the smallest H(x,y), in general an impossible task, but an H(x,y) smaller than our trading competitors. The chance of finding |x|/2 xors coincidentally in two random sequences of size x is 1/2^(|x|/2). From this we can compute risk profiles, p(ruin), etc. Here we can see formally Occam's Razor, what both scientists and traders have long suspected: why a more parsiminous model is also a more predictive model. The chance that we can compress a randomly generated sequence is very small -- we are much more likely to find uncompressed "brilliant pennies". The larger is the difference |x1| + |x2| + ... - H(x1,x2,...), the greater the chance that the component thus compressed by Htm(f(x1,x2,...) is not random, but rather contains a compression that will also compress the component similarly in the future. (In future work I will endevour to prove this formally, also providing a formal model for the parsimony and predictive capability of arbitrage strategies). It's easy to extend this theory from binary prices to arbitrary price ranges (but of course it's too big to fit in the margins). The binary model above predicts 100% of the sample with risk a function of the 1/2^(|x|/2). In real-world models compression is only statistically correct, that is to say the models include large components, some hedgeable and some not. To lower risk as much as possible, we create synthetic assets to hedge away as many random components as possible, thus making our trading strategy correspond to the best compression we can find with the information at hand. Nick Szabo szabo@netcom.com ------------------------------ Date: Sat, 23 Oct 1993 10:49:20 -0400 From: "Perry E. Metzger" Subject: Beating the Stock Market? Ray says: > Timothy C. May () writes: > > I was skeptical of his reported annual rates of return (> 30% on > > average, for the past N years), having read Burton Malkiel's "A Random > > Walk Down Wall Street" at an impressionable age. [...] > Statistically, if you do your homework (e.g. follow technology, follow > trends) I believe you can beat the market. You won't be able to beat it > all the time, but it will be enough to multiply some wealth. [...] Some traders like Paul Jones and Lewis Bacon have managed to produce double and even occassionally treble digit returns regularly for years. It is certainly possible to do it, and do it for prolonged periods. Its very unusual, however -- we are talking maybe a dozen to two dozen managers in the world. Whenever you are shown such returns, a red warning flag should immediately go up. I immediately reach for Nexis/Lexis and start searches. If the guy is the subject of Wall Street Journal articles on "Joe Foo does it for another year running!" then he might, might, be trustworthy. More likely you will find out that the guy has been in and out of court for years for fraud. In any case, such investments are very high risk, and shouldn't be made with money you can't afford to lose. Perry ------------------------------ Date: Sat, 23 Oct 1993 12:53:31 -0500 (CDT) From: derek@cs.wisc.edu (Derek Zahn) Subject: Beating the Stock Market? Nick Szabo: > I made some doodles this evening towards AIT theories of > time series compression and arbitrage. Neat! Let's see... > Now let's create two random temporal sequences x and y. The ground > rule for doing trading here is that we can read any > prices at time T the time it takes information to propagate from one market > to another. I'm not sure what you mean here. If you can trade and read both markets instantaneously, doesn't that mean that delta-t is zero? Or do you mean some kind of information propagation channel that you have access to but others don't? > By themselves both x and y are random, but > we notice that for even bits in y, y(t) = x(t-1) xor 1. So > for sufficiently large |x|, H(x,y) < H(x) + H(y). Yes... In _Mathematical foundations of information theory_ (p8), Khinchin puts it this way: "The amount of information given by the realization of two finite schemes A and B equals the amount of information given by the realization of scheme A plus the mathematical expectation of the amount of additional information given by the realization of scheme B after the realization of the scheme A." Or, H_a(B) <= H(B), where H_a(B) is the entropy of the scheme B given that the scheme A has been realized. This is elementary info theory... I suppose that you're driving toward the idea of using something like Genetic Programming with a high parsimony bias to construct arbitrage models for the two sequences. This seems promising to me... However, in the case of trading strategies, it's not clear to me that: > an H(x,y) smaller than our trading competitors. ... > a more parsiminous model is also a more predictive > model. My doubt comes from the fact that every trader in the market is building models of the market and the actual market prices are emergent phenomena of their ensemble. I'm not sure that Occam's razor is as compelling in such a circumstance. > In real-world > models compression is only statistically correct, This I think is a crucial issue -- parsimony vs accuracy. Suppose I have a compression program for x that is 90% accurate and its size is 0.2 * |x|. Suppose I have another that is 93% accurate but its size is 0.4 * |x|. Which is better? It seems like there should be a principled answer to this sort of question. > To lower risk as much as possible, we create synthetic assets > to hedge away as many random components as possible, thus making our > trading strategy correspond to the best compression we can find with > the information at hand. Could you elaborate on this? I'm not sure what you mean by synthetic assets as hedges. derek ------------------------------ Date: Sat, 23 Oct 1993 13:10:03 -0500 (CDT) From: derek@cs.wisc.edu (Derek Zahn) Subject: Beating the Stock Market? Tim May: > So, multiplying wealth by analysis seems pretty Extropian to me. Data, data, we need data! It's unreasonable (IMO) to think that we can train some automated system to do market prediction without letting it have the same information that its competitors (professional analysts) can use. I get exhausted just thinking about such a data collection and massaging task, not to mention that much of it (e.g. product classification, political turmoil) is difficult to formalize well. On the other hand, I see these toupeed "technical analysts" on CNBC with their strangely-drawn "support" lines and their bollinger bands and other heuristic aids handed down from generation to generation... those formulae are usually based on simple trading price data, which is easily given to automated regression engines. Some of those things are kind of interesting.... for instance, the "support" idea -- that there is a natural range surrounding a trend that if broken through likely indicates that _something_ is happening and it's a good thing to jump on such a new (hopefully) trend. I wonder to what extent the wisdom that such trends will continue is caused by self-fulfilling prophecy as many traders using similar strategies create its validity by bidding stocks up or down... derek football might be easier ------------------------------ Date: Sat, 23 Oct 1993 13:15:21 -0500 (CDT) From: derek@cs.wisc.edu (Derek Zahn) Subject: Beating the Stock Market? Oh, also note that the Santa Fe Institute does a time series prediction contest I think yearly; some of the finest minds in the field take part. One of the more interesting and successful ideas is to take a hodgepodge of different techniques, train them all, and then train a "supervisor" to decide how to weight their opinions in a particular circumstance. If I remember right, the guy who won the contest with this approach is named Chen or Chien or something like that and works for Thinking Machines. derek ------------------------------ Date: Sat, 23 Oct 93 12:01:50 PDT From: tcmay@netcom.com (Timothy C. May) Subject: Beating the Stock Market? (Good to see the old AIT crowd, Nick S. and Derek Z., back at work.) Derek Zahn writes: > Tim May: > > > So, multiplying wealth by analysis seems pretty Extropian to me. > > Data, data, we need data! It's unreasonable (IMO) to think that > we can train some automated system to do market prediction without > letting it have the same information that its competitors (professional > analysts) can use. I get exhausted just thinking about such a > data collection and massaging task, not to mention that much of it > (e.g. product classification, political turmoil) is difficult to > formalize well. The hedge fund manager I mentioned subscribes to several hundred dollars a month of price quotes, and routinely loads 50 megabyte data sets into RAM for crunching. (On a mix of IBM RS/6000s, 486 boxes, and Mac Quadra 950s.) I don't know if this is "enough." Nick's simple mutual entropy calculation for a random sequence was at one extreme (one or a few price vectors), the stuff Wall Street has is at the other side (firehoses of data). How much can be done "in between"? Does the amount of data--the price of slippers in Tangiers, the yak population in San Diego--make a real difference? > On the other hand, I see these toupeed "technical analysts" on > CNBC with their strangely-drawn "support" lines and their bollinger > bands and other heuristic aids handed down from generation to > generation... those formulae are usually based on simple trading > price data, which is easily given to automated regression engines. These are pretty easily dismissed: the "angles" they talk about (e.g., 45 degree lines drawn) and whatnot are not invariant to time-scaling. I gave watching CNBC, except for the running ticker, even before I had my cable t.v. subscription cancelled. These market mavens consistently fail to report true indicators of their past accuracy, just the selective revealing of winners (selective exposure of information is of course the explanation of the "brilliant penny" scam and all the similar variations mentioned here lately). The techniques I'm talking about, which may or may not work, are a whole lot more sophisticated than finding "candlestick formations" in mostly random data. Time-series analysis is only a tiny part of the picture. (Hint: automate a lot of "fundamental" ideas into thousands of if-then rule sets...like an expert system. Mix with neural nets, genetic programming to "crawl" through huge amounts of data, combine with sophisticated statistical analyis, season with hedging and portfolio insurance to adjust the beta to taste.) > Some of those things are kind of interesting.... for instance, the > "support" idea -- that there is a natural range surrounding a trend > that if broken through likely indicates that _something_ is happening > and it's a good thing to jump on such a new (hopefully) trend. I > wonder to what extent the wisdom that such trends will continue is > caused by self-fulfilling prophecy as many traders using similar > strategies create its validity by bidding stocks up or down... Well, this is what has me interested. When I first started investing in stocks, in the mid-70s, I went through a phase of keeping charts on stocks I followed, on computing simple moving averages (computed with exponential smoothes on my then-new H-P 25 programmable calculator!), and so forth. This lasted for about 3 months, then I gave it up. My friend's work with using a panoply of induction techniques (every trendy word, from neural nets to genetic programming to Black-Scholes portfolio methods to chaos to expert systems to Bayesian statistics....) crunching on huge data sets on fast computers has me very interested. Like I said, I'm skeptical, but intrigued. Since I'm about to start the Kolmogorov-Chaitin book anyway... And I can imagine how even in a random, chaotic data set a "strategy" can be found: An amazingly simple strategy: Suppose Malkiel and the efficient market random walkers are right, that there is no mechanistic strategy (hence _no_ strategy??). But then by betting _against_ all those who _do_ think there is a strategy, one will profit from their misplaced faith. That is, "no strategy" implies there _is_ a strategy! I obviously don't claim this is trivial to implement, or necessarily exploitable. But it's an area that combines a lot of interesting algorithms, gets into areas of predictability and determinism in a more concrete way (for me at least) than speculations about "free will" do, and could be lucrative if in fact these methods work. I'm skeptical, but who knows? > derek > football might be easier Actually, I just saw an ad for a "neural net football predictor" program. It has scores of the last n seasons already in it, and the user enters new scores as they happen. Then the "neural net" supposedly learns the patterns and predicts future scores. Obviously, "fundamental analysis," such as injuries to key players and the like, is not reflected in this model (so far as I know). But this football example makes the same points. Can crunching on a lot of data (horse races, football and sports, stocks) with a lot of approaaches (as mentioned earlier) give an "edge"? As home users will have access to more and more such data, there may also be a market for software which "helps" them make bets with their freinds or bookies on point spreads and the like. (Scenario: Joe Sixpack loads his "Joe Montana Picks the Winners" CD-ROM into his 3DO machine on top of his t.v., opens a Bud, and sits down to crunch the latest scores, possibly downloaded automatically, via subscription. Farfetched? What about track betters armed with Newtons? Could be a market, even if the techniques are dubious.) I'll keep you posted. --Tim May (BTW, my friend the hedge fund manager is a friend of John Koza's and is acknowledged in "Genetic Programming." He's been running such models for several years now and is also in contact with "Chaos Cabal" guru Doyne Farmer, formerly of Los Alamos and the Santa Fe Institute and now with his own company, Prediction, Inc.) -- .......................................................................... Timothy C. May | Crypto Anarchy: encryption, digital money, tcmay@netcom.com | anonymous networks, digital pseudonyms, zero 408-688-5409 | knowledge, reputations, information markets, W.A.S.T.E.: Aptos, CA | black markets, collapse of governments. Higher Power: 2^756839 | Public Key: PGP and MailSafe available. Note: I put time and money into writing this posting. I hope you enjoy it. ------------------------------ Date: Sat, 23 Oct 1993 14:35:22 -0500 From: "Phil G. Fraering" Subject: Genetic Programming and stocks: the "data" question I'd like to raise a couple points about doing market analysis of any kind and the "data" question that Derek Zahn and Tim May were just talking about. Simply put, a lot of the bets and analyses that are being discussed here seem to be immune to any sort of algorithmic system applied to any sort of large commercial data dump. Or so it seems to me. For instance: the 3DO/Atari/Commodore/ect. example: there doesn't seem to me to be any way for 3DO's future stock performance to be modeled as a function of a) its competitors past performance, b), its past performance, general consumer spending, durable goods spending, et cetara.... one of the main factors seems to be based on, "Is the company going to deliver?" As someone at least a little skeptical about 3DO/CD^32/Jag/whatever's market prospects, I didn't really have any doubt that those companies were actually going to _deliver_ those products, but it also seems to me that a lot of the fluctuation in 3DO Corp's price was due to the doubts of others in precisely this area, and in a way rather artificial. Hmm... thinking... is it possible that each company has a "confidence" function, how much its stock is going to rise/fall based on whether or not it meets a given performance benchmark (delivers a product, etc)... Microsoft, for instance, didn't seem to suffer much at all no matter how many times Windows NT was delayed. How expressable as an algorithm could a confidence function be? I know that company stocks aren't the target per se of most of the AIT applications to market prediction that have been discussed both here on the list and elsewhere. But for non-resource-extraction companies, and/or other non-commodity companies, is there anything for the algorithm engines to evaluate? +-----------------------+---------------------------------------+ |Phil Fraering | "...drag them, kicking and screaming, | |pgf@srl03.cacs.usl.edu | into the Century of the Fruitbat." | +-----------------------+-Terry Pratchett, _Reaper Man_---------+ ------------------------------ Date: Sat, 23 Oct 93 13:03:32 PDT From: tcmay@netcom.com (Timothy C. May) Subject: Genetic Programming and stocks: the "data" question Phil Fraering writes: > For instance: the 3DO/Atari/Commodore/ect. example: there doesn't seem > to me to be any way for 3DO's future stock performance to be modeled .... Indeed, and no one I know thinks that a particular stock can be predicted in this way. The time series of a stock is underconstrained, in a sense, as a vast number of factors affect the future price. The kind of "prediction" we are talking about here is more related to whether a slight statistical "edge" can be found. > Hmm... thinking... is it possible that each company has a "confidence" > function, how much its stock is going to rise/fall based on whether or > not it meets a given performance benchmark (delivers a product, etc)... Well, there are a zillion such "heuristics" which may in fact enter in to a model, a "machine" that crunches data of all sorts and outputs buy and sell orders. In fact, my friend's system presumably looks for this sort of correlation. Maybe too much data to crunch. Maybe it's all futile. But it can't hurt to take another look at algorithmic trading. --Tim -- .......................................................................... Timothy C. May | Crypto Anarchy: encryption, digital money, tcmay@netcom.com | anonymous networks, digital pseudonyms, zero 408-688-5409 | knowledge, reputations, information markets, W.A.S.T.E.: Aptos, CA | black markets, collapse of governments. Higher Power: 2^756839 | Public Key: PGP and MailSafe available. Note: I put time and money into writing this posting. I hope you enjoy it. ------------------------------ Date: Sat, 23 Oct 93 13:17:22 -0700 From: kwatson@netcom.com (Kennita Watson) Subject: scams 2^N possible con victims N days Call the victims and tell them each day your prediction of which way a stock with equiprobable likelyhood of rising or falling will move. Tell half one story, half the other. At the end of N days, you will have convinced one investor that you really really can predict the market. Don't you need 3^N, to take care of the case when the stock stays the same? Kennita ------------------------------ Date: Sat, 23 Oct 93 13:37:06 -0700 From: kwatson@netcom.com (Kennita Watson) Subject: OUTREACH: Magazines I don't tend to keep my old copies of _Extropy_ or _Cryonics_ magazines around for reference, but it seems a shame to toss them immediately. I was considering dropping them off at my doctors' offices (clinic, chiropractor, dentist, optometrist). I can see plusses and minuses. What do you think? Kennita Watson | Do I want to live forever? Maybe not, but give kwatson@netcom.com | me one or two hundred thousand years to think HEx: KNNTA | it over. - KLW, 1993 ------------------------------ Date: Sat, 23 Oct 1993 16:50:26 -0500 From: "Phil G. Fraering" Subject: scams >Don't you need 3^N, to take care of the case when the stock stays the >same? Nah, that can be explained away: "See? There was no activity, which is because the investors know the stock was overvalued/undervalued." "Considering that the whole market went up, in relative terms, stock X went down." "See? A fraction of a cent decreace! And that guy in New York was selling short! I told you it would fall!" After all, with this scam, you're looking for the people who have abandoned a scientific mindset to begin with and can often make useful "fudges" to garner more than the classical 1 out of 2 ^ N seekers... Phil ------------------------------ Date: Sun, 24 Oct 93 00:31:33 GMT From: sjw@liberty.demon.co.uk (Stephen J. Whitrow) Subject: Beating the Stock Market? A few years ago when I was trying out some futures trading, my brokers were providing me with recommendations from a "computer trading system" they had, which "tracks the movements in thirty commodity futures markets and uses over forty analytical tools in order to produce its portfolio or recommendations. On a day to day basis the system generates various buying and selling recommendations, accompanied by a relevant stop loss." They also claimed that it was already being used by "one of the four major clearing banks in the U.K." Well, the system was producing an average gross profit per trade of about $60 judging by the trades I'd been advised. The trouble was I was charged $100 commission a round turn, but it might have had some potential for an owner of the system utilising a discount broker. I suspect the risks of holding open positions overnight would outweigh potential gains, though. It was called *The Intelligent Trading System* by Providence Research Group, Inc. Anyone heard any more on this? I designed a system for day trading silver futures on the COMEX Exchange, using a very simple strategy. Over a period of 47 days the system _had_ made average gross profits of $218 per day with only ten losing days, and a minimum profit of $475 every week. (This was the optimised 'with hindsight' version, with carefully chosen targets and stops, but this choice wasn't too critical.) Naturally, it failed later when there was a sudden slide in the precious metals. It shows how easy it is to design simple mechanical systems which fail to live up to initial promise. With the more sophisticated methods, I'd expect any falling off in performance to be a more gradual process as others discover them. Steve Whitrow sjw@liberty.demon.co.uk ------------------------------ Date: Sat, 23 Oct 93 21:35:24 EDT From: mpeel@ccs.carleton.ca (Michelle Peel) Subject: OUTREACH: Magazines Kennita Watson writes: > > I don't tend to keep my old copies of _Extropy_ or _Cryonics_ > magazines around for reference, but it seems a shame to toss them > immediately. I was considering dropping them off at my doctors' > offices (clinic, chiropractor, dentist, optometrist). > > I can see plusses and minuses. What do you think? > > Kennita Watson | Do I want to live forever? Maybe not, but give > kwatson@netcom.com | me one or two hundred thousand years to think > HEx: KNNTA | it over. - KLW, 1993 > Make sure that the doctor (etc.) doesn't just throw them out for fear of being linked to the philosophies described therein. Me. ------------------------------ Date: Sat, 23 Oct 1993 23:09:56 -0400 From: Duncan Frissell Subject: QUERY: off shore accounts P >1) How is one established? Calling faxing, writing, telexing banks taken from one of the international bank directories in your local library. P >2) What countries are good to establish one in? Depends. For most US resident's needs, even the UK is an "offshore" country. Switzerland remains good unless you are a very large-scale criminal type. I like the Isle of Man and the Channel Islands myself. P >3) I understand the tax advantage reason why one would be established P >and the fact that the assets are basically untraceable by the USG & US P >companies. Are there any other reasons why it might be a good idea? Judgment proofing, investment diversity, currency diversity, better service, to get a Eurocheque card you can use while traveling in Europe, personal growth, political risk diversity, and higher interest. P >4) What is an approximate amount of assets that one must have before P >it is worth establishing such an account? I know someone who opened one with $300 once. P >5) How are transactions handled? By mail, fedex, fax and telex. P >6) What is the legality of such an account? What must you legally P >tell the USG about? As legal as church on a Sunday. You are required to report foreign accounts totaling more than $10,000 on which you have signature authority. You have to report interest income received from all accounts. P >7) Do you lose out by not receiving interest? What makes you think offshore accounts don't pay interest? P >8) How safe are those assets? As safe (on average) as US deposits. P >9) Is there any references about this topic? Book, ftp sites, ... Thousands (though not many online). Someone wrote a FAQ a while ago. Duncan Frissell --- WinQwk 2.0b#1165 ------------------------------ Date: Sat, 23 Oct 1993 23:45:21 -0400 (EDT) From: ptbast@ivy.WPI.EDU (Peter F Bastien) Subject: QUERY: off shore accounts [Stuff deleted] > > P >2) What countries are good to establish one in? > > Depends. For most US resident's needs, even the UK is an "offshore" > country. Switzerland remains good unless you are a very large-scale > criminal type. I like the Isle of Man and the Channel Islands myself. > How about Canada? Or is it too "close" to the USG? > P >3) I understand the tax advantage reason why one would be established > P >and the fact that the assets are basically untraceable by the USG & US > P >companies. Are there any other reasons why it might be a good idea? > > Judgment proofing, investment diversity, currency diversity, better > service, to get a Eurocheque card you can use while traveling in Europe, > personal growth, political risk diversity, and higher interest. > [Stuff deleted] > > P >8) How safe are those assets? > > As safe (on average) as US deposits. Does this mean the deposits are issured as they are in the US? [Stuff deleted] > > Duncan Frissell > --- WinQwk 2.0b#1165 > Just some follow up questions. Thanks, Pete ------------------------------ Date: Sun, 24 Oct 93 00:10:02 EDT From: The Hawthorne Exchange Subject: Nightly Market Report The Hawthorne Exchange - HEx Nightly Market Report For more information on HEx, send email to HEx@sea.east.sun.com with the Subject info. News Summary as of: Sun Oct 24 00:10:02 EDT 1993 Newly Registered Reputations: (None) New Share Issues: (None) Share Splits: (None) Market Summary as of: Sun Oct 24 00:00:02 EDT 1993 Reputations of members of the Extropians mailing list: [ Note: Contact hex-request to have a reputation placed on this list. ] Total Shares Symbol Bid Ask Last Issued Outstanding AMARA .10 .50 - 10000 - ANTON .61 .63 .63 10000 1943 ARKU .30 .31 .30 10000 5301 BLAIR .01 1.20 .01 10000 26 BROOX .01 1.00 - 10000 - DEREK .06 .19 .19 100000 18430 DROSE - .15 .10 10000 3000 DRS - .15 .15 10000 2600 DVDT .75 1.75 1.70 10000 10000 E .80 1.00 .90 10000 8011 ESR - - - - - FCP .06 1.30 1.50 80000 15345 GHG .02 .30 .20 10000 8180 GODII .01 1.00 - 10000 - GOEBEL .01 .25 1.00 10000 767 H .40 .76 .76 30000 10290 HAM .60 .90 .90 20000 15918 HANNO .15 .24 - 10000 500 HFINN 1.50 6.00 .01 10000 1005 IMMFR .25 .70 .80 10000 1838 JFREE .02 .50 .50 10000 3200 JOHN .30 .40 .35 10000 600 JPP .26 .29 .26 10000 3500 KARL 1.00 1.50 2.00 10000 1000 KNNTA .12 .19 .26 100000 10100 LEFTY .31 .40 .40 10000 4751 MARCR - - - - - MEEKS - - - - - MLINK - .01 .01 1000000 102602 MWM - 1.50 .01 10000 1260 N 5.00 9.00 9.00 10000 4750 P 22.50 25.00 1.50 1000000 94 PETER - .01 1.00 10000000 600 PRICE - .01 .01 10000000 1410 R .40 .80 .70 10000 6100 RJC .65 2.00 1.00 10000 5200 ROMA - - - - - RWHIT - - - - - SAMEER .30 .75 .61 10000 9810 SHAWN .55 .55 .01 10000 25 SWANK - 1.00 - 10000 - TIM .10 .60 .50 10000 2104 WILKEN 1.00 10.00 1.00 10000 102 ----------------------------------------------------------------------------- Other reputations: Total Shares Symbol Bid Ask Last Issued Outstanding 1000 .05 .40 .20 10000 5000 110 .01 .10 .10 10000 1750 150 .01 .10 .10 10000 1750 1E6 .20 - .20 10000 8825 1E9 .01 .09 .20 10000 7000 200 .02 .20 .10 10000 5075 80 .01 - - 10000 - 90 .01 - .10 10000 2000 ACS .10 .15 .12 10000 3223 AI .01 .09 .10 10000 2400 ALCOR .25 .85 .60 10000 3675.00 ALTINST - .25 .05 10000 4000 ANARCHY .20 .90 1.00 10000 1100 BIOPR .01 .09 .05 10000 3000 BITD .01 1.00 - 10000 - BLACK - .10 .10 100000 6000 CHUCK - - - - - CYPHP .20 .40 .30 10000 10000 D&M - - - 10000 - DC1000 - .10 - 10000 - DC200 - .15 .10 10000 1500 DC7000 - .10 - 10000 - DCFLOP .15 - .15 10000 6000 DRXLR .75 .90 .80 10000 4545 EXI .10 .25 1.54 10000 3025 FAB - - - - - GOD - .10 .10 10000 3000 GUNS - .90 1.00 10000 3900 HART - 1.99 2.00 10000 9000 HEINLN .28 .30 .30 10000 6600 HEX 100.00 101.00 100.00 10000 4188 KLAUS - .45 .45 100000 36004 KPJ - - - - - LEARY .01 .50 .20 10000 1000 LEF .10 .35 .10 10000 5214 LIST .40 10.00 .75 10000 5000 LP .25 .30 .50 10000 5625 LSOFT .50 1.00 .50 10000 9550 LURKR - .01 - 100000 - MED21 .01 .30 .30 10000 5399 MMORE - 1.25 .10 10000 3000 MNSKY - 1.80 - 10000 - MORE .38 1.25 .75 10000 2660 NEWTON - .50 .20 10000 1000 NLAW - .50 - 10000 - NNLAW - .50 - 10000 - NSS .02 .03 .01 10000 25 OCEAN .15 .18 .20 10000 6600 OOMPH - 15.53 22.00 20000 - PENNY - .08 1.50 10000 2500 PGP - 1.00 1.00 100000 2100 PLANET .01 .02 .02 10000 4000 PPL .30 .45 .30 10000 4600 RAND .18 .20 .20 10000 3900 RAW - .05 - 10000 - SHECKY - - - - - SSI .15 .20 .20 10000 5200 TCMAY .38 .40 .38 10000 6200 TRANS .01 .90 .60 10000 3211 VINGE .01 1.00 .75 10000 3449 ----------------------------------------------------------------------------- ------------------------------ Date: Sun, 24 Oct 1993 00:10:28 -0400 From: Duncan Frissell Subject: Beating the Stock Market? I take it that everyone has read the Economist's October 9, 1993 Finance Survey on this very topic? "The crossing of economic and mathematical theory has yielded a bumper harvest of jargon. Does the stockmarket show generalized auto-regressive conditional heteroskedasticity? Can time-delayed embedding explain and profit from the leptokurtosis of stock trades, or will the limit set by the Lyapunov exponent forever make it impossible? Does the back-propagation of errors lead inevitably to overfitting?" Here are the abstracts of the individual survey articles: Title: On the edge. (computer prediction of capital markets) (Finance Survey) Authors: Ridley, Matt Subjects: Program trading (Securities)_Research Nonlinear theories_Usage Reference #: A14513797 Abstract: Intelligent computer software which employs recent mathematical research in non-linear statistics, can now produce approximate predictions of financial markets. Title: Efficiency and after. (classic economic theory versus computer prediction of capital markets) (Finance Survey) Authors: Ridley, Matt Subjects: Time and economic reactions_Research Program trading (Securities)_Economic aspects Reference #: A14513807 Abstract: Currency arbitrage markets can be predicted by computer programs because no market is perfectly efficient in the classical economics sense. Slight inefficiencies in the uses of the same market information can be predicted for short time periods. Title: A dose of empirical evidence. (computer prediction of capital markets) (Finance Survey) Authors: Ridley, Matt Subjects: Program trading (Securities)_Analysis Nonlinear theories_Usage Reference #: A14513813 Abstract: Non-linear, or unexplained currency market behavior can be discovered with large amounts of data and modern, powerful computing. Such behavior can be generalized inductively and used to inform trading decisions. Title: A tale of fat tails. (computer prediction of capital markets) (Finance Survey) Authors: Ridley, Matt Subjects: Program trading (Securities)_Analysis Gaussian distribution_Research Random noise theory_Research Reference #: A14513825 Abstract: Random statistical 'noise' which does not fit the two-tailed random Gaussian distribution is the key to understanding current research in computer prediction of markets. The herd behavior of human traders is being quantified into statistical patterns. Title: The proof of the pudding. (Olsen and Associates - computer prediction of capital markets) (Finance Survey) (Company Profile) Authors: Ridley, Matt Subjects: Program trading (Securities)_Analysis Time and economic reactions_Research Companies: Olsen and Associates_Evaluation Abstract: Five companies which use extensive non-linear computer trading programs are described. A sixth, Olsen, is very open about its research and methods. Founder Richard Olsen explains how he analyzes huge amounts of data to predict how markets will react to world or market events. Title: Neural networking. (computer prediction of capital markets) (Finance Survey) (Company Profile) Authors: Ridley, Matt Subjects: Program trading (Securities)_Analysis Neural networks_Usage Reference #: A14513847 Abstract: Neural network computer programs recognize patterns in data. With the right data, commercial neural networks such as NeuralWare's Professional II+ or Right Information System's Forethought can be effective investment tools. Title: Cash machines? (computer prediction of capital markets) (Finance Survey) Authors: Ridley, Matt Subjects: Program trading (Securities)_Forecasts Foreign exchange_Forecasts Reference #: A14513849 Abstract: Forecasting software will change capital markets forever. However, no current software can apply human trading intuition. The impact of computer forecasting will be in easier access to markets by more traders, increasing liquidity and spurring general economic growth. ************ I have full text. If anyone can suggest a place where I could dump it for short time personal FTPing by list members, I'd be happy to. Duncan Frissell --- WinQwk 2.0b#1165 ------------------------------ End of Extropians Digest V93 #296 *********************************