Monkeys, StrataSearch & Shakespeare

Strategies might be attractive on the surface but fail moving forward. Why? Here we discuss the qualities of effective trading strategies, and the many traps one should avoid.

Monkeys, StrataSearch & Shakespeare

Postby Dacamic » Thu Jan 25, 2007 1:28 am

It is not uncommon for people to believe that, figuratively speaking, an invisible hand guides StrataSearch when it builds systems. On the contrary, its strategy building process is substantially void of intelligence (like some of my posts). StrataSearch in its basic form uses a brute force approach by randomly combining formulas and parameter sets and makes no judgment about potential performance. Along those lines, one member of our forum -- Barnabyj -- reminded us of a lingering, relevant question:
    If you gave an infinite number of monkeys an infinite number of typewriters, would they eventually produce the complete works of Shakespeare?
I clipped from the Internet one answer to this question (written by David Adam, Science Correspondent of The Guardian):
    "The answer to this, mathematicians assure us, is yes. But now someone has attempted to put the theory to the test. Admittedly the British academics involved in this unusual project did not have an infinite number of typewriters, nor monkeys, nor time, but they did have six Sulawesi crested macaque monkeys, and one computer, and four weeks for them to get creative.

    "The results of this trial at Paignton zoo in Devon were more Mothercare than Macbeth. The macaques - Elmo, Gum, Heather, Holly, Mistletoe and Rowan - produced just five pages of text between them, primarily filled with the letter S.

    "There were greater signs of creativity towards the end, with the letters A, J, L and M making fleeting appearances, but they wrote nothing even close to a word of human language.

    "The macaques should not feel too bad about their lack of productivity, however. Assuming each monkey typed a steady 120 characters a minute, mathematicians have calculated it would take 10 <+>813 (10 followed by 813 zeros) monkeys about five years to knock out a decent version of Shakespeare's Sonnet 3, which begins: 'Look in thy glass and tell the face thou viewest, Now is the time that face should form another.' And that's if they had a computer each."
Our tools and tasks are not the same as aspiring simian playwrights. I still find parallels between this anecdote and brute force that help underpin the paradigm by which I view strategy development in StrataSearch. Without doubt, StrataSearch will create systems equivalent to pages filled with the letter S; I don't claim otherwise. The distinction we need to remember is that such failed outcome is rooted in random chance, not intelligent design. There is little chance a wayward scientist will corrupt our tests by typing a few paragraphs while the monkeys sleep.
Steve
Dacamic
 
Posts: 457
Joined: Wed Nov 30, 2005 12:40 pm

Postby Overload » Thu Jan 25, 2007 9:59 am

See, that's actually a pretty interesting perspective, and it brings up an interesting point about how artificial intelligence (AI) works. While that brute force approach to creating brand new systems has its limitations, it actually has some intelligent design already within it. In particular, the trading rules themselves are based on logical formulas already in use in the industry, and they're assembled according to methods already considered logical. So... that's more like the monkeys having "word buttons" instead of "letter buttons", which greatly increases the likelihood of success.

But more importantly, we're adding an intelligent selection criteria onto the end of the process. This takes us right into the heart of the issue: how does one quantitatively select systems that will perform well into the future? As the OneClick was being created, I asked a number of traders to provide me with an exact formula (or series of formulas) that could be used to guarantee a system would work moving forward. Not one of them could provide me with this, and most in fact couldn't even manage to get their thoughts on paper at all. This is the monumental hurdle we're trying to overcome, and little by little we're getting closer. But it will take a number of additional enhancements to get us fully there.

When we've achieved this objective, we'll indeed be able to use the phrase "so simple a monkey could do it." In the meantime, however, it still takes human input to get it right. Still, I guess we could currently say "so simple a 40-year old could do it", which is probably a lot more honest and true than can be claimed by the "manual" technical analysis software packages.

That reminds me of a story when I worked at a software company in Chicago years ago. Someone in the office was creating a company t-shirt for use in what was then called the Manny Hanny, a 50,000 person 5k run through The Loop. She had put the phrase "Software That Runs Around the World" as a clever advertising logo on the back. One of the managers, however, thought that gave the competition too much credit. Instead, he said, "Software That Runs" already puts you at the front of the pack.

Pete
Overload
 
Posts: 2248
Joined: Wed Nov 30, 2005 12:14 pm

Postby taowave » Thu Jan 25, 2007 10:01 am

Hi Steve,

While we are on the topic of monkeys and randonmess,I would like to bring up an article I just read in Traders World by a trader who has returned 100%+(leveraged and 2 years running) in the World Cup Trading championships.

He trades breakout systems combined with relative strength and EOD.What I found interesting is that he is in the camp that Entries are way overrated and devotes most of his work on exits and running monte carlo's to "optimise position sizing". He stated that he could run successful systems with random entries.

Any thoughts on Random entries,Monte Carlos and position sizing??

Allan
taowave
 
Posts: 584
Joined: Sat Dec 02, 2006 12:39 pm

Postby Barnabyj » Thu Jan 25, 2007 5:07 pm

It's our National Day today over here today so I've had a bit of time to scan the forum. Congratulations on these well crafted and thoughtful contributions. Any timetable for simple 58 years olds?
Barnabyj
 
Posts: 118
Joined: Wed Aug 02, 2006 5:58 am

Postby Overload » Thu Jan 25, 2007 5:19 pm

Any timetable for simple 58 years olds?


Big laughs. But I meant anyone 40 or older can do it. Those under 40 may need to gather a bit more experience first. Sorry, but you're already in the zone.

Pete
Overload
 
Posts: 2248
Joined: Wed Nov 30, 2005 12:14 pm

Postby Dacamic » Fri Jan 26, 2007 1:49 pm

taowave wrote: ... He trades breakout systems combined with relative strength and EOD.What I found interesting is that he is in the camp that Entries are way overrated and devotes most of his work on exits and running monte carlo's to "optimise position sizing". He stated that he could run successful systems with random entries.

Any thoughts on Random entries,Monte Carlos and position sizing??

Although he says he could be successful using random entries, I don't think he's saying we should trade that way. Instead, I'm guessing he's trying to re-emphasize the importance he perceives in those aspects of his systems. In other words, his profitable "random entry" systems could probably be improved by using properly tested non-random entries. As discussed below, however, he might be calling a rose by another name, even though it's still a rose.

Equity prices move randomly at times, and during those times an uptick is as likely as a downtick. Entry strategies are designed to filter that noise and detect favorable movement that is no longer random. Without those filters, we would enter positions having about a 50/50 chance of moving the wrong way; thus, after transaction costs, we would have negative expectancy at the outset. At this point, position sizing won't help us 'cuz we are as likely as not to bump our bet on a losing position. So, exit strategies are our only hope for getting onto the right track. In a random entry system, though, exit strategies perform the initial task of culling randomly moving and adversely moving positions. Of course this is the primary task of entry strategies. Without them, exit strategies have to take on that role and, hence, the basis of my rose is a rose remark.

Whether discussing entries or exits, we're trying to solve "if x, then y, else z". As said above, position sizing offers no advantage while prices are in a "coin flip" stage, because y (good) is as likely a z (not good). So, when using random entries, we're again relying on our "exit strings" to let us know when non-random, favorable movement is occurring and position size should be increased. In other words, a secondary entry assessment must be made before position size can be properly adjusted.

Monte Carlo simulations are very useful for defining a broad range of outcomes, which can help us better understand "if x, then y, else z". I lean heavily on Monte Carlo when evaluating systems built by StrataSearch. I have not used Monte Carlo to evaluate position sizing. It nonetheless seems like it would be useful for understanding probable outcomes at various stages of a position's life.
Steve
Dacamic
 
Posts: 457
Joined: Wed Nov 30, 2005 12:40 pm

Re: Monkeys, StrataSearch & Shakespeare

Postby mandelmus » Thu Feb 23, 2012 6:24 pm

Dacamic wrote:"The macaques should not feel too bad about their lack of productivity, however. Assuming each monkey typed a steady 120 characters a minute, mathematicians have calculated it would take 10 <+>813 (10 followed by 813 zeros) monkeys about five years to knock out a decent version of Shakespeare's Sonnet 3, which begins: 'Look in thy glass and tell the face thou viewest, Now is the time that face should form another.'

... even with that number of monkeys, it is unlikely that they will come up with even the first line of Shakespeare's Sonnet. To understand this challenge properly, we need to consider what actually needs to happen for this to work out. We are looking at the likelihood that any monkey starts typing with the letter "L" (highly likely if we don't consider case), followed by the letter "o" (still highly likely), followed by the letter "o" (likely), followed by the letter "k" (still likely), followed by a space " "(still likely), followed by the letter "i" (still likely, but getting less so) ... the correct outcome is dependent upon each correct next letter typed, the longer the monkeys type, the less likely they will follow the Sonnet exactly. This is the same reason that while the probability of getting any number combination in a lottery is the same, the likelihood is not. For example, for a pick-5 lottery, which number set is more likely: "7-16-17-39-51" or "1-2-3-4-5-6" ... the probabilities are the same, but the likelihoods are not.

If you have a lot of time and patience, more detail explained here ... http://saliu.com/
mandelmus
 
Posts: 141
Joined: Fri Dec 23, 2011 1:51 pm

Re:

Postby mandelmus » Thu Feb 23, 2012 6:34 pm

taowave wrote:Hi Steve,

While we are on the topic of monkeys and randonmess,I would like to bring up an article I just read in Traders World by a trader who has returned 100%+(leveraged and 2 years running) in the World Cup Trading championships.

He trades breakout systems combined with relative strength and EOD.What I found interesting is that he is in the camp that Entries are way overrated and devotes most of his work on exits and running monte carlo's to "optimise position sizing". He stated that he could run successful systems with random entries.

Any thoughts on Random entries,Monte Carlos and position sizing??

Allan


Found the article at Kevin Daley's site ... http://www.kjtradingsystems.com/Interview.pdf (opens a pdf) ... although he does say, in the article, that his "best performing system has a breakout indicator and a relative strength indicator for entry and average daily range calculation for exits."
mandelmus
 
Posts: 141
Joined: Fri Dec 23, 2011 1:51 pm


Return to Curve-Fitting and Other Pitfalls

cron