Further thoughts on Backtesting

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
Post Reply
Chris67
Roundtable Knight
Roundtable Knight
Posts: 1046
Joined: Tue Dec 16, 2003 2:12 pm
Location: London

Further thoughts on Backtesting

Post by Chris67 » Wed Jun 30, 2004 9:32 am

I put a piece on here a few weeks ago about walk forward testing .. I would love to hear more feedback and myself have some further thoughts ...

I still dont see how walk forward testing can be too heavily criticised. If you take a 10 year stretch of data to be tested , for example , 1990 - 2000 .. you then design a system that seems to work well .. perhaps an MAR of 2. You then roll it forward and test it on 2000-2004 and you get an mar of say 0.6 and the performance seems to have fallen dramatically away.. is this still a good system and if you hadnt of tested it on an out of sequence length of data surely you'd never really know if you optimised and if the sysytem works going forward.

What I am trying to do is to design something on 1990 - 2000 and then try it on an out of sample 4-5 year data period and see if i get a semblence of robustness .. i.e perhpas mar declined to 1.2 .. it my book this could still be a good system.

One other final thought .. sure this wll stirr a few people up .. I seriously think its a bad idea to backtest any form of trendfollowing system on the years 1980 - 1989 ... i think a one armed , blind lobotomised monkey couild have made money trendfollowing with any kind of system in this period since the whole nature of markets has now changed with the onset of computers and trendfollowing and communications .. Agreed people human nature nver changes and thats what drives markets but the characteristics of markets when you have several hundread thousand trendfollowers in the market do change I believe every 10 years or so .. this would back up Richard Dennis in his assesment that its only worth backtesting over a 10 year period and not 20 which is percieved wisdom ..

Any thoughts warmly appreciated.

Chris

Ted Annemann
Roundtable Knight
Roundtable Knight
Posts: 118
Joined: Tue Apr 15, 2003 7:44 pm
Location: Arizona

Post by Ted Annemann » Wed Jun 30, 2004 11:36 am

One thing you can do is ask others for their opinion about your opinion.

Another thing you can do is formulate a hypothesis, design an experiment to test your hypothesis, perform the experiment, analyse the resulting experimental data, and draw a conclusion.

For example, you can hypothesize that a trading system which scores well on the Yhaatzup Mqyy test protocol, will score poorly on the Bwuso test protocol. Your experiment could be to generate 100 trading systems at random (perhaps by choosing 100 sets of the 2 parameters in Veritrader's Dual Moving Average system) and then test these 100 systems on both the Yhaatzup Mqyy test protocol, and also the Bwuso test protocol. Now you can analyze your data to see whether a statistically significant number of poor performers on Bwuso, scored well on Yhaatzup Mqyy.

Perhaps Yhaatzup Mqyy = "from 1980 to 2004"
Perhaps Bwuso = "from 1994 to 2004"
but this is totally up to you. It's your experiment, you get to define it however you wish.

verec
Roundtable Knight
Roundtable Knight
Posts: 162
Joined: Mon Jun 09, 2003 7:04 pm
Location: London, UK
Contact:

Re: Further thoughts on Backtesting

Post by verec » Wed Jun 30, 2004 6:30 pm

Chris67 wrote:I seriously think its a bad idea to backtest any form of trendfollowing system on the years 1980 - 1989 ... i think a one armed , blind lobotomised monkey couild have made money trendfollowing with any kind of system in this period [...]
It may be so for the very reasons you outline, but then you have to question whether 1990-2000 is any better! Think: a bull market you see twice a century! Do you really think that limiting your data to those 10 years is going to help your system, even for the known 2001-2003 time frame?

Ted has a point when he says (roughly paraphrased): "test and test again", though opinions are interesting when it comes to decide what to test.

The idea that 10 years back testing may provide better "tuning" than 20 years is interesting, precisely because it is counter intuitive to most very long term trend followers. This doesn't make it right, but a good candidate for testing.

imhO :roll:

ksberg
Roundtable Knight
Roundtable Knight
Posts: 208
Joined: Fri Jan 23, 2004 1:39 am
Location: San Diego

Time Periods

Post by ksberg » Wed Jun 30, 2004 7:31 pm

One could also counter that if you write self-adapting systems, then any timeframe is fair game.

I think the question of period has to be seen in context of which markets. The S&P made a significant change in 1997 (outside the scope of 80-89). Bonds characteristics have also undergone change. At a recent, finer-grain level, the DAX has picked up day-trading volume as S&P and NASDAQ volatility has declined.

I've recently heard the term "commodity long wave" to describe the shift in markets from equities to commodities. Many believe equities will look more like 1965-1975, and commodities are likely to have higher volatility. (recall that the last huge commodity boom took place in the 70's). So, heck, I may want to know how well my system does on 1970's type data (some 30 years back).

I have one question: why would the next 10 years be anything like the prior 10 years? Yes, the most recent periods share technology (ECN's, DAT, global pricing), but there are likely to be very different fundamentals at work.

Maybe it makes sense to synthesize data based on chronological proximity and fundamental market characteristics, and then back test. This would be the time-series equivalent of morphing two images. For instance, the future might bring long trends of significant price levels, but with added daily volatility, fake-outs and price extensions.

As in, just when you thought it was safe to enter the water, the market will be there lurking to eat your trading system for lunch.

Cheers,

Kevin

damian
Roundtable Knight
Roundtable Knight
Posts: 815
Joined: Tue Apr 15, 2003 8:43 pm
Location: dusseldorf

Post by damian » Wed Jun 30, 2004 7:53 pm

Ted Annemann wrote:One thing you can do is ask others for their opinion about your opinion.

Another thing you can do is formulate a hypothesis, design an experiment to test your hypothesis, perform the experiment, analyse the resulting experimental data, and draw a conclusion.

[.....blah blah, roll eyes, blah.......]
Chris, another thing you can do is continue to ask perfectly fair questions on a perfectly open public forum about topics that are perfectly within the bounds of forum policy compliance without the risk opf being trodden on with a well seasoned and tired line delivered by the all too predictable members of the forum (whom have obvious talent in this area yet choose to pounce on anyone that dares to ask a question). I learn by trying, asking, trying, failing, asking, trying, failing, asking.... and also hearing what others ask. It seems that this is not a popular technique here. Luckily it was encouraged at my university and my excellent place of employment. But then I am not MIT alumni....

I have more than slowly gone off this forum for no other reason than quite fair discussion is crushed by elitists who try to make a reasonable point yet sacrifice an openly conversational atmosphere in the meantime. It is a shame as many of the more 'elite' members have much to offer. It is quite fair that they choose not to offer but please, stop crushing legitimate contributions in the meantime.

Chris, Verec quoted a part of your post above and I partially agree with you. What follows is purely discusional in nature, not psuedo science.

I have lost money for a number of reasons. One is that I was conditioning (optimising?) my system to make money from trends. Is this not logical? As a trend follower I want to profit from trends and that is exactly what my system does very well, so long as 'the markets' actually trend as many did in the time period that you quote. During other time periods I was trying to surf wave type A with surfboard type B... and wiped out. Most forms of pure backtesting are postured to highlight how your system can make money when 'the markets' perform in the way that you knew they did.

I am usually forced to run a one man campaign so I understand if others think little of my opening paragraphs in this post. No problems, so be it.

damian

Kiwi
Roundtable Knight
Roundtable Knight
Posts: 513
Joined: Wed Apr 16, 2003 1:18 am
Location: Nowhere near

Post by Kiwi » Wed Jun 30, 2004 8:50 pm

Heck Damian,

I do have difficulty understanding some of your posts. This may be caused by a shortening of my attention span as I apply discretionary trading to shorter and shorter term markets. I am now paper trading Hang Seng using 13 tick bars -- which take about 20 seconds to complete. If system trading is like chess and day trading is like flying a flight simulator then trading the Hang Seng is like playing Doom. Live and Die quickly :)

Damian's point is good though. Its really easy for those of us who have been on this and other forums for a while to get dismissive of the questions that get asked repeatedly. Often they have been answered. Often to those who have done a lot of work they seem to be lazy. But I think we forget how little we once knew.

We need to be kind to new entrants and not move this forum in the direction of tradersclub or, god forbid, Elite (puke) Trader where monstrous egos prowl the board trying to build themselves by attacking others and their ideas.

So let us apply our undoubted intelligence to being kinder to others rather than following that vicious spiral of ever increasingly clever put downs.

John


PS. Ted, Not that I thought your post was anything more than a little edgy, you might want to join me over at Elite Trader. My sport is attacking guys with names like Nitro and Jack Rabid --- they really are quite a challenge. It does make me wonder about my own psyche but its addictive. Then I come back here to get clean again.

Kiwi
Roundtable Knight
Roundtable Knight
Posts: 513
Joined: Wed Apr 16, 2003 1:18 am
Location: Nowhere near

Post by Kiwi » Wed Jun 30, 2004 10:47 pm

Ok ... perhaps I should try and give my opinion on Chris's question.

I personally have taken the approach of dividing the market time up into segments. I take Damians point and like to have segments that represent different market conditions. If I was really serious about this (I am not now as I trade only one system now and all my other money is discretionary which is the ultimate in walk forward continuous adaptation system) then I would take this approach.

1) I would decide what kind of markets I wanted my strategy to make money in.
2) I would decide how I would stop it hemoraging in the other kinds of markets ... or what level of blood loss I would accept as part of a diversified system portfolio.
3) I would construct contracts that represented the type of markets I planned to make money in (by splicing backadjusted contracts together in excel that were representative of those periods). I would construct perhaps 5 different types of artificial markets using historical data.

4) I would build my system on the good data fine tuning it as much as seemed reasonable.

5) Then I would test it on real markets and on 20 year periods and everything in between .... but I would save some of them for later testing. This would tell me how good or bad things were likely to be and I would adjust the basics.

6) I would repeat these cycles and then try walk forward testing on my reserved datasets.

At the end of this process I would have a strategy which was tested on historical, artificial and then walk forward data. I would also know what market conditions had what impact on my system (at least historic examples). I could then monitor performance against these to determine if the real time performance was similar to historic situational performance. This is one reason why I would never ever trade another black box system like Synergy.

Getting back to my original statement about being discretionary. I am a rule follower and my rules were tested in this fashion ... so I stop trading when the short term market behaves in certain manners and start again when a defined event happens. I think that my discretionary trading is much better for starting as a systems trader because it taught me about rules and about money management. But most importantly it taught me about testing and about ideas for taking money from the money tree.

My thoughts :)

ksberg
Roundtable Knight
Roundtable Knight
Posts: 208
Joined: Fri Jan 23, 2004 1:39 am
Location: San Diego

Re: Further thoughts on Backtesting

Post by ksberg » Wed Jun 30, 2004 10:47 pm

Chris67 wrote:I still dont see how walk forward testing can be too heavily criticised. If you take a 10 year stretch of data to be tested , for example , 1990 - 2000 .. you then design a system that seems to work well .. perhaps an MAR of 2. You then roll it forward and test it on 2000-2004 and you get an mar of say 0.6 and the performance seems to have fallen dramatically away.. is this still a good system and if you hadnt of tested it on an out of sequence length of data surely you'd never really know if you optimised and if the sysytem works going forward.

What I am trying to do is to design something on 1990 - 2000 and then try it on an out of sample 4-5 year data period and see if i get a semblence of robustness .. i.e perhpas mar declined to 1.2 .. it my book this could still be a good system.
Any thoughts warmly appreciated.

Chris
Chris,

Have you looked into Start-Trade analysis (also called Worst-Case Analysis)? WCA is essentially micro walk-forward testing with overlap. Then, you can examine the distribution of statistics (Net Profit, CAGR, DD, MAR, etc). I find it offers far more insight into system behavior than comparing two periods. Not that the two-period walk-forward isn't useful, I just find the WCA type of testing more informative.

Cheers,

Kevin

Chris67
Roundtable Knight
Roundtable Knight
Posts: 1046
Joined: Tue Dec 16, 2003 2:12 pm
Location: London

Post by Chris67 » Thu Jul 01, 2004 1:33 am

Thank you all for your opinions and responses .. it all helps ...

Cheers , Chris

Forum Mgmnt
Roundtable Knight
Roundtable Knight
Posts: 1842
Joined: Tue Apr 15, 2003 11:02 am
Contact:

Post by Forum Mgmnt » Thu Jul 01, 2004 3:23 am

Interesting thread on two accounts.

First, having worked with many different types of people over the years, I'm pretty used to criticism, strong opinions, and the diverse communication styles. There is much to be learned from what may first appear to be harsh or unwanted direction.

Those of us who've been around could benefit from examining, yet again :roll:, the same issues with the eyes of the neophyte.

Those of us who are learning could try to see the reasons behind the seemingly curt or dismissive responses of the more experienced.

I fairly regularly read Elite Trader myself. I find this place to be a refreshing change. Even so, lest we get too far down that same path, we need to actively fight the human tendencies that led to the demise of that and other forums. The one quality we can all develop which will help is tolerance. :)

If we try to look behind the words and think first that someone who spends the time to type a response to our message is probably trying to help us learn something they feel is valuable, something they probably spent a lot of time and effort to learn, we might get closer to the truth.

Those who are forceful and constant advocates of testing are so for a reason, some of us may be less empiricistic and more comfortable with discussion and debate first, but we should not for one minute discount the truths of those who advocate testing. After all, in the end if you are the one who has to pull the trigger on a trade, you need to have your own personal reasons to be confident in your approach.

As in most areas of conflict, intellectual and otherwise, I see value in both perspectives.

Second, back to the matter at hand. :idea:

Often it is easy to get so caught up in the method or approach that one loses sight of the goal itself.

There is certainly benefit to using an approach that involves some out-of-sample testing. However, I think it's more important to understand the reasons why you might want to do out-of-sample testing to begin with as there might be other, perhaps even better, approaches for accomplishing these goals.

Out-of-sample testing is used to make sure that one isn't drawing conclusions based on a nonrepresentative sample. This kind of statistical approach grew out of the need to draw conclusions when it is completely impractical to test the entire population.

When a quality engineer wants to make sure that bolts are strong enough, they can't test every single bolt or they won't have any left to sell, so they need to test a sample of the bolts. When a pharmaceutical company wants to test a cancer drug, they can't test it on every cancer patient in the world, the need to test it on a sample.

Out-of-sample testing is a way of measuring whether or not the conclusions one might draw about one particular sample is likely to apply to the population as a whole (i.e. all the bolts, or all the cancer patients).

Now before we get back to trading, I want to ask a question. Overlooking the practicality of performing the test, which would let you draw a more accurate conclusion about future bolts produced a factory or the effect of drugs on new cancer patients; a test made of one sample which is then confirmed with a second out-of-sample test; or a test of the entire population of all bolts produced to date or all cancer patients that have ever had a given form of cancer?

Look at this another way. Tests with large sample sizes are better than tests with smaller sample sizes. Given identical sample sizes, a test with an out-of-sample confirmation is better than one without. Taking this to it's logical conclusion, the best possible test is one with the largest possible sample size with an out-of-sample confirmation. The largest possible test therefore includes testing the entire population.

Back to trading and the point of all this, we have to ask ourselves the question: Why am I doing all this? What is this testing for? Why am I spending so much time reading and posting on forums?

For my testing, the point is to improve my chances that I will be able to make money, no matter what the future brings.

So while I believe the concept of out-of-sample testing is valid, I think there is a better approach, to test the entire population available.

So while I understand the rationale behind it, I'm not a big fan of "walk-forward" testing.

This is especially true in the context of what's happened in the last year or so. I believe you should test for the worst periods and since the last year has been tough for most systems, excluding the last few years in your system development will almost always cause you to get systems that perform well and degrade in the out-of-sample case (i.e. in the last few years).

Here's what I do instead:

1) I test and build systems over the last 10 to 12 years or so depending on the markets, I want to capture all of the markets various phases. Also, the last 10 to 12 years have been tougher on trading systems than the previous 10 to 12 years and I'd rather design for adversity.

2) I then test the same system parameters over the previous 10 to 12 years to see if it is much different. If it is, that throws up a big red flag.

3) Test over the entire history (for most markets this includes both of the above 20 to 25 years) to see if it is much different. Again, If I do 1) and 2), it would be rare for this to be different but I like to run the test anyway.

So in some sense, I'm doing a sample, an out of sample-confirmation, and a test of the entire population.

Now back to the goal, this is just one part of a complete regimen for robustness testing, the others would include:

1) Parameter Sensitivity Testing - Checking how sensitive the results are to the particular values that were optimized in the test. Randomly varying the test parameters and checking the impact on results is an important check and fodder for another topic.

2) Market and Trader Order Testing - Whether you are talking about Monte-Carlo analysis, Worst Case Analysis, etc. these basically come down to creating an artificial set of market conditions or trade outcomes from the original markets and trades to help give one a better picture of the potential variability of results due to market order.

Okay, enough of my pontifications.

Thanks Chris for stimulating an interesting thread.

- Forum Mgmnt
Last edited by Forum Mgmnt on Thu Jul 01, 2004 11:24 am, edited 1 time in total.

Hiramhon
Roundtable Fellow
Roundtable Fellow
Posts: 98
Joined: Fri May 09, 2003 12:45 am

Post by Hiramhon » Thu Jul 01, 2004 8:22 am

The chasm between word and deed is so remarkable that I shall remark upon it. For the former we have
I learn by trying, asking, trying, failing, asking, trying, failing, asking.... Luckily it was encouraged at my university and my excellent place of employment.
We need to be kind to new entrants
Those of us who've been around could benefit from examining, yet again , the same issues with the eyes of the neophyte
For the latter we have the thread "What system is best" in the Total Newbie area. Strikingly, none of the above authors have contributed anything at all to this thread. Here's a chance to demonstrate devotion to the ideals of sharing and teaching, but no one has practiced what they preach. Interesting.

Forum Mgmnt
Roundtable Knight
Roundtable Knight
Posts: 1842
Joined: Tue Apr 15, 2003 11:02 am
Contact:

Post by Forum Mgmnt » Thu Jul 01, 2004 9:16 am

Hiramhon,

Well, there's always room for a good curmudgeon in any healthy debate. :shock: I don't mean offense by that, I truly value the curmudgeon's role as a catalyst for rethinking one's positions.

I lay no claim to sainthood, selflessness, or pure altruism, never have, never will.

I'm more of a synergy or sybiosis person. I like the mutually beneficial. After all, I do what I want because I enjoy it or because I have to; put another way, I never do what I don't enjoy unless I have to.

If you'll reread the comment you quoted from my post you'll notice that I said we "would benefit" from revisiting old topics, I didn't say it would be fun all the time, and I certainly never claimed to be the savior and respondent for all who ask the same questions we've answered over and over again.

Like many people, I prefer new discussion and interesting questions. I know, however, that I benefit and learn by reexamining my own assumptions and beliefs and by challenging their core. We can only do this if we try to think things through. The process of explaining things to neophytes is helpful here, not necessarily always enjoyable.

Sometimes I like helping new people, sometimes I don't. I tend to enjoy helping those who exhibit some level of effort and who are attempting to learn on their own more than helping those who appear to be looking to others to do their work for them.

I also tend to want to help those who are asking interesting questions.

I find it interesting that you find hypocrisy in our perspectives :?

- Forum Mgmnt
Last edited by Forum Mgmnt on Thu Jul 01, 2004 11:21 am, edited 1 time in total.

Jake Carriker
Site Admin
Site Admin
Posts: 1493
Joined: Fri Sep 12, 2003 10:32 am
Location: Austin, Texas

Post by Jake Carriker » Thu Jul 01, 2004 10:27 am

Here is an interesting qoute from Michael Clarke's description of his trading programs. If someone is looking for opinions from those that have been there and done that, then this is one. Clarke certainly qualifies as a succesful trader that is extremely versed in mechanical system testing. He has a very respectable track record and manages better than $275 million.
CCM's philosophy when developing a model is to test it using a large pool of commodity interests, approximately 105, with data as far back as 1945. In order for the model to be accepted into CCM's portfolio of available models, it must trade all 105 markets using the same rules and parameters and the results should indicate excellent performance characteristics for the vast majority (at least 90%) of the markets and for the group of markets as a whole. Also no model is accepted unless it shows stability of performance during tests involved with shifting parameters and altering rules. Much effort has been expended in developing tools to assist in this effort to assure robustness of the models. I consider the software we have developed in this area to be one of our edges in the markets. With regards to out-of-sample testing CCM no longer performs this test. I noticed that if the length of the out-of-sample period was longer that the average length of time that the particular model might be expected to have a mediocre or poor performance, then the effective parameters and rules were very similar to the test sample.
You can find the rest of the piece at www.iasg.com in the manager database. An interesting perspective.

Best,
Jake Carriker

leonardo
Roundtable Knight
Roundtable Knight
Posts: 100
Joined: Wed Apr 30, 2003 2:27 pm
Location: Minneapolis, MN
Contact:

Post by leonardo » Thu Jul 01, 2004 11:56 pm

I wish to include my few cents worth.

I believe in mechanical methodologies for attacking markets, and believe that you can create a fairly durable system which will work over long periods of time.

In addition to my own usual forms of trading I trade a very old system which I received when I was 25, back in the early '80's. I got it from an quaint 85 year-old oat trader. Having successfully traded oats (it was a point of pride with him) and other grains with it for more than 50 years, he allegedly had gotten the rules from another old trader who traded on the cotton exchange in New York in the first part of the century.

Six months before he died, and after closing out a particularly good oat trade--he came into the brokerage house where we traded and said he was done trading. After jokingly mentioning that the oat market would never be the same again without him; he quietly told me that he had something to show me. He had a copy of his rules that he traded by and he wanted me to have it. He also said; very seriously, that I could take over for him in the oats. :)

Here we had tons of great markets in the '80's and he thought that I might prefer oats. (There is a lesson in this story for this aspect alone. The gentleman had been trading for most of his life in markets which were on the wane by the eighties. Are we trading for money or ego? Or just because we are comfortable in trading something which might not be appropriate any longer?)

He explained that he figured that I was the only person he had met in years that he thought could or would use it. When I asked him, "why me?", he said it was because I was the only other person he saw brazenly buying new highs/selling new lows on any and all markets. His system was a unique breakout system which, of course, depended on buying new highs.

While this true story appears that it would make a great infomercial, the Oat Trader's Breakout System wasn't/isn't the holy grail. No system is. But then again, it's not a bad system.

The rules are based on an observation many years ago of market action in a special situation which happens every couple of years in any market, and uses hard and fast rules to trap the market into giving up profits when the situation develops.

Fortunately, it works just as well on our new markets as it does on oats. It doesn't trade that often, but when it does---you are dealing with a market which is going somewhere. After programming it, I've found that it has good and "less-good" times in all markets and you have to be extremely consistent in your trading of it if you want to get to the profits.

Just like the Turtle system and any other good durable system.

The point is: systems which are based on a correct theory of market action tend to be robust. They tend to work equally good/bad over most markets. The more data you have tends to reinforce the point. A system with exceptional results just hasn't been trading long enough.

Live long and prosper.

Leonardo---

enigma
Roundtable Fellow
Roundtable Fellow
Posts: 99
Joined: Tue Apr 29, 2003 6:56 am
Location: UK

Post by enigma » Fri Jul 02, 2004 4:53 am

Leonardo, always enjoy reading your posts :) Love the trading stories and insights... thanks!

Forum Mgmnt
Roundtable Knight
Roundtable Knight
Posts: 1842
Joined: Tue Apr 15, 2003 11:02 am
Contact:

Post by Forum Mgmnt » Fri Jul 02, 2004 7:28 am

I wanted to clarify one thing.

The reason that I generally test in two parts, i.e. the last 12 years, then the previous 10 to ? (depending on the markets history) is NOT because I want an out-of-sample test.

It is because a) the more recent markets are slightly worse and I believe a more conservative indicator of what we might get next year, b) it is faster doing simulations over a shorter timeframe.

If tests were instantaneous, I'd test over all history every time and merely report the results for the last 10 years as additional detail for the larger test, much like I get annual summaries right now.

I believe that testing over all history available is the best way to get a system that will hold up over time. Any strategy that works over many markets and many years will of necessity gain its edge from market behavior based on a human psychological characteristic that is unlikely to go away, and for this reason such a strategy will be very likely to continue to work in the future.

- Forum Mgmnt

verec
Roundtable Knight
Roundtable Knight
Posts: 162
Joined: Mon Jun 09, 2003 7:04 pm
Location: London, UK
Contact:

Post by verec » Sun Jul 04, 2004 8:15 pm

Forum Mgmnt wrote:Any strategy that works over many markets and many years will of necessity gain its edge from market behavior based on a human psychological characteristic that is unlikely to go away, and for this reason such a strategy will be very likely to continue to work in the future.
Isn't this beleif based on the assumption that most trading is discretionary and not system based?

In other words, what is your level of confidence that your testing results will still give you some reasonable edge the day that most trading will occur through mostly never overriden computer based systems, assuming that such a day will come?

si
Senior Member
Senior Member
Posts: 48
Joined: Sun May 30, 2004 4:39 pm
Location: Austin, USA

Post by si » Mon Jul 05, 2004 12:27 am

verec wrote:... the day that most trading will occur through mostly never overriden computer based systems, assuming that such a day will come?
"Large" traders are ultimately constrained by client behaviour. Can client deposit and withdrawal of funds be controlled or modelled by computers? Or is it likely to be psychological? What about institutions where a manager does not want to stick his neck out in appearing to take large risks and would rather be considered conservative and safe? Would every such manager have the freedom to let his automated program stay on course, even after say a 40% drawdown?

The markets change, the market mechanics change, the participants change but on the average, (IMHO) the net behaviour remains the same.

verec
Roundtable Knight
Roundtable Knight
Posts: 162
Joined: Mon Jun 09, 2003 7:04 pm
Location: London, UK
Contact:

Post by verec » Mon Jul 05, 2004 5:32 am

si wrote:Can client deposit and withdrawal of funds be controlled or modelled by computers?
That's an excellent point.

Although, at least rhetorically, one could argue that that is precisely what trading systems are all about ...

But I do get your point, and share the beleif, that the day every single investor will use and apply his own trading system, as opposed to trusting his money with a fund manager, is probably never going to happen.

Sorry I missed that :oops:

Forum Mgmnt
Roundtable Knight
Roundtable Knight
Posts: 1842
Joined: Tue Apr 15, 2003 11:02 am
Contact:

Post by Forum Mgmnt » Tue Jul 06, 2004 8:57 am

Human psychology decides how much money to give computers, when to increase allocations, when to decrease allocations, how much data to use during testing and training, etc.

Humans turn on the switches, not vice-versa.

Another factor is fund performance and the money flows at the macro level. Investors, especially institutional investors, do what worked well last year and the last few years, this generally means that lots of money tends to come in at the wrong time causing cyclical performance patterns for strategies at the macro level.

- Forum Mgmnt

Post Reply