Thoughts for setting a One Click "envirnoment"

StrataSearch will find what you tell it to, but what should that be? Will the same criteria work across different Sectors? Different time periods? Here we discuss the ins and outs of OneClick Searches.

Thoughts for setting a One Click "envirnoment"

Postby TomHam » Sat Mar 14, 2009 3:32 pm

First I must state that I Love StrataSearch; wish I would have found it sooner. :)

I thought I would share my obsevations of using S-S over the past 2 months, and maybe get a discussion going with other users. My first thought(s) are on what type of envirnoment to set-up with a "typical" One Click, or atleast things to consider when doing so. (If interest prevails, perhaps we can next discuss the actual criteria next.)

So, a few observations:

1) the Time Period -At first I thought it would be "clever" to select an evaluation time period that covered everything. That is, both Up and Down markets. I really have stopped trying to run this because I found significant instablity going foward.

What I mean by that is setting an evaluation period and leaving out, say, the last 6 months; it would be set for my Out of Sample (OS) period. What I found was great results during the evaluation period (In Sample, IS) but slowly degrading performance (to mediocore) in the OS period.

What I do now is determine the market envirnoment first (Bull/Bear) and set my evaluation period (IS period) accordingly. So far, this seems much more robust going forward (both real time and in OS testing).

Thoughts or comments from other users ?

2) The Sector / Family to evaluate-

This is a more wide open and subjective topic. Obviously, everyone has a different objective, so I'm being rather general here. Bear with me.
How many securities to include ? I think much of this comes down to how collelated they are to eachother and, are they liquid ?

Correlation: We don't need 3 BioTech ETF's, or (likely) 10 software compaynys in most sector / families (yes, there are exceptions). If we are looking for general market sector movements this holds, if we're looking in a more focused methodology, it may not. With a more focus methodology you already have a good "handle" on the specifics of what you're looking for, and probabaly have done a "pre-screen" of candidates before submitting them to S-S anyway.

Liquidity: Having a reasonable idea of how orders get filled (and how orders get a poor fill) I want to chose securities that have liquidity. If the market can not regularly absorb my order without affecting the price, then I better use an order desk to meter in my order or prepare to spend time doing it myself (slowly) or find a more active security. I define "liquidity" as what the average dollar volume is (price x volume), not just the number of shares traded.

So S-S community, here are some thoughts. Please share yours, especially those obtained by testing or actual experices. S-S is such a neat tool, that it takes ideas and suppositions, and turns them (atleast) toward reality. I'm learning that a number of things that I thought were valid, really don't hold up very well when subjected to exhaustive testing.
Cool ! (no, it's way cool!) Take Care & Good Trading ! ............. Tom ...............
TomHam
 
Posts: 18
Joined: Sun Jan 18, 2009 1:41 pm
Location: near Detroit, Michigan

Re: Thoughts for setting a One Click "envirnoment"

Postby Overload » Sun Mar 15, 2009 9:59 am

I've got a few comments I can add to that:

1) the Time Period

I agree with you on this one. I think it's important to keep in mind that even the most successful back test doesn't mean a system is prepared to handle scenarios that weren't even in the back test. If your system has been back tested against a 100% bull market, and it back tested exceptionally well, one shouldn't assume that it will perform well in a bear market. This is an extreme example, but I think it describes it well. There are many types of markets, or unique market scenarios, and it's impossible to know how they'll perform moving forward unless something similar has been viewed in a back test.

With this in mind, it's important that the out-of-sample data not present any surprise scenarios that weren't in the in-sample data. While there are certainly reasons to test your system against as many scenarios as possible, doing this in the out-of-sample test is really quite meaningless. During the out-of-sample test, you should really only be testing whether the system performs well in scenarios in which is was trained to perform.

As a brief example, it was tempting for a while there to build a system on 2004-2007 data, and then forward test it into 2008. In retrospect, we can see that this doesn't make much sense. 2008 data is entirely different than 2004-2007 data, so it wouldn't be expected that a system would perform well in that out-of-sample test. If one wanted to use 2008 as the out-of-sample period, building the system against 2000-2002 data might be a better choice.

2) The Sector / Family to evaluate

I tend to like a larger sector, as more symbols provides more scenarios to exploit. But a larger sector can also mean a slower search. For example, searching for a system against the Russell 2000 might take weeks to reach completion. However, there is a trick that can be used to increase the speed when evaluating larger sectors.

In a basic setup using my example, a search is done against the Russell 2000 sector itself, and each and every new combination must be run against those 2000 stocks. As we've seen, however, the majority of the combinations do not create profitable systems and are immediately discarded. Thus, there is a great deal of inefficiency running those ineffective combinations against the entire 2000 stocks.

The trick is to instead use a sampling. Using the Screener, you can create a random sampling of the Russell 2000 that is perhaps only 200 stocks instead of 2000. The basic OneClick Search can then be run against these 200 stocks, getting a quick indication of whether the combination may be worthwhile or not. If it passes that test, an Alternate Data Analysis can then be run against the entire Russell 2000 to see if the system performs as well against the entire sector as it does against the sample. This could even be made a more sophisticated by creating several Russell 2000 samplings, and running Alternate Data tests against each before finally testing the complete sector.

While this approach can greatly speed a search, there is at least one caveat. In particular, evaluating the complete sector as an Alternate Data test means that there won't be as many evaluation fields available against that test. And that means it will take a little more manual confirmation when all is said and done. To make this a little easier, it's probably worthwhile to compare the sampling of 200 stocks with the complete sector prior to even beginning the search, just to be sure the sampling is representative. If it turns out that the sampling isn't truly representative of the sector as a whole, then the technique might not prove very effective. Regardless, if done properly, this is a technique that can greatly increase the speed of a search against a large sector. Hope it helps.

Pete
Overload
 
Posts: 2246
Joined: Wed Nov 30, 2005 12:14 pm

Re: Thoughts for setting a One Click "envirnoment"

Postby Dacamic » Sun Mar 15, 2009 4:45 pm

Maybe we should coin a new phrase: discretionary systematic trader. This oxymoronic label can be our admission that significant consideration must be given to when and where trading systems are used. Rumors about systems that perform well in all markets and conditions occasionally circulate within an Ivory Tower located near the unicorn petting zoo on Atlantis. Beyond that mythical realm, though, us mere mortals are faced with questions of the nature discussed by Pete and Tom.

Time Period

I agree with what has already been said in this thread: test long systems with "bull market" data and short systems with "bear market" data. The dilemma created by focusing system development in this manner is traders might need to manually activate and de-activate systems in live trading depending upon their forecasts about market conditions (which is what I do). To mitigate risk associated with less-than-perfect timing, systems should be tested in unfavorable conditions to understand how they perform when the wind is blowing against them. The ideal probably is for systems to be quiet when conditions are not favorable for them; naturally, significant drawdowns are at the spectrum's other end. Where a system falls within that range is a factor in deciding how much effort a trader should dedicate to evaluating near-term market conditions.

Sector

To comment upon the easier point first, I use two filters related to liquidity:

    // Minimum Price Requirement (Supporting Entry)
    (close > 3); and,

    // Minimum Liquidity Requirement (Supporting Entry)
    (mov(volume, 20, simple) * close >= 200000).
Certainly, these two filters could be combined to create a dollar-volume-based filter, but I want my trades to clear these two hurdles separately.

As a general rule, I use stocks in the Russell 2000 for my long system and only the SPY ETF for my short system. The liquidity, diversity and simplicity associated with Index- and sector-based ETF's are attractive; however, they tend to be highly correlated with each other, and so using more than one such ETF seems unnecessary. My choice to settle on SPY for my short system (actually, SH in live trading) is driven by personal preference, a result of my obsession over the years with the S&P 500 index. I could likewise use SPY for my long system, but instead seek additional return from stocks moving faster than their index. The Russell 2000 satisifies my subjective (possibly irrational) criteria for depth and diversity, while still tending to be highly correlated with the S&P 500.
Steve
Dacamic
 
Posts: 457
Joined: Wed Nov 30, 2005 12:40 pm

Re: Thoughts for setting a One Click "envirnoment"

Postby TomHam » Sun Mar 15, 2009 5:42 pm

Thanks Steve & Pete for your opening comments. Hopefully, we'll get a few more in here.

Steve does direct us to a topic I'd like to comment about later; namely when & how do we know that a (previously developed) strategy is "broken". It likely will "break" because markets, and condidtions change from those that were used to develop the original strategy set.
But, let's give this topic a change to grow with other comments first. Then, perhaps we can take on the selection of criteria to let S-S evaluate (that will be intersting & subjective) :)
Take Care & Good Trading ! .............. Tom ..............
TomHam
 
Posts: 18
Joined: Sun Jan 18, 2009 1:41 pm
Location: near Detroit, Michigan

Re: Thoughts for setting a One Click "envirnoment"

Postby taowave » Mon Mar 16, 2009 10:11 am

Hi Steve and all,

As a "systematic discretionary" trader,I must ask the "discretionary systematic trader" why he/she would feel compelled to manually activate and deactivate systems based on their forecasts.It would appear to me that at that point,you have moved into my camp,and if one is successful it is due to having the ability to call market direction.

If one is a "capable" systems trader,why not run a portfolio of long short systems and eliminate the "forecasting" aspect?? Personally,it makes my hair stand up on end to think of system traders calling market direction:)

Hopefully,the combining of systems will lead to a sharply lower volatilty of returns coupled with a somewhat lower return.At that point leverage may be prudently applied.

I havent visited this issue for some time,but if I remember correctly,SS does not have the capability to assign/force a ratio of long systems to short systems within a portfolio.If I was a pure and capable systems trader,I would not employ discretion but instead I would run Multi systems,long and short simulaneously.It does not need to be a market neutral book,but there must be a mix of long and short positions.In a peverse way,SS has led me to the conclusion that anything other than that is an outright market bet,and IMHO,your activation/deactivation concept is another way of saing the same thing

Allan





Dacamic wrote:Maybe we should coin a new phrase: discretionary systematic trader. This oxymoronic label can be our admission that significant consideration must be given to when and where trading systems are used. Rumors about systems that perform well in all markets and conditions occasionally circulate within an Ivory Tower located near the unicorn petting zoo on Atlantis. Beyond that mythical realm, though, us mere mortals are faced with questions of the nature discussed by Pete and Tom.

Time Period

I agree with what has already been said in this thread: test long systems with "bull market" data and short systems with "bear market" data. The dilemma created by focusing system development in this manner is traders might need to manually activate and de-activate systems in live trading depending upon their forecasts about market conditions (which is what I do). To mitigate risk associated with less-than-perfect timing, systems should be tested in unfavorable conditions to understand how they perform when the wind is blowing against them. The ideal probably is for systems to be quiet when conditions are not favorable for them; naturally, significant drawdowns are at the spectrum's other end. Where a system falls within that range is a factor in deciding how much effort a trader should dedicate to evaluating near-term market conditions.

Sector

To comment upon the easier point first, I use two filters related to liquidity:

    // Minimum Price Requirement (Supporting Entry)
    (close > 3); and,

    // Minimum Liquidity Requirement (Supporting Entry)
    (mov(volume, 20, simple) * close >= 200000).
Certainly, these two filters could be combined to create a dollar-volume-based filter, but I want my trades to clear these two hurdles separately.

As a general rule, I use stocks in the Russell 2000 for my long system and only the SPY ETF for my short system. The liquidity, diversity and simplicity associated with Index- and sector-based ETF's are attractive; however, they tend to be highly correlated with each other, and so using more than one such ETF seems unnecessary. My choice to settle on SPY for my short system (actually, SH in live trading) is driven by personal preference, a result of my obsession over the years with the S&P 500 index. I could likewise use SPY for my long system, but instead seek additional return from stocks moving faster than their index. The Russell 2000 satisifies my subjective (possibly irrational) criteria for depth and diversity, while still tending to be highly correlated with the S&P 500.
taowave
 
Posts: 584
Joined: Sat Dec 02, 2006 12:39 pm

Re: Thoughts for setting a One Click "envirnoment"

Postby Overload » Mon Mar 16, 2009 3:05 pm

Personally,it makes my hair stand up on end to think of system traders calling market direction:)

Allan, you are absolutely correct. In fact, I'd been thinking of bringing something similar up earlier in response to Tom's comments, but decided not to jump the gun.

I _still_ haven't had the chance to begin my next round of searching for trading systems since V4 was released, but having learned something in the last couple years, there is a new approach that I will implement. And that is to find a reasonable indicator that identifies these broad market conditions. For example, something clearly changed in 2007 that indicated the bull run was coming to an end. Likewise, something changed in 2002 that indicated that bear was coming to an end. I haven't yet experimented to identify what a good "reversal" indicator would be in those cases, but something as simple as an S&P crossover will not cut it. It will probably have to be a combination of several indicators, and I'm guessing that the VIX will somehow be a part of it. When I have a chance to play with this, I'll post what I'm able to come up with.

The logic behind such an indicator is that a "long system" can systematically be turned on during bullish periods and a "short system" can likewise be turned on during bearish periods. It is, essentially, a switch that flips between the 5 bullish strategies and the 5 bearish strategies. The 5 strategies in each category can be a combination of long and short systems as desired, but they still activate accordingly as per the broad market conditions. Thus, it becomes a systematic switch rather than a discretionary one.

Having not actually tried to create such an indicator, my fear is that the best indication of such a reversal may be based on fundamental values rather than technical ones. This might be okay, but will require a greater evaluation of indexes such as interest rates, average P/E ratios, perhaps even the strength of the dollar. Since such reversals don't happen often, there will also be limited ability to back test the theory. With these limitations in mind, I can only suggest that even system traders have no choice but to read a few tea leaves.

Pete
Overload
 
Posts: 2246
Joined: Wed Nov 30, 2005 12:14 pm

Re: Thoughts for setting a One Click "envirnoment"

Postby Dacamic » Mon Mar 16, 2009 3:29 pm

Allan,

My trading criteria varies depending upon time frame. I rely on criteria outside of StrataSearch to create a macro-level forecast to help identify cycles typically several years in length (and no shorter than six months). Within those long-term cycles, I use StrataSearch-built strategies to improve returns and drawdowns compared to buy-and-hold.

Counter-trend moves are too quick for me to trade profitably; thus, I choose to trade almost exclusively in the direction of the prevailing long-term trend.
Steve
Dacamic
 
Posts: 457
Joined: Wed Nov 30, 2005 12:40 pm

Re: Thoughts for setting a One Click "envirnoment"

Postby TomHam » Mon Mar 16, 2009 3:48 pm

Thoughts on a Bull / Bear switch indicator-

My feeling that just about any simple indcator to time the macro market is tuff. A short term (30-50) MA crossing a longer term (200-250) just has to much lag, plus a likely problem with whip-saws.

What would be terriffic, is the breaking of a (defined) trend line, perhaps drawn on a longer (weekly) time frame. My (limited) experience is that formula-izing such a trend line is difficult; beyond my ability I'm afraid. But . . . if Steve and Pete see value, and could be talked into it . . . . I'm sure it's with in their expansive mathamatical abilities. (Did I also mention how good looking they were ? :) ) Hey guys, you can't blame me for trying, it's worth a shot.
............ Tom ...........
TomHam
 
Posts: 18
Joined: Sun Jan 18, 2009 1:41 pm
Location: near Detroit, Michigan

Re: Thoughts for setting a One Click "envirnoment"

Postby taowave » Mon Mar 16, 2009 4:16 pm

Interesting responce,especially the occasional implementation of tea leaves,gann lines and planetary alignments.

I guess my question is essentially why not spend the resources on developing Multis that are "forced to run long vs short systems in some predetrmined perhaps optimised ratio?? Isnt that a bit closer to the pure systems approach,and takes away the need for picking direction??If I had the ability,that is the direction I would head in

I certainly like the approach you and Steve are suggesting,but you sound a bit like MOI!!!



Overload wrote:
Personally,it makes my hair stand up on end to think of system traders calling market direction:)

Allan, you are absolutely correct. In fact, I'd been thinking of bringing something similar up earlier in response to Tom's comments, but decided not to jump the gun.

I _still_ haven't had the chance to begin my next round of searching for trading systems since V4 was released, but having learned something in the last couple years, there is a new approach that I will implement. And that is to find a reasonable indicator that identifies these broad market conditions. For example, something clearly changed in 2007 that indicated the bull run was coming to an end. Likewise, something changed in 2002 that indicated that bear was coming to an end. I haven't yet experimented to identify what a good "reversal" indicator would be in those cases, but something as simple as an S&P crossover will not cut it. It will probably have to be a combination of several indicators, and I'm guessing that the VIX will somehow be a part of it. When I have a chance to play with this, I'll post what I'm able to come up with.

The logic behind such an indicator is that a "long system" can systematically be turned on during bullish periods and a "short system" can likewise be turned on during bearish periods. It is, essentially, a switch that flips between the 5 bullish strategies and the 5 bearish strategies. The 5 strategies in each category can be a combination of long and short systems as desired, but they still activate accordingly as per the broad market conditions. Thus, it becomes a systematic switch rather than a discretionary one.

Having not actually tried to create such an indicator, my fear is that the best indication of such a reversal may be based on fundamental values rather than technical ones. This might be okay, but will require a greater evaluation of indexes such as interest rates, average P/E ratios, perhaps even the strength of the dollar. Since such reversals don't happen often, there will also be limited ability to back test the theory. With these limitations in mind, I can only suggest that even system traders have no choice but to read a few tea leaves.

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

Re: Thoughts for setting a One Click "envirnoment"

Postby Overload » Mon Mar 16, 2009 5:31 pm

I guess my question is essentially why not spend the resources on developing Multis that are "forced to run long vs short systems in some predetrmined perhaps optimised ratio?? Isnt that a bit closer to the pure systems approach,and takes away the need for picking direction??If I had the ability,that is the direction I would head in

I certainly like the approach you and Steve are suggesting,but you sound a bit like MOI!!!

I'm still only theorizing at this point, and that explains my need for a few tea leaves. Unless (or until) I can come up with an indicator that properly handles this, I admit there may be a certain need for some discretionary input.

But the general theory goes like this. Long strategies in a bull market work differently than long strategies in a bear market. Likewise for short strategies. Either market can be traded with a combination of long and short strategies to capture the minor swings and individual stocks heading in their independent directions. But, apart from the direction, the speed and volatility of a bull versus bear market are significantly different and therefore are best traded with systems unique to those environments.

Part of my belief in this theory is that, like Tom said in his original post, I also haven't been succesful at finding a single system that works effectively in both bull and bear markets. When looking for a profitable system from 2004-2007, the best I could find was a system that was basically flat for 2000-2002. I wasn't able to find a system that was profitable in both the independent periods of 2000-2002 and 2004-2007, or at least not a system that took the best advantage of those 2004-2007 years.

So, my theory is just to add one more layer in the system building. Rather than my multi-system having 5 strategies that work across all markets, my multi-system will have 5 strategies for bull markets and 5 strategies for bear markets.

The question, as I said, is whether I can come up with an indicator that helps me determine whether we're in a bull or bear market. My Magic 8-Ball says... Reply Hazy, Try Again.

Pete
Overload
 
Posts: 2246
Joined: Wed Nov 30, 2005 12:14 pm

Re: Thoughts for setting a One Click "envirnoment"

Postby TomHam » Mon Mar 16, 2009 6:21 pm

O.K., here's a thought; I'll stick my neck out on this one.
A theory (as yet untried) to use a rolling period of time (say 12 months or ?), where S-S continually rolls forward every 2 to 4 weeks, searching for a better multi-system than it had found in the previous period. In that way, maybe, just maybe, it "self adjusts" to any new market condition; Bull, Bear or in Transistion. Thus, no particular need to find a really good method of defining what detects a Bull or Bear envirnoment.

Quite a concept; just a thought. (BTW, there is some basis for this. A number of acedemic studies on the time period for "relative strength" of trending securities. i.e. ones that do well in the future, and how they behaved in the immediate past.) The good news is, S-S can easily do that ! it sure would be interesting to talk to someone who has tested this concept. But a lot of this depends on what time frame you wish to trade on, holding periods, etc. But, if it worked on say, SPY, it could be that market indicator for other sectors. Which strategy to apply and when; hence becoming the market indicator itself. Sorry, I got carried away.
TomHam
 
Posts: 18
Joined: Sun Jan 18, 2009 1:41 pm
Location: near Detroit, Michigan

Re: Thoughts for setting a One Click "envirnoment"

Postby taowave » Mon Mar 16, 2009 6:59 pm

Tom,maybe I am missing something,but that sounds like a multi system walk foward.

I think the reality is Pete and Steve are probably correct and there may have to be an element of tea leaves..

I have much respect for William Oneil of IB Daily,and he certainly incorporates market direction with technicals,fundamentals and relative strength.









TomHam wrote:O.K., here's a thought; I'll stick my neck out on this one.
A theory (as yet untried) to use a rolling period of time (say 12 months or ?), where S-S continually rolls forward every 2 to 4 weeks, searching for a better multi-system than it had found in the previous period. In that way, maybe, just maybe, it "self adjusts" to any new market condition; Bull, Bear or in Transistion. Thus, no particular need to find a really good method of defining what detects a Bull or Bear envirnoment.

Quite a concept; just a thought. (BTW, there is some basis for this. A number of acedemic studies on the time period for "relative strength" of trending securities. i.e. ones that do well in the future, and how they behaved in the immediate past.) The good news is, S-S can easily do that ! it sure would be interesting to talk to someone who has tested this concept. But a lot of this depends on what time frame you wish to trade on, holding periods, etc. But, if it worked on say, SPY, it could be that market indicator for other sectors. Which strategy to apply and when; hence becoming the market indicator itself. Sorry, I got carried away.
taowave
 
Posts: 584
Joined: Sat Dec 02, 2006 12:39 pm

Re: Thoughts for setting a One Click "envirnoment"

Postby Overload » Mon Mar 16, 2009 7:08 pm

A theory (as yet untried) to use a rolling period of time (say 12 months or ?), where S-S continually rolls forward every 2 to 4 weeks, searching for a better multi-system than it had found in the previous period. In that way, maybe, just maybe, it "self adjusts" to any new market condition; Bull, Bear or in Transistion. Thus, no particular need to find a really good method of defining what detects a Bull or Bear envirnoment.

We're a few steps ahead of you on this one. In StrataSearch, this is available under the processing known as Dynamic Trading Strategies. Enter the OneClick Setups, update the setup of your choice, and navigate to the Dynamic Strategies tab. There, you'll find the setups needed to perform that "walk-forward" analysis, where a new trading system is created and used at various intervals along the way. Click the Help on the OneClick Setups window for more information.

By the way, while there are numerous software packages that allow you to run a walk-forward analysis, StrataSearch is the only one that lets you search for entirely new trading systems at each interval.

Pete
Overload
 
Posts: 2246
Joined: Wed Nov 30, 2005 12:14 pm

Re: Thoughts for setting a One Click "envirnoment"

Postby TomHam » Mon Mar 16, 2009 7:59 pm

Hank Pruden has a very good book called "The Three Skills of Top Trading". The first section of the book discusses "system building" and the role of right brain & left brain functions as part of the entire decision process. Systems range from black box, mechanical, decision suport, descretionary to seat of the pants types.

One point Hank makes (to support his own philosphy) is that if trading successfully could be done 100% quantitatively, then we may all be in trouble. . . . because the guy on the other side of your trade is a software engineer in India, who has studied trading for the past 10 years. (Can you complete ?) The idea is that some decretion is absolutely fine. After all, we have the best computer and softwate between our ears. :)

(note, the rest of Hank's book is on the Wyckoff method, which many either love or hate; I find of interest.)
Good Trading. ........ Tom ..........
TomHam
 
Posts: 18
Joined: Sun Jan 18, 2009 1:41 pm
Location: near Detroit, Michigan

Re: Thoughts for setting a One Click "envirnoment"

Postby JLG » Tue Apr 14, 2009 12:43 pm

I am very leery of dividing testing environments into bull and bear periods. I think it hides the risk associated with a change in envrionment and takes away focus from the problem of anticipating a change in the market leading to false expectations. It seems to me that it locks you into a paradigm that assumes an available look forward bias. The tails of financial markets can happen quickly and are critical to both risk and return.

I am more optimistic about working on the transitions with Stratasearch than others. I have focused my work on Supporting Entry and Exit rules for market, sector and indexes with some success. The availability of just three indexes is very limiting, but once a relationship is developed for a given symbol it can be coded directly into trading rules making these variables open for other testing of any symbol as a basis for market filtering.

I consider a back test only helpful in generating ideas. For me a successful walk forward test is required before I form a hypothesis. Only after walkforward success in other markets, countries or exchanges, do I then consider a hypothesis robust enough for paper or live testing or use in building multisystems. I spend about 75% of my time in walk forward mode. This supports my preference for robust trading rules that have limited parameter variables and stability rather than those which have higher natural variance or parameter complexity which can lead to adverse selection based on random results or larger degrees of curve fitting. Complexity in my systems is more related to the number of trading rules, not the number of parameters. But I test my systems to eliminate any complexity that does not add value walk forward.

I pursue long and short systems in tandem but developed separately. I target to be in the market 30-40% with long strategies, 20-30% with short and out of the market at other times. Each strategy has its own filter to determine when it should be in the market. With this approach I am better prepared to handle the transitions.

I am less comfortable with the combination of discretion and systems than others. I think the competitive issue is real although I worry more about the likes of D.E. Shaw, Rennaisance, CTAs and Goldman than programmers in India. But I suspect there will always be room for smaller traders in niches. I do see a place for the addition of discretion with Stratasearch systems for those that can day trade profitably to add return and reduce risk on Stratasearch's end of day signals. This is may not be for everyone but it has worked for me. In fact I think an Intraday product could represent a major advance for Stratasearch.
JLG
JLG
 
Posts: 100
Joined: Thu Oct 25, 2007 3:32 pm

Next

Return to OneClick Searches and Scoring

cron