Ive spent a bit of time test-driving C2 and thinking about how to build a hedge fund from multiple moderate to high risk systems (thereby reducing risk). Putting it all down on paper has been quite helpful.
The following two lists are aspects of trading systems and a total fund that I consider most important for me in order of priority. If you think Ive got my priorities badly out of order, or there is some issue I have neglected, Id be grateful for anyones feedback.
Individual system requirements
--------------------------------------
-System age (higher)
-Number of trades system has made and continues to make (higher)
-Max DD 20% (ideally < 10%)
-System that is not overly trend correlated, or strongly correlated to the market chart.
-System takes long and short positions
-System is not based solely on one instrument (eg EURUSD only)
-Annualised return > 25%
Overall hedge fund requirements
---------------------------------------
-Minimum 3 systems
-Low degree of correlation between systems
-Each system based on different instrument class (eg. forex, futures, shares/index)
-Each system from a different vendor (protection against unfortunate circumstances affecting the vendor)
Some discussion:
Expectancy
--------------
Not sure how much emphasis to place on % winning trades vs. % losing. Some systems derive their positive expectancy by having a low % of winning trades that are highly profitable compared to the losing trades (type A) while other systems have a high % percentage of winners but the win and loss amounts are more equal (type B). Im just not sure if one is necessarily better than the other. Im wondering if type A and type B are equally prone to slippage, but not sure how to figure that out.
Sharpe, APD, and profit factor
-------------------------------------
Im not sure how much emphasis to place on these. While these derivations might be a useful screening tool, ultimately it seems wiser to me to assess the components of the formula individually (based on their relevance to my personal requirements) rather than the ratio the formula spits out.
Realism factor
-----------------
This is a C2 black box so I dont know how much faith you can put in it. At the end of the day simulation can only do so much, and you have to test the system with a real account. In terms of fills, I wonder if the most popular systems traded with the most popular brokers are going to suffer compared with less popular systems traded through less popular brokers.
Other concerns
------------------
forex systems with a high number of trades - for some reason C2s comms on forex are higher than on everything else plus they have higher transaction costs than futures.
comments:
"System that is not overly trend correlated, or strongly correlated to the market chart."
don’t see the point. It is not bad to attach to trends. The markets are usually in trends about some 35-40% of the time, and that is often when you make most of the profits. sideways markets often grind up your profits.
"Im not sure how much emphasis to place on these. Sharpe, APD, and profit factor "
This tells me that you don’t understand what to look for in a system. These are the most important things, once a system is established. Many vendors refuse to take a loss or just Hold&Hope it turns around. Then suddenly, your leveraged holdings will get hit with a catastrophic loss, as happens here every time we get a sudden drop in market. Sharpe and APD will help show this. The profit factor helps show you whether or not the system outperforms. They are not something to “discuss” or put into the appendix!
"Realism factor …This is a C2 black box so I dont know how much faith you can put in it. "
Again, this worries me, as I don’t know if you grasp the point of this. We have vendors who will put on a position that reaches a large fraction of the daily volume of an instrument, nad then advise OTHERS to do the same. You have more learning to do, before getting in…
Zsolt,
You might find Science Trader blog (http://scitra.blogspot.com/) useful for a start. He is already doing it.
Eu
-System takes long and short positions
-System is not based solely on one instrument (eg EURUSD only)
I’m not convinced that these points are important. Extreme-os takes only long positions but I think it is a good system (the slippage is a problem, but that has nothing to do with the direction of the trades). TMG trades only QQQQ but so far it looks like sound system (I know the opinons differ on that, but that has nothing to do with the instrument).
Max DD 20% (ideally < 10%)
I also take a look in the advanced statistics to learn more about the DDs. The max DD is only one observation. Sometimes a vendor is not used to the interface (at least this is what they say) and then the first DD may be very relevant. It can also be informative to ask if there were special reasons for the DD. For example, if the vendor says that he changed the system after this then you know that all statistics that are based on the full history of the system are suspect because you don’t know the contributions of the old and the new system.
-Annualised return > 25%
I think I would require more, certainly with a DD of 20%, unless I am sure that there is almost no slippage.
Not sure how much emphasis to place on % winning trades vs. % losing.
An extremely high percentage winning trades (say, 90%) usually means that the trade is hold until it becomes a winner, and that deep drawdowns are accepted. I wouldn’t go for that. I think that about 70% winners is the maximum that is realistically possible in that case.
Im wondering if type A and type B are equally prone to slippage, but not sure how to figure that out.
Slippage has no relation with this. To assess the slippage it is better to consider (1) the P/L per unit (2) the average length of the trade, and (3) the slippage that is reported for autotraders (in the question mark behind Cumu $ corrected for commissions and slippage). Generally, systems that trade on a small time scale and with small P/L per unit can be expected to have larger slippage. Complicated signals like stops and bracket orders can also lead to a form of slippage in gen 1 autotrading.
Sharpe, APD, and profit factor - Im not sure how much emphasis to place on these.
For Sharpe ratio I would use the lower bound of the confidence interval, which is reported in the advanced statistics. I use both Sharpe and APD (among other things) to decide how much weight the system gets in my portfolio.
Realism factor - This is a C2 black box …
Indeed. I use it as a reason for further inquiry, and no more than that. This may depend on the size of your trading capital though.
In addition to what you mentioned, I would consider:
- the leverage that the system uses.
- the maximum position size that the system has used, and what the vendor says about this. Some systems start trading 1 contract at a time, obtain a perfect equity curve that entices you to increase leverage, and then they suddenly make a trade with 500 contracts. A sort of “one strike and you’re dead” system.
- the symbols that the system trades. E.g. penny stocks have more risk than other stocks.
- the trading theories that the vendor advocates, particularly his opinion about stop losses and averaging down. I’m not saying that he should use stop losses, but at least he should have a sophisticated and preferably tested opinion about these matters.
edit: Sometimes a vendor is not used to the interface (at least this is what they say) and then the first DD may not be very relevant.
Zsolt,
I won’t go over all your indicators one by one, but I do think it’s a good insight that you are looking for a low degree of correlation between systems. The problem is that C2 doesn’t provide these correlations, so it’s not easy to figure out.
I also want to point out that you can adjust a system sometimes quite easily if it doesn’t entirely fit one of your parameters. For example, a system with 40% drawdown can be traded with half the leverage the vendor uses, so that historical drawdown then will fit your 20% limit. For other parameters e.g. the Sharpe ratio, this is a lot harder (if not impossible).
Eu: Thanks for bringing up my blog. I hope it’s helpful to others.
I took a look at your blog. Another Interesting blog:
http://www.cxoadvisory.com/blog/external/blog10-22-04/
and
http://www.cxoadvisory.com/blog/internal/blog2-07-06/
-System takes long and short positions
-System is not based solely on one instrument (eg EURUSD only)
I’m not convinced that these points are important. Extreme-os takes only long positions but I think it is a good system (the slippage is a problem, but that has nothing to do with the direction of the trades). TMG trades only QQQQ but so far it looks like sound system (I know the opinons differ on that, but that has nothing to do with the instrument).
Yes, they are Jules. Using multiple instruments, hopefully mostly uncorrelated, and goind long & short helps to diversify the system and smooth the equity curve and reduce the sudden drops in system equity
C2 will soon start providing inter-system correlation data. I’ll make an announcement about this new feature shortly.
Recent research indicates that the ‘average expert’ (producing around 25% annualized return with around 20% DD) has little to offer investors/traders.
Finding, ‘consistently exceptional’ advisers/systems/methods (with annualized return atleast twice the historical DD) is no easier than identifying outperforming stocks.
Indiscriminately seeking the output of several so-called ‘average expert’ is probably a waste of time.
Learning what makes one ‘exceptional expert’, accurate and consistently outperforming the market is worthwhile; and so much the better if he offers advise not just on stocks but also on forex and futures.
If you find five "average experts" who each produce 25 percent annual returns with 20 percent draw-downs and they are all uncorrelated with each other (there equity curves are correlated at an R of 0) then you can divide your trading capital by 5 and invest equally in each. At the end of an "average" first year, you will receive a return of 25% (5% from each of the five advisors), and your max draw-down that year would be 8.94% (sqrt(445))
>If you find five "average experts" who each produce 25 percent annual returns with 20 percent draw-downs and they are all uncorrelated with each other (there equity curves are correlated at an R of 0)…
In his 2007 book Evidence-Based Technical Analysis: Applying the Scientific Method and Statistical Inference to Trading Signals, David Aronson says:
There are biases in our thinking processes, especially with respect to complex and uncertain information, that undermines the validity of subjective technical analysis including correlations. italics mine
Our brains are so strongly inclined to find patterns in nature, perhaps as evolutionary compensation for limited processing power, that we often see patterns where none really exist. This tendency toward spurious correlations, evident in subjective chart analysis, is maladapted to modern financial markets.
Erroneous knowledge (superstition) is resilient due to biases in our thinking processes such as:
1. Overconfidence;
2. Optimism;
3. Confirmation (discounting contradictory data);
4. Self-attribution (smart when right and unlucky when wrong); and,
5. Hindsight (overstating past successes and understating past failures).
Good stories well told can make people misweight or ignore facts.
People are not naturally rigorous logicians and statisticians. A need to simplify complexity and cope with uncertainty makes us prone to seeing and accepting unsound correlations. We tend to overweight vivid examples, recent data and inferences from small samples.
In summary, the scientific method (logic) is a reliable alternative path to validity, mitigating the misleading effects of our cognitive biases including correlations. italics mine
One of the errors that tends to be made is the assumption that correlation between systems or trading instruments is static. In truth you have to analyze dynamic market conditions of significant stress to find out which instruments are correlated.
For example, during the 1987 stock market crash, the stock market was highly correlated to Cattle futures…
>In truth you have to analyze dynamic market conditions of significant stress to find out which instruments are correlated.
Agree. Also, in his book, Aronson says:
* Rigorous testing of a rule requires that it be objective, implementable as a computer program (deterministic logic) that generates unambiguous long, short or neutral positions.
* A rule is good only if it beats a reasonable benchmark with a statistically significant margin of victory.
* Detrending the test data set (for example, daily returns for the S&P 500 index) is a consistent approach to benchmarking.
* Accurate historical testing requires: (1) avoiding look-ahead bias ("leakage of future information" into the analysis); and, (2) accounting for trading costs.
In summary, a rigorously logical method is essential to transform TA from subjective opinion to objective knowledge.
I agree that correlation is not static in most cases. However in conditions of significant stress correlations between stocks, index futures etc. tend go to 1 anyways (perhaps with a few exceptions). In those cases I think its better to rely on a hedge (e.g. put options to cope with crashes) rather than correlations. IOW correlations can be useful for normal market conditions, but their value during stress events is limited. Also, because stress events are typically rare, it’s very hard to come up with a reliable estimate in case you have a prior belief that the correlation should be different from 1 during such an event.
Brian, Keith:
Thank you for some very "non"-average contributions. Gilbert
"In those cases I think its better to rely on a hedge"
I agree with using options if you can tolerate the added insurance expenses. Another possibility is to reduce the level of leverage down to something that you are comfortable with.
The best alternative is to play with someone else’s money 
“Yes, they are Jules. Using multiple instruments, hopefully mostly uncorrelated, and goind long & short helps to diversify the system and smooth the equity curve and reduce the sudden drops in system equity”.
Let me clarify what I mean. I agree that these two points deserve attention because it is better to be diversified. But I don’t consider them important enough to dismiss a system entirely on one of these points when all other points are excellent. As you say, the most important point is to have a smooth equity curve. Diversification is only a means to this end imho. If a smooth equity curve is obtained and maintained with only long positions, and the other criteria (long history, small correlation with the S&P, etc.) are also satisfied, then I think it is worthwhile to consider the system seriously and that it may be a better choice than a diversified system with a less attractive equity curve. But perhaps the next market correction will prove me wrong 
Sounds like we have new Pal. Or are you the old Pal?
2 Zsolt:
Im new here, but have narrowed my criterion for a possibly successful system for moving beyond the 'hedge"s…
From (my thread on) ET:
"Now that I’ve looked at the best of the best (at C2) sorted as follows…
“1- length of track record with 2yrs+ (precious few)
2- annual percent return greater than 50% (much fewer)
3- reasonable uptrend (narrow further)
4- contrast this with Realism Factor and APD “Average Profit to Drawdown”
”…and you are left with a handful.
"paysense"
Hence, I am developing notation on ‘My Analyst’ page as well as cred as C2 system provider.Gilbert
