| Index | Recent Threads | Who's Online | User List | Search |
|
|
 |
Portfolio123 » List all forums » Forum: Simulations and Portfolios » Thread: Decide SIM as RELIABLE PORT for ACTUAL TRADING |
|
Total posts in this thread: 2 |
[Request a Feature]
[Post new Thread] [Add To My Favorites] [Watch this Thread] |
| Author |
|
|
essam
Advanced Member
|
DEAR ALL, I would like to elaborate on subject "MOVING SIMULATION & DECIEDE TO BE AS RELIABLE PORT FOR ACTUAL TRADING": 1) What are the major factors (i.e. drawdown,...) and advisable limits (i.e. drawdown <=25,....) to decide that a specific SIM is a RELIABLE PORT FOR ACTUAL TRADING ? 2) In principal if any BUY OR SELL RULE (TECHNICAL or FUNDMENTAL) INCREASES the Gain/Stk/Day for a specific SIM, then this BUY OR SELL RULE, then could/should be used in the SIM/PORT, and vice versa is correct ? 3) If using "BarChart - Signals Top 100" increases the Gain/Stk/Day in a PORT1 and decreases Gain/Stk/Day in a PORT2, then in ACTUAL TRADING we should use it for PORT1 and not for PORT2 ? 4) In general if any SIM is decided to be a RELIABLE PORT for Actual Trading, then we should stick to the PORT RECOMMENDATION ONLY, except for TRAIL STOP ? 5) Before MOVING a SIM to be a RELIABLE PORT for ACTUAL TRADING, how long we should paper trading before ACTUAL TRADING ? 6) We should SIMULATE EXACTLY THE PORT'S BUY /SELL RULES AND CONDITIONS into the real trading (i.e., if PORT1 uses Signals Top 100, trail stop, weekly rebalance, then in ACTUAL TRADING we should use the SAME use Signals Top 100, trail stop, weekly rebalance) ? 7) Optimizing a SIM for LONG POSITIONS, we should consider the FULL HISTORICAL PERIOD (3/2001 till 4/2009), or only the BEAR MARKET PERIOD to be minimize the risk ? 8) How to get from Search the SIMs / PORTs that have the highest Gain/Stk/Day ? 9) for Ranking Systems a minimum Rebalance Frequency is Week, the question is that Performance of the Ranking System could be used for DAILY REBALANCE SIM/PORT with same reliability (i.e. Rank>99.5 for WEEKLY SIM the performance 100%, is this could be used as Rank>99.5 for DAILY SIM) ? 10) Is the Ranking Systems are updated daily or only weekly Best Regards, Essam |
||
|
|
o806
Advanced Member
|
Hi: My general rule is to make the portfolio exactly like the sim I have developed. Since I know I will do the exact match I only test "reasonable" rules in my sims. So there is no value in developing a sim that has a minimum price of $1.00 if I know I really will only buy stocks that are over $2.00. The same for minimum volume in the sim. Why test a sim with a low volume rule if I know that I will be buying lot sizes that will move the price if I were to try to trade the sim. I also exclude OTC stocks in my sims. So, I begin by testing the sim on a universe of tradeable stocks. P123's custom universe function is great for this type of testing. Your other question is how to know if the sim is reliable enough to trade. Many people test sim variations to find the "optimal" settings. Optimal varies from person to person. Some only care about highest gains. Others want the draw downs to be low. Still others look at the annual gains divided by the worst draw down to get a risk reward measurement. Like everyone else, I test variations looking for settings that produce optimal results. In addition to testing to find optimal settings, I also test the sim rules for "robustness". How well do the rules handle hardship and changing conditions. One way to test for robustness is to test that sim with small and medium variations in all its rules. Of course, one setting will be "optimal" but I give just as much attention to the results of the non-optimal settings. If the sim is still OK with sub-optimal setting, that gives me confidence that if the future is slightly different from the past the sim will still work, at least to some degree. But what if the future is a lot different than the past? To be prepared I like to test the sim on different time periods. Here are my test periods March 2001 - March 2003 (overall bear with a couple rallies) April 2003 - Dec 2003 (a huge rally - everything succeeds) 2004-2006 (general bull) 2007-2008 (or Feb 2009) (a very hasty bear) Feb 2009 - present Rarely will I see my sim out perform the market benchmark in all the periods, but that would be the ideal. I am content to have a sim that did no worse than the market in the bear periods if it did much better in the bull sections. I like to see how the sim behaves in each of the time periods with the minor variations in rules I spoke about above. Here is an example. When considering whether or not to add a new rule to my sim, I will test the rule over the entire data history. If it adds no value, it is thrown away. If it helps, then I test the rule and slight variations of the rule for each time period. If the rule helps in all time periods, it is a keeper for sure. If it helps in the majority and does not hurt in the rest, it is also a keeper. If it helps a lot in one time period and hurts a little in the rest, it is highly suspect of being curve fit and needs more careful study. If a rule helps in half the time periods and hurts in the other half, it is random and is discarded. Well that gives the idea of how I test my rules. Regards, Brian |
||
|
| [Show Thread Printable Version] [Post new Thread] |
