| Index | Recent Threads | Who's Online | User List | Search |
|
|
 |
Portfolio123 » List all forums » Forum: Simulations and Portfolios » Thread: Simulate a new ranking system |
|
Total posts in this thread: 11
|
[Request a Feature]
[Post new Thread] [Add To My Favorites] [Watch this Thread] |
| Author |
|
|
azouz
Advanced Member
|
HI all, I am developping some new ranking systems and come up with differents ones that have an annualized return of +100% for rank>99.5 i was thinking it was a good ranking system comparing that to available ranking systems. But each time i do a simulation using one of theses +100% ranking systems with differents simulations rules, i am getting a very bad result (generally less than 50% annualized) and a very big drawdown even with a stop loss. Have anyone experienced that ? and how to resolve that ? |
||
|
|
visankar
Advanced Member
|
I think that is expected - anytime a simulation gives 100% , that means that it has been curve fitted. There would be no guarantees that is should behave that way in the future. A stable system should give stable returns when parameters change within reasonable limits. Also microcaps have been in favor over last few years - should that change, a lot of the high return sims are going to get affected. |
||
|
|
azouz
Advanced Member
|
I means that the ranking systems i developped show a bigger return that most of available ranking systems. But when running a simulation the return didn't follow. Available ranking system like IReturn or Bompus... show greater return than mine when using them for simulation but not when running a performance test for ranking system. |
||
|
|
tonysf
Advanced Member
|
May be you can try simulations that the buy and sell rules have little important. May be try a weekly rebalance with no buy rule and a sell rule of rank<100. And you can try 2, 3 and 4 week rebalance. This way you can really "test drive" your new ranking system free of influence from buy/sell rules. Hope this helps. |
||
|
|
Barn
Advanced Member
|
Hey I know exactly where you're coming from and it's frustrating. I made a Test ranking system with just Price/Book and Price/Sales and the results were amazing but then when I tried to make a sim with it I could barely keep up with the S&P. I can't quite figure out how a ranking system can perform so well until you simulate with it and it falls apart. I have tried lots of rules and no rules and both perform poorly. ---------------------------------------- Just my $0.02 ... Actually I need that back for my port. Barn ---------------------------------------- [Edit 1 times, last edit by Barn at Mar 13, 2006 10:54:30 AM] |
||
|
|
azouz
Advanced Member
|
Thats exacly what happen to me everytime i came up with a good ranking system. I also tried differents rules in the simulation each time but always getting poor return. any portfolio123 user experienced this and found a solution to that problem ? |
||
|
|
jerrodmason
Advanced Member UNITED STATES Joined: Jan 14, 2005 Posts: 600 Status: Offline |
Azouz, there are two key factors included in a sim that do not show up in the ranking system performance evaluation. Either one can make a huge difference:
Cheers, J ---------------------------------------- The smart money was on Goliath. "He's not the Messiah. He's a very naughty boy." ---------------------------------------- [Edit 1 times, last edit by jerrodmason at Mar 13, 2006 6:07:26 PM] |
||
|
|
azouz
Advanced Member
|
ok J but my question was why others ranking system with lower return than mine show a great annualized return when using in a sim with commission and next open price. while my ranking systems (which show greater buckets return) are always doing bad when using with a sim. |
||
|
|
BJS
Advanced Member
|
azouz, There must be a disconnect between your ranking system and the buy/sell rules you are using. I would try this: 1. Look and see what rebalance period you were using when you checked the performance chart for your ranking system (4 weeks, weekly, etc.) 2. Look to see how you were handling NA's when you ran your performance chart.(keep NA's or use previous data). 3. When you design your sim, use the same rebalance period and NA rule. 4. Run Ranks in your ranking system, and see how many stocks have a rank value above 99.5. I usually find there's around 35. 5. Now design a sim this way: a. Set commission and slippage to zero b. Set number of stocks to match how many there are in your highest bucket (i.e. for a bucket 99.5-100, about 35 stocks). So for this case, you would use about 3% of equity per stock. c. Set the rebalance period to match what you used in your ranking system. d. Your only buy rule is Rank>99.5 & Price>3 e. Your only sell rule is Rank<99.5 This should get you a performance that is close to what your ranking system showed. After you've confirmed that a sim can match the performance of a ranking system, you can then change number of stocks, add commissions & slippage, and add buy and sell rules. But you should make sure your buy and sell rules make sense for your ranking system. For example, if your ranking system is valuation-driven with very few technical factors, the turnover should be pretty low. So maybe you should design a sim with low turnover in mind. But if your ranking system has many technical factors, the turnover might be a lot higher, and your sim might need to be designed with a high turnover. Hope this helps. Brian |
||
|
|
DennyHalwes
Advanced Member UNITED STATES Joined: Apr 28, 2004 Posts: 1532 Status: Offline |
azouz, If you are still having trouble developing a good Sim after trying Brian's suggestions don't give up on a good Ranking System. Make your system public and ask for help developing a good Sim. The members here are always looking for more investment options, and will be glad to help. A good public Ranking System and Sims is worth a lot more to you than a good private Ranking System without any good Sims. Denny  ---------------------------------------- "The significant problems we face cannot be solved at the same level of thinking that we were at when we created them". Albert Einstein |
||
|
|
[Show Thread Printable Version] [Post new Thread] |
Re: Simulate a new ranking system
