| Index | Recent Threads | Who's Online | User List | Search |
|
|
 |
Portfolio123 » List all forums » Forum: Data related issues » Thread: PctFromHi doesn't seem to simulate trailing stop well |
|
Total posts in this thread: 8 |
[Request a Feature]
[Post new Thread] [Add To My Favorites] [Watch this Thread] |
| Author |
|
|
kbellare
Advanced Member
|
Hi, I have a PctFromHi<-15 Sell rule in my portfolios, but it never seems to execute the sell. I use 15% trailing stops in my brokerage account, and for the last 15 positions i've sold under this condition, I've had to manually sell in my port123 portfolio. - At re-balance time, does the PctFromHi only look at the previous day closing price and not the range of prices since the last re-balance? That could explain why it doesn't triiger. - What's worse is even though my portfolios simulate well (consistently outperformed S&P) for last 5 years, the actual performance over last month has been terrible (underperformed S&P). I typically trade during market hours (not next-day open), so some difference in entry price and volume - but that shouldn't matter. - Any suggestions on why actual performance falls behind simulations? Is this a common problem with new users of port123, and how to avoid this? Here are a couple of portfolios i've used over the last month http://www.portfolio123.com/port_summary.jsp?portid=500121 http://www.portfolio123.com/port_summary.jsp?portid=502884 thanks Kiran |
||
|
|
Tomyani
Advanced Member
|
HI, Pct. from hi sell rule only checks this condition at rebalance dates. So if you have weekly or monthly rebalance, port checks if stock is down 15% or more only at that point. Trailing stop checks all the time. It will lead to more exits. It can't be modeled with Port123. As far as why you have way underperformed of late -- you have relatively small stock portfolios. They are going to have very erratic short term performance. You also likely have smaller stocks -- I don't think you use any liquidity filters. So, when market goes up a lot you will likely make more, when it goes down a lot likely will lose more. IF you traded low beta large cap, you would see this reversed (make less in up markets and lose less in down markets). Best thing I can suggest is more models with more entry and exit rules so some portion of port is in and some is out in different market conditions. Best, Tom |
||
|
|
Isaac
Advanced Member
|
Why can't trailing stop loss rules be incorporated into P123? This capability would be a great enhancement. Isaac |
||
|
|
marco
|
A timely stoploss function (instead of only at rebalances) is one of the feature requests in our radar map for next releases. ---------------------------------------- Portfolio123 Staff |
||
|
|
dwpeters
Advanced Member
|
Hi, You ask a number of good questions:
Already answered by Tomyani
There can be a number of reasons for this. 1. P123 ports/sims are often more volatile than the market. So you may gain more when the market is up and lose more when the market is down. 2. Data mining is MUCH easier with P123 than I ever thought it would be. Strategies that perform well in the backtesting don't always hold up out of sample. Great care must be taken when developing strategies to avoid this. 3. Since April this has been a very difficult market to trade. I have 8, 20 stock benchmark portfolio's based on the best strategies I have seen in P123 using minimal buy/sell rules and no market timing. All but 1 are down over the past 4 and 13 weeks. 4. This has been a very difficult market to time. I notice your timing rules have not gotten you out of the market, but with the ups and downs I expect that there are few rules that would have been out since May and stayed out, without whipsaws.
Entry and exit prices can make a significant difference, depending on slippage, trading frequency, etc., especially when you have no liquidity rules as previously pointed out.
Yes, I think this is a common problem to new users. See above and prior answers. Also, I would add that you are using a trailing stop that you did not (could not) test in P123. My experience is that stop loss rules generally hurt returns. They may still be useful to limit risks, but they add an entirely new variable to the results. Don |
|||||||||
|
|
o806
Advanced Member
|
Hi Kiran: Tom is right about the small number of stocks in your portfolios. You will need to have 50 or 100 closed trades in a portfolio to overcome the noise of the market. So with 2 stocks at a time, its going to be up to a year before you can tell if the recent divergence of reality from expectation is due to random noise in the market affecting the few stocks you hold at a time, or if the difference is more serious due to a flaw in the strategy itself. Also Don is right about data mining (also known as curve fitting) being a real danger that might be at least part of the reason your results differ from what your back tests may have lead you to expect. It would have been great if Portfolio123 would have completed its plan to add data for the years prior to 2001 since that would reduce the extend of unintentional data mining/curve fitting. But nothing has been heard about that project for so I would not be surprised if it has been abandoned. Regards, Brian |
||
|
|
marco
|
Not at all. But the Point In Time, in-house, ratio engine needs to be completed first. First step will be to launch the new engine and only go back to 01. We can go further back, but only for fin. statement derived ratios. Other ratios will be missing, like estimates, insider, institutional, etc. To extend those there is no other way other than buying the data. From what I see extending data further back w/o the other dats-sets might not make too much sense for P123 users as systems seem to rely on the whole spectrum of databases. ---------------------------------------- Portfolio123 Staff ---------------------------------------- [Edit 1 times, last edit by marco at Aug 29, 2010 12:06:49 PM] |
|||
|
|
o806
Advanced Member
|
Marco: I am pleased to hear that extending data back prior to 2001 is still a possibility. Since data prior to 2001 would be used for things like "proof of concept" and "out of sample" testing we would not need to have thkis for all stocks. The penny stocks and low liquidity ones could be skipped in my opinion. Having financial data for largest 3,000 or 4,000 stocks would be sufficient for testing purposes (e.g, the R3000 or similar). Even if some data were missing (no earnings estimates for example), having the financial ratios, etc. available would be quite helpful. Regards, Brian |
||
|
| [Show Thread Printable Version] [Post new Thread] |
