 |
benwm
Registered: Mar 2008
Posts: 1311 |
11-01-11 10:31 AM
That's a pretty good comeback Rol. Well done.
I'm sure you've learnt a lot during the process to improve your system.
|
| |
|
Edit/Delete • Quote • Complain |
clarodina
Registered: Feb 2007
Posts: 348 |
11-01-11 11:13 AM
Quote from Rol:
Hi clarodina,
It is always nice to see a stock return a high PF in backtesting, but I don’t limit stocks to only those with a high PF. That would be too easy to exploit (I have tried it). Past equity curves of a stock do not predict future success. Unforeseen events could always strike a company. The edge, I believe, comes from overlaying all of the individual equity curves to produce a “net equity curve,” and trying to determine if that curve is realistic and acceptable.
Not talking about having min price and vol filter on your live filtering but rather on your testing rules on historical data. Some stks have OHLC having good equity curve on testing but they are not realistic on live trading due to liquidity problem and low price but they have good curve. Do you have min vol rule on your testing?
You don't choose stks with high pf or equity curve? What do you use? You would choose a stk with side way equity on your testing for live trading?
Isn't layering the sideway curve reduce the good curve of net equity curve?
|
| |
|
Edit/Delete • Quote • Complain |
Rol
Registered: Jul 2009
Posts: 425 |
11-01-11 12:29 PM
Quote from clarodina:
Not talking about having min price and vol filter on your live filtering but rather on your testing rules on historical data. Some stks have OHLC having good equity curve on testing but they are not realistic on live trading due to liquidity problem and low price but they have good curve. Do you have min vol rule on your testing?
You don't choose stks with high pf or equity curve? What do you use? You would choose a stk with side way equity on your testing for live trading?
Isn't layering the sideway curve reduce the good curve of net equity curve?
I use the same price and volume filters on portfolio backtesting as well as live trading. Unrealistic fills is another reason I exclude low price, low volume stocks. It is hard to explain without looking at many individual equity curves. Over many years, some curves are in smooth uptrends, some are in sideways trends, but may then "breakout" to the upside, and some are in downtrends. You see every shaped curve imaginable. In the end, however, there are more winners than losers. I could remove the net losers, but as trades occur the equity curves are changing, and stocks may move from the net winner column to net loser column, and vice versa. I don't know how a stock will perform when I buy it. This unknowingness of which stocks will perform and which ones will not can be an asset, I believe, and allow me to work "under the radar." Many are looking for a sure thing. I am only looking for ballpark results and for what is good enough for me. You could try removing the worst performers if you think it helps.
|
| |
|
Edit/Delete • Quote • Complain |
clarodina
Registered: Feb 2007
Posts: 348 |
11-02-11 09:41 PM
Quote from Rol:
I could remove the net losers, but as trades occur the equity curves are changing, and stocks may move from the net winner column to net loser column, and vice versa. I don't know how a stock will perform when I buy it. This unknowingness of which stocks will perform and which ones will not can be an asset
Rol, have you consider the smooth curves are due to random factors on testing individual stk? you mention about as trades occur the equity curves are changing, and stocks may move from the net winner column to net loser column. Isn't this suggesting that the rules do not have an edge. If the rules have an edge on the particular individual stk, smooth curve should continue at least not much less compare to sideway curve from another indivudual stk. And layering of many different smooth curve should be better than layering of many smooth curve and not so smooth curve on live trading. This bring to the issue of should the sideway or downtrend curves be eliminated.
|
| |
|
Edit/Delete • Quote • Complain |
Rol
Registered: Jul 2009
Posts: 425 |
11-03-11 12:36 AM
Quote from clarodina:
Rol, have you consider the smooth curves are due to random factors on testing individual stk? you mention about as trades occur the equity curves are changing, and stocks may move from the net winner column to net loser column. Isn't this suggesting that the rules do not have an edge. If the rules have an edge on the particular individual stk, smooth curve should continue at least not much less compare to sideway curve from another indivudual stk. And layering of many different smooth curve should be better than layering of many smooth curve and not so smooth curve on live trading. This bring to the issue of should the sideway or downtrend curves be eliminated.
I have thought of and tried before what you suggest of removing stocks that are losers, but in live trading it did not appear to help any. I am not able to take every trade anyway a stock triggers, due to BP limitations. I would have to take every trade to be able to benefit from any smooth uptrend. My strategy is portfolio based so some of the logic does not apply as if you were using a strategy applied to a futures contract. I don’t really ask why my strategy works at a particular time on a particular stock, and not another stock and another time. It just appears to me in live trading that given enough BP this is a strategy that works for me. I would say only about 10% of the equity curves could be called smooth. Most are not. I mentioned earlier in my journal that individual curves are often jagged. When would you determine that a curve is smooth enough to accept while going through a few thousands symbols? It is only the portfolio equity curve that becomes smooth over a long enough time period, and even it has occasional large DD. Newer stocks may not have much backtesting data to generate a meaningful curve. Would I eliminate those too? Sure I could optimize for an individual stock to produce a nice smooth equity curve, but it would be subject to curve fitting. I have tried to optimize for all stocks with a “best fit”. Given enough time, maybe decades, I think all stocks would become net profitable with my strategy. I just don’t have the luxury of waiting that long. The trick to interpreting backtesting data comes when applying it to the real world. Why don’t you post 6 random stocks you would like to see equity curves on, and later I will show them with my current settings? Maybe it will give you an idea of how difficult it might be to do what you suggest.
|
| |
|
Edit/Delete • Quote • Complain |
clarodina
Registered: Feb 2007
Posts: 348 |
11-04-11 08:39 PM
Quote from Rol:
I have thought of and tried before what you suggest of removing stocks that are losers, but in live trading it did not appear to help any. I am not able to take every trade anyway a stock triggers, due to BP limitations. I would have to take every trade to be able to benefit from any smooth uptrend. My strategy is portfolio based so some of the logic does not apply as if you were using a strategy applied to a futures contract. I don’t really ask why my strategy works at a particular time on a particular stock, and not another stock and another time. It just appears to me in live trading that given enough BP this is a strategy that works for me. I would say only about 10% of the equity curves could be called smooth. Most are not. I mentioned earlier in my journal that individual curves are often jagged. When would you determine that a curve is smooth enough to accept while going through a few thousands symbols? It is only the portfolio equity curve that becomes smooth over a long enough time period, and even it has occasional large DD. Newer stocks may not have much backtesting data to generate a meaningful curve. Would I eliminate those too? Sure I could optimize for an individual stock to produce a nice smooth equity curve, but it would be subject to curve fitting. I have tried to optimize for all stocks with a “best fit”. Given enough time, maybe decades, I think all stocks would become net profitable with my strategy. I just don’t have the luxury of waiting that long. The trick to interpreting backtesting data comes when applying it to the real world. Why don’t you post 6 random stocks you would like to see equity curves on, and later I will show them with my current settings? Maybe it will give you an idea of how difficult it might be to do what you suggest.
 1.jpg
This has been downloaded 47 time(s).
|
| |
|
Edit/Delete • Quote • Complain |
| Receive
an email whenever a new post is added to this thread by subscribing
to it. |
|
|
|
|
