Page 1 of 1
Posted: Sat Jul 05, 2003 8:58 am
How does one usually deal with the issue of backtesting a stock system when shorting is affected by the up-tick rule?
Let's say my system uses daily data only and I run the system everyday when the markets are closed. Would the following work?
1. At the end of day 1 I get a signal to short a stock.
2. I use the high of day 2 as my entry price if high > open. Or perhaps I can use a percentage of the distance between the high and the open, i.e. entry price = 0.5*(high-open)+open.
3. If high = open then I use the close of day 2 as the entry price if close > low.
4. If high = open and low = close, then can I use any price (apart from the low) within the range of day 2, since this price would be > low. But this might not be realistic as the stock's price may be declining the whole day without having an up-tick. If that's the case, I would need to use the prices from day 3.
Would appreciate any suggestions. Thanks.
Posted: Tue Jul 08, 2003 7:26 pm
To estimate conservatively, you should use the lowest price the following day for your entry point. It is rare not to get an execution all day long in equities (unless it's thin, or you're trading futures), so it's hard to imagine you'll need to go to a third day.
Maybe you can use Single Stock Futures data, and use that as your vehicle. At least you can short right away without an uptick, and your pricing for testing will be more realistic. You give up a few cents sometimes, but I've noticed I have RECEIVED a few cents more often than I have given up. For example, HP stock has been rising for the past few months, but the futures have always been trading at a discount in the back months. See how SSFs work for your testing for short positions.
As for the lock-out problem with futures, read this post:
Posted: Wed Jul 09, 2003 6:48 am
I have some prejudices here, but I think with reason. It seems to me that shorting stocks with a mechanical system would be difficult to do effectively, because the information flow on the short side is different from the information flow on the long side. Buying momentum longs tends to work because good news is diffused gradually. On the short side, however, everyone tends to get the bad news--which the company has usually been trying to suppress--all at once, and therefore you don't have a chance to catch the meat of the move unless you were in for fundamental reasons.
Posted: Wed Jul 09, 2003 7:43 am
I'm not trading stocks now but when I was I didnt find that this was the case. I recall a number of wonderful examples where the technicals (simple lower highs and lows even) revealed a bad report a couple of weeks before it came out (unisys comes to mind).
What was true was that panic is faster than optimism so you have to be quicker on the trigger both entering and exiting. Also if you wait for the fundamentals to measure the start of your decline then your statement would appear true.
I'm still trading indexes and short vs long seems to make little impact there.
Posted: Wed Jul 09, 2003 10:43 am
I currently use the turtle rules for stocks, from the risk management aspect, ATR calculations, to pyramiding....the only thing I've changed up on the rules are my entry indicators. I use two indicators to define the start of a trend...one of the most important requirements are that the stock has a high volume rate of 1 millon+ volume shrs traded each day with utilizing the two percent stop on the first day of securing my position...so if I get stopped out I wait till the next day to enter another short position or long position depending on the indicators....let's say my position doesnt hit my 2% stop then I use the N stops initated by the ATR...but one of the most important thing I believe is the entry it has to be timed just right I avoid buying at the open...I wait for thirty minutes after the opening bell. I truly believe the turtle rules can apply to any market...but it has to fit the person's pysche to work.
Posted: Wed Jul 09, 2003 11:22 am
translating Unit value to Stocks postion.
stock position = Account size * percentage/ share price
stock position = 450000 * .10 / 22.00
1st stock position = 2045 shrs
hopefully this helps out.
Re: Shorting Stocks
Posted: Wed Jul 09, 2003 4:37 pm
bagherra wrote:I have some prejudices here, but I think with reason. It seems to me that shorting stocks with a mechanical system would be difficult to do effectively, because the information flow on the short side is different from the information flow on the long side. Buying momentum longs tends to work because good news is diffused gradually. On the short side, however, everyone tends to get the bad news--which the company has usually been trying to suppress--all at once, and therefore you don't have a chance to catch the meat of the move unless you were in for fundamental reasons.
Aside from what John said, I think that especially for a short or medium term mechanical system you do not have to worry too much about their entries synchronizing with momentum on the short side. Just a thought. I think that's just how short term systems work in general.
Posted: Fri Jul 11, 2003 8:50 am
Again, thanks for the reply everyone.
On a separate issue, I tested the breakout system (long only) given in Tharp's TYWTF with the fixed fraction and % volatility on S&P 500 stocks and found that generally, I can't trade more than 3-4 stocks at anyone time because of the high prices of the stock even with an initial capital of $1m. With 50% margin, the max. number naturally becomes 6-8 stocks at any given time. The preliminary implication is that trading stocks with these two money management methods might not be feasible for stocks if one wants good diversification.
Moreover, using the % volatility model when the stop-loss strategy is not a function of the ATR (it is the 21 day channel for this case) means that the bet fraction varies with each trade. One of my results show that trading on a 1% vol had a trade with a bet size of 11%! This is pretty high risk ... and unless I made a mistake with my test and assumptions, Tharp didn't seem to have mentioned this issue in his book.
Lastly, stock-splits and (paying or receiving) dividends on shorted stocks gave me some problems. My understanding at the moment is that one needs alot of information such as the size of the dividend and when they are due anda also stock split dates in order to have realistic back-tests. Trading SSF will be more suitable, but historical data is very limited. Does anyone have any suggestions on overcoming these problems? Thanks.