Instead of selecting stocks that are extended upward or downward at the end of the day, why don't you select stocks that statistically tend to gap down or up overnight and short or buy such stocks near the close of the session? AAPL for example, more than half the time, opens at least $0.50 on average above its close of the previous day. That might be an edge you could exploit.
Now we got all kinds of funny remarks but nobody has apparently tested this strategy or doesn't want to publish the results. Has anybody come up with more than jr07 the guy that started this thread? Felix
right, the guy had a question... grabbing a cursory gander at several stocks, it seems rather random (not juxtaposed with the s&p though), like some days that they were killed, the next day opened up or gapped down again. i'd suspect the news would have a big part to play in it. for example, if it killed its earnings and guidance and gapped up, it would be more likely to gap up further the next day due to the ensuing analyst upgrades, etc. or if it missed, then the next couple of days would be more likely to be down for the same reasons. i like your idea of buying the oversold and shorting the overbought at the end of each day perhaps simultaneously so that you are at least market neutral to combat what you were saying regarding the market gapping down the next morn' and you getting jacked because you are just long. some time ago, i had backtested 20 high volume stocks, both NYSE and NASDAQ, to see if i could short the ones that were down the most at different time periods like 10:00, 10:30, 11:00 and go long the ones that were up the most but it wasn't reliable enough. then tried the opposite, kinda like your idea, to countertrend the stocks, which proved unreliable the same. anyway, good luck with your potential system.
I did try something similiar and I really don't think my comments can add anything specific. You have a simple idea, then you paper trade it for several weeks, meanwhile checking some historical data, then you trade it live and it works somewhat till it doesn't then you decide to "tweak". Now comes the hard part as the op noted --------------- "I then noticed that if the general market context is in a certain direction," ---------------- You now are tweaking your original idea to fit a changing market, not the static point where you first began when it worked. Soon your original idea is unrecognizable and so is the market where you first began. What to do? Your simple idea which worked at some point in time, is a tool and now relegated to the tool box because you do not need it anymore. You need a different size wrench for this market. Not to despair, if you hang in there long enough and stay in tune to the market you might recoginze an opportunity in the future when it has value.
I have never backtested a system or idea before how can I do this? I totally agree with the comments about watching the news behind the moves, for instance if a stock is up xx% because it was taken over, it probably wont move much from there. J
as a slacker shortcut, what i was doing to backtest your idea was to just pull up the daily chart of stocks looking for the longer green or red bars indicating the biggest down or up days then just looked at the next day's open and bar to see if it had gapped and what it did. to tighten it up a bit, you might check the close price from prior day then the big move day's closing price and then the next day's opening price (to test for your hypothesized gap) and move for that day and write down the data for each stock over the past 6 months (when i scanned the few, each one only had a few of these relatively "big" days, which could have been 5-15% moves depending on the sector and volatility, like banks will move 15% whereas say CVX only moved 5% on its biggest days). to add a little validity by comparing it to the overall market, for the days that have the big moves, perhaps write a simple excel spread something like: date CVX SPX 3-23 +4.47(+6%) +7% 3-24(Open) -.90 -1% 3-24(Close) -.20 -1.9% from this one sample, you can see that there was a gap down at the open but it showed strength against the entire market closing up 70 cents from the open when the S&P lost another .9%. there are at least 47 ways to skin a cat and i am sure some are not as painfully circumlocutionary as this one but hey, it's just an idea of a backtest that may work for testing your particular system. good luck dude
I have a system like that that I have been tracking for 5 years now. It works. Consider buying deep in the money puts for expansion days up and deep in the money calls for expansion days down. Don't short anything cheaper then $50 bucks. Ok to buy cheap stocks that have cratered. You need to understand why the stock is up or down on the day. Study the news. Here is site for you to get a grip on what news is important. http://thepatternsite.com/eventpatterns.html Also don't short stocks up big on day if they are NOT heavily owned by institutions. Reason is if the stock is up on good news but not heavily owned by funds then they will typically come in and start buying. The price they are paying now looks stupid but they don't care. They think its a deal NOW. If already heavily by institutions then they will be taking profits and selling to other institutions. Also, don't short any low float stocks. Stick with 30 million or more. NO Biotechs. Also, on up stocks with news the high is using 3 days later. It takes about 3 days for stock brokers to call all their clients and get them roped in. Also takes about 3 days for people to take the plunge if they are on their own. John
If you backtest be very carefull of opening prices, they are not reliable very often. Sometimes there is a 100 shares trade 1$ away from the "open price" and in the data you see that ridiculous price.
depends on what you are looking at as an ultra liquid NYSE stock is VERY rarely going to get an OPG print that is out by $1, even 50 cents is freakish on a $70 stock like Exxon (i do some OPG trades with the envelope at just 20 cent intervals of late and rarely get the second level). also depends on the charting software you are using as some are better at eradicating any "tails" caused by aberrant fills. if your software happens to suck at this, you could bring your period down to 5 minutes to get a better idea as to approximately where the actual open was and where you would have been able to realistically get out though that could be waaaaay tedious if you are looking at info from 3 months ago and your software (as it sucks;-) requires you to scroll back through 90 days of info. good point but if you want to backtest, you gotta weigh research time versus the probability of data error and how much an error would really matter to the system you are testing.