The idea behind the blog would be to use a simple stats test (binomial test), on various entry strategy to see if the entries are actually better then random.
Generally, I'd test for...
The much higher and much lower breakpoints would be based on a % of the average true range.
The time frames after entry tested would be for short, medium, and long term, those being relative to the strategy itself. (EG: For an opening range breakout, I'd test 5 minutes after breakout, an hour after breakout, and end of day)
Now I don't subscribe to the idea that a chart is a chart is a chart, and that the same patterns that work on a daily chart work on a 5 minute chart. I also don't subscribe to the idea that the same patterns that work on one trading instrument work on all instruments.
Consequently, for each signal I'm testing, I would test it under a range of conditions, with a couple different symbols for the sake of thoroughness.
Currently, my candidates would be...
Index ETFS SPY QQQ IWM
Non-equity ETF TLT GLD USO
High Beta NYSE TEX MTW MTG YOKU SUNE
High Beta Nasdaq KERX JASO PEIX ODP ACAD
Low Beta NYSE SO ED DUK CMS DPS
Low Beta Nasdaq NBIX ARCP AMAG AGNC MEMP
While this is already a big list of things to be running tests on, it so far likely underweights small caps, which could conceivably behave differently. It is biased towards high volume instruments, which would likely move differently as well.
Testing setups across multiple categories ensures that if a pattern works for only certain types of stocks, it can still be found.
Testing setups across multiple stocks inside the categories is similar to a mini-replication.
Note though that there is still a bit of a bias if we check over the same timeframe, and it turns out the data is correlated, and so anything that looks particularly promising would be replicated later using data from a different time period.
Here are some quick ideas to get a scope of the kind of things I might test. Most of these are daytrading ideas, although I could test longer term position/swing trading ideas if people express interest.
- Opening Range breakout, 20 minute opening range.
- Opening Range breakout, 20 minute opening range, breakout between 10:00 and 11:00.
- Opening Range breakout, 20 minute opening range, breakout between 11:00 and 2:30.
- Bounce off even dollar levels among stocks between $100-$200, using VWAP to filter for trend direction. Test $30-$80 stocks with half dollar levels, test $250+ stocks with levels ending in 5 or 0. ($255, $260, $265 ...)
- Breakout of last 5 days high
- Breakout of last 5 days high when high wasn't set yesterday. (signifying a need for it to 'cool off')
- Breakout of last 5 days high when high WAS set yesterday. (Does momentum from yesterday continue?)
- Breakout of last 5 days high when high was set yesterday, and close was above the high.
- last 5 days high when... Breakout via Gap
- when... Breakout happens between 9:30 and 10:00
- when... Breakout between 10:00 and 11:00
- when... Breakout between 11:00 and 2:30
- when... Breakout between 2:30 and 4:00
- Entry in direction of breakout if volume in first 5 minutes of breakout is greater then 5% of Average Daily Volume.
- idea 14 but filtered by time
- idea 14 but filtered by index movement. (take trade only if ES Futures are within 4 ticks of their high of day)
Etc. What I would likely try to do is cover a specific theme every week. So perhaps 5 day high/low breakouts is one week. Opening Range breakouts is another week. Gaps might be a third week, and fade strategies based on ATR or Crabel-style stretch values could be a 4th week.
During the week, I'd cover a number of variations on the idea, and with each variation I'd run the tests on the lists of candidates above. This might tell us things like, 'opening range breakouts on high volume are generally much higher one hour later on high beta NYSE stocks (p<0.01), high beta NASDAQ stocks (p<0.05) but not low beta stocks or major index etfs'.
Is this kind of blog you would be interested in reading? Is there anything specifically you would want to see covered?
-
Commodities?
-
Bonds?
-
Small caps?
-
Low volume stocks (<500,000 average daily volume)?
-
Low price stocks (trading at <$5.00)?
-
News plays (I'd probably filter for gap>10%)?
tldr: If I start a trading blog focused purely on entry signals, and testing those entry signals for statistical significance, would you read it?
Submitted May 13, 2015 at 06:21PM by Khaos1125 http://ift.tt/1cy0IPc