My reading of it is that the 'bearish three-line strike' is ranked number one overall, I think the low frequency of the three stars hurts the overall rating. Can anyone figure out what criteria he uses to determine if a pattern is successful or not?
My reading of it is that the 'bearish three-line strike' is ranked number one overall, I think the low frequency of the three stars hurts the overall rating. Can anyone figure out what criteria he uses to determine if a pattern is successful or not?
Albert Carey
Member
- Messages
- 88
- Likes
- 3
APS can work with NT daily data right after you export files. If you want to work with intraday data you need to do some extra work, which is a pain in the neck because NT does not place a delimiter after the date and you have to edit the file and do it yourself using an editor. APS support has detailed instructions for that here
http://www.tradingpatterns.com/Support/ninja/ninja.html
kilotrader
Junior member
- Messages
- 12
- Likes
- 1
I completed looking at the Bullish Harami, Bullish Engulfing Pattern, Hammer and the Doji and found the doji as the most reliable. Results are at :
http://blog.kilotrader.com/2010/05/summary-of-candlestick-backtesting.html
http://blog.kilotrader.com/2010/05/summary-of-candlestick-backtesting.html
What was your criteria for 'success'? How far would it have to reverse to be considered one of the successful signals? I use PIN bars which are similar to hammers and you can scalp them with a high success rate even though they don't always turn out to be trend reversals.
kilotrader
Junior member
- Messages
- 12
- Likes
- 1
For this analysis, I looked at next day close to establish the success criteria.
I am tempted to think that the PIN bar is more like a doji. The required that Open and Close are very close to each other. For this analysis, I used the criteria that the distance between open and close is that within 15% of the total day's trading range. I would think that in a PIN BAR you would want to tighten this criterion for less than 5% or may be 2%.
Details @ http://blog.kilotrader.com. Specific posts of interest :
http://blog.kilotrader.com/2010/05/summary-of-candlestick-backtesting.html
http://blog.kilotrader.com/2010/05/do-candlestick-patterns-work.html
I have found in testing that nothing seems to work without some discretion being used. I trade PINs but only at certain support and resistance levels. I programmed PINs into tradestation some time back and I could always beat the automated results by using my own discretion and picking which ones to enter.
Having said that I think you could improve your automated results by adjusting you profit levels. It's not really about percentages it's about how many points you make, some systems with less than a 50% hit rate can be successful and likewise some with 90% hit rates can fail. You only have to look at the recent holy Grail thread to see that
Having said that I think you could improve your automated results by adjusting you profit levels. It's not really about percentages it's about how many points you make, some systems with less than a 50% hit rate can be successful and likewise some with 90% hit rates can fail. You only have to look at the recent holy Grail thread to see that
wackypete2
Legendary member
- Messages
- 10,211
- Likes
- 2,058
kilotrader
Junior member
- Messages
- 12
- Likes
- 1
I tried to do some runs to figure out if PIN bars are better/worse. I changed the criteria for the open to be equal to close. So, I was looking for 5 down days and the sixth day where open=close. I found NO instances of this condition on SP500 stocks for the last 15 years. So I did a parametric study where I changed the gap between open and close in relation to high and low from 0 to 20%. The results are at : http://blog.kilotrader.com/2010/05/doji-parametric-study.html
I am actually surprised at the results. The trend reversal rates where ope-close gaps is less than 10% of the candle length are actually quite low ( less than 50%).
Essentially this means that a doji after 5 down days is NOT important and you could do just as well be looking at a up day after 5 down days...
wackypete2
Legendary member
- Messages
- 10,211
- Likes
- 2,058
Interesting stuff. Well done. Your blog site made for some interesting reading.
Peter
Peter
I look for PINs at swing high or low rather than at a set number of up or down bars, if the open equals the close then normally they are good but it doesn't happen much even in 5 or 15 minute charts. As I said before I found that anything I programmed wasn't as good as manually choosing the trades but then I only have basic programming skills
ps you might want to try using slow stochastic as a guide to identifying the swing high/lows,I have it on my charts.
ps you might want to try using slow stochastic as a guide to identifying the swing high/lows,I have it on my charts.
If anyone has a used copy for sale please pm
