Once upon a time I wrote up something on candlesticks and larned a bunch of neato words, like "spinning tops" and "hammers" and "dojis", each of which seems to make some Buy/Sell decisions by looking at a candlestick pattern for a single day. Then I decided to look at the evolution of the candlestick patterns for two weeks. In particular, I looked for interesting patterns preceding stock price minima or maxima. Now that has been fun! As usual, the Buy & Sell signals generated are sometimes excellent ... and sometimes lousy. However, if anybuddy would like to play, the stuff is here: http://www.gummy-stuff.org/gBS.htm
I believe candles are a very powerful tool. However, and this is a big however, they are not a system in it of itself. You must use candles as PART of your system (eg not going long just because you see a hammer) There must be some other contributing factors that will implore you to buy. When people just simply buy a hammer, and it doesn't work, they blame the candles for not working. Thats the problem with people that backtest candle patterns and say they dont work. I will agree 100% that if you back test candle patterns it will show they dont work. This is assuming a computerized back test. But if you do a manual backtest and use candles in conjunction with whatever else you want to use, I find that they work very well. Just my take.
I think the point 273 was making was you CAN backtest candles, BUT NOT with a computer program. Candles are very 'back testable' (if that's a word) when done properly, which means basically doing it manually. There are not many computer programs out there that can properly detect true candle patterns. For example, Steve Nison teaches candles as a potential signal of the end of a trend. Therefore, in order to trade like that, you must have a trend. Programming a trend into a computer language is not easy and what I call a trend and what you call a trend can be very different things. So, backtesting that through a simple computer program will not work.
Actually I think you could backtest candle patterns if you really wanted too ... 10 years ago I set up a simple scan to look for hammer bottoms; I used quotes plus at the time; and I really wasnt a good programer; but I figured it out ... I'm sure there are programs out there that can scan for certain candle patterns or someone I'm sure has written one for the various charting packages ... I Agree with the above posters thou; candles can be decieving without additional things to back them up ... But they are a powerful tool. A trick I learned in one of Steve Nison's books was if your studying candles and you cant figure out what you are looking at; combine 2 of the candles together to see what candle it forms. Alot of times the combined candles form something more obvious like a engulfing pattern or/dark cloud cover/hammer/shooting star etc ...
Kut2k2, I fully agree with you.I dont understand this "mantra" that candlesticks fail on their own,but are a great secondary indicator... Candlesticks are one of the few "indicators" that can be easily programmed.Yes,there are various definitons of trends,but that too can be fairly easily programmed. If they dont work in a backtest,the only reason they work in actual trading is due to the fact that the user is a talented discretionary trader.He may owe his success to the candlestick formation,but in reality it is no more than the placebo effect
tao - I disagree completely. Candlesticks cannot be easily programmed, so I am not sure where you heard they were easy to program. Yes, you can easily code a 'hammer' but a hammer in and of itself is not a legit buy/sell signal. Plus, candles, esp intraday are not a perfect science. So if your 'coded hammer' does not trigger, that does not mean a great hammer is not staring you in the face. The 'candlestick guru' Steve Nison just recently designed a screening software package that takes into account a variety of factors. Outside of this, I have not seen anything that truly looks at candles and how they are supposed to be traded. And this software he now has only does end of day as intraday is not available (at least not yet). As for this whole backtesting thing, by your theory, if it can backtest fine, then it should be great in real-time. If that was the case, every trader here would be making millions with their great, curve fitted, backtested systems. I love how some traders think something must backtest great or it's not worth looking at, yet the majority of these wonderful back tested programs fail in real-time. Maybe not right away, but eventually they almost always fail. Knowing that, why would you put so much emphasis on a computer generated backtesting? Oh yeah, there's no 'placebo' effect when trading with candlesticks. Once learned properly, candles can tell you a lot of what is going on in the markets.
Brownsfan prettimuch sums up a near all of it. I think any candlestick trader will tell you to always use candles formations to either confirm your signal - Or use candles for signals, but always confirm by an alternate system. In general â as I think the overall point of this post drives for â candle formations are awesome. Consistently the biggest and best moves on daily and weekly charts are preceded by textbook candlestick formations. My recent favorite was <a herf="http://www.fxwords.com/b/bearish-evening-doji-star-candlestick.html">bearish evening start doji â short NZDJPY way back on a daily chart 12/7/2005</a> Just classic confirmation of an overextended move.
This discussion is a mystery (to me). Can a computer be be programmed to display candlesticks? Yes. Can a computer be programmed to recognize particular patterns, like hammers, dojis, etc. Yes. Can a computer be programmed to generate buy/sell signals? Yes. Can the buy/sell signals be backtested? Yes. Are the buy/sell signals any good? Sometimes yes ... sometimes no. http://www.gummy-stuff.org/candlesticks.htm