Best Definition of a Pullback

JJJJ

When scanning for a pullback, I have read that different people define a pullback differently. I have seen the following:
1) 3 consecutive black bars or
2) 3 consecutive lows or
3) 3 consecutive highs or
4) 3 consecutive closes or
5) A combination of the above

What is the feeling out there as to the best indication o?f a true pullback that I can incorporate in my scans ?

markd

Well, I suppose you could say that a down tick is pull back.

That's kind of extreme, but it's true, depending on your perspective. Also, it seems it would make a difference whether you were in a daily, weekly or monthly time frame, or five minutes or an hour. The amount you have at risk in each time frame is very different.

Or, you could say a pull back is where you start feel pain if you are long, or start to sense opportunity if you are waiting in cash.

I don't think an objective definition is possible. I'm assuming what you want to know is, when does the selling generally stop and the buying return. In other words, you want to find the points of support that make good entries.

The usual definition of an up trend is higher highs and higher lows. So if an up trend is in place, you should expect to see support above the previous low. But where, and how would you scan for it? I think Fast Stochastics is useful. It tracks where the close is in relation to the range of prices in the time period specified in the parameters - 5 days, 10 days, 21 days, whatever (for all of these, the second parameter is 1, so there is no signal line). For some reason, time, and not just price, seems to be an important element influencing buyers and sellers to act. When a stock is in a strong up trend, a dip below 20 on the Fast Stochastic scale is very often a very good set up.

The master of this approach is Joanne Klein. She has a Public Chart List called Above the Green Line. If you haven't, you should check it out. It's been rated number 1 for a very long time. I don't trade it, but that's because I like to go my own way. (P.S. - I have zero connection to JK - don't know her, never met her, never talked or corresponded; I just think she does good work worth recommending.)

If you want to test this idea for yourself, you could write a scan that requires a strong SCTR score - in other words, a strong up trend - and Fast Stochastics 5,1 < 20. I think it probably works best on big or mid cap stocks, especially if they are in the major indexes - the SP100, 500, 400 or NDX. In other words, look for institutional quality stocks that have a pretty regular looking chart most of the time. (note: JK actually use IBD RS scores, but you can't scan for those on Stockcharts - pretty sure you will get plenty of hits with SCTR instead).

You could also try Fast K 10, 1, or 21, 1 or any other number x, 1 and compare the results. In general, the shorter the time period (e.g. 5, 10) , the more hits, but also more duds. Both the good hits and not so good hits seem to run in batches. You should probably avoid trades that had very heavy volume and long red candles in the down leg that produced the hit, and/or an up leg on low volume and candles with small range and/or small bodies (more wick and tail than body). That could indicate a change of trend, although sometimes it doesn't. It might help to overlay fib lines and take the trades nearest, say, the 50 or 62 - just an idea, I haven't tested it.