Scan for "downward thrust" (Dave Landry style)

Comments

• Options

  markd mod

  December 2019 edited December 2019

  I don't keep up with Dave Landry, but here's a suggestion:
  
  and [Long Body is true] // open and close near extremes of candles, so selling only ( or buying if up close)
  and [range > ATR(14)*2] // wider than normal range (you can adjust parameters 14, 2)
  and [close < 1 day ago close] // be sure to get a down day
  and [low = min(10,low)] // less likely to be a retracement of strong up day if a new low (you can adjust the parameter)
  
  You could add volume, too:
  
  and [volume > 1 day ago volume] // higher volume, so lower prices attracting new sellers
  and [volume > sma(21, volume)] // or whatever average you like
  
  

  0
• Options

  lmkwin ✭✭

  December 2019

  What do you consider "significant"? A stock that goes down 5%? In a day/ week/ month? I think if you can answer that, it will be easier to "fine tune".
  
  I'd look at the PctChange or maybe a change in SCTR or something like that. On the homepage of StockCharts, you can view the "Additional Data Panels" and look at the SCTR changes. It shows in a "top ten" kind of format. The ones that are down the most might be a good place for you to think about it? Also shows the top % changes for various indexes and exchanges.
  
  You also might want to take your "alert" and put it into the Advanced Scan Workbench and then run it against prior dates to see what results it throws off. You can "back test" the scan by changing the date at the top of the Workbench to a starting date that is back from 0 or today.

  0
• Options

  cillyElph ✭

  January 2020

  Thanks for the responses to this. Apologies for not responding. I plan to try some of the tips above from Markd on this. Looks promising. And to IMKWIN, you hit the nail on the head. Wouldn't it be nice if "significant" here could be quantified. No such luck (at least not yet). FYI - Landry encourages his followers to have relatively few indicators - he suggests that we simply look at 100's (maybe 1000's ?) of charts per day until you can eyeball a chart and see a trend (without your glasses) or a downward thrust that should be very obvious, If it isn't obvious to the eye, he doesn't recommend using it (for his TKO method anyway). I like a lot of his commentary so I plan to continue following him and trying to learn his trading strategies but that doesn't mean I won't be trying to find shortcuts using indicators. Thanks again.

  0
• Options

  markd mod

  January 2020 edited January 2020

  Well, it's good advice to look at 100's or 1000's of charts. But, it will be a lot more productive if you do it in an orderly - rather than random - way.
  
  One way to do this is to specialize, at least for a while, in one or two or three sectors, or even one or two or three industries in different sectors - one sector at the the top of the SCTR rankings - so maybe sctr.industry > 75, one in the middle - maybe 45 - 75, and one under-performing - less than 45.
  
  Then rank the symbols in each of these industries by sctr (some won't have a sctr) and divide them up in a similar way.
  
  Then see how similar the charts are (or are not) within a sctr range, notice which ones tend to move together, which independently, which lead, which lag, and what clues do they give (if any) before they start to move up or down.
  
  For instance, you might notice that for weak stocks, a rally might have a hard time getting going, then surge. That looks like a breakout, but in fact its often a buying climax. A similar pattern in an uptrend is more likely to be a genuine breakout, and it might be worth an entry if other signs are positive. So, apparently similar behaviors, but different outcomes because the trends are different.
  
  A good way to look at lots of stocks at once is candle glance. You can have only one candle glance style in use at a time, but you can replace the Candleglance style whenever you want. So have at least two - CG Daily - maybe a three or four month daily chart, and CG Weekly - maybe a two year weekly chart. You could even have a CG Monthly chart - maybe 10 years of monthly (the length will depend on how much detail will reasonably fit into the candleglance thumbnail). Switch between them and see how charts look in different time frames.
  
  

  0