[pcabc]

Hi Pcabc, no trend lines. I could have used "slope" i could have used MA > MA[1] i just happened to choose the linear regression as that also signifies turning points. no other reason

Sorry, I typed trend-lines but I meant regression. Still, if it works it could be interesting, it should detect when a moving average is turning up rather than flattening. This may be of use at the ends of stages?

if you use close is greater than 30, then you are too quickly going from one stage to another
take the S&P again with a recent rising 30SMA. if you are in a stage 3 when it closes below, you would be in a stage 2 a couple of weeks later, where ive added the text in my chart. 
the book and method looks to break a "high" an area of some previous resistance. i haven't yet come across in any charting package this concept of a high (almost like a zig zag), so ive used the donchian 20. you can see that the recent donchian break of the 20 again does seem to be doing a good job.
thats all it does, a good job, by no means perfect. 

You've prompted me to have a think about this myself. I think I will have a go at a few methods and have a plasy with channels for that purpose. I effectively use channels to detect new highs in my screener.

thanks for the comments though. my code could indeed be simpler using "slope". I just did it last night having been using amibroker previously which is better at backtesting and yes, you could use my method slavishly although i only did it to backtest the method 
if you can think of other entries happy to look at all of them
thanks again pcabc. I may open another thread simply for this topic