The threads about random trading started me thinking about this. It would be possible to define trendiness in terms of the profitability of random entry trading with fixed target and stops. It is easy to see that a in a strongly trending market, the win/loss rate of random entry, fixed stop/target trading will approach 50% even with a profit/loss ratio greater than 100%, resulting in a positive expectancy. (50% of the time the random entry will put you in the right direction). This definition of trendiness would be dependent on the size of the moves targeted, so with a little thought it should be possible to create an indicator which takes three parameters - a stop, a target, and time frame. The indicator would look back at the price action over the desired time frame, and give a result based on the probability of hitting the target before hitting the stop, assuming a random entry time and direction. It could be normalized so that 100% represents maximum trendiness (50% win/loss rate of random entry), 50% represents a random market [causes the random entry system to break even -> win/loss rate = loss/(profit+loss)], and approaching 0% means the market is much choppier than random for the size moves you are targetting and you are better off with a wide stop and a tight target.
Picking a target is junk, at least for me. As IF I could tell how far the move will go. Target. Ha ha. If you can pick target, you can predict the future. Better to do as Nicholas Darvas did, and stay with the trade as long as it is acting right. When it quits acting right, run like a theif. Subjective? Yes. That is why it is called experienced. All this target to risk ratio, while necessary from a large perspective both in trading and in life (i.e., I weigh the risk of me having to drive to the reward of me wanting to get home) is pretty much, for me anyway, crap. Quantifying trend would also be, for me, a waste of time or way to lose money.
Thanks! The target and the stop parameters of the indicator are just for measurement purposes. You don't have to actually trade with them, but they should be typical of the kind of moves you go for if you were planning to use the indicator. You said your performance depends on trendiness. If such a trend quantifying indicator proved to be a good short term predictor of future trendiness, don't you think that would help you? Perhaps not if your time frame for individual trades is a big chunk of the trading day - you wouldn't have enough data for the indicator at the start of the trade, and the prior day's data would be useless I'd imagine.
It seems not many would fully understand this whole thread, including myself - from the very 1st post!
I need that also . The definition depends on the usefulness. The usefulness for me - not perhaps for others - is to heriarchise days according to the existence of an entry point that must be the nearest in time to the opening as possible (so that one can just go away after setting the protective stop) and trade profitable at close with pullback as far as possible from the entry.
My exact purpose is to know if ALL the causes for such trending days is due exclusively to upper scale or not. I know for example -see below - that the big rallye on 18th was due to scale 3 which projected 10258/10276 and then 10294/10334 on my model so that we hit straight away the theorical top of 9252.9 (marked automatically as Feynman Local Top BEFORE the market opens of course I didn't add it after the fact ) - in real 9254 for that session (the day Saddam was captured the model seems to project before it was known publicly ). I want to make a check of that. <IMG SRC=http://www.elitetrader.com/vb/attachment.php?s=&postid=393806>
This is not easy. Volatility, momentum and time frame are the constituent factors, I think, but no one study or indicator combines them all. Wouldn't the intraday stochastic and MACD give a pretty good quantification? I once cobbled together a homemade indicator that I called the REI (Range Extension Index). It's nothing but a moving total of upward range extensions, bar by bar, minus the moving total of downward range extensions. The lookback period is arbitrary. I used nine, then later used Ehlers' MESA to find the dominant cycle in the data and made that the lookback period.
This gives an idea of what some trending days have in common - remember the model on the left is calculated BEFORE the market opens so it is aimed to forecast such kind of day in advance: Can see below a bearish pic on daily scale (middle column not first column which is hourly scale) which is due to short term bearish consolidation but the global trend is bullish (global trend in the future not in the past since the model is a predictive one not a descriptive one). <IMG SRC=http://www.elitetrader.com/vb/attachment.php?s=&postid=418698>