You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
range vs trend and how to predict it
- Thread starter travis
- Start date
Thanks for sharing your work, newbie2006 and flip. It sounds like you have done more work than I did, and more competent. I hope everyone else can now add to this topic here, because I have ran out of statistical material to back up my assertions on this subject. I realize now that mine were most of all intuitions, probably right but not statistically proven, plus some statistical work, which I am too ignorant and tired to finish as I should.
I think you really need to look at intraday day data to find a stock or SIF really trended (i.e opended near H/L and closed near L/H). I am currently working on trying to measure the intraday volatility along these same lines of thinking. Just looking at the EOD data isn't enough. A stock could have a nice trend in the morning, then reverse with a nice trend in the other direction and close near the open (all in all a very tradeable setup) vs a stock that just gets whipped round all day, but the EOD data would look identical to the former.
If you want another edge, look at the ATR for the last 10 days and plot an 8EMA on it. Look for breakouts above and below to carry some follow through for at least a couple days if not more. Remember it does not always mean you will have a sustained directional move, but you will definately be getting range expansion days and more volatility. Also look for narrow range days, inside days, doji's, structure and cyclical patterns to help with your analysis. Thats all I am giving away, but just with those few techniques you can predict these days with greater than 70% if you apply them correctly and spend some time studying the data.
Quote from giggollo:
Interesting, maybe with this measure of trendiness, some stocks can be categorized as "trendy" and others as "not trendy", and different strategies can be applied , depending on whether a stock is trendy or not..good work
it would indeed be very interesting, whether there are stocks that are on average more trendy than others. a possible strategy would be in such a case to form 2 groups of stocks with equal size (10 or more stocks each). then apply only a breakout strategy (e.g. ORB) to the trendy group and only a mean reversion strategy to the less trendy group. when the overall markets trends strongly up or down, the breakout group will make profits while the mean reversion group loses. however, the profits of the first group should outweigh the losses of the second group if the trend/nontrend effect remains robust.
but thats just a first thought, have not tested something like that yet. the question is also, whether such an effect exits at all. i suspect, the trendiness / non-trendiness changes rather randomly across stocks, but who knows...
do you trade stocks?
Quote from Prevail:
flip, keep those numbers coming. it sounds like predicting those 16 percent trend days is still elusive.
yes, in my experience it's still rather elusive, but maybe different conditions - like the suggestions from apex82 - can help to produce better results.
two other ideas that come to my mind:
1) include other time series / other markets. for instance, if today crude oil (or commodities in general), or bonds etc. are trending / have a day with a large range, there might be an influence on the trendiness of s&p the next day.
2) use some econometric methods (e.g. garch) to forecast the range and trendiness, or at least use such forecasts as additional factors.
is there anyone who has done some tests or has experience concerning such alternative approaches?
Quote from apex82:
If you want another edge, look at the ATR for the last 10 days and plot an 8EMA on it. Look for breakouts above and below to carry some follow through for at least a couple days if not more. Remember it does not always mean you will have a sustained directional move, but you will definately be getting range expansion days and more volatility. Also look for narrow range days, inside days, doji's, structure and cyclical patterns to help with your analysis. Thats all I am giving away, but just with those few techniques you can predict these days with greater than 70% if you apply them correctly and spend some time studying the data.
apex82, thanks for your comments. iâve a question regarding the ATR EMA breakout suggestion: with âbreakout aboveâ, do you mean that the ATR of the last 10 days is higher than its 8EMA for an âabove breakoutâ? so, when (ATR,10) > EMA((ATR,10),8) then expect high range and trendiness on the next days?
Quote from flip:
it would indeed be very interesting, whether there are stocks that are on average more trendy than others. a possible strategy would be in such a case to form 2 groups of stocks with equal size (10 or more stocks each). then apply only a breakout strategy (e.g. ORB) to the trendy group and only a mean reversion strategy to the less trendy group. when the overall markets trends strongly up or down, the breakout group will make profits while the mean reversion group loses. however, the profits of the first group should outweigh the losses of the second group if the trend/nontrend effect remains robust.
but thats just a first thought, have not tested something like that yet. the question is also, whether such an effect exits at all. i suspect, the trendiness / non-trendiness changes rather randomly across stocks, but who knows...
do you trade stocks?
Interesting ideas, yes i trade stocks
Hey when y'all are finished figuring out the market trendiness, don't forget to look at specific sectors, too.
As for the daily trendiness, rather than take the open (MO) and close (MC) you might make use of the average high and average low measured at various times during the day. For example:
Market open at 10,000, high is 10,200, low is 9,800, close is 10,000. Trendiness would be 0 but it was obviously a very trendy day. Instead of MO and MC, the average high (only > MO, MC) measured at 10, 10:30, 11, 11:30...3:30, 4:00 might be more instructive. Same for the low (only low < MO, MC).
Instead of MO-MC you would have something like AVG(10050, 10100, 10200, 10100, 10050) - AVG (9975, 9825) which would give you a much higher number, close to 1 rather than 0. Since the day was trendy that would be more accurate.
Chabah
As for the daily trendiness, rather than take the open (MO) and close (MC) you might make use of the average high and average low measured at various times during the day. For example:
Market open at 10,000, high is 10,200, low is 9,800, close is 10,000. Trendiness would be 0 but it was obviously a very trendy day. Instead of MO and MC, the average high (only > MO, MC) measured at 10, 10:30, 11, 11:30...3:30, 4:00 might be more instructive. Same for the low (only low < MO, MC).
Instead of MO-MC you would have something like AVG(10050, 10100, 10200, 10100, 10050) - AVG (9975, 9825) which would give you a much higher number, close to 1 rather than 0. Since the day was trendy that would be more accurate.
Chabah