It seems that some hedge funds and prop desks try to make profit from intraday trends on stocks. I made some simple calculations on a 50 days intraday data set : I computed the occuring frequency of - let's say - eight up moves out of ten moves. I did it on several stocks, and with different sampling rules : on a time basis, volume basis (after N shares exchanged), and different parameters for this rules. I found that during those 50 days, this event (8 up moves out of 10 moves) occurs 4% of the time which is very close to the theoric binomial probability of 4.4% for a random walk. This would suggest that such "trends" are only a chance phenomenon. Now if some guys are able to make money with trends, it means that : 1) Deterministic trends exist (different from chance ones) 2) Deterministic mean-reverting period exist, in order to compensate from point 1) patterns 3) They can identify these two regimes Well, this result may not be big news but are you ok with it or do you see things differently ?

You mean eight consecutive up moves out of ten consecutive moves? Did you take samples of ten moves in a sliding fashion? (like 1 to 10, 2 to 11... or 1 to 10, 11 to 20... )And what is the data interval?

I mean 8 up moves (not necessarily consecutive) out of ten consecutive moves. That was my first idea to define a trend : a high density of same direction moves.

And the time interval goes from 10 seconds up to 20 min.

i did some similar study study over 10 stocks with 2 moths tick data.instead of "8-out-10" trend definition I used:- linear fit (interpreting bid-ask spread+accumulated error as the current error band) and - simple rule for trend breaking point, when algorithm compares two hypotheses: 1-line vs 2-linesdata were ticks, no intervals.using that I studied the effect of bid-ask volume asymmetry on forward trend and found that everything is neutral and there is no strong correlation, which could be used in the trade-trigger... so, i would confirm your results from the other point.

a point maybe that although direction is random walk, magnitude is not. thus you might be right only at random level, but this might be offset by fat tails in magnitude, enabling to make money.

For what it's worth, regarding intraday volume trends, the first and last hours of the day are the busiest. Lunchtime is the slowest time of the day. Knowing that, intraday price trends early and late in the day have more power behind them. They tend to run farther and faster. Midday price trends are more likely to reverse because of their usual lack of volume. Try to incorporate that into your research.

QuoteOriginally posted by: nazzdackKnowing that, intraday price trends early and late in the day have more power behind them.But do they really exist ?QuoteOriginally posted by: secondMana point maybe that although direction is random walk, magnitude is not. thus you might be right only at random level, but this might be offset by fat tails in magnitude, enabling to make money.Could you give some more details please ? If direction is random walk, how can you make money ?

Come on ! I'm sure that many smart guys out there have a lot to say about this !

there is sufficient information here to dig further.