No serious trading algo compresses data into bars. Bars are just a visual aid for fitting large timeseries datasets into charts so they can be eyeballed. They should definitely not be used to make trading decisions. Get yourself an unfiltered datafeed (which includes the exchange's timestamps) and feed your algos the raw data (i.e. individual transaction prints and order book updates). Anything less and you are cheating yourself.
Thanks. I trade US stocks, and currently use my broker's data feed (MBT) and eSignal's. - According to eSignal support, the time-stamps in their data are not those from the exchange; instead, they are eSignal's own time-stamps. - I have asked MBT whether their data is time-stamped (which I doubt), and if so whose time-stamp it is (theirs, or the exchange's). Subject to MBT's answer, I suspect I can enhance my existing situation (that is by obtaining time-stamped input data, even if it is not the exchange's time-stamp) simply by using eSignal as my primary real-time data feed (which I haven't been doing until now, depending instead on eSignal just for historical data). As thstart rightly points out, the weakness of this intended approach will be that my actual trades will be executed off a different set of bids and offers (those available through MBT) from the data that my signals are generated from. But this will have to do for now ... It's my plan to review (and potentially upgrade) all my trading infrastructure (trading software, broker platform, data feeds, internet connectivity) once I've reached target performance milestones (and I'm happy with progress so far ... i.e. I am making money!). In the meantime, I will have to accept that my set-up is sub-optimal, and start to plan what to change once circumstances permit. I must cheat myself for a little longer ... Thanks everyone for their input to this thread. I have found all your comments and suggestions very helpful.
Okay, your trading software isn't broken, it's brain damaged for trying to work with a broken data feed. Retail trading is already a rigged game. Why compound this by using ticks that aren't timestamped at the source?
If you're using time-based bars then you can start anywhere and maintain bar synchronization. But, filter_sweep was referring to volume/tick bars, which aren't time-based. All bars that are not based on time have to start at the beginning of a trading session in order to be properly synchronized.
It actually doesn't matter if the timestamps are from the exchange or the 3rd party supplying the feed...just as long as they are consistent and accurate. Good luck.
Dear Newman On behalf of myself - and all the other scum and idiots like me who use ET - could I take this brief opportunity to say how privileged we all feel to spend these few, precious moments online with you each day? Your trading might and experience cast all else in shadow. Your trading brilliance is a bright beacon that shines light for us in dark places. We are blessed that you so recently (March 2010) joined our errant flock. It is not for us to speculate why you have never been among us before. It is certainly not for us to ponder why you no longer work as a junior analyst in the third-rate investment bank/hedge-fund that for a few brief months allowed you to dream you too could be a great trader ...
All I am trying to say is that there is no advantage to OHLC data. And I hope you get to read this before it is removed by ET police. There is no advantage to trying to learn on ET either. Some guy shows up here saying LOOK into volume. Oh yeahh volume based charts. And you jump on it as a possibility. Another guy shows up and goes "you are on the right track" as if he knows. What 99% here don't know and how can they due to lack of work or time to put in the work,... is that an edge that works on say MSFT, will not necessarily work on IBM. This boggles the mind of noobs for obvious reasons. They know how hard it was to come up with an edge in one stock and now they are asked to do it again on another. WHOOPTEDO. Time to cry and look up to god in agony. The only long term success in trading is to understand market as a single equation from which all behavior can be modeled. One must move beyond the stock market and FX market and realize this modeling equation THAT WORKS FOR ALL THINGS THAT GENERATE CHART, WATER PRESSURE PUMP MEASUREMENTS, RANDOM GENERATORS, CPU TEMPERATURE OSCILLATIONS. YOU WANNA SEE FUCKING BRILLIANCE CAN YOU TRADE CPU TEMPERATURE OSCILLATIONS IF YOU HAD TO YOU BETTER BE BE ABLE TO SAY YES, if you want to make it in this field. ok I won't bother you little people anymore. Aufedersein
Could you please enlighten me as to how a single equation can accurately model these disparate systems? Don't be concerned about being too technical for, as someone with a background in physics, computer hardware and financial markets, I should hopefully be able to get the jist. Or are you perchance just the latest troll to frequent these boards?
Oh you'd like that wouldn't you. You'd like that I draw the line from A to B to C. Explain it neatly and even foresee your questions and answer them too. After all why wait for you to ask. I can pre empt your questions. Maybe you think trading is like academia. I shall pour my heart into a paper that my peers will read. And they will clap clap clap how brilliant I am. And I will get offered a new University position. Isn't that how you think life works Mr Educated.