Data Feeds that Match Bar Opens and Closes?

Discussion in 'Data Sets and Feeds' started by achilles28, Oct 31, 2012.

  1. In my experience, most futures data providers don't match bar opens and closes. Either I'm missing something, or even "tick" data providers, like Barchart and TT, are out to lunch.

    In my humble experience, with time bars, when one closes, immediately, the next opens. Mismatched O/C's between adjacent bars tells me that the data feed is not true tick, large gaps exist in the data, or it's a charting issue. I see this all the time with other traders too, that post snapshots of their charts. This is a dangerous way to trade, imo = crap data.

    What's the best way to solve this?
  2. Could you post an example, where you consider tick data not to match OHLC of a bar?

  3. Here are two examples (Barchar and IQFeed).

    If you look closely, many opens and closes of adjacent bars do not line up.

  4. These are time (5-min) bars, constructed from the underlying tick data.

    The tick data is a record of last traded price and volume at each time a trade is transacted. As trades occur whenever a marketable order crosses a resting limit order, ticks will occur unevenly spaced in time (i.e. whenever two "traders" trade) and potentially with gaps between adjacent ticks (e.g. in the simplest case, if bid and offer stay the same, and if a trade at the bid is followed by a trade at the offer, then there will be a gap between adjacent ticks equal to the spread, etc).

    A time bar's OHLC reference the price of the first tick (open), the price of the highest tick (high), the price of the lowest tick (low) and the price of the last tick (close) between two points in time. As these ticks can occur with gaps between them, there is no reason that the price of the last tick of one bar (close) has to be at the same price as the first tick of the next bar (open).

    Indeed, these ticks can occur at any time within this interval (i.e. they don't have to be evenly spaced out over the time interval), or even not at all (in which case, there will be no time bar for that interval). For example, if a time bar just had a single tick in it, then that bar's OHLC would be a single price. And there is no obligation for that price to be the same as the previous time bar's close.

    One can construct different bar types (e.g. range bar, renko brick, median renko brick, etc) that have different rules, which might specify that an adjacent bar/brick must begin where the previous bar/brick ended... but I haven't seen that done with time bars (which is not to say that it isn't done ... just that I haven't seen it).
  5. Right - there is no reason that they should match up.

    The last tick of the previous interval will in general not match the first tick of the next.
  6. For tick bars, I understand that (bad example, on my behalf).

    For time bars, no, that's incorrect. The interval between time bars have nothing to do with ticks. When one closes, the next opens. It's that simple. With time bars, there should be no gap between sequential closes and opens. They should be exact. The close of one and the open of the next is exactly the same point because it's based on time, not ticks.
  7. 2rosy


    create your own bars from the tick data.
  8. 1. I don't see this as a significant issue... maybe if you're "scalping for a tic or two"... but you probably shouldn't be doing that anyway.. it's too difficult.

    2. If there are a few ms between the closing of one bar and the opening of another, there can be a few cents/tics price change during that interval.

    As I watch a tic chart on CNBS, the price changes sometimes are so fast that the numbers are blurred.... we're talking "hundredths of a second"... I use eSignal where I believe they report only "one tic out of 6 or 10 actual"... even that shows fast price changes.
  9. What data service do you guys and girls like for CME futures?
  10. The thing is, I trade patterns off the 10s, 30s, 1min and 5 min chart. I don't trade bar to bar, but accurate ohlc of the entire forest is necessary to get a good picture. If 50% of opens and closes don't match on the 10s, that's a boatload of "missing" data for a tick feed, on a time chart, ya?
    #10     Oct 31, 2012