The Base Interval Decision

There is a combo box in the Database Settings dialog labelled Base time interval. Setting it takes four seconds. It decides which questions this database will be able to answer for as long as it exists, and there is no setting, licence, subscription or upgrade that changes that afterwards.

This lesson is about making that choice deliberately: what the control governs, what it costs at each setting, and why the obvious escape route — “I will pick something coarse now and change it if I need to” — leads to re-collecting your entire history rather than to a different database.

The grain, and everything above it

An AmiBroker database stores bars at exactly one interval. That interval is the base time interval, and the User’s Guide’s word for it is the grain: the smallest unit of price history the database physically holds.

Every other interval you see in AmiBroker is produced from that grain by time compression, calculated when you ask for it and thrown away afterwards. You never download five-minute bars and one-minute bars separately. You store one-minute bars, and when you switch the chart to five minutes, AmiBroker groups them five at a time and draws the result.

The consequence is stated plainly in the charting chapter: set the base interval to 5-minute and every periodicity from 5-minute upwards becomes available. Set it to end-of-day and the intraday menu entries are disabled entirely — the guide’s own wording is that you will not be able to use intraday charting or analysis at all until the base interval is set to something below end-of-day.

✕Common mistake

The most common first hour with AmiBroker: a reader creates a database, accepts the default end-of-day base interval, imports data, then finds every intraday entry in the View → Intraday menu greyed out. Nothing is broken and nothing is missing a licence. The database has no sub-daily grain, so there is nothing for AmiBroker to compress into a five-minute bar, and it disables the menu rather than drawing you something false.

Compression upward is free. The other direction does not exist.

Take five consecutive one-minute bars and compress them into one five-minute bar. The rule is mechanical: the open of the first, the highest high, the lowest low, the close of the last, and the sum of the volumes.

Five one-minute bars compressed into one five-minute bar (illustrative values)

Synthetic numbers, used only to show the arithmetic. The rightmost column is the compressed bar.

Bar	09:30	09:31	09:32	09:33	09:34	→ 09:30–09:34
Open	20.00	20.06	20.11	20.04	20.02	20.00
High	20.08	20.14	20.12	20.05	20.09	20.14
Low	19.97	20.03	20.02	19.99	20.01	19.97
Close	20.06	20.11	20.04	20.02	20.07	20.07
Volume	4,200	3,100	2,700	1,900	3,400	15,300

Synthetic numbers, used only to show the arithmetic. The rightmost column is the compressed bar. Prices in this diagram are invented for the illustration. They are not market data and nothing should be inferred from them.

Now try to work backwards. From the five-minute bar alone you know the session opened at 20.00, traded as high as 20.14 and as low as 19.97, and closed the five minutes at 20.07 on 15,300 shares. You do not know which minute the high happened in. You do not know whether the low came before or after it. You do not know whether price went up and then down or down and then up, whether the volume arrived evenly or all in one minute, or whether there was a minute in there with no trading at all.

That information is not hidden, compressed or locked. It was never written down. Compression is a projection: many inputs map to the same output, so the output cannot identify its input.

Everything else about the base interval follows from this one fact.

What “you cannot change it later” actually means

It is worth being precise here, because the folklore and the documentation say slightly different things.

The Base time interval control sits in the Data source half of the Database Settings dialog. The guide describes the General half — the folder name and the Create button — as being enabled only at database creation time, and the Data source half as remaining usable afterwards through File → Database Settings. So the guide does not say the combo box is permanently greyed out, and this course will not tell you it is. What the guide does say, about real-time sources, is that the base interval should be set once at the database creation time, because a real-time source collects ticks and packs them into interval bars, and this setting defines the minimum grain it packs to.

The reason the advice is so firm is physical rather than a matter of the user interface. Suppose you built a five-minute database, ran it for six months, and now change the setting to one minute. Nothing in your six months of stored bars becomes finer. Those bars were compressed on the way in and the discarded detail is gone. From the moment of the change, new bars arrive at the new grain, so you now own a database with a coarse past, a fine present and a discontinuity in the middle — which is worse than either.

What actually happens when you change the base interval on a populated database

1. 1Six months of 5-minute bars are already storedEach one is a lossy summary of five minutes of trading
2. 2You set Base time interval to 1-minuteThe control accepts it
3. 3The stored history does not changeThere is nothing to un-compress; the detail was never written
4. 4New bars arrive at 1-minuteFrom now on only
5. 5Your database now has two grains with a seam between themAny study spanning the seam is measuring the seam
6. 6The real fix: a new database at the finer grain, re-populatedWhich is what you would have done anyway

This is why the decision is treated as permanent even though the control is not disabled.

The practical rule: treat the base interval as a property of the database, not a setting of it. If you need a finer grain, you need a different database.

Choosing: 1-minute, 5-minute or tick

Three questions decide it, in this order.

What is the finest thing you want to ask about? Not the finest chart you imagine looking at — the finest measurement your rules will make. An opening range defined as the first fifteen minutes needs bars that fit inside fifteen minutes; one-minute bars give you fifteen of them, five-minute bars give you three, and three is enough to define the range but not enough to see how it was built. A rule that reacts to the first one-minute close above the range needs one-minute bars by definition.

What can your data source actually deliver, and what does your edition allow? These are two separate constraints and both bite.

What are you willing to store and wait for? Covered in the next section.

Base interval	What it supports	Edition	Typical use
Tick	Every interval above it, plus N-tick and N-second charts, Time & Sales style study	Professional only	Order-flow and microstructure study, very short symbol lists
5-second / 15-second	Everything from 5 or 15 seconds up	Professional only	Fast intraday work where the minute is too coarse
1-minute	Everything from 1 minute up: 1, 5, 15, 30, 60 minutes, daily and above	Standard and Professional	The default choice for intraday trading research
5-minute	Everything from 5 minutes up	Standard and Professional	Longer-horizon intraday and swing work, much smaller database
End-of-day	Daily and above only. No intraday charts at all	Standard and Professional	Position and swing research, long histories

✱AmiBroker Professional required

The Standard edition supports intraday starting from the 1-minute interval. Tick, 5-second and 15-second base intervals, N-second charts and N-tick charts require the Professional edition. This is stated on AmiBroker’s own order and features pages and it is the constraint that matters most for this lesson: if you are on Standard, the tick row of the table above is not available to you, and 1-minute is your floor.

There is also a source-side constraint that has nothing to do with your preferences. AmiBroker’s plug-in documentation is explicit that not every feed offers every grain: the myTrack page states that plug-in supports only 1-minute and daily, and the IQFeed setup walkthrough selects 1-minute. A base interval finer than your feed can deliver produces a database with a grain it can never fill.

!Important

Data-vendor capabilities change and AmiBroker’s provider pages are old — the provider catalogue page states of itself that it is current as of March 2023, and the IQFeed setup page still links a client version from an earlier era. Treat every per-vendor grain and depth in this course as what the official documentation says, and confirm the current figure with the vendor before you commit a database to it.

What each choice costs

The arithmetic is worth doing once, on paper, before you touch the dialog.

A United States equity regular session runs 09:30 to 16:00, which is 390 minutes. So one regular session is:

Base interval	Bars in one 390-minute session
1-minute	390
5-minute	78
15-minute	26
Hourly	7, depending on how the last part-hour aligns

AmiBroker’s performance chapter gives the storage constant: one data bar is 40 bytes, and the Number of bars to load setting causes that memory to be actually allocated, per symbol. So 100,000 bars is about 4 MB per symbol, and with a 500-symbol in-memory cache the guide notes you can reach around 2 GB. The chapter’s advice is to keep the working set inside the CPU’s on-chip cache, and with a 4 MB cache it advises strongly against exceeding 100,000 bars.

Turn that into history:

Base interval	Bars for 100 regular sessions	30,000 bars is roughly	100,000 bars is roughly
1-minute	39,000	77 sessions, about 15 weeks	256 sessions, about a year
5-minute	7,800	384 sessions, about 18 months	1,282 sessions, about five years

Those are regular-hours numbers. If the feed also stores pre-market and after-hours bars, each calendar day consumes far more of your bar budget. The guide’s own IQFeed figures illustrate this: it recommends 100,000 bars on a 1-minute database to obtain the maximum history available, which it describes as about eight months. Eight months is roughly 170 sessions, and 100,000 divided by 170 is around 588 bars per session — considerably more than the 390 minutes of a regular session. That arithmetic is an inference from the guide’s own two numbers, not a vendor statement, but it points at something you should check directly on your own data: how many bars per day is your feed actually storing?

✦Tip

The Database Settings dialog displays the equivalent number of days next to the Number of bars field, computed for the base interval you have chosen. Read it before you press OK. It is the fastest sanity check in the program, and it is the one people skip.

One database per purpose

There is a temptation to build one database that does everything: ten years of daily history and one-minute intraday, one place, one set of symbols. AmiBroker’s own documentation pushes back on this in two different ways.

The first is the Allow mixed EOD/Intraday data option in Intraday Settings, which lets a single database hold both true end-of-day bars and intraday bars. In intraday view modes the daily bars are dropped on the fly; in daily mode the real daily bar is preferred over one compressed from intraday. The guide names IQFeed and eSignal 1.7.0 or higher as plug-ins supporting it — and dates that statement “as of June 2008”, which tells you how long ago that page was last revised.

The second, repeated on several of the provider pages for feeds that do not support mixed mode, is blunter: run two databases, one end-of-day and one intraday, in two AmiBroker instances if you need both open at once. The same advice appears in the author’s guidance about bar counts — keep a separate database for real-time trading and another for long-history analysis, rather than one enormous real-time database.

This is not a workaround. It is the correct architecture, and it follows directly from the first half of this lesson. The two databases answer different questions, need different grains and need different bar counts. A database is cheap: it is a folder. Your history is not.

A workable split for most readers:

• A long-history end-of-day database. Base interval end-of-day, a large universe, years of bars. This is where Parts 12 to 16 and 28 to 35 live.
• A short-history intraday database. Base interval 1-minute, a small, deliberately chosen watch list, weeks or months of bars. This is where Parts 20 to 26 live.

If you have no live feed

Nothing above depends on a subscription. The base interval of a local database — one whose Data source is (local), fed by imported files rather than a plug-in — behaves identically: it sets the grain, it enables every interval at or above itself, and it cannot be un-compressed either.

So the Level A route through this part is:

1. Create a local database with base interval 1-minute (or 5-minute if your files are five-minute bars — importing one-minute-grained data into a five-minute database is not possible, but importing five-minute bars into a one-minute database is fine, since five-minute bars are simply bars stamped five minutes apart).
2. Import historical intraday files for a handful of symbols. The lab at the end of this part gives the exact import definition.
3. Use Bar Replay to move through those bars as if they were arriving live. Bar Replay truncates the whole database at a playback position, so a formula sees only bars up to that point — which is what a live feed does, without the feed.

You will not have streaming quotes, and Parts 22 and 23 will tell you honestly where that matters. You will have a real intraday database, at a grain you chose, with all of the storage arithmetic and every maintenance problem that a paid database has.

Recap

The base interval is the grain: the finest bar the database physically stores. Every coarser interval is compressed from it on demand, which is free; the reverse operation does not exist, because compression discards the information that would be needed to invert it. The control is not disabled after creation, but changing it does not retroactively refine stored bars, so in practice the decision is permanent and the real remedy is a second database.

Choose the grain from the finest measurement your rules make, constrained by what your edition allows — tick and second intervals are Professional — and by what your source can deliver. Then size the bar count from arithmetic: 390 bars per regular session at one minute, 40 bytes per bar, and a dialog that will tell you the days-equivalent if you read it.

The next lesson goes through the Database Settings and Intraday Settings dialogs control by control, including the several controls that do nothing whatsoever in a local database.

Check your understanding

Question 1. A database has a 5-minute base interval and two years of stored bars. You change the base interval to 1-minute. What happens to the two years of history?

• It is re-compressed to 1-minute bars over the next few minutes
• It stays exactly as it is; only bars arriving from now on use the finer grain
• It is deleted, because the grain no longer matches
• AmiBroker refuses the change until the database is emptied

Show the answer and why

Answer: It stays exactly as it is; only bars arriving from now on use the finer grain

The stored bars are already-compressed summaries. There is no finer detail inside them to recover, so the change affects only future bars and leaves a seam in the middle of the history. This is why the base interval is treated as a property of the database rather than a setting.

Question 2. Which of these are genuinely available from a database whose base interval is 5-minute? Select all that apply.

• A 15-minute chart
• A 1-minute chart
• A daily chart
• An hourly exploration

Show the answer and why

Answer: A 15-minute chart, A daily chart, An hourly exploration

Every interval at or above the base interval is available by on-the-fly compression. Nothing below it is, at any price. A 1-minute chart would require detail that was discarded before the bars were written.

Question 3. You are on the Standard edition and want to study order flow using tick bars. What is your position?

• Set the base interval to tick; the edition affects only the number of Analysis threads
• Tick and N-second base intervals require the Professional edition, so 1-minute is your floor on Standard
• Tick data requires a database rebuild but no licence change
• Tick bars are available but limited to 10 symbols

Show the answer and why

Answer: Tick and N-second base intervals require the Professional edition, so 1-minute is your floor on Standard

AmiBroker’s edition comparison puts intraday "starting from the 1-minute interval" in Standard, and all tick and second intervals, N-second charts and N-tick charts in Professional. The 10-symbol Standard limit is a separate restriction, and it applies to the Real-Time Quote window.

Question 4. Your feed stores pre-market and after-hours bars as well as the regular session. What does that do to a fixed "number of bars to load" setting?

• Nothing; extended-hours bars are stored separately
• It covers fewer calendar days, because each day consumes more of the bar budget
• It covers more calendar days, because extended-hours bars are smaller
• It has no effect until session filtering is switched on

Show the answer and why

Answer: It covers fewer calendar days, because each day consumes more of the bar budget

The setting caps bars, not days. A day with 390 regular-session minutes plus extended-hours bars uses more of the cap than a regular-session-only day, so the same bar count reaches less far back. This is one of the root causes of the incomplete-history challenge later in this part.

Question 5. Which is the better reason to keep separate end-of-day and intraday databases?

• AmiBroker cannot open a database larger than 2 GB
• They answer different questions and therefore need different grains and bar counts
• Intraday data corrupts daily data when stored together
• Licensing requires one database per interval

Show the answer and why

Answer: They answer different questions and therefore need different grains and bar counts

The split is architectural, not a limitation workaround. A long-history daily database and a short-history fine-grained database have different requirements at every layer. AmiBroker does document a mixed EOD/intraday mode for some plug-ins, but the guide still recommends two databases for feeds that lack it.

Check my answersClear