台指全 explained: all-sessions TAIEX futures data for backtests
台指全 joins the TAIEX futures day and night sessions into one continuous series, rolled from one contract month to the next. Use it when your strategy trades or reacts to the night session, and check the roll method and the bar count per day before you test.
By Soren Huang · Published · updated · Editorial policy
Contents
What does 台指全 mean?
台指全 is a label many Taiwan charting and quote tools use for a TAIEX futures series that joins the day session (08:45–13:45) and the night session (15:00–05:00) into one chart. It is usually built from the nearest contract month and switched to the next month at each roll. Use it when your strategy trades at night.
The character 全 means "all" or "whole", so the name says "TAIEX futures, all sessions". Naming differs between data vendors. Some tools pair it with a day-only series, and some add the contract month to the name. Open the symbol settings in your tool and confirm two things: which sessions are included, and how the contract months are joined.
The underlying contract is the TAIEX futures contract (TX) on the Taiwan Futures Exchange (TAIFEX), where one index point is worth NT$200. The Mini-TAIEX contract (MTX) trades in the same sessions at NT$50 per point, so everything below applies to it as well.
What exactly does an all-sessions series combine?
An all-sessions series combines two things: both trading sessions of each trading day, and a chain of contract months joined end to end.
The night session is TAIFEX's after-hours session. TAIFEX added it in 2017. Its trades count toward the next trading day, not the calendar day on which they happen. So a trading day begins at 15:00 on the previous afternoon, pauses at 05:00, resumes at 08:45 and ends at 13:45. The Friday night session runs until 05:00 on Saturday and belongs to Monday's trading day.
The second part is the contract-month chain. Several TX contract months trade at the same time, and each one expires on the third Wednesday of its month. A continuous contract is a single price series made by taking one contract month at a time and switching to the next before or at expiry. This is what 台指期連續月資料 (continuous-month TAIEX futures data) means. Without it, you would have a separate short chart for every month and no long history to test on.
How is a continuous series stitched at the roll?
The roll is where the continuous series is cut and rejoined, and the method used there changes the prices your backtest sees. Rolling is the switch from the expiring contract month to the next one. At that moment the two months usually trade at different prices. For TAIEX futures, the gap between months is often largest in the summer dividend season, when the index is expected to drop as stocks pay dividends.
Vendors handle that gap in one of two main ways:
| Method | What it does | Good for | What it distorts |
|---|---|---|---|
| Unadjusted (raw) | Joins the actual prices of each month as they traded | Real price levels, round-number levels, matching old charts | Creates a jump at every roll that no trader actually experienced |
| Back-adjusted | Shifts all older prices by the roll gap so the series has no jump | Point-based profit and loss, stops and indicators across rolls | Older price levels no longer match what traded on that day |
Back-adjustment is the process of moving every price before a roll up or down by the size of the gap at that roll. In plain words:
Adjusted price = raw price + the sum of all roll gaps after that bar
Here the "raw price" is the price that actually traded on that bar. A "roll gap" is the next month's price minus the expiring month's price at the moment of the switch, and it can be negative. The newest prices are left as they are; only the older history moves.
Hypothetical example
At the roll, the expiring month trades at 20,000 and the next month at 19,900, a gap of −100 points. In an unadjusted series the chart drops 100 points in one bar. A long position with an 80-point stop would be stopped out in the backtest at a loss of 80 points, or NT$16,000 per TX contract, although no real trader holding and rolling the position lost that money. In a back-adjusted series, every older price is lowered by 100 points, the jump disappears, and the stop is not hit.
The roll date also differs between vendors. Some switch on expiry day, others a few days earlier when trading volume moves to the next month. Find out which rule your series uses, because a strategy that holds positions over expiry will show its roll-day trades on different dates.
How does the night session change daily bars?
Adding the night session changes the open, high and low of a daily bar. If the bar boundary follows the trading day, it does not change the close.
In day-only data, the daily open is the first trade at 08:45. In 台指全, the daily open is the first trade at 15:00 the previous afternoon. The close is the 13:45 last trade in both versions, as long as the vendor ends the daily bar at the end of the trading day. The high and low include every move made overnight, for example during US trading hours.
Hypothetical example
Day-only bar for Tuesday: open 20,150, high 20,220, low 20,100, close 20,180. The all-sessions bar for the same trading day starts at 15:00 on Monday at 20,050. Overnight it falls to 19,980 and rallies to 20,160. The all-sessions bar reads: open 20,050, high 20,220, low 19,980, close 20,180. The close is the same. The daily range grows from 120 to 240 points. The morning "gap" from Monday's close is mostly gone, because the night session has already traded through it.
That shift affects common rules in predictable ways:
- Gap strategies that trade the difference between yesterday's close and today's open mostly lose their signal, because the night session fills the gap before the morning.
- Opening-range breakouts built on "the first 30 minutes" now refer to 15:00–15:30, not 08:45–09:15. Set the session time in the rule if you mean the morning open.
- Range-based indicators such as the Average True Range (ATR) read larger on all-sessions daily bars, so ATR-based stops and position sizes change too.
- Close-based indicators such as a moving average of daily closes stay nearly the same, because the closes match.
- Volume per bar is higher, but night bars on intraday charts are usually thinner than day bars, which matters for slippage.
Some tools end daily bars at midnight instead of at the end of the trading day. Then a night session is split across two daily bars, and the close is no longer the 13:45 price. Check where your daily bar starts before you compare it with any other chart. The guide to day-only vs. full-session bars in futures backtests covers this for other markets.
Should you backtest on all-sessions or day-only data?
Choose the series that matches the hours your strategy will actually trade. Testing on hours you will never trade either adds trades you cannot take or removes moves you will have to sit through.
| Your situation | Use | Reason |
|---|---|---|
| You enter and exit only between 08:45 and 13:45 | Day-only | Signals and fills come only from hours you trade |
| You trade or place orders in the night session | All-sessions (台指全) | The backtest must see the bars where your orders fill |
| You hold positions overnight but trade only in the day | All-sessions for risk, day-only rules for entries | Overnight moves hit your open position and your stops, even if you do not trade them |
| Your rules use daily highs, lows or ranges | Test both | The two versions give different daily values, and the strategy should hold up in the one you will trade |
One practical setup is an all-sessions series with a trading-session setting that only allows entries in chosen hours. That keeps overnight price moves in the equity curve and limits entries to the hours you watch. In AlphaPrime (Valynox), the backtesting research scope includes trading sessions, fees and slippage for this reason.
How do you check a 台指全 series for gaps before testing?
Check the series bar by bar against the session calendar before the first test, because a missing night or a doubled roll day changes results without any error message. Work through these steps in order:
- Find where the night data starts. The night session began in 2017. A 台指全 series that starts earlier contains day-only trading days before that date. Either start your test after the night session began, or treat the early years as a separate period.
- Count bars per trading day. On 1-minute bars, a full day session gives about 300 bars and a full night session about 840 bars. The exact number depends on whether your vendor time-stamps bars at the start or the end. Days with far fewer bars point to missing data.
- Look for missing nights. List trading days that have day bars but no night bars. Compare them with the exchange holiday calendar and with closures such as typhoon days. Anything left over is a data gap.
- Inspect every roll date. Look for unexpected jumps in a back-adjusted series and for duplicated or missing bars on the switch day.
- Check the time zone and timestamps. All times should be Taiwan time (UTC+8), with no duplicate timestamps and no bars inside the 05:00–08:45 or 13:45–15:00 breaks.
- Compare a few daily closes with the exchange's published prices for the contract month your series held on those days.
Long holidays such as Lunar New Year create real multi-day gaps that are not errors. The article on market holidays in backtests explains how to tell real closures from missing data. For bad ticks and other data problems, see backtest data quality: gaps, bad ticks, time zones and rolls. The five settings to check first when backtesting TAIEX futures lists the remaining settings.
What this means if you use AlphaPrime (Valynox): TAIEX futures are a supported market, and trading sessions, fees and slippage are part of each backtest, so you can test the same rules on 台指全 and day-only data and see the difference directly. Check the series first, because every strategy in a search is only as accurate as the bars it was tested on.
Frequently asked questions
What is 台指全?
台指全 is a common chart label for a continuous TAIEX futures series that includes both the day session and the night session. The day session runs 08:45–13:45 and the night session 15:00–05:00, Taiwan time. The series is usually built from the nearest contract month and rolled to the next month around expiry.
Which trading day does the TAIEX futures night session belong to?
TAIFEX counts night session trades toward the next trading day. A night session that starts at 15:00 on Monday is part of Tuesday's trading day. The Friday night session belongs to Monday.
Does the night session change the daily close in 台指全 data?
If the daily bar follows the trading day, the close is still the 13:45 last trade, the same as in day-only data. The open, high, low and volume change, because the bar now starts at 15:00 the previous afternoon. Some tools cut daily bars at midnight instead, and then the close changes too.
Should I use back-adjusted or unadjusted continuous TAIEX futures data?
For backtests that count results in points and use stops, back-adjusted data is usually better, because it removes the price jump between contract months at each roll. Unadjusted data keeps the real price levels but shows jumps no trader actually experienced. Many traders keep both and use each for its own purpose.
Why does my 台指全 history have no night bars in early years?
TAIFEX introduced the after-hours session in 2017, so earlier trading days have only day session bars. Results from those years are effectively day-only results. Start the test after the night session began, or analyse the two periods separately.
What AlphaPrime does here
AlphaPrime's backtesting evaluates strategies with trading sessions, fees and slippage set for the data you test on. The strategy databank lets you classify all-sessions and day-only results as separate batches and revalidate strategies when you update or change the series.