TAIEX settlement day: does it move prices, and how to test it
Traders often report more volume and sharper moves around TAIEX settlement, but the size of the effect changes over time. Measure it in your own data against comparable days, then split your strategy's trades by settlement day before you add any filter.
By Soren Huang · Published · updated · Editorial policy
Contents
- Does settlement day change how TAIEX futures move?
- What happens on a TAIEX futures settlement day?
- What settlement-day effects do traders usually report?
- How do you measure the settlement-day effect in your own data?
- How do you check whether settlement days help or hurt your strategy?
- Should you exclude settlement days from a backtest?
- Why are settlement-day results so hard to trust?
- Frequently asked questions
- References
Does settlement day change how TAIEX futures move?
Often it does, but not by a fixed amount. Traders commonly report higher volume and sharper moves near the settlement window on TAIEX futures settlement day. The effect differs from year to year. The only reliable answer for your strategy is to measure settlement days in your own data and compare them with similar ordinary days.
What happens on a TAIEX futures settlement day?
On settlement day the expiring monthly contract stops trading at 13:30 and its last value is fixed from the cash index, so the times below are the ones to mark in your data. A settlement day (結算日) is the last trading day of a futures contract; for monthly TX it is the third Wednesday of the contract month. Holidays move some settlement days, and the TAIEX futures settlement dates tool lists the actual dates you need for tagging.
The final settlement price is the price used to close every open position in the expiring contract. TAIFEX computes it from the TAIEX cash index during the last 30 minutes before the stock market closes, which is why the 13:00–13:30 window is the part of the day to measure separately.
These are the key times on a normal settlement day:
- 08:45 – the futures day session opens.
- 09:00 – the cash stock market opens.
- 13:00–13:30 – the settlement window. The index values in this window are averaged to make the final settlement price.
- 13:30 – the cash market closes, and the expiring contract stops trading.
- 13:45 – the day session closes for the other contract months, which now include the new nearest month.
Exchanges can change contract rules and trading hours. Before you build a test on these times, check the current TX specifications on the TAIFEX contract page.
What settlement-day effects do traders usually report?
The effects traders mention most often are about volume, the roll and the last 30 minutes. They are reasons to test, not proven rules. The table lists them with a plausible cause for each.
| Reported effect | Where you would see it | Why it can happen |
|---|---|---|
| Higher volume | Expiring and next-month contracts during the days before settlement | Traders roll positions. To roll is to close the expiring contract and open the same position in a later month. |
| Larger moves in the settlement window | Cash index and futures between 13:00 and 13:30 | Large hedging or unwinding orders tied to positions that settle on the average can concentrate in this window. |
| Price drifting toward large option strikes | Futures price near strikes with high TXO open interest | Option sellers and buyers hedge positions that settle at the same price. |
| Calmer or wider trading the next day | The first day of the new nearest month | Positions are reset and liquidity moves to the new contract. |
None of these effects happens on every settlement day, and their size changes when the market changes.
How do you measure the settlement-day effect in your own data?
Compare settlement days with ordinary days that are as similar as possible, and use a measure that adjusts for the general volatility level. A raw point range is misleading. A 300-point day means something different when the index is calm than when it is swinging hard.
A useful measure is the day's range divided by the recent average range:
Range ratio = (High − Low) ÷ ATR
Here, High and Low are the highest and lowest prices of the day. ATR (average true range) is the average of the daily true range, which includes overnight gaps, over the previous 20 trading days. A ratio near 1.0 means a typical day. A ratio of 1.3 means a day 30% wider than usual. Follow these steps:
- Tag every settlement day in your history from the official calendar. Include the days that were moved because of holidays.
- Check your contract series. A continuous series that switches contracts on settlement day can create an artificial jump in that day's range. Our article on backtest data quality, gaps and rolls shows how to find this.
- Choose placebo days. A placebo day is an ordinary day that is similar to the event day in every way except the event. For TAIEX, the Wednesday one week before each settlement is a good choice.
- Compute the range ratio for settlement days, placebo Wednesdays and all other days. Then do the same for the 13:00–13:30 window alone.
- Compare medians, not averages. The median is the middle value, so one crash day cannot pull it far.
- Repeat for each half of your history to see whether the effect is stable.
Hypothetical example
Ten years of 5-minute TX data give about 120 settlement days and about 2,300 other days. The median range ratio is 1.12 on settlement days, 1.05 on placebo Wednesdays and 1.00 on all other days. In the first five years the settlement median is 1.18. In the last five years it is 1.06. Most of the difference therefore comes from the earlier period, and only part of it is specific to settlement.
How do you check whether settlement days help or hurt your strategy?
Split your strategy's trades by date and compare the groups after costs. A higher market range does not tell you how your rules behave. A trend strategy may profit from a sharp settlement move that a mean-reversion strategy loses on. The table lists three tests, each answering a different question. The backtesting fee and slippage settings in AlphaPrime (Valynox) apply to every trade alike, so the settlement-day group and the ordinary-day group are compared after the same costs.
| Test | How | What it answers |
|---|---|---|
| Split | Label each trade "settlement day" or "other day" and compare the average trade, the number of trades and the worst trade. | Does the strategy behave differently on these days? |
| Exclude | Rerun the backtest with no entries on settlement days, or with positions closed before 13:00. | What would the whole result look like without them? |
| Isolate | Rerun the backtest using only settlement days. | Is there something on these days worth trading on purpose? |
In AlphaPrime's Performance Report, the trades list shows each trade's dates. You can match those dates against the settlement calendar to do the split. Commission and slippage should be in every version of the test. Settlement-day spreads and fills can differ from normal days, so a cost assumption that is fine on ordinary days may be optimistic in the settlement window.
Hypothetical example
A strategy has 800 trades after costs, and 40 of them fall on settlement days. The other-day trades average +6 points. The settlement-day trades average −4 points, a total of −160 points. One settlement-day trade lost 120 points. Without it, the other 39 trades average about −1 point. Most of the "settlement problem" is one bad trade.
Should you exclude settlement days from a backtest?
Exclude them only if the effect passes several independent checks. Adding a filter after you have seen the results is an extra optimization step. Each filter you try is one more trial. The more trials you run, the more likely the best result is luck, as Harvey and Liu (2015) and Bailey and López de Prado (2014) explain. Our article on overfitting in strategy generation covers this risk in practice.
Keep a settlement-day filter only if all of these are true:
- The settlement-day difference is larger than the difference on placebo Wednesdays.
- It appears in both halves of your history, not only in one period.
- It does not disappear when you remove the one or two largest trades.
- It holds after realistic costs, including wider slippage in the settlement window.
- It survives walk-forward analysis. Walk-forward analysis is a test that repeatedly tunes a strategy on one period and then checks it on the following, unseen period. In each window, choose the filter on the in-sample data, which is the period used for tuning, and use it unchanged on the out-of-sample period, which is the later period the strategy has not seen.
AlphaPrime's strategy validation includes walk-forward analysis and custom thresholds. You can run the version with the filter and the version without it through the same checks and compare them on equal terms.
Why are settlement-day results so hard to trust?
There are few settlement days, and the market around them keeps changing. Monthly settlement gives about 12 days a year. Ten years of data therefore hold about 120 settlement days. A strategy may trade on only a fraction of them. With 30 or 40 trades, one gap or one news day can decide the average, as the second example showed.
Behaviour also changes over time. New contract types, changes in who trades the market and changes in option open interest can all shift how the settlement window behaves. A pattern from the earlier years of your data may be gone today. For this reason, the split by period in the measurement steps matters as much as the overall comparison.
To read small samples honestly:
- Report the number of settlement-day trades next to every average.
- Look at the worst and best settlement-day trades, not only the average.
- Prefer a filter that makes sense from the mechanics, such as closing before the 13:00–13:30 window. Avoid filters you only found by searching many variations.
- Recheck the effect each year as new settlement days are added to your data.
What this means if you use AlphaPrime: tag settlement days in the trade dates the Performance Report shows, and compare them with placebo days. Only then run the version with a filter through the same validation as every other variant before you trust it.
Frequently asked questions
When is TAIEX futures settlement day?
Monthly TAIEX futures settle on the third Wednesday of the contract month. If that day is not a trading day, settlement moves to the next business day. Check the TAIFEX contract page or a settlement calendar for exact dates, because holidays shift some months.
Is TAIEX futures volatility higher on settlement day?
Traders often report higher volume and sharper moves near the 13:00–13:30 settlement window, but the size of the effect changes from year to year. Measure it by comparing the range on settlement days with the range on similar ordinary Wednesdays, adjusted for recent volatility. If both groups look alike, settlement is not what drives the pattern.
How is the TAIEX futures final settlement price calculated?
TAIFEX uses the simple average of the TAIEX cash index during the last 30 minutes before the stock market closes on settlement day. All open positions in the expiring contract are closed at that price. Monthly TAIEX options expiring on the same day use the same price.
Should a systematic strategy avoid trading on settlement day?
That depends on your strategy, not on a general rule. Split your trades into settlement-day and other-day groups and compare them after costs. Add a filter only if the difference is larger than on placebo days, holds in both halves of your history and survives walk-forward analysis.
References
- Harvey, C. R., & Liu, Y. (2015). Backtesting. The Journal of Portfolio Management, 42(1), 13–28.
- Bailey, D. H., & López de Prado, M. (2014). The Deflated Sharpe Ratio: Correcting for Selection Bias, Backtest Overfitting and Non-Normality. The Journal of Portfolio Management, 40(5), 94–107.