When an index showed an old close as today's value for five days: why Jupick added a second data source

In September 2026, the data source Jupick used for Korean indexes stopped updating from September 18. The last close was the value for September 17, yet for the next five days the screen showed that old close as if it were today's value. The market commentary archive stalled for the same period.

What made it worse was that the freeze was hard to notice. The data source did not return an error or an empty response; it kept returning the value from a past day. The numbers themselves looked fine.

An index holding the same value for several days might suggest a quiet market, but in fact the data had stopped updating. From the outside, the two could not be told apart.

A data pipeline has several safeguards. It checks whether a value is empty, whether it makes no sense, such as zero or a negative number, and whether no response arrived at all. This freeze matched none of them.

The value returned was a valid closing price, the format was normal and the response arrived on time. It was simply yesterday's value, or the day before's. The problem was not “the value is wrong” but “the value is stale,” so checks that look only at values did not catch it.

In this case the only signal was the date, not the value. Only by looking at when the last close was from did it become clear that the data had stopped.

If one data source can stop quietly, relying on that one source alone is risky. So Jupick changed how it collects the main Korean indexes to take them from two places: the original data source and another data source that provides the same indexes.

The two data sources each compile the same KOSPI and KOSDAQ through their own routes. When one stops, the other usually keeps updating, so looking at both shows which one holds the more recent value.

The purpose of this redundancy is not to choose the more accurate value but to choose the value that has not stopped. Jupick does not mix the two sets of numbers; it takes the more recent source as a whole.

Which of the two data sources to use is decided by the date of the last close. If the backup source has a more recent last close than the primary source, Jupick switches to the backup. Otherwise it keeps the primary source it was using.

If both sources have the same last date, or if either one's date is unknown, Jupick keeps the primary source. It switches only when it can be sure the other is more recent, and stays with the original source when in doubt.

This way, if a five-day freeze like the earlier one happens again, the source that has not stopped is moved onto the screen automatically.

Before switching to the backup source, Jupick checks one more thing: whether the two sources really point to the same index. If a symbol is mapped incorrectly, the value of an entirely different index could be shown as KOSPI.

The check is whether the two sources' last closes are of similar size. If one is between half and double the other, they are treated as the same index; if it lies outside that range, something is treated as off and no switch is made.

For the same index, the two sources' values cannot differ by much. This range check is a safeguard against pulling in a completely different value just because it is “more recent.”

Redundancy does not prevent every freeze. If both sources stop at the same time, or both freeze on the same day, there is no more recent source and the stale value still shows.

So Jupick attaches an observation date to each index, so that when the value is from cannot be hidden. Cases the fix cannot fully prevent are left honestly, with the date shown.

Data freshness is a separate issue from data accuracy. A correct value can still be yesterday's, and that cannot be seen from the value alone.

The second data source exists only to show a fresher value, not to say anything about where an index is headed. Whether an index goes up or down has nothing to do with this processing.

A more recent value does not mean a better market either. This article is about how data is kept up to date, not about the direction of the market.

This content is for information only and is not investment advice. Jupick does not forecast future stock prices or index levels and does not encourage any trade. Investment decisions and responsibility for them rest with the user.