# Why "test" data?

**URL:** <https://forum.numer.ai/t/why-test-data/4997>\
**Category:** Tournament\
**Created:** [February 28, 2022, 5:57pm UTC](https://forum.numer.ai/t/why-test-data/4997 "2022-02-28T17:57:16Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![eleele](https://avatars.discourse-cdn.com/v4/letter/e/ce7236/32.png) [@eleele](https://forum.numer.ai/u/eleele)\
**Post date:** [February 28, 2022, 5:57pm UTC](https://forum.numer.ai/t/why-test-data/4997/1 "2022-02-28T17:57:16Z")

</div>

According to the description of the latest dataset, data with `data_type == 'test'`

> …is used for internal testing, but is not part of the tournament scoring and payouts

So, why internal testing if what matters is only the live performance? Any explanation/suggestion about internal testing?

Thanks!

---

<div class="post-metadata">

**Author:** ![luee](https://avatars.discourse-cdn.com/v4/letter/l/c67d28/32.png) [@luee](https://forum.numer.ai/u/luee)\
**Post date:** [February 28, 2022, 6:04pm UTC](https://forum.numer.ai/t/why-test-data/4997/2 "2022-02-28T18:04:54Z")

</div>

Seems to be legacy at this point, I believe the data will be released in full with targets sometime in March

---

<div class="post-metadata">

**Author:** ![ambrul11](https://avatars.discourse-cdn.com/v4/letter/a/e68b1a/32.png) [@ambrul11](https://forum.numer.ai/u/ambrul11)\
**Post date:** [April 8, 2022, 1:43pm UTC](https://forum.numer.ai/t/why-test-data/4997/3 "2022-04-08T13:43:47Z")

</div>

I think they’re testing your submission to make sure it scores (based on test part) above some benchmark. If test scores of your submission are below benchmark, then they may not want to include your model into the mix. I’m a newbie to numerai, so, don’t take it as a fact.

---

<div class="post-metadata">

**Author:** ![themicon](https://yyz1.discourse-cdn.com/flex009/user_avatar/forum.numer.ai/themicon/32/2487_2.png) [@themicon](https://forum.numer.ai/u/themicon)\
**Post date:** [April 8, 2022, 9:49pm UTC](https://forum.numer.ai/t/why-test-data/4997/4 "2022-04-08T21:49:18Z")

</div>

They used to use the test data for internal stuff.

With the release of data V4, the test data has been added to the validation data and the targets are now also provided.

---

<div class="post-metadata">

**Author:** ![rigrog](https://avatars.discourse-cdn.com/v4/letter/r/dec6dc/32.png) [@rigrog](https://forum.numer.ai/u/rigrog)\
**Post date:** [April 10, 2022, 3:18am UTC](https://forum.numer.ai/t/why-test-data/4997/5 "2022-04-10T03:18:52Z")

</div>

It seems MikeP accepted the suggestion I made, and you upvoted (thanks!), so ‘train’ and ‘validation’ go up to era 1000, and include target numbers. After that, there are still about 27,000 rows of ‘test’ data, currently eras 1001 - 1005, with empty targets.

Next week, era 1001 will become ‘validation’ and have target numbers, era 1006 will become ‘test’. And so forth, each week.