# New data V4 gives the target for test data. Sounds good but I am worried **URL:** \ **Category:** Tournament\ **Created:** [March 29, 2022, 3:34am UTC](https://forum.numer.ai/t/new-data-v4-gives-the-target-for-test-data-sounds-good-but-i-am-worried/5166 "2022-03-29T03:34:01Z")\ **Posts on this page:** 3\ **Page:** 1
**Author:** ![climber\_1](https://avatars.discourse-cdn.com/v4/letter/c/57b2e6/32.png) [@climber\_1](https://forum.numer.ai/u/climber_1)\ **Post date:** [March 29, 2022, 3:34am UTC](https://forum.numer.ai/t/new-data-v4-gives-the-target-for-test-data-sounds-good-but-i-am-worried/5166/1 "2022-03-29T03:34:01Z")
For years I thought Numerai uses your model’s prediction on test data to determine the meta-model contribution. That’s why you need to submit a huge file each round. No one knows the target on the test set, perfect plan. Now it’s no longer required on model submission. How are they gonna know how good the model is? My answer to this is we are no longer needed. Their model is good enough and doesn’t need our model anymore. ---
**Author:** ![richai](https://yyz1.discourse-cdn.com/flex009/user_avatar/forum.numer.ai/richai/32/2493_2.png) [@richai](https://forum.numer.ai/u/richai)\ **Post date:** [March 29, 2022, 3:36am UTC](https://forum.numer.ai/t/new-data-v4-gives-the-target-for-test-data-sounds-good-but-i-am-worried/5166/2 "2022-03-29T03:36:57Z")
completely wrong 🙂 We weight models based on their stakes. We don’t need to know how good a model is on the test set. It’s not as strong of a signal as stake. We trust you. ---
**Author:** ![wigglemuse](https://yyz1.discourse-cdn.com/flex009/user_avatar/forum.numer.ai/wigglemuse/32/3094_2.png) [@wigglemuse](https://forum.numer.ai/u/wigglemuse)\ **Post date:** [March 29, 2022, 3:37am UTC](https://forum.numer.ai/t/new-data-v4-gives-the-target-for-test-data-sounds-good-but-i-am-worried/5166/3 "2022-03-29T03:37:04Z")
They never used it to determine metamodel contribution – they use the size of your stake.