# Violin plots are inaccurate

**URL:** <https://community.wandb.ai/t/violin-plots-are-inaccurate/1644>\
**Category:** W&B Help\
**Tags:** wandb\
**Created:** [December 30, 2021, 8:43am UTC](https://community.wandb.ai/t/violin-plots-are-inaccurate/1644 "2021-12-30T08:43:55Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![bznm](https://sea2.discourse-cdn.com/flex020/user_avatar/community.wandb.ai/bznm/32/586_2.png) [@bznm](https://community.wandb.ai/u/bznm)\
**Post date:** [December 30, 2021, 8:43am UTC](https://community.wandb.ai/t/violin-plots-are-inaccurate/1644/1 "2021-12-30T08:43:55Z")

</div>

I really like the violin plots but there is something strange in my plots.

 ![screen_violin](https://us1.discourse-cdn.com/flex020/uploads/wandb/original/1X/19acb7ed3ad4d2c1f2fc8f2037098205c753e63a.png)  
All of these distributions have the lower bound at 0.0, however they appear to be ‘randomly’ translated of some shift. Moreover, I have realized that the maximum value that you can estimate by looking at the extremum of the plot is clearly not corresponding with the maximum value you get from the tabular data.  
Is this intended or not?

Thanks in advance

---

<div class="post-metadata">

**Author:** ![ramit\_goolry](https://avatars.discourse-cdn.com/v4/letter/r/85e7bf/32.png) [@ramit\_goolry](https://community.wandb.ai/u/ramit_goolry)\
**Post date:** [December 30, 2021, 7:05pm UTC](https://community.wandb.ai/t/violin-plots-are-inaccurate/1644/2 "2021-12-30T19:05:08Z")

</div>

Hi @bznm,

Could you share a link to the page where you see this behaviour? It will be very helpful to get a reproduction of this issue.

Thanks,  
Ramit

---

<div class="post-metadata">

**Author:** ![bznm](https://sea2.discourse-cdn.com/flex020/user_avatar/community.wandb.ai/bznm/32/586_2.png) [@bznm](https://community.wandb.ai/u/bznm)\
**Post date:** [December 31, 2021, 10:45am UTC](https://community.wandb.ai/t/violin-plots-are-inaccurate/1644/3 "2021-12-31T10:45:35Z")

</div>

Please, have a look [here](https://wandb.ai/bznm/pytorch-demo?workspace=user-bznm).  
If you look at the tabular values, you can see that both the arch have some experiments with accuracy 0.0, so I expect the two violin plots to have the same extremum. Instead, they appear to be shifted. Moreover, you can see from the table that the maximum accuracy reached by arch4 is about 0.81. However, looking at the violin plot you would say that the max value is about 0.9. In so doing the plots are misleading. Can you help?

---

<div class="post-metadata">

**Author:** ![ramit\_goolry](https://avatars.discourse-cdn.com/v4/letter/r/85e7bf/32.png) [@ramit\_goolry](https://community.wandb.ai/u/ramit_goolry)\
**Post date:** [January 4, 2022, 12:25am UTC](https://community.wandb.ai/t/violin-plots-are-inaccurate/1644/4 "2022-01-04T00:25:49Z")

</div>

Hi @bznm ,

I have created an internal ticket on this issue for our engineering team to investigate this. I will notify you as soon as there is some movement on this issue.

Thanks,  
Ramit

---

<div class="post-metadata">

**Author:** ![system](https://us1.discourse-cdn.com/flex020/uploads/wandb/original/1X/366b649231631dbab896843020da0056074ac79d.png) [@system](https://community.wandb.ai/u/system)\
**Post date:** [March 5, 2022, 12:26am UTC](https://community.wandb.ai/t/violin-plots-are-inaccurate/1644/5 "2022-03-05T00:26:28Z")

</div>

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.