# #16 Solution

Thank you Kaggle for hosting this awesome competition and congratulations to every one that participated & fought hard in this competition! Especially to
[Markus F](https://www.kaggle.com/friedchips)
,
[Thomas Rohwer](https://www.kaggle.com/trohwer64)
, and
[Nanashi](https://www.kaggle.com/jesucristo)
for their incredibly helpful contribution in discussions and kernels! I personally would have not made it #16 if it wasnt for their contribution on
[The Missing Link](https://www.kaggle.com/friedchips/the-missing-link)
,
["The Orientation Sensor" or "Science vs. Alchemy" discussion](https://www.kaggle.com/c/career-con-2019/discussion/87239#latest-512162)
, and
[Smart Robots. Complete Notebook](https://www.kaggle.com/jesucristo/1-smart-robots-complete-notebook-0-73)
.
[Here's the kernel for my solution](https://www.kaggle.com/ilhamfp31/16-solution-0-76)
TL DR:
- I link the runs between train and test data with Markus F's kernel
- Manually correct train runs that have multiple surface
- I combine the known surface from test with my best submission (LB 0.73)