{ "id": 388094, "title": "Who Took Teddy, Harry or Frank? - EDA", "url": "/competitions/predict-student-performance-from-game-play/discussion/388094", "author_name": "Chris Deotte", "post_date": "2023-02-16T00:46:41.216000", "votes": 64, "comment_count": 9, "views": 0, "content": "

Exploring where users click is fun and interesting. In the examples below, the right image is a scatter plot of users' clicks from columns room_coor_x and room_coor_y. The left image is the corresponding screen shot captured during gameplay.

\n

Who Took Teddy, Harry or Frank?

\n

During gameplay, we find a pair of glasses (evidence for who kidnapped our pet Teddy) and are asked who they belong to. We click the directory on the wall and must compare the glasses to images on the directory (and click which person matches). See below. By analyzing where all 11k users click, we see that many game players thought the glasses matched Harry instead of Frank. This is interesting insight into how users perform during click type questions. The final 18 questions (that we are predicting if users answer correctly) are all similar click type questions. (The correct answer here is Frank).
\n\"\"

\n

Which Entry Matches the Slip?

\n

Here's another interesting example. We find an old dry cleaner slip with serial number and are asked to compare it with the dry cleaner's logbook to identify the owner. (We must click the serial number in logbook that matches the slip.) Most users incorrectly choose the first entry as matching the slip. Or perhaps users were confused and clicked each entry one by one starting from the top. (The correct answer is of course the 5th row down).
\n\"\"

\n

What is Useful on the Plague?

\n

The dry cleaner slip is supposedly for an old basketball shirt. However when we compare it with the basketball plaque we see something odd. The slip says 1916 but the basketball team was formed in 1974. When users are asked \"Anything userful on here?\", referring to clicking something on the plaque, many users are confused about what is useful as illustrated that they click all sorts of things. (Before clicking the correct thing which was the date \"1974\").
\n\"\"

\n

Starter Notebook

\n

I published a starter notebook here displaying click EDA for all 19 game rooms. The starter notebook displays navigate_clicks and person_clicks and labels items of interest in each room

", "messages": [ { "id": 2146535, "postDate": "2023-02-16T00:46:41.217Z", "content": "

Exploring where users click is fun and interesting. In the examples below, the right image is a scatter plot of users' clicks from columns room_coor_x and room_coor_y. The left image is the corresponding screen shot captured during gameplay.

\n

Who Took Teddy, Harry or Frank?

\n

During gameplay, we find a pair of glasses (evidence for who kidnapped our pet Teddy) and are asked who they belong to. We click the directory on the wall and must compare the glasses to images on the directory (and click which person matches). See below. By analyzing where all 11k users click, we see that many game players thought the glasses matched Harry instead of Frank. This is interesting insight into how users perform during click type questions. The final 18 questions (that we are predicting if users answer correctly) are all similar click type questions. (The correct answer here is Frank).
\n\"\"

\n

Which Entry Matches the Slip?

\n

Here's another interesting example. We find an old dry cleaner slip with serial number and are asked to compare it with the dry cleaner's logbook to identify the owner. (We must click the serial number in logbook that matches the slip.) Most users incorrectly choose the first entry as matching the slip. Or perhaps users were confused and clicked each entry one by one starting from the top. (The correct answer is of course the 5th row down).
\n\"\"

\n

What is Useful on the Plague?

\n

The dry cleaner slip is supposedly for an old basketball shirt. However when we compare it with the basketball plaque we see something odd. The slip says 1916 but the basketball team was formed in 1974. When users are asked \"Anything userful on here?\", referring to clicking something on the plaque, many users are confused about what is useful as illustrated that they click all sorts of things. (Before clicking the correct thing which was the date \"1974\").
\n\"\"

\n

Starter Notebook

\n

I published a starter notebook here displaying click EDA for all 19 game rooms. The starter notebook displays navigate_clicks and person_clicks and labels items of interest in each room

", "rawMarkdown": "Exploring where users click is fun and interesting. In the examples below, the right image is a scatter plot of users' clicks from columns `room_coor_x` and `room_coor_y`. The left image is the corresponding screen shot captured during gameplay.\n\n# Who Took Teddy, Harry or Frank?\nDuring gameplay, we find a pair of glasses (evidence for who kidnapped our pet Teddy) and are asked who they belong to. We click the directory on the wall and must compare the glasses to images on the directory (and click which person matches). See below. By analyzing where all 11k users click, we see that many game players thought the glasses matched Harry instead of Frank. This is interesting insight into how users perform during click type questions. The final 18 questions (that we are predicting if users answer correctly) are all similar click type questions. (The correct answer here is Frank).\n![](https://raw.githubusercontent.com/cdeotte/Kaggle_Images/main/Feb-2023/teddy.png)\n\n# Which Entry Matches the Slip?\nHere's another interesting example. We find an old dry cleaner slip with serial number and are asked to compare it with the dry cleaner's logbook to identify the owner. (We must click the serial number in logbook that matches the slip.) Most users incorrectly choose the first entry as matching the slip. Or perhaps users were confused and clicked each entry one by one starting from the top. (The correct answer is of course the 5th row down).\n![](https://raw.githubusercontent.com/cdeotte/Kaggle_Images/main/Feb-2023/laundry.png)\n\n# What is Useful on the Plague?\nThe dry cleaner slip is supposedly for an old basketball shirt. However when we compare it with the basketball plaque we see something odd. The slip says 1916 but the basketball team was formed in 1974. When users are asked \"Anything userful on here?\", referring to clicking something on the plaque, many users are confused about what is useful as illustrated that they click all sorts of things. (Before clicking the correct thing which was the date \"1974\").\n![](https://raw.githubusercontent.com/cdeotte/Kaggle_Images/main/Feb-2023/basketball.png)\n\n# Starter Notebook\nI published a starter notebook [here][1] displaying click EDA for all 19 game rooms. The starter notebook displays `navigate_clicks` and `person_clicks` and labels items of interest in each room\n\n[1]: https://www.kaggle.com/code/cdeotte/game-room-click-eda", "votes": 63 }, { "id": 2149104, "postDate": "2023-02-18T00:21:21.627Z", "content": "

Thanks for sharing @cdeotte, I'm studying your awesome EDA skills by reading your notebooks and discussions. Really helpful:)

", "rawMarkdown": "Thanks for sharing @cdeotte, I'm studying your awesome EDA skills by reading your notebooks and discussions. Really helpful:)", "votes": 1, "replies": [ { "id": 2149107, "postDate": "2023-02-18T00:25:15.383Z", "content": "

Thanks William!

", "rawMarkdown": "Thanks William!", "votes": 1 }, { "id": 2149110, "postDate": "2023-02-18T00:28:52.487Z", "content": "

The code to reproduce these plots is

\n
IDX = ((train.fqid=='directory.closeup.archivist')|(train.fqid=='directory'))&(train.level==15)\\\n    &(train.event_name!='navigate_click')\ntmp = train.loc[IDX]\nplt.scatter(tmp.room_coor_x,tmp.room_coor_y)\n
\n

The trick is finding the correct fqid and level and excluding navigate_click events which are the first click which brings us into the scenario and has weird coordinates.

", "rawMarkdown": "The code to reproduce these plots is \n\n IDX = ((train.fqid=='directory.closeup.archivist')|(train.fqid=='directory'))&(train.level==15)\\\n &(train.event_name!='navigate_click')\n tmp = train.loc[IDX]\n plt.scatter(tmp.room_coor_x,tmp.room_coor_y)\n\nThe trick is finding the correct `fqid` and `level` and excluding `navigate_click` events which are the first click which brings us into the scenario and has weird coordinates. ", "votes": 4, "replies": [ { "id": 2149131, "postDate": "2023-02-18T01:08:53.990Z", "content": "

Thanks for sharing @cdeotte , let me try:)

", "rawMarkdown": "Thanks for sharing @cdeotte , let me try:)", "votes": 1 }, { "id": 2177935, "postDate": "2023-03-11T23:17:27.460Z", "rawMarkdown": "", "isDeleted": true } ] } ] }, { "id": 2148745, "postDate": "2023-02-17T16:53:12.610Z", "content": "

@cdeotte Ohhh, clean plots and notebook!

", "rawMarkdown": "@cdeotte Ohhh, clean plots and notebook!", "votes": 1 }, { "id": 2219998, "postDate": "2023-04-13T04:08:16.137Z", "content": "

Thanks for the great EDA! I am a python newbie, is this code something to think about before creating the model? I am not sure when I have to think about this and would like to know.

", "rawMarkdown": "Thanks for the great EDA! I am a python newbie, is this code something to think about before creating the model? I am not sure when I have to think about this and would like to know.", "votes": 2 }, { "id": 2182194, "postDate": "2023-03-15T01:59:51.117Z", "content": "

I made a notebook doing the EDA for each clues the player has to find, see here!

", "rawMarkdown": "I made a notebook doing the **EDA for each clues the player has to find**, see [here](https://www.kaggle.com/code/janmpia/clues-clicks-eda-features)!", "votes": 2 }, { "id": 2221052, "postDate": "2023-04-13T23:47:46.067Z", "content": "

Thank you for great visualization.
\nReally helpful:)
\nI'll also check your notebook!

", "rawMarkdown": "Thank you for great visualization.\nReally helpful:)\nI'll also check your notebook!" } ], "comments": [ { "id": 2149104, "author_name": "william.wu", "author_url": "", "post_date": "2023-02-18T00:21:21.627000", "content": "

Thanks for sharing @cdeotte, I'm studying your awesome EDA skills by reading your notebooks and discussions. Really helpful:)

", "votes": 1, "replies": [ { "id": 2149107, "author_name": "Chris Deotte", "author_url": "", "post_date": "2023-02-18T00:25:15.383000", "content": "

Thanks William!

", "votes": 1, "replies": [] }, { "id": 2149110, "author_name": "Chris Deotte", "author_url": "", "post_date": "2023-02-18T00:28:52.487000", "content": "

The code to reproduce these plots is

\n
IDX = ((train.fqid=='directory.closeup.archivist')|(train.fqid=='directory'))&(train.level==15)\\\n    &(train.event_name!='navigate_click')\ntmp = train.loc[IDX]\nplt.scatter(tmp.room_coor_x,tmp.room_coor_y)\n
\n

The trick is finding the correct fqid and level and excluding navigate_click events which are the first click which brings us into the scenario and has weird coordinates.

", "votes": 4, "replies": [ { "id": 2149131, "author_name": "william.wu", "author_url": "", "post_date": "2023-02-18T01:08:53.990000", "content": "

Thanks for sharing @cdeotte , let me try:)

", "votes": 1, "replies": [] }, { "id": 2177935, "author_name": "", "author_url": "", "post_date": "2023-03-11T23:17:27.460000", "content": "", "votes": 0, "replies": [] } ] } ] }, { "id": 2148745, "author_name": "Minh T. Nguyen", "author_url": "", "post_date": "2023-02-17T16:53:12.610000", "content": "

@cdeotte Ohhh, clean plots and notebook!

", "votes": 1, "replies": [] }, { "id": 2219998, "author_name": "riicon", "author_url": "", "post_date": "2023-04-13T04:08:16.137000", "content": "

Thanks for the great EDA! I am a python newbie, is this code something to think about before creating the model? I am not sure when I have to think about this and would like to know.

", "votes": 2, "replies": [] }, { "id": 2182194, "author_name": "JEANMPIA", "author_url": "", "post_date": "2023-03-15T01:59:51.117000", "content": "

I made a notebook doing the EDA for each clues the player has to find, see here!

", "votes": 2, "replies": [] }, { "id": 2221052, "author_name": "waticson", "author_url": "", "post_date": "2023-04-13T23:47:46.067000", "content": "

Thank you for great visualization.
\nReally helpful:)
\nI'll also check your notebook!

", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "2146535": "Exploring where users click is fun and interesting. In the examples below, the right image is a scatter plot of users' clicks from columns `room_coor_x` and `room_coor_y`. The left image is the corresponding screen shot captured during gameplay.\n\n# Who Took Teddy, Harry or Frank?\nDuring gameplay, we find a pair of glasses (evidence for who kidnapped our pet Teddy) and are asked who they belong to. We click the directory on the wall and must compare the glasses to images on the directory (and click which person matches). See below. By analyzing where all 11k users click, we see that many game players thought the glasses matched Harry instead of Frank. This is interesting insight into how users perform during click type questions. The final 18 questions (that we are predicting if users answer correctly) are all similar click type questions. (The correct answer here is Frank).\n![](https://raw.githubusercontent.com/cdeotte/Kaggle_Images/main/Feb-2023/teddy.png)\n\n# Which Entry Matches the Slip?\nHere's another interesting example. We find an old dry cleaner slip with serial number and are asked to compare it with the dry cleaner's logbook to identify the owner. (We must click the serial number in logbook that matches the slip.) Most users incorrectly choose the first entry as matching the slip. Or perhaps users were confused and clicked each entry one by one starting from the top. (The correct answer is of course the 5th row down).\n![](https://raw.githubusercontent.com/cdeotte/Kaggle_Images/main/Feb-2023/laundry.png)\n\n# What is Useful on the Plague?\nThe dry cleaner slip is supposedly for an old basketball shirt. However when we compare it with the basketball plaque we see something odd. The slip says 1916 but the basketball team was formed in 1974. When users are asked \"Anything userful on here?\", referring to clicking something on the plaque, many users are confused about what is useful as illustrated that they click all sorts of things. (Before clicking the correct thing which was the date \"1974\").\n![](https://raw.githubusercontent.com/cdeotte/Kaggle_Images/main/Feb-2023/basketball.png)\n\n# Starter Notebook\nI published a starter notebook [here][1] displaying click EDA for all 19 game rooms. The starter notebook displays `navigate_clicks` and `person_clicks` and labels items of interest in each room\n\n[1]: https://www.kaggle.com/code/cdeotte/game-room-click-eda", "2149104": "Thanks for sharing @cdeotte, I'm studying your awesome EDA skills by reading your notebooks and discussions. Really helpful:)", "2148745": "@cdeotte Ohhh, clean plots and notebook!", "2219998": "Thanks for the great EDA! I am a python newbie, is this code something to think about before creating the model? I am not sure when I have to think about this and would like to know.", "2182194": "I made a notebook doing the **EDA for each clues the player has to find**, see [here](https://www.kaggle.com/code/janmpia/clues-clicks-eda-features)!", "2221052": "Thank you for great visualization.\nReally helpful:)\nI'll also check your notebook!" } }