{ "id": 274066, "title": "Thanks for joining the 2022 Big Data Bowl! ", "url": "/competitions/nfl-big-data-bowl-2022/discussion/274066", "author_name": "Michael Lopez", "post_date": "2021-09-23T19:59:54.832000", "votes": 41, "comment_count": 27, "views": 0, "content": "

We're excited for our fourth Big Data Bowl, which focuses on using the NFL's Next Gen Stats (NGS) to undercover trends in the \"third\" part of the game of football -- special teams.

\n

As a heads up, a few of the NFL's data scientists will be here throughout this contest to provide help and answer any questions you may have. We encourage you to reach out with questions and we'll do our best to assist.

\n

Here are a few things to note:

\n\n

That's it to get you started -- check out the data, and hope folks have fun!

", "messages": [ { "id": 1522075, "postDate": "2021-09-23T19:59:54.833Z", "content": "

We're excited for our fourth Big Data Bowl, which focuses on using the NFL's Next Gen Stats (NGS) to undercover trends in the \"third\" part of the game of football -- special teams.

\n

As a heads up, a few of the NFL's data scientists will be here throughout this contest to provide help and answer any questions you may have. We encourage you to reach out with questions and we'll do our best to assist.

\n

Here are a few things to note:

\n\n

That's it to get you started -- check out the data, and hope folks have fun!

", "rawMarkdown": "We're excited for our fourth Big Data Bowl, which focuses on using the NFL's Next Gen Stats (NGS) to undercover trends in the \"third\" part of the game of football -- special teams.\n\nAs a heads up, a few of the NFL's data scientists will be here throughout this contest to provide help and answer any questions you may have. We encourage you to reach out with questions and we'll do our best to assist.\n\nHere are a few things to note:\n\n- There are three general types of special teams plays: punts, kickoffs, and field goals/extra points. In general, strategies on each of these types of plays are distinct.\n\n- We've shared NGS data from each of the 2018-2020 seasons, which includes player location, speed, acceleration and orientation information for all players on the field. We've also included PFF scouting data, which adds additional football context to player and team behaviors.\n\n- New this year to the Big Data Bowl: \"Coaches Corner\". We've partnered with a few NFL coaches to do virtual film sessions, covering the in's and out's of team and player strategy. Stay tuned to the forums for more info.\n\n- Two Twitter accounts to follow for more info: @StatsbyLopez and @DataWithBliss will both be posting Big Data Bowl related analysis and charts. For those who want to follow along, please use the hashtag #BigDataBowl\n\n- The first step in many NFL analyses with player tracking data is to standardize locational info. Tom has a Notebook from the 2021 Big Data Bowl here that has some ideas to get you started: https://www.kaggle.com/tombliss/tutorial\n\n- Like the 2021 Big Data Bowl, the 2022 version does not have a target metric for folks to aim for. \n\nThat's it to get you started -- check out the data, and hope folks have fun!", "votes": 40 }, { "id": 1525532, "postDate": "2021-09-27T13:37:23.207Z", "content": "

Thank you for helping me understand the AbsoluteYardline Number. I wasn't taking into account teams switch sides after every qtr.
\nSo, for locations on the field, qtrs 1 and 3 are directly comparable, but not directly with 2 and 4.

\n

thanks.

", "rawMarkdown": "Thank you for helping me understand the AbsoluteYardline Number. I wasn't taking into account teams switch sides after every qtr.\nSo, for locations on the field, qtrs 1 and 3 are directly comparable, but not directly with 2 and 4.\n\nthanks.", "votes": 1, "replies": [ { "id": 1525776, "postDate": "2021-09-27T16:08:15.227Z", "content": "

Happy to help!

\n

Teams always switch sides at the end of the 1st and 3rd quarters. However depending on the coin toss decision they may also switch during the 2nd half as well.

\n

So either the 1st / 4th and 2nd / 3rd will have a given team kicking the same direction or the 1st / 3rd and 2nd / 4th as you said.

", "rawMarkdown": "Happy to help!\n\nTeams always switch sides at the end of the 1st and 3rd quarters. However depending on the coin toss decision they may also switch during the 2nd half as well.\n\nSo either the 1st / 4th and 2nd / 3rd will have a given team kicking the same direction or the 1st / 3rd and 2nd / 4th as you said.", "votes": 1 } ] }, { "id": 1524021, "postDate": "2021-09-26T03:08:58.827Z", "content": "

I believe there is an inconsistency on the 'Plays' data set. (Unless there is a penalty), A kickoff takes place from the 35 yard line, and if there is a 'touchback' the receiving team takes possession at their own 25 yard line.

\n

However, in this data set, the value for \"Absolute Yardage Number\" is being coded two different ways (45 yards, and 75 yards).
\nBoth of these could make sense

\n
\n

a kick from the 35 yard line, and the opponent takes over at their own 25 yard line, is a field possession difference of 45 yards.
\n If the receiving team takes over at the 25, they are 75 yards away from the end zone.

\n
\n

But, both can't be correct. What should be used?
\n45, or 75?

\n

Thanks

", "rawMarkdown": "I believe there is an inconsistency on the 'Plays' data set. (Unless there is a penalty), A kickoff takes place from the 35 yard line, and if there is a 'touchback' the receiving team takes possession at their own 25 yard line. \n\nHowever, in this data set, the value for \"Absolute Yardage Number\" is being coded two different ways (45 yards, and 75 yards).\nBoth of these could make sense \n> a kick from the 35 yard line, and the opponent takes over at their own 25 yard line, is a field possession difference of 45 yards.\n> If the receiving team takes over at the 25, they are 75 yards away from the end zone.\n\nBut, **both** can't be correct. What should be used? \n45, or 75? \n\nThanks", "votes": 1, "replies": [ { "id": 1525440, "postDate": "2021-09-27T12:41:05.487Z", "content": "

Hello Glenn,

\n

Good question! I believe you are asking about the absoluteYardlineNumber variable. That variable is the yard line in tracking data coordinates. It is in the same direction as the x coordinate and represents where the ball was snapped / kicked off. View the image below that gives a map of what each coordinate represents.

\n

An absoluteYardlineNumber coordinate of 0 to 10 would represent the first endzone. Although, the ball could never be snapped or kicked off from an endzone.

\n

An absoluteYardlineNumber coordinate of 11 - 109 represents the field of play. The values will always be between these two numbers.

\n

An absoluteYardlineNumber coordinate of 110-120 would represent the second endzone. Although, as said before, the ball could never be snapped or kicked off from an endzone.

\n

Thus, the values of 45 and 75 represent the 35 yard line on the field, but kicking away from endzones.

\n

The coordinate of 45 is 35 yards from the first endzone while the coordinate of 75 is 35 yards from the second endzone.

\n

I hope that answers your question. I also will change the data description to make this more clear.

\n

\"image\"

", "rawMarkdown": "Hello Glenn,\n\nGood question! I believe you are asking about the `absoluteYardlineNumber` variable. That variable is the yard line in tracking data coordinates. It is in the same direction as the x coordinate and represents where the ball was snapped / kicked off. View the image below that gives a map of what each coordinate represents.\n\nAn `absoluteYardlineNumber` coordinate of 0 to 10 would represent the first endzone. Although, the ball could never be snapped or kicked off from an endzone.\n\nAn `absoluteYardlineNumber` coordinate of 11 - 109 represents the field of play. The values will always be between these two numbers.\n\nAn `absoluteYardlineNumber` coordinate of 110-120 would represent the second endzone. Although, as said before, the ball could never be snapped or kicked off from an endzone.\n\nThus, the values of 45 and 75 represent the 35 yard line on the field, but kicking away from endzones.\n\nThe coordinate of 45 is 35 yards from the first endzone while the coordinate of 75 is 35 yards from the second endzone.\n\nI hope that answers your question. I also will change the data description to make this more clear.\n\n![image](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3258%2F820e86013d48faacf33b7a32a15e814c%2FIncreasing%20Dir%20and%20O.png?generation=1572285857588233&alt=media)" } ] }, { "id": 1636193, "postDate": "2022-01-02T17:10:46.593Z", "content": "

Is the appendix included in the 2000 word limit for notebook submissions?

", "rawMarkdown": "Is the appendix included in the 2000 word limit for notebook submissions?", "replies": [ { "id": 1636375, "postDate": "2022-01-02T20:04:01.197Z", "content": "

No it is not.

", "rawMarkdown": "No it is not." } ] }, { "id": 1569833, "postDate": "2021-11-03T18:33:57.220Z", "content": "

I have a couple of questions on two types of data:
\nkickLength: \"Kick length in air of kickoff, field goal or punt (numeric):. Is this kick length the length projected along the x-line? Say if the ball is kicked diagonally from the 35 yards line to the 50 yards line is the kick length 15 yards or the absolute diagonal length? Actually I think it is much easier for some observers to get the projected length, as they have the lines on the field to help.
\nkickReturnYardage is this yardage an absolute value of yards gained? Say in a kickoff the offense catches the ball at their 5 yards line and runs with it up to their 25 yards line where the returner is tackled. They have gained no yards but the returner ran with the ball 20 yards back. Is in this scenario the kickReturnYardage 0 or 20? I hope my questions make sense.

", "rawMarkdown": "I have a couple of questions on two types of data:\nkickLength: \"Kick length in air of kickoff, field goal or punt (numeric):. Is this kick length the length projected along the x-line? Say if the ball is kicked diagonally from the 35 yards line to the 50 yards line is the kick length 15 yards or the absolute diagonal length? Actually I think it is much easier for some observers to get the projected length, as they have the lines on the field to help. \nkickReturnYardage is this yardage an absolute value of yards gained? Say in a kickoff the offense catches the ball at their 5 yards line and runs with it up to their 25 yards line where the returner is tackled. They have gained no yards but the returner ran with the ball 20 yards back. Is in this scenario the kickReturnYardage 0 or 20? I hope my questions make sense.", "replies": [ { "id": 1569848, "postDate": "2021-11-03T18:55:27.670Z", "content": "

Yes - kickLength is along the x-line (or downfield). If the ball is kicked 60 yards along the x-line (downfield) and 11 yards along the y-line (crossfield), the value of kickLength will be 60 as opposed to sqrt(60^2 + 11^2) = 61.

\n

kickReturnYardage is the yards downfield (along the x-line) that the returner(s) gained from when they received the ball to when the play ends. Relation to the 25 yard line is irrelevant for this column. In your example, the kickReturnYardage column would have the value of 20.

", "rawMarkdown": "Yes - `kickLength` is along the x-line (or downfield). If the ball is kicked 60 yards along the x-line (downfield) and 11 yards along the y-line (crossfield), the value of `kickLength` will be 60 as opposed to sqrt(60^2 + 11^2) = 61.\n\n`kickReturnYardage` is the yards downfield (along the x-line) that the returner(s) gained from when they received the ball to when the play ends. Relation to the 25 yard line is irrelevant for this column. In your example, the `kickReturnYardage` column would have the value of 20." } ] }, { "id": 1556782, "postDate": "2021-10-25T06:26:08.790Z", "content": "

Hey I'm trying to identify the returner on any given play but I'm finding that the 'returnerId column' doesn't match any of the numbers in the 'nflId' column - even when a return is made (i.e not just on fair catch situations). Any reason why this might occur? How would you recommend labelling the returner (other than just taking the player the farthest downfield)

", "rawMarkdown": "Hey I'm trying to identify the returner on any given play but I'm finding that the 'returnerId column' doesn't match any of the numbers in the 'nflId' column - even when a return is made (i.e not just on fair catch situations). Any reason why this might occur? How would you recommend labelling the returner (other than just taking the player the farthest downfield)", "replies": [ { "id": 1557125, "postDate": "2021-10-25T12:46:22.103Z", "content": "

I just tested within Kaggle on my end and was able to have a lossless join between the returnerId column and the nflId column.

\n

Be aware that returnerId is of variable type character / string as it can have multiple players (and in some situations will have multiple nflIds listed separated by semicolons) and nflId in the players / tracking data is a numeric as it is always one number.

\n

Try separating returnerId at ';' or dropping the plays with a ';' in returnerId and setting then setting returnerId as a numeric. After that you should be able to merge without issue.

", "rawMarkdown": "I just tested within Kaggle on my end and was able to have a lossless join between the `returnerId` column and the `nflId` column.\n\nBe aware that `returnerId` is of variable type character / string as it can have multiple players (and in some situations will have multiple nflIds listed separated by semicolons) and `nflId` in the players / tracking data is a numeric as it is always one number.\n\nTry separating `returnerId` at ';' or dropping the plays with a ';' in `returnerId` and setting then setting `returnerId` as a numeric. After that you should be able to merge without issue." }, { "id": 1558045, "postDate": "2021-10-26T03:06:39.743Z", "content": "

I'm aware of the ; separated values. Does the returner id only refer to the kickoff returners (I'm looking for punt returners). I find when I do it the values are slightly off (like the nflId of the returner will be 40560 and the returnerId will be 40660)

", "rawMarkdown": "I'm aware of the ; separated values. Does the returner id only refer to the kickoff returners (I'm looking for punt returners). I find when I do it the values are slightly off (like the nflId of the returner will be 40560 and the returnerId will be 40660)" }, { "id": 1558075, "postDate": "2021-10-26T03:39:57.267Z", "content": "

Hello,

\n

returnerId can be used on both kickoff and punt plays.

\n

I tested again and specifically looked at punt plays and still found no issues joining to other datasets by returnerId = nflId.

\n

Here is the code I used:

\n

https://www.kaggle.com/tombliss/returnerid-variable-test

\n

Please notice that the displayName that is merged to the plays data set from the players data by returnerId = nflId matches the player referenced as the returner in the playDescription variable.

", "rawMarkdown": "Hello,\n\n`returnerId` can be used on both kickoff and punt plays.\n\nI tested again and specifically looked at punt plays and still found no issues joining to other datasets by `returnerId` = `nflId`.\n\nHere is the code I used:\n\nhttps://www.kaggle.com/tombliss/returnerid-variable-test\n\nPlease notice that the `displayName` that is merged to the plays data set from the players data by `returnerId` = `nflId` matches the player referenced as the returner in the `playDescription` variable." } ] }, { "id": 1537015, "postDate": "2021-10-07T07:33:57.203Z", "content": "

May I ask you a question? Superbowl or American Football is my first time to know through your competition in my lifetime, i know it is your biggest gala in a year! my question is that the 'frameId' in your player.csv, whether 'frameId' is video data, through it we can draw a visual movement? or a player's movement track ? thanks!!!

", "rawMarkdown": "May I ask you a question? Superbowl or American Football is my first time to know through your competition in my lifetime, i know it is your biggest gala in a year! my question is that the 'frameId' in your player.csv, whether 'frameId' is video data, through it we can draw a visual movement? or a player's movement track ? thanks!!! ", "replies": [ { "id": 1537551, "postDate": "2021-10-07T15:10:51.717Z", "content": "

frameId is the frame identifier for each measurement of player locations in the tracking data. It is not something specific to NFL football.

\n

I.e.

\n

0 seconds into play typically is frame 1
\n0.1 seconds into play typically is frame 2
\nect.

", "rawMarkdown": "`frameId` is the frame identifier for each measurement of player locations in the tracking data. It is not something specific to NFL football.\n\nI.e.\n\n0 seconds into play typically is frame 1\n0.1 seconds into play typically is frame 2\nect." } ] }, { "id": 1533966, "postDate": "2021-10-04T14:13:52.973Z", "content": "

Hi! I was wondering if we are allowed to incorporate weather data, such as wind speed, temperature, and precipitation, in our models so long as we obtain that data from a public database. Similarly, is there any public database that tells us the maximum decibels reached during a game?

", "rawMarkdown": "Hi! I was wondering if we are allowed to incorporate weather data, such as wind speed, temperature, and precipitation, in our models so long as we obtain that data from a public database. Similarly, is there any public database that tells us the maximum decibels reached during a game?", "replies": [ { "id": 1534042, "postDate": "2021-10-04T14:50:22.940Z", "content": "

Yes you are allowed to incorporate weather data as long as it is free and publicly available to all competition participants. See rule 7C: https://www.kaggle.com/c/nfl-big-data-bowl-2022/rules.

\n

One possible source for weather data is my GitHub repo which I just updated to be complete for 2020:
\nhttps://github.com/ThompsonJamesBliss/WeatherData. I also posted this data on Kaggle here: https://www.kaggle.com/tombliss/weather-data.

\n

There is no public data that I am aware of that contains maximum decibels reached or anything else to do with stadium sound during a game.

", "rawMarkdown": "Yes you are allowed to incorporate weather data as long as it is free and publicly available to all competition participants. See rule 7C: https://www.kaggle.com/c/nfl-big-data-bowl-2022/rules.\n\nOne possible source for weather data is my GitHub repo which I just updated to be complete for 2020:\nhttps://github.com/ThompsonJamesBliss/WeatherData. I also posted this data on Kaggle here: https://www.kaggle.com/tombliss/weather-data.\n\nThere is no public data that I am aware of that contains maximum decibels reached or anything else to do with stadium sound during a game." }, { "id": 1536140, "postDate": "2021-10-06T13:49:59.087Z", "rawMarkdown": "", "isDeleted": true } ] }, { "id": 1533558, "postDate": "2021-10-04T06:26:47.513Z", "content": "

I see a lot of the play data labelled with \"Down ==0\"; what does this refer to?

", "rawMarkdown": "I see a lot of the play data labelled with \"Down ==0\"; what does this refer to?\n", "replies": [ { "id": 1533790, "postDate": "2021-10-04T11:26:12.910Z", "content": "

\"Down == 0\" refers to plays that are not from scrimmage. These are plays that are not part of a traditional set of downs where the team with the ball is attempting to gain a first down/touchdown. All Kickoff / PAT plays are considered plays not from scrimmage.

", "rawMarkdown": "\"Down == 0\" refers to plays that are not from scrimmage. These are plays that are not part of a traditional set of downs where the team with the ball is attempting to gain a first down/touchdown. All Kickoff / PAT plays are considered plays not from scrimmage." } ] }, { "id": 1532285, "postDate": "2021-10-02T19:27:45.603Z", "content": "

Question: does there need to be code written for every piece of analysis? For example, there is no column in the 'play' data set indicating if a turnover took place. It's fairly easy to use text analysis on the PlayColumn to identify the word 'fumble', and then to read these handful of rows to see which team recovered, if there was a penalty, if the play was reviewed or not, etc.
\nIt's just so much quicker and easier to read these and 'manually' create the indicator column.

\n

Thanks

", "rawMarkdown": "Question: does there need to be code written for every piece of analysis? For example, there is no column in the 'play' data set indicating if a turnover took place. It's fairly easy to use text analysis on the PlayColumn to identify the word 'fumble', and then to read these handful of rows to see which team recovered, if there was a penalty, if the play was reviewed or not, etc.\nIt's just so much quicker and easier to read these and 'manually' create the indicator column.\n\nThanks\n\n", "replies": [ { "id": 1533800, "postDate": "2021-10-04T11:31:42.633Z", "content": "

There is a free public play-by-play NFL data set from the nflfastR package in R that includes clean turnover information and merges easily to the Big Data Bowl data. If you are using R you can install/load the nflfastR package: https://www.nflfastr.com/ or if you are using Python you can download the raw data from their GitHub: https://github.com/nflverse/nflfastR-data/tree/master/data

", "rawMarkdown": "There is a free public play-by-play NFL data set from the nflfastR package in R that includes clean turnover information and merges easily to the Big Data Bowl data. If you are using R you can install/load the nflfastR package: https://www.nflfastr.com/ or if you are using Python you can download the raw data from their GitHub: https://github.com/nflverse/nflfastR-data/tree/master/data" } ] }, { "id": 1531164, "postDate": "2021-10-01T17:17:28.093Z", "content": "

Thank you for the opportunity to get acquainted with football. It is very funny to start deep dive at the point when you know absolutely nothing about this sport:)

\n

I also wanted to clarify one question. Am I right that in this competition there is absolute freedom about the subject of research and the only thing I should take care of is correspondence of my research to #Evaluation rules of this competition?

", "rawMarkdown": "Thank you for the opportunity to get acquainted with football. It is very funny to start deep dive at the point when you know absolutely nothing about this sport:)\n\nI also wanted to clarify one question. Am I right that in this competition there is absolute freedom about the subject of research and the only thing I should take care of is correspondence of my research to #Evaluation rules of this competition?", "replies": [ { "id": 1531205, "postDate": "2021-10-01T18:09:24.560Z", "content": "

Hello,

\n

Yes the challenge is to analyze the player tracking data that corresponds to NFL special teams plays and within that there is no specific topic of research that you need to analyze.

", "rawMarkdown": "Hello,\n\nYes the challenge is to analyze the player tracking data that corresponds to NFL special teams plays and within that there is no specific topic of research that you need to analyze.", "votes": 1 }, { "id": 1532812, "postDate": "2021-10-03T12:07:00.173Z", "content": "

alright, thanks for your answer!

", "rawMarkdown": "alright, thanks for your answer!" } ] }, { "id": 1523980, "postDate": "2021-09-26T00:15:31.653Z", "content": "

Question: If we find other useful data sources, can we use them to augment the analysis? Or are we limited exclusively to the data sets that were provided?

", "rawMarkdown": "Question: If we find other useful data sources, can we use them to augment the analysis? Or are we limited exclusively to the data sets that were provided?", "replies": [ { "id": 1525442, "postDate": "2021-09-27T12:43:34.993Z", "content": "

Hello,

\n

Good question! You are allowed to use other data sources as long as they are free and publicly available to all competition participants. See rule 7C: https://www.kaggle.com/c/nfl-big-data-bowl-2022/rules.

", "rawMarkdown": "Hello,\n\nGood question! You are allowed to use other data sources as long as they are free and publicly available to all competition participants. See rule 7C: https://www.kaggle.com/c/nfl-big-data-bowl-2022/rules." } ] }, { "id": 1575447, "postDate": "2021-11-08T12:09:11.620Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 1525532, "author_name": "Glenn A Clark", "author_url": "", "post_date": "2021-09-27T13:37:23.207000", "content": "

Thank you for helping me understand the AbsoluteYardline Number. I wasn't taking into account teams switch sides after every qtr.
\nSo, for locations on the field, qtrs 1 and 3 are directly comparable, but not directly with 2 and 4.

\n

thanks.

", "votes": 1, "replies": [ { "id": 1525776, "author_name": "Tom Bliss", "author_url": "", "post_date": "2021-09-27T16:08:15.227000", "content": "

Happy to help!

\n

Teams always switch sides at the end of the 1st and 3rd quarters. However depending on the coin toss decision they may also switch during the 2nd half as well.

\n

So either the 1st / 4th and 2nd / 3rd will have a given team kicking the same direction or the 1st / 3rd and 2nd / 4th as you said.

", "votes": 1, "replies": [] } ] }, { "id": 1524021, "author_name": "Glenn A Clark", "author_url": "", "post_date": "2021-09-26T03:08:58.827000", "content": "

I believe there is an inconsistency on the 'Plays' data set. (Unless there is a penalty), A kickoff takes place from the 35 yard line, and if there is a 'touchback' the receiving team takes possession at their own 25 yard line.

\n

However, in this data set, the value for \"Absolute Yardage Number\" is being coded two different ways (45 yards, and 75 yards).
\nBoth of these could make sense

\n
\n

a kick from the 35 yard line, and the opponent takes over at their own 25 yard line, is a field possession difference of 45 yards.
\n If the receiving team takes over at the 25, they are 75 yards away from the end zone.

\n
\n

But, both can't be correct. What should be used?
\n45, or 75?

\n

Thanks

", "votes": 1, "replies": [ { "id": 1525440, "author_name": "Tom Bliss", "author_url": "", "post_date": "2021-09-27T12:41:05.487000", "content": "

Hello Glenn,

\n

Good question! I believe you are asking about the absoluteYardlineNumber variable. That variable is the yard line in tracking data coordinates. It is in the same direction as the x coordinate and represents where the ball was snapped / kicked off. View the image below that gives a map of what each coordinate represents.

\n

An absoluteYardlineNumber coordinate of 0 to 10 would represent the first endzone. Although, the ball could never be snapped or kicked off from an endzone.

\n

An absoluteYardlineNumber coordinate of 11 - 109 represents the field of play. The values will always be between these two numbers.

\n

An absoluteYardlineNumber coordinate of 110-120 would represent the second endzone. Although, as said before, the ball could never be snapped or kicked off from an endzone.

\n

Thus, the values of 45 and 75 represent the 35 yard line on the field, but kicking away from endzones.

\n

The coordinate of 45 is 35 yards from the first endzone while the coordinate of 75 is 35 yards from the second endzone.

\n

I hope that answers your question. I also will change the data description to make this more clear.

\n

\"image\"

", "votes": 0, "replies": [] } ] }, { "id": 1636193, "author_name": "Liam Dao", "author_url": "", "post_date": "2022-01-02T17:10:46.593000", "content": "

Is the appendix included in the 2000 word limit for notebook submissions?

", "votes": 0, "replies": [ { "id": 1636375, "author_name": "Tom Bliss", "author_url": "", "post_date": "2022-01-02T20:04:01.197000", "content": "

No it is not.

", "votes": 0, "replies": [] } ] }, { "id": 1569833, "author_name": "Adrian Muresan", "author_url": "", "post_date": "2021-11-03T18:33:57.220000", "content": "

I have a couple of questions on two types of data:
\nkickLength: \"Kick length in air of kickoff, field goal or punt (numeric):. Is this kick length the length projected along the x-line? Say if the ball is kicked diagonally from the 35 yards line to the 50 yards line is the kick length 15 yards or the absolute diagonal length? Actually I think it is much easier for some observers to get the projected length, as they have the lines on the field to help.
\nkickReturnYardage is this yardage an absolute value of yards gained? Say in a kickoff the offense catches the ball at their 5 yards line and runs with it up to their 25 yards line where the returner is tackled. They have gained no yards but the returner ran with the ball 20 yards back. Is in this scenario the kickReturnYardage 0 or 20? I hope my questions make sense.

", "votes": 0, "replies": [ { "id": 1569848, "author_name": "Tom Bliss", "author_url": "", "post_date": "2021-11-03T18:55:27.670000", "content": "

Yes - kickLength is along the x-line (or downfield). If the ball is kicked 60 yards along the x-line (downfield) and 11 yards along the y-line (crossfield), the value of kickLength will be 60 as opposed to sqrt(60^2 + 11^2) = 61.

\n

kickReturnYardage is the yards downfield (along the x-line) that the returner(s) gained from when they received the ball to when the play ends. Relation to the 25 yard line is irrelevant for this column. In your example, the kickReturnYardage column would have the value of 20.

", "votes": 0, "replies": [] } ] }, { "id": 1556782, "author_name": "Elijah Cavan", "author_url": "", "post_date": "2021-10-25T06:26:08.790000", "content": "

Hey I'm trying to identify the returner on any given play but I'm finding that the 'returnerId column' doesn't match any of the numbers in the 'nflId' column - even when a return is made (i.e not just on fair catch situations). Any reason why this might occur? How would you recommend labelling the returner (other than just taking the player the farthest downfield)

", "votes": 0, "replies": [ { "id": 1557125, "author_name": "Tom Bliss", "author_url": "", "post_date": "2021-10-25T12:46:22.103000", "content": "

I just tested within Kaggle on my end and was able to have a lossless join between the returnerId column and the nflId column.

\n

Be aware that returnerId is of variable type character / string as it can have multiple players (and in some situations will have multiple nflIds listed separated by semicolons) and nflId in the players / tracking data is a numeric as it is always one number.

\n

Try separating returnerId at ';' or dropping the plays with a ';' in returnerId and setting then setting returnerId as a numeric. After that you should be able to merge without issue.

", "votes": 0, "replies": [] }, { "id": 1558045, "author_name": "Elijah Cavan", "author_url": "", "post_date": "2021-10-26T03:06:39.743000", "content": "

I'm aware of the ; separated values. Does the returner id only refer to the kickoff returners (I'm looking for punt returners). I find when I do it the values are slightly off (like the nflId of the returner will be 40560 and the returnerId will be 40660)

", "votes": 0, "replies": [] }, { "id": 1558075, "author_name": "Tom Bliss", "author_url": "", "post_date": "2021-10-26T03:39:57.267000", "content": "

Hello,

\n

returnerId can be used on both kickoff and punt plays.

\n

I tested again and specifically looked at punt plays and still found no issues joining to other datasets by returnerId = nflId.

\n

Here is the code I used:

\n

https://www.kaggle.com/tombliss/returnerid-variable-test

\n

Please notice that the displayName that is merged to the plays data set from the players data by returnerId = nflId matches the player referenced as the returner in the playDescription variable.

", "votes": 0, "replies": [] } ] }, { "id": 1537015, "author_name": "Sophie", "author_url": "", "post_date": "2021-10-07T07:33:57.203000", "content": "

May I ask you a question? Superbowl or American Football is my first time to know through your competition in my lifetime, i know it is your biggest gala in a year! my question is that the 'frameId' in your player.csv, whether 'frameId' is video data, through it we can draw a visual movement? or a player's movement track ? thanks!!!

", "votes": 0, "replies": [ { "id": 1537551, "author_name": "Tom Bliss", "author_url": "", "post_date": "2021-10-07T15:10:51.717000", "content": "

frameId is the frame identifier for each measurement of player locations in the tracking data. It is not something specific to NFL football.

\n

I.e.

\n

0 seconds into play typically is frame 1
\n0.1 seconds into play typically is frame 2
\nect.

", "votes": 0, "replies": [] } ] }, { "id": 1533966, "author_name": "Gabby Creneti", "author_url": "", "post_date": "2021-10-04T14:13:52.973000", "content": "

Hi! I was wondering if we are allowed to incorporate weather data, such as wind speed, temperature, and precipitation, in our models so long as we obtain that data from a public database. Similarly, is there any public database that tells us the maximum decibels reached during a game?

", "votes": 0, "replies": [ { "id": 1534042, "author_name": "Tom Bliss", "author_url": "", "post_date": "2021-10-04T14:50:22.940000", "content": "

Yes you are allowed to incorporate weather data as long as it is free and publicly available to all competition participants. See rule 7C: https://www.kaggle.com/c/nfl-big-data-bowl-2022/rules.

\n

One possible source for weather data is my GitHub repo which I just updated to be complete for 2020:
\nhttps://github.com/ThompsonJamesBliss/WeatherData. I also posted this data on Kaggle here: https://www.kaggle.com/tombliss/weather-data.

\n

There is no public data that I am aware of that contains maximum decibels reached or anything else to do with stadium sound during a game.

", "votes": 0, "replies": [] }, { "id": 1536140, "author_name": "", "author_url": "", "post_date": "2021-10-06T13:49:59.087000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 1533558, "author_name": "Elijah Cavan", "author_url": "", "post_date": "2021-10-04T06:26:47.513000", "content": "

I see a lot of the play data labelled with \"Down ==0\"; what does this refer to?

", "votes": 0, "replies": [ { "id": 1533790, "author_name": "Tom Bliss", "author_url": "", "post_date": "2021-10-04T11:26:12.910000", "content": "

\"Down == 0\" refers to plays that are not from scrimmage. These are plays that are not part of a traditional set of downs where the team with the ball is attempting to gain a first down/touchdown. All Kickoff / PAT plays are considered plays not from scrimmage.

", "votes": 0, "replies": [] } ] }, { "id": 1532285, "author_name": "Glenn A Clark", "author_url": "", "post_date": "2021-10-02T19:27:45.603000", "content": "

Question: does there need to be code written for every piece of analysis? For example, there is no column in the 'play' data set indicating if a turnover took place. It's fairly easy to use text analysis on the PlayColumn to identify the word 'fumble', and then to read these handful of rows to see which team recovered, if there was a penalty, if the play was reviewed or not, etc.
\nIt's just so much quicker and easier to read these and 'manually' create the indicator column.

\n

Thanks

", "votes": 0, "replies": [ { "id": 1533800, "author_name": "Tom Bliss", "author_url": "", "post_date": "2021-10-04T11:31:42.633000", "content": "

There is a free public play-by-play NFL data set from the nflfastR package in R that includes clean turnover information and merges easily to the Big Data Bowl data. If you are using R you can install/load the nflfastR package: https://www.nflfastr.com/ or if you are using Python you can download the raw data from their GitHub: https://github.com/nflverse/nflfastR-data/tree/master/data

", "votes": 0, "replies": [] } ] }, { "id": 1531164, "author_name": "Arkadiy Synovets", "author_url": "", "post_date": "2021-10-01T17:17:28.093000", "content": "

Thank you for the opportunity to get acquainted with football. It is very funny to start deep dive at the point when you know absolutely nothing about this sport:)

\n

I also wanted to clarify one question. Am I right that in this competition there is absolute freedom about the subject of research and the only thing I should take care of is correspondence of my research to #Evaluation rules of this competition?

", "votes": 0, "replies": [ { "id": 1531205, "author_name": "Tom Bliss", "author_url": "", "post_date": "2021-10-01T18:09:24.560000", "content": "

Hello,

\n

Yes the challenge is to analyze the player tracking data that corresponds to NFL special teams plays and within that there is no specific topic of research that you need to analyze.

", "votes": 1, "replies": [] }, { "id": 1532812, "author_name": "Arkadiy Synovets", "author_url": "", "post_date": "2021-10-03T12:07:00.173000", "content": "

alright, thanks for your answer!

", "votes": 0, "replies": [] } ] }, { "id": 1523980, "author_name": "Glenn A Clark", "author_url": "", "post_date": "2021-09-26T00:15:31.653000", "content": "

Question: If we find other useful data sources, can we use them to augment the analysis? Or are we limited exclusively to the data sets that were provided?

", "votes": 0, "replies": [ { "id": 1525442, "author_name": "Tom Bliss", "author_url": "", "post_date": "2021-09-27T12:43:34.993000", "content": "

Hello,

\n

Good question! You are allowed to use other data sources as long as they are free and publicly available to all competition participants. See rule 7C: https://www.kaggle.com/c/nfl-big-data-bowl-2022/rules.

", "votes": 0, "replies": [] } ] }, { "id": 1575447, "author_name": "", "author_url": "", "post_date": "2021-11-08T12:09:11.620000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1522075": "We're excited for our fourth Big Data Bowl, which focuses on using the NFL's Next Gen Stats (NGS) to undercover trends in the \"third\" part of the game of football -- special teams.\n\nAs a heads up, a few of the NFL's data scientists will be here throughout this contest to provide help and answer any questions you may have. We encourage you to reach out with questions and we'll do our best to assist.\n\nHere are a few things to note:\n\n- There are three general types of special teams plays: punts, kickoffs, and field goals/extra points. In general, strategies on each of these types of plays are distinct.\n\n- We've shared NGS data from each of the 2018-2020 seasons, which includes player location, speed, acceleration and orientation information for all players on the field. We've also included PFF scouting data, which adds additional football context to player and team behaviors.\n\n- New this year to the Big Data Bowl: \"Coaches Corner\". We've partnered with a few NFL coaches to do virtual film sessions, covering the in's and out's of team and player strategy. Stay tuned to the forums for more info.\n\n- Two Twitter accounts to follow for more info: @StatsbyLopez and @DataWithBliss will both be posting Big Data Bowl related analysis and charts. For those who want to follow along, please use the hashtag #BigDataBowl\n\n- The first step in many NFL analyses with player tracking data is to standardize locational info. Tom has a Notebook from the 2021 Big Data Bowl here that has some ideas to get you started: https://www.kaggle.com/tombliss/tutorial\n\n- Like the 2021 Big Data Bowl, the 2022 version does not have a target metric for folks to aim for. \n\nThat's it to get you started -- check out the data, and hope folks have fun!", "1525532": "Thank you for helping me understand the AbsoluteYardline Number. I wasn't taking into account teams switch sides after every qtr.\nSo, for locations on the field, qtrs 1 and 3 are directly comparable, but not directly with 2 and 4.\n\nthanks.", "1524021": "I believe there is an inconsistency on the 'Plays' data set. (Unless there is a penalty), A kickoff takes place from the 35 yard line, and if there is a 'touchback' the receiving team takes possession at their own 25 yard line. \n\nHowever, in this data set, the value for \"Absolute Yardage Number\" is being coded two different ways (45 yards, and 75 yards).\nBoth of these could make sense \n> a kick from the 35 yard line, and the opponent takes over at their own 25 yard line, is a field possession difference of 45 yards.\n> If the receiving team takes over at the 25, they are 75 yards away from the end zone.\n\nBut, **both** can't be correct. What should be used? \n45, or 75? \n\nThanks", "1636193": "Is the appendix included in the 2000 word limit for notebook submissions?", "1569833": "I have a couple of questions on two types of data:\nkickLength: \"Kick length in air of kickoff, field goal or punt (numeric):. Is this kick length the length projected along the x-line? Say if the ball is kicked diagonally from the 35 yards line to the 50 yards line is the kick length 15 yards or the absolute diagonal length? Actually I think it is much easier for some observers to get the projected length, as they have the lines on the field to help. \nkickReturnYardage is this yardage an absolute value of yards gained? Say in a kickoff the offense catches the ball at their 5 yards line and runs with it up to their 25 yards line where the returner is tackled. They have gained no yards but the returner ran with the ball 20 yards back. Is in this scenario the kickReturnYardage 0 or 20? I hope my questions make sense.", "1556782": "Hey I'm trying to identify the returner on any given play but I'm finding that the 'returnerId column' doesn't match any of the numbers in the 'nflId' column - even when a return is made (i.e not just on fair catch situations). Any reason why this might occur? How would you recommend labelling the returner (other than just taking the player the farthest downfield)", "1537015": "May I ask you a question? Superbowl or American Football is my first time to know through your competition in my lifetime, i know it is your biggest gala in a year! my question is that the 'frameId' in your player.csv, whether 'frameId' is video data, through it we can draw a visual movement? or a player's movement track ? thanks!!! ", "1533966": "Hi! I was wondering if we are allowed to incorporate weather data, such as wind speed, temperature, and precipitation, in our models so long as we obtain that data from a public database. Similarly, is there any public database that tells us the maximum decibels reached during a game?", "1533558": "I see a lot of the play data labelled with \"Down ==0\"; what does this refer to?\n", "1532285": "Question: does there need to be code written for every piece of analysis? For example, there is no column in the 'play' data set indicating if a turnover took place. It's fairly easy to use text analysis on the PlayColumn to identify the word 'fumble', and then to read these handful of rows to see which team recovered, if there was a penalty, if the play was reviewed or not, etc.\nIt's just so much quicker and easier to read these and 'manually' create the indicator column.\n\nThanks\n\n", "1531164": "Thank you for the opportunity to get acquainted with football. It is very funny to start deep dive at the point when you know absolutely nothing about this sport:)\n\nI also wanted to clarify one question. Am I right that in this competition there is absolute freedom about the subject of research and the only thing I should take care of is correspondence of my research to #Evaluation rules of this competition?", "1523980": "Question: If we find other useful data sources, can we use them to augment the analysis? Or are we limited exclusively to the data sets that were provided?", "1575447": "" } }