{
  "id": 132624,
  "title": "NCAA(M) March Madness® - One Stop Wiki : Know everything in a glance.",
  "url": "/competitions/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/discussion/132624",
  "author_name": "",
  "post_date": "2020-02-27T00:21:59.074738500Z",
  "votes": 18,
  "comment_count": 2,
  "views": 0,
  "content": "<h3>-----------------------------------------------------------------------------</h3>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<p><strong>This post is prepared for beginners or new competition joiners as the target audience. It will save much of their time. Even, those who are familiar of these things can use it to cross-check with stuffs and timelines.</strong></p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-media/competitions/march-madness-2018/lockup_cloud.png\" alt=\"banner\"></p>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<h2>This competition will be conducted in 2 phases:</h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<p><strong><code>First Stage of Competition</code></strong></p>\n\n<p><strong>Deadline : 14th March --&gt; Historical prediction ends.</strong>\n* In general you will have a proper model built in your bag by this date.\n* Rely on past tournaments to build and test models.\n* if you find any external stats/resources useful, Post it publicly as a data-set. check this [thread](March Madness®).</p>\n\n<p><strong><code>Second Stage of Competition</code></strong></p>\n\n<p><strong>Deadline : 16th March  --&gt; 2020 matches predictions ends.</strong>\n* Competitors will forecast outcomes of all possible match-ups in 2020 NCAA games.\n* This is the REAL test in this competition - to predict 2020 match results.</p>\n\n<h2><a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview\">Migrated Data for the Event:</a></h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<p>--&gt; A huge bulk of data-set has been migrated to <strong>GCP</strong> for these events.\n* 80+ years of data ( this year it is 83rd NCAA event in 2020)\n* 90 Championships\n* 24 different sports.</p>\n\n<h2><a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/evaluation\">Submission File:</a></h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<ul>\n<li>Different for Stage 1 (<strong>Historical Model</strong>) and Stage 2 (<strong>2020 forecasting</strong>).</li>\n<li>submission file would contain the list of all possible match-ups between teams.</li>\n<li>Max <strong>50 Submissions</strong> per day.</li>\n<li>Upto <strong>2 final Submissions</strong> for Judging.</li>\n</ul>\n\n<p>Since, \n<strong><code>Team 1 v/s Team 2 == Team 2 v/s Team 1</code></strong> ==&gt; <strong><code>include only game pairs</code></strong></p>\n\n<p>VIZ:</p>\n\n<p><code>TEAM 1 --&amp;gt; Lower team id      |   TEAM 2 --&amp;gt; Higher team id</code></p>\n\n<p>Thus:\n<code>1 tournament --&amp;gt; 68 teams ( 64 + 4 )</code>\n<code>Total Match-ups = (68 * 67) / 2 = 2,278</code></p>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<p>*<strong>Unique Game id : 2015_1107_1112 ==&gt;2015 : Year of match played | 1107 : team 1 | 1112 : team 2*</strong></p>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<h2><a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/discussion/130591\">Team Request MegaThread !</a></h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<h2><a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/evaluation\">Evaluation : </a></h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<p>Submissions are scored on the log loss:</p>\n\n<p><code>LogLoss=−1/n * (∑[yilog(y^i)+(1−yi)log(1−y^i)]),</code>\nwhere</p>\n\n<ul>\n<li><strong>n is the number of games played</strong></li>\n<li><strong>y^i is the predicted probability of team 1 beating team 2</strong></li>\n<li><strong>yi is 1 if team 1 wins, 0 if team 2 wins</strong></li>\n<li><strong>log() is the natural (base e) logarithm</strong></li>\n</ul>\n\n<p>--&gt; The use of the logarithm provides extreme punishments for being both confident and wrong. In the worst possible case, a prediction that something is true when it is actually false will add an infinite amount to your error score. In order to prevent this, predictions are bounded away from the extremes by a small value.</p>\n\n<h2><a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/timeline\">Detailed Timeline : </a></h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<p>1.) <code>STAGE 1 - Not counted towards final Ranking</code>.\n==&gt; 14 March ( Saturday)--&gt; Build and test on historical Data --&gt; Leaderboard also on Historical Data.\n2.) <code>STAGE 2 - Final Rankings and Ratings rewarded based on this Leaderboard score</code>.\n==&gt; 15 March ( Sunday) --&gt; 68 teams Announced.\n==&gt; 16 March ( Monday) --&gt; 2020 prediction Submission Start. --&gt; Release of up-to-date 2019-2020 season data.\n==&gt; 19 March ( Thursday) : 03:00 PM UTC --&gt; Final Deadline to submit 2020 predictions.\n==&gt; March 19 onward - April 6 --&gt; Leaderboard keeps refreshing daily --&gt; Watch Game results of payouts. --&gt; Wish for your luck also. 😄 </p>\n\n<p><em><strong><code>Result of Competition Announced</code></strong></em> </p>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<h2><a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview\">Prizes:</a></h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<ul>\n<li><strong>1st Place - $10,000</strong></li>\n<li><strong>2nd Place - $7,000</strong></li>\n<li><strong>3rd Place - $5,000</strong></li>\n<li><strong>4th Place - $2,000</strong></li>\n<li><strong>5th Place - $1,000</strong></li>\n</ul>\n\n<h2><a href=\"https://www.kaggle.com/c/mens-machine-learning-competition-2019/discussion/130624\">Last year Top winning Solutions Thread</a>.</h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<h2><a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/faqs\"> Important FAQs : </a></h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<p><em><strong><code>Q : Can I update my solution after tournament( regular games) begin ?</code></strong></em></p>\n\n<p><strong>A :</strong> No changes are permitted once tournament begins. Kaggle Competition finishes by then. Tournament Dates = 17 March - 6 April.</p>\n\n<p><em><strong><code>Q : How Leaderboard is gonna work when event is in Future ?</code></strong></em></p>\n\n<p><strong>A :</strong> Kaggle will update Solution file as the games occur, which will cause the ranks on the leaderboard to change. Those play-outs which won't happen ( due to team disqualifications in earlier stages) will not be considered for scoring and leaderboard ranking. Only played matches would be considered for scoring.</p>\n\n<p><em><strong><code>Q : More weights given to later ( important) games ?</code></strong></em></p>\n\n<p><strong>A :</strong> All games are counted equally. Similar weights given to all matches.</p>\n\n<p><em><strong><code>Q : Is it mandatory to participate in both stage of competition ?</code></strong></em></p>\n\n<p><strong>A :</strong> NO need to participate in the 1st stage also, to participate in the 2nd stage.</p>\n\n<p>&gt; <strong><em><code>Q : Is it fine if i use LEAKS and stay on top of leaderboard with 0.0000 Score?</code></em></strong></p>\n\n<p><strong>A :</strong> Everybody knows the reason of that behaviour on leaderboard. Its so easy to get that Leaky score. But, that won't gonna help your model for 2020 Forecasting results. So, Its better to make a generalised model based on historical events. Moreover this competitions ranking is little-bit based on luck also, like what is the outcome of each play-outs. Read this <a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/discussion/131163\">thread</a> for more detail on this.</p>\n\n<h1>Thanks! Upvote if you liked it and found useful.</h1>",
  "messages": [
    {
      "id": "757608",
      "postDate": "02/27/2020 00:21:59",
      "content": "<h3>-----------------------------------------------------------------------------</h3>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<p><strong>This post is prepared for beginners or new competition joiners as the target audience. It will save much of their time. Even, those who are familiar of these things can use it to cross-check with stuffs and timelines.</strong></p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-media/competitions/march-madness-2018/lockup_cloud.png\" alt=\"banner\"></p>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<h2>This competition will be conducted in 2 phases:</h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<p><strong><code>First Stage of Competition</code></strong></p>\n\n<p><strong>Deadline : 14th March --&gt; Historical prediction ends.</strong>\n* In general you will have a proper model built in your bag by this date.\n* Rely on past tournaments to build and test models.\n* if you find any external stats/resources useful, Post it publicly as a data-set. check this [thread](March Madness®).</p>\n\n<p><strong><code>Second Stage of Competition</code></strong></p>\n\n<p><strong>Deadline : 16th March  --&gt; 2020 matches predictions ends.</strong>\n* Competitors will forecast outcomes of all possible match-ups in 2020 NCAA games.\n* This is the REAL test in this competition - to predict 2020 match results.</p>\n\n<h2><a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview\">Migrated Data for the Event:</a></h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<p>--&gt; A huge bulk of data-set has been migrated to <strong>GCP</strong> for these events.\n* 80+ years of data ( this year it is 83rd NCAA event in 2020)\n* 90 Championships\n* 24 different sports.</p>\n\n<h2><a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/evaluation\">Submission File:</a></h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<ul>\n<li>Different for Stage 1 (<strong>Historical Model</strong>) and Stage 2 (<strong>2020 forecasting</strong>).</li>\n<li>submission file would contain the list of all possible match-ups between teams.</li>\n<li>Max <strong>50 Submissions</strong> per day.</li>\n<li>Upto <strong>2 final Submissions</strong> for Judging.</li>\n</ul>\n\n<p>Since, \n<strong><code>Team 1 v/s Team 2 == Team 2 v/s Team 1</code></strong> ==&gt; <strong><code>include only game pairs</code></strong></p>\n\n<p>VIZ:</p>\n\n<p><code>TEAM 1 --&amp;gt; Lower team id      |   TEAM 2 --&amp;gt; Higher team id</code></p>\n\n<p>Thus:\n<code>1 tournament --&amp;gt; 68 teams ( 64 + 4 )</code>\n<code>Total Match-ups = (68 * 67) / 2 = 2,278</code></p>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<p>*<strong>Unique Game id : 2015_1107_1112 ==&gt;2015 : Year of match played | 1107 : team 1 | 1112 : team 2*</strong></p>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<h2><a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/discussion/130591\">Team Request MegaThread !</a></h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<h2><a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/evaluation\">Evaluation : </a></h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<p>Submissions are scored on the log loss:</p>\n\n<p><code>LogLoss=−1/n * (∑[yilog(y^i)+(1−yi)log(1−y^i)]),</code>\nwhere</p>\n\n<ul>\n<li><strong>n is the number of games played</strong></li>\n<li><strong>y^i is the predicted probability of team 1 beating team 2</strong></li>\n<li><strong>yi is 1 if team 1 wins, 0 if team 2 wins</strong></li>\n<li><strong>log() is the natural (base e) logarithm</strong></li>\n</ul>\n\n<p>--&gt; The use of the logarithm provides extreme punishments for being both confident and wrong. In the worst possible case, a prediction that something is true when it is actually false will add an infinite amount to your error score. In order to prevent this, predictions are bounded away from the extremes by a small value.</p>\n\n<h2><a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/timeline\">Detailed Timeline : </a></h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<p>1.) <code>STAGE 1 - Not counted towards final Ranking</code>.\n==&gt; 14 March ( Saturday)--&gt; Build and test on historical Data --&gt; Leaderboard also on Historical Data.\n2.) <code>STAGE 2 - Final Rankings and Ratings rewarded based on this Leaderboard score</code>.\n==&gt; 15 March ( Sunday) --&gt; 68 teams Announced.\n==&gt; 16 March ( Monday) --&gt; 2020 prediction Submission Start. --&gt; Release of up-to-date 2019-2020 season data.\n==&gt; 19 March ( Thursday) : 03:00 PM UTC --&gt; Final Deadline to submit 2020 predictions.\n==&gt; March 19 onward - April 6 --&gt; Leaderboard keeps refreshing daily --&gt; Watch Game results of payouts. --&gt; Wish for your luck also. 😄 </p>\n\n<p><em><strong><code>Result of Competition Announced</code></strong></em> </p>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<h2><a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview\">Prizes:</a></h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<ul>\n<li><strong>1st Place - $10,000</strong></li>\n<li><strong>2nd Place - $7,000</strong></li>\n<li><strong>3rd Place - $5,000</strong></li>\n<li><strong>4th Place - $2,000</strong></li>\n<li><strong>5th Place - $1,000</strong></li>\n</ul>\n\n<h2><a href=\"https://www.kaggle.com/c/mens-machine-learning-competition-2019/discussion/130624\">Last year Top winning Solutions Thread</a>.</h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<h2><a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/faqs\"> Important FAQs : </a></h2>\n\n<h3>-----------------------------------------------------------------------------</h3>\n\n<p><em><strong><code>Q : Can I update my solution after tournament( regular games) begin ?</code></strong></em></p>\n\n<p><strong>A :</strong> No changes are permitted once tournament begins. Kaggle Competition finishes by then. Tournament Dates = 17 March - 6 April.</p>\n\n<p><em><strong><code>Q : How Leaderboard is gonna work when event is in Future ?</code></strong></em></p>\n\n<p><strong>A :</strong> Kaggle will update Solution file as the games occur, which will cause the ranks on the leaderboard to change. Those play-outs which won't happen ( due to team disqualifications in earlier stages) will not be considered for scoring and leaderboard ranking. Only played matches would be considered for scoring.</p>\n\n<p><em><strong><code>Q : More weights given to later ( important) games ?</code></strong></em></p>\n\n<p><strong>A :</strong> All games are counted equally. Similar weights given to all matches.</p>\n\n<p><em><strong><code>Q : Is it mandatory to participate in both stage of competition ?</code></strong></em></p>\n\n<p><strong>A :</strong> NO need to participate in the 1st stage also, to participate in the 2nd stage.</p>\n\n<p>&gt; <strong><em><code>Q : Is it fine if i use LEAKS and stay on top of leaderboard with 0.0000 Score?</code></em></strong></p>\n\n<p><strong>A :</strong> Everybody knows the reason of that behaviour on leaderboard. Its so easy to get that Leaky score. But, that won't gonna help your model for 2020 Forecasting results. So, Its better to make a generalised model based on historical events. Moreover this competitions ranking is little-bit based on luck also, like what is the outcome of each play-outs. Read this <a href=\"https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/discussion/131163\">thread</a> for more detail on this.</p>\n\n<h1>Thanks! Upvote if you liked it and found useful.</h1>",
      "rawMarkdown": "### -----------------------------------------------------------------------------\n### -----------------------------------------------------------------------------\n**This post is prepared for beginners or new competition joiners as the target audience. It will save much of their time. Even, those who are familiar of these things can use it to cross-check with stuffs and timelines.**\n\n![banner](https://storage.googleapis.com/kaggle-media/competitions/march-madness-2018/lockup_cloud.png)\n### -----------------------------------------------------------------------------\n### -----------------------------------------------------------------------------\n\n## This competition will be conducted in 2 phases:\n### -----------------------------------------------------------------------------\n**`First Stage of Competition`**\n\n**Deadline : 14th March --&gt; Historical prediction ends.**\n* In general you will have a proper model built in your bag by this date.\n* Rely on past tournaments to build and test models.\n* if you find any external stats/resources useful, Post it publicly as a data-set. check this [thread](March Madness®).\n\n**`Second Stage of Competition`**\n\n**Deadline : 16th March  --&gt; 2020 matches predictions ends.**\n* Competitors will forecast outcomes of all possible match-ups in 2020 NCAA games.\n* This is the REAL test in this competition - to predict 2020 match results.\n\n## [Migrated Data for the Event:](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview)\n### -----------------------------------------------------------------------------\n--&gt; A huge bulk of data-set has been migrated to **GCP** for these events.\n* 80+ years of data ( this year it is 83rd NCAA event in 2020)\n* 90 Championships\n* 24 different sports.\n\n\n## [Submission File:](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/evaluation)\n### -----------------------------------------------------------------------------\n* Different for Stage 1 (**Historical Model**) and Stage 2 (**2020 forecasting**).\n* submission file would contain the list of all possible match-ups between teams.\n* Max **50 Submissions** per day.\n* Upto **2 final Submissions** for Judging.\n\nSince, \n**`Team 1 v/s Team 2 == Team 2 v/s Team 1 `** ==&gt; **`include only game pairs`**\n\nVIZ:\n\n`TEAM 1 --&gt; Lower team id      |   TEAM 2 --&gt; Higher team id`\n\nThus:\n`1 tournament --&gt; 68 teams ( 64 + 4 )`\n`Total Match-ups = (68 * 67) / 2 = 2,278`\n\n\n### -----------------------------------------------------------------------------\n***Unique Game id : 2015_1107_1112 ==&gt;2015 : Year of match played | 1107 : team 1 | 1112 : team 2***\n### -----------------------------------------------------------------------------\n\n## [Team Request MegaThread !](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/discussion/130591)\n### -----------------------------------------------------------------------------\n\n\n## [Evaluation : ](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/evaluation)\n### -----------------------------------------------------------------------------\nSubmissions are scored on the log loss:\n\n`LogLoss=−1/n * (∑[yilog(y^i)+(1−yi)log(1−y^i)]),`\nwhere\n\n* **n is the number of games played**\n* **y^i is the predicted probability of team 1 beating team 2**\n* **yi is 1 if team 1 wins, 0 if team 2 wins**\n* **log() is the natural (base e) logarithm**\n\n--&gt; The use of the logarithm provides extreme punishments for being both confident and wrong. In the worst possible case, a prediction that something is true when it is actually false will add an infinite amount to your error score. In order to prevent this, predictions are bounded away from the extremes by a small value.\n\n## [Detailed Timeline : ](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/timeline)\n### -----------------------------------------------------------------------------\n1.) `STAGE 1 - Not counted towards final Ranking`.\n==&gt; 14 March ( Saturday)--&gt; Build and test on historical Data --&gt; Leaderboard also on Historical Data.\n2.) `STAGE 2 - Final Rankings and Ratings rewarded based on this Leaderboard score`.\n==&gt; 15 March ( Sunday) --&gt; 68 teams Announced.\n==&gt; 16 March ( Monday) --&gt; 2020 prediction Submission Start. --&gt; Release of up-to-date 2019-2020 season data.\n==&gt; 19 March ( Thursday) : 03:00 PM UTC --&gt; Final Deadline to submit 2020 predictions.\n==&gt; March 19 onward - April 6 --&gt; Leaderboard keeps refreshing daily --&gt; Watch Game results of payouts. --&gt; Wish for your luck also. 😄 \n\n***`Result of Competition Announced`*** \n\n### -----------------------------------------------------------------------------\n\n## [Prizes:](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview)\n### -----------------------------------------------------------------------------\n* **1st Place - $10,000**\n* **2nd Place - $7,000**\n* **3rd Place - $5,000**\n* **4th Place - $2,000**\n* **5th Place - $1,000**\n\n## [Last year Top winning Solutions Thread](https://www.kaggle.com/c/mens-machine-learning-competition-2019/discussion/130624).\n### -----------------------------------------------------------------------------\n\n### -----------------------------------------------------------------------------\n##[ Important FAQs : ](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/faqs)\n### -----------------------------------------------------------------------------\n\n***`Q : Can I update my solution after tournament( regular games) begin ?`***\n\n**A :** No changes are permitted once tournament begins. Kaggle Competition finishes by then. Tournament Dates = 17 March - 6 April.\n\n\n***`Q : How Leaderboard is gonna work when event is in Future ?`***\n\n**A :** Kaggle will update Solution file as the games occur, which will cause the ranks on the leaderboard to change. Those play-outs which won't happen ( due to team disqualifications in earlier stages) will not be considered for scoring and leaderboard ranking. Only played matches would be considered for scoring.\n\n\n***`Q : More weights given to later ( important) games ?`***\n\n**A :** All games are counted equally. Similar weights given to all matches.\n\n\n***`Q : Is it mandatory to participate in both stage of competition ?`***\n\n**A :** NO need to participate in the 1st stage also, to participate in the 2nd stage.\n\n&gt; ***`Q : Is it fine if i use LEAKS and stay on top of leaderboard with 0.0000 Score?`***\n\n**A :** Everybody knows the reason of that behaviour on leaderboard. Its so easy to get that Leaky score. But, that won't gonna help your model for 2020 Forecasting results. So, Its better to make a generalised model based on historical events. Moreover this competitions ranking is little-bit based on luck also, like what is the outcome of each play-outs. Read this [thread](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/discussion/131163) for more detail on this.\n\n# Thanks! Upvote if you liked it and found useful.",
      "votes": null
    },
    {
      "id": "759929",
      "postDate": "02/29/2020 16:17:09",
      "content": "<p>\"80+ years of data ( this year it is 83rd NCAA event in 2020)\"\nWhat's this 80 year dataset contain?  Can I access it outside GCP?</p>",
      "rawMarkdown": "\"80+ years of data ( this year it is 83rd NCAA event in 2020)\"\nWhat's this 80 year dataset contain?  Can I access it outside GCP?",
      "votes": null
    },
    {
      "id": "760785",
      "postDate": "03/01/2020 18:08:08",
      "content": "<p>Since, I haven't tried them by myself. I couldn't answer this. \nThis might help you find it.\n1.) <a href=\"https://cloud.google.com/blog/topics/inside-google-cloud/turning-data-into-ncaa-march-madness-insights\">https://cloud.google.com/blog/topics/inside-google-cloud/turning-data-into-ncaa-march-madness-insights</a>\n2.) <a href=\"https://cloud.google.com/blog/products/gcp/architecting-live-ncaa-predictions-from-archives-to-insights\">https://cloud.google.com/blog/products/gcp/architecting-live-ncaa-predictions-from-archives-to-insights</a>\n3.) <a href=\"https://www.neudesic.com/blog/neudesic-data-scientists-used-google-cloud-platform-predict-outcomes-ncaa-march-madness-2019/\">https://www.neudesic.com/blog/neudesic-data-scientists-used-google-cloud-platform-predict-outcomes-ncaa-march-madness-2019/</a></p>",
      "rawMarkdown": "Since, I haven't tried them by myself. I couldn't answer this. \nThis might help you find it.\n1.) https://cloud.google.com/blog/topics/inside-google-cloud/turning-data-into-ncaa-march-madness-insights\n2.) https://cloud.google.com/blog/products/gcp/architecting-live-ncaa-predictions-from-archives-to-insights\n3.) https://www.neudesic.com/blog/neudesic-data-scientists-used-google-cloud-platform-predict-outcomes-ncaa-march-madness-2019/",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 759929,
      "author_name": "zachmayer",
      "author_url": "",
      "post_date": "02/29/2020 16:17:09",
      "content": "<p>\"80+ years of data ( this year it is 83rd NCAA event in 2020)\"\nWhat's this 80 year dataset contain?  Can I access it outside GCP?</p>",
      "votes": null,
      "replies": [
        {
          "id": 760785,
          "author_name": "atulanandjha",
          "author_url": "",
          "post_date": "03/01/2020 18:08:08",
          "content": "<p>Since, I haven't tried them by myself. I couldn't answer this. \nThis might help you find it.\n1.) <a href=\"https://cloud.google.com/blog/topics/inside-google-cloud/turning-data-into-ncaa-march-madness-insights\">https://cloud.google.com/blog/topics/inside-google-cloud/turning-data-into-ncaa-march-madness-insights</a>\n2.) <a href=\"https://cloud.google.com/blog/products/gcp/architecting-live-ncaa-predictions-from-archives-to-insights\">https://cloud.google.com/blog/products/gcp/architecting-live-ncaa-predictions-from-archives-to-insights</a>\n3.) <a href=\"https://www.neudesic.com/blog/neudesic-data-scientists-used-google-cloud-platform-predict-outcomes-ncaa-march-madness-2019/\">https://www.neudesic.com/blog/neudesic-data-scientists-used-google-cloud-platform-predict-outcomes-ncaa-march-madness-2019/</a></p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "757608": "### -----------------------------------------------------------------------------\n### -----------------------------------------------------------------------------\n**This post is prepared for beginners or new competition joiners as the target audience. It will save much of their time. Even, those who are familiar of these things can use it to cross-check with stuffs and timelines.**\n\n![banner](https://storage.googleapis.com/kaggle-media/competitions/march-madness-2018/lockup_cloud.png)\n### -----------------------------------------------------------------------------\n### -----------------------------------------------------------------------------\n\n## This competition will be conducted in 2 phases:\n### -----------------------------------------------------------------------------\n**`First Stage of Competition`**\n\n**Deadline : 14th March --&gt; Historical prediction ends.**\n* In general you will have a proper model built in your bag by this date.\n* Rely on past tournaments to build and test models.\n* if you find any external stats/resources useful, Post it publicly as a data-set. check this [thread](March Madness®).\n\n**`Second Stage of Competition`**\n\n**Deadline : 16th March  --&gt; 2020 matches predictions ends.**\n* Competitors will forecast outcomes of all possible match-ups in 2020 NCAA games.\n* This is the REAL test in this competition - to predict 2020 match results.\n\n## [Migrated Data for the Event:](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview)\n### -----------------------------------------------------------------------------\n--&gt; A huge bulk of data-set has been migrated to **GCP** for these events.\n* 80+ years of data ( this year it is 83rd NCAA event in 2020)\n* 90 Championships\n* 24 different sports.\n\n\n## [Submission File:](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/evaluation)\n### -----------------------------------------------------------------------------\n* Different for Stage 1 (**Historical Model**) and Stage 2 (**2020 forecasting**).\n* submission file would contain the list of all possible match-ups between teams.\n* Max **50 Submissions** per day.\n* Upto **2 final Submissions** for Judging.\n\nSince, \n**`Team 1 v/s Team 2 == Team 2 v/s Team 1 `** ==&gt; **`include only game pairs`**\n\nVIZ:\n\n`TEAM 1 --&gt; Lower team id      |   TEAM 2 --&gt; Higher team id`\n\nThus:\n`1 tournament --&gt; 68 teams ( 64 + 4 )`\n`Total Match-ups = (68 * 67) / 2 = 2,278`\n\n\n### -----------------------------------------------------------------------------\n***Unique Game id : 2015_1107_1112 ==&gt;2015 : Year of match played | 1107 : team 1 | 1112 : team 2***\n### -----------------------------------------------------------------------------\n\n## [Team Request MegaThread !](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/discussion/130591)\n### -----------------------------------------------------------------------------\n\n\n## [Evaluation : ](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/evaluation)\n### -----------------------------------------------------------------------------\nSubmissions are scored on the log loss:\n\n`LogLoss=−1/n * (∑[yilog(y^i)+(1−yi)log(1−y^i)]),`\nwhere\n\n* **n is the number of games played**\n* **y^i is the predicted probability of team 1 beating team 2**\n* **yi is 1 if team 1 wins, 0 if team 2 wins**\n* **log() is the natural (base e) logarithm**\n\n--&gt; The use of the logarithm provides extreme punishments for being both confident and wrong. In the worst possible case, a prediction that something is true when it is actually false will add an infinite amount to your error score. In order to prevent this, predictions are bounded away from the extremes by a small value.\n\n## [Detailed Timeline : ](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/timeline)\n### -----------------------------------------------------------------------------\n1.) `STAGE 1 - Not counted towards final Ranking`.\n==&gt; 14 March ( Saturday)--&gt; Build and test on historical Data --&gt; Leaderboard also on Historical Data.\n2.) `STAGE 2 - Final Rankings and Ratings rewarded based on this Leaderboard score`.\n==&gt; 15 March ( Sunday) --&gt; 68 teams Announced.\n==&gt; 16 March ( Monday) --&gt; 2020 prediction Submission Start. --&gt; Release of up-to-date 2019-2020 season data.\n==&gt; 19 March ( Thursday) : 03:00 PM UTC --&gt; Final Deadline to submit 2020 predictions.\n==&gt; March 19 onward - April 6 --&gt; Leaderboard keeps refreshing daily --&gt; Watch Game results of payouts. --&gt; Wish for your luck also. 😄 \n\n***`Result of Competition Announced`*** \n\n### -----------------------------------------------------------------------------\n\n## [Prizes:](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview)\n### -----------------------------------------------------------------------------\n* **1st Place - $10,000**\n* **2nd Place - $7,000**\n* **3rd Place - $5,000**\n* **4th Place - $2,000**\n* **5th Place - $1,000**\n\n## [Last year Top winning Solutions Thread](https://www.kaggle.com/c/mens-machine-learning-competition-2019/discussion/130624).\n### -----------------------------------------------------------------------------\n\n### -----------------------------------------------------------------------------\n##[ Important FAQs : ](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/overview/faqs)\n### -----------------------------------------------------------------------------\n\n***`Q : Can I update my solution after tournament( regular games) begin ?`***\n\n**A :** No changes are permitted once tournament begins. Kaggle Competition finishes by then. Tournament Dates = 17 March - 6 April.\n\n\n***`Q : How Leaderboard is gonna work when event is in Future ?`***\n\n**A :** Kaggle will update Solution file as the games occur, which will cause the ranks on the leaderboard to change. Those play-outs which won't happen ( due to team disqualifications in earlier stages) will not be considered for scoring and leaderboard ranking. Only played matches would be considered for scoring.\n\n\n***`Q : More weights given to later ( important) games ?`***\n\n**A :** All games are counted equally. Similar weights given to all matches.\n\n\n***`Q : Is it mandatory to participate in both stage of competition ?`***\n\n**A :** NO need to participate in the 1st stage also, to participate in the 2nd stage.\n\n&gt; ***`Q : Is it fine if i use LEAKS and stay on top of leaderboard with 0.0000 Score?`***\n\n**A :** Everybody knows the reason of that behaviour on leaderboard. Its so easy to get that Leaky score. But, that won't gonna help your model for 2020 Forecasting results. So, Its better to make a generalised model based on historical events. Moreover this competitions ranking is little-bit based on luck also, like what is the outcome of each play-outs. Read this [thread](https://www.kaggle.com/c/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/discussion/131163) for more detail on this.\n\n# Thanks! Upvote if you liked it and found useful.",
    "759929": "\"80+ years of data ( this year it is 83rd NCAA event in 2020)\"\nWhat's this 80 year dataset contain?  Can I access it outside GCP?",
    "760785": "Since, I haven't tried them by myself. I couldn't answer this. \nThis might help you find it.\n1.) https://cloud.google.com/blog/topics/inside-google-cloud/turning-data-into-ncaa-march-madness-insights\n2.) https://cloud.google.com/blog/products/gcp/architecting-live-ncaa-predictions-from-archives-to-insights\n3.) https://www.neudesic.com/blog/neudesic-data-scientists-used-google-cloud-platform-predict-outcomes-ncaa-march-madness-2019/"
  },
  "source": "meta"
}