{
  "id": 583299,
  "title": "Studying the leaderboard transition at the end of the competition",
  "url": "/competitions/birdclef-2025/discussion/583299",
  "author_name": "",
  "post_date": "2025-06-06T00:54:14.820208300Z",
  "votes": 27,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hello all,</p>\n<p>We have a moderate level of churn here as expected, given the nature and size of the dataset. Let's delve into this in detail and study the regions of interest here-</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8273630%2F01b2a152efe9a0a3e0f8fba2b4b7cc98%2FLB.png?generation=1749169944723520&amp;alt=media\" alt=\"\"></p>\n<h2>Regions of interest</h2>\n<table>\n<thead>\n<tr>\n<th>Region Label</th>\n<th>Comments</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>R1</td>\n<td>Indicates the rough account of top 10% rankers in the competition sustaining across the leaderboards. <br> Congrats to these teams for sustaining through the churn to win a medal here! <br> We observe a lot of <strong>local churn</strong> here and shall delve into this element separately below</td>\n</tr>\n<tr>\n<td>R2</td>\n<td>Indicates the teams scoring in the top 10% in the public leaderboard and falling off the perch on the private leaderboard. This pool is a bit thick, indicating some level of churn here.</td>\n</tr>\n<tr>\n<td>R3</td>\n<td>Congrats to these teams for beating the churn! These teams moved up the leaderboard and won a medal here on the private leaderboard from a modest public leaderboard rank!</td>\n</tr>\n<tr>\n<td>R4</td>\n<td>This is very similar to R2 but shakeup occurs in a region outside the medal zone</td>\n</tr>\n</tbody>\n</table>\n<h2>Stability analysis of the top 10% finishers</h2>\n<table>\n<thead>\n<tr>\n<th>Region Label</th>\n<th>Last Rank</th>\n<th>Submissions - min, mean, max</th>\n<th>Score band</th>\n<th>Stability analysis</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>Gold medal</td>\n<td>14</td>\n<td>242, 337 , 439</td>\n<td>0.93 - 0.918</td>\n<td>Stable teams = 10 <br> Entering from below = 4</td>\n</tr>\n<tr>\n<td>Silver medal</td>\n<td>108</td>\n<td>1, 153.84, 428</td>\n<td>0.917 - 0.894</td>\n<td>Stable teams = 49 <br> Entering from below = 41  <br> Entering from above = 4</td>\n</tr>\n<tr>\n<td>Bronze medal</td>\n<td>216</td>\n<td>1, 54.64, 302</td>\n<td>0.894 - 0.893</td>\n<td>Stable teams = 21 <br> Entering from below = 78  <br> Entering from above = 9</td>\n</tr>\n</tbody>\n</table>\n<p><br>We observe 2 important facts here-</p>\n<ul>\n<li>A lot of the bronze medal teams have churned either way - fallen off the perch and risen into the medal zone</li>\n<li>Average submissions drastically reduce across gold -&gt; silver -&gt; bronze regions</li>\n</ul>\n<h2>Other adjutant analysis</h2>\n<table>\n<thead>\n<tr>\n<th>Event Label</th>\n<th>Public LB rank</th>\n<th>Private LB rank</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>Biggest riser across leaderboards - overall</td>\n<td>673</td>\n<td>62</td>\n</tr>\n<tr>\n<td>Biggest riser across leaderboards - R1</td>\n<td>181</td>\n<td>47</td>\n</tr>\n<tr>\n<td>Highest fall across leaderboards - overall</td>\n<td>418</td>\n<td>1925</td>\n</tr>\n<tr>\n<td>Highest fall across leaderboards - R1</td>\n<td>65</td>\n<td>156</td>\n</tr>\n</tbody>\n</table>\n<h2>Score clusters</h2>\n<p>The below table highlights the top-3 score clusters across the entire leaderboard (private leaderboard score)</p>\n<table>\n<thead>\n<tr>\n<th>Score value</th>\n<th>Number of teams</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>0.893</td>\n<td>299</td>\n</tr>\n<tr>\n<td>0.892</td>\n<td>89</td>\n</tr>\n<tr>\n<td>0.809</td>\n<td>78</td>\n</tr>\n</tbody>\n</table>\n<h2>Final thoughts</h2>\n<p>Wishing the medal winners a hearty congratulations and best of luck to one and all! Happy learning and best regards for the future Kaggle competitions and otherwise!</p>\n<p>Regards,<br>\nRavi R</p>",
  "messages": [
    {
      "id": "3218194",
      "postDate": "06/06/2025 00:54:14",
      "content": "<p>Hello all,</p>\n<p>We have a moderate level of churn here as expected, given the nature and size of the dataset. Let's delve into this in detail and study the regions of interest here-</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8273630%2F01b2a152efe9a0a3e0f8fba2b4b7cc98%2FLB.png?generation=1749169944723520&amp;alt=media\" alt=\"\"></p>\n<h2>Regions of interest</h2>\n<table>\n<thead>\n<tr>\n<th>Region Label</th>\n<th>Comments</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>R1</td>\n<td>Indicates the rough account of top 10% rankers in the competition sustaining across the leaderboards. <br> Congrats to these teams for sustaining through the churn to win a medal here! <br> We observe a lot of <strong>local churn</strong> here and shall delve into this element separately below</td>\n</tr>\n<tr>\n<td>R2</td>\n<td>Indicates the teams scoring in the top 10% in the public leaderboard and falling off the perch on the private leaderboard. This pool is a bit thick, indicating some level of churn here.</td>\n</tr>\n<tr>\n<td>R3</td>\n<td>Congrats to these teams for beating the churn! These teams moved up the leaderboard and won a medal here on the private leaderboard from a modest public leaderboard rank!</td>\n</tr>\n<tr>\n<td>R4</td>\n<td>This is very similar to R2 but shakeup occurs in a region outside the medal zone</td>\n</tr>\n</tbody>\n</table>\n<h2>Stability analysis of the top 10% finishers</h2>\n<table>\n<thead>\n<tr>\n<th>Region Label</th>\n<th>Last Rank</th>\n<th>Submissions - min, mean, max</th>\n<th>Score band</th>\n<th>Stability analysis</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>Gold medal</td>\n<td>14</td>\n<td>242, 337 , 439</td>\n<td>0.93 - 0.918</td>\n<td>Stable teams = 10 <br> Entering from below = 4</td>\n</tr>\n<tr>\n<td>Silver medal</td>\n<td>108</td>\n<td>1, 153.84, 428</td>\n<td>0.917 - 0.894</td>\n<td>Stable teams = 49 <br> Entering from below = 41  <br> Entering from above = 4</td>\n</tr>\n<tr>\n<td>Bronze medal</td>\n<td>216</td>\n<td>1, 54.64, 302</td>\n<td>0.894 - 0.893</td>\n<td>Stable teams = 21 <br> Entering from below = 78  <br> Entering from above = 9</td>\n</tr>\n</tbody>\n</table>\n<p><br>We observe 2 important facts here-</p>\n<ul>\n<li>A lot of the bronze medal teams have churned either way - fallen off the perch and risen into the medal zone</li>\n<li>Average submissions drastically reduce across gold -&gt; silver -&gt; bronze regions</li>\n</ul>\n<h2>Other adjutant analysis</h2>\n<table>\n<thead>\n<tr>\n<th>Event Label</th>\n<th>Public LB rank</th>\n<th>Private LB rank</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>Biggest riser across leaderboards - overall</td>\n<td>673</td>\n<td>62</td>\n</tr>\n<tr>\n<td>Biggest riser across leaderboards - R1</td>\n<td>181</td>\n<td>47</td>\n</tr>\n<tr>\n<td>Highest fall across leaderboards - overall</td>\n<td>418</td>\n<td>1925</td>\n</tr>\n<tr>\n<td>Highest fall across leaderboards - R1</td>\n<td>65</td>\n<td>156</td>\n</tr>\n</tbody>\n</table>\n<h2>Score clusters</h2>\n<p>The below table highlights the top-3 score clusters across the entire leaderboard (private leaderboard score)</p>\n<table>\n<thead>\n<tr>\n<th>Score value</th>\n<th>Number of teams</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>0.893</td>\n<td>299</td>\n</tr>\n<tr>\n<td>0.892</td>\n<td>89</td>\n</tr>\n<tr>\n<td>0.809</td>\n<td>78</td>\n</tr>\n</tbody>\n</table>\n<h2>Final thoughts</h2>\n<p>Wishing the medal winners a hearty congratulations and best of luck to one and all! Happy learning and best regards for the future Kaggle competitions and otherwise!</p>\n<p>Regards,<br>\nRavi R</p>",
      "rawMarkdown": "Hello all,\n\nWe have a moderate level of churn here as expected, given the nature and size of the dataset. Let's delve into this in detail and study the regions of interest here-\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8273630%2F01b2a152efe9a0a3e0f8fba2b4b7cc98%2FLB.png?generation=1749169944723520&alt=media)\n\n## Regions of interest\n\n|Region Label|Comments|\n| ------------ | ----------- |\n|R1 | Indicates the rough account of top 10% rankers in the competition sustaining across the leaderboards. <br> Congrats to these teams for sustaining through the churn to win a medal here! <br> We observe a lot of **local churn** here and shall delve into this element separately below |\n|R2| Indicates the teams scoring in the top 10% in the public leaderboard and falling off the perch on the private leaderboard. This pool is a bit thick, indicating some level of churn here. |\n|R3 | Congrats to these teams for beating the churn! These teams moved up the leaderboard and won a medal here on the private leaderboard from a modest public leaderboard rank! |\n|R4 | This is very similar to R2 but shakeup occurs in a region outside the medal zone|\n\n## Stability analysis of the top 10% finishers \n\n|Region Label|Last Rank|Submissions - min, mean, max| Score band | Stability analysis |\n|------ | -------- | --------- | ----------- | ---------- |\n|Gold medal   | 14 | 242, 337 , 439 | 0.93 - 0.918 | Stable teams = 10 <br> Entering from below = 4 |\n|Silver medal | 108 |1, 153.84, 428 | 0.917 - 0.894 | Stable teams = 49 <br> Entering from below = 41  <br> Entering from above = 4|\n|Bronze medal | 216 | 1, 54.64, 302 | 0.894 - 0.893| Stable teams = 21 <br> Entering from below = 78  <br> Entering from above = 9 |\n\n<br>We observe 2 important facts here-\n- A lot of the bronze medal teams have churned either way - fallen off the perch and risen into the medal zone\n- Average submissions drastically reduce across gold -> silver -> bronze regions\n\n## Other adjutant analysis \n\n|Event Label| Public LB rank|Private LB rank|\n|------ | -------- | --------- | \n| Biggest riser across leaderboards - overall | 673 | 62 |\n| Biggest riser across leaderboards - R1| 181| 47 |\n| Highest fall across leaderboards - overall| 418 | 1925 |\n| Highest fall across leaderboards - R1|65 | 156 |\n\n## Score clusters \n\nThe below table highlights the top-3 score clusters across the entire leaderboard (private leaderboard score)\n\n|Score value| Number of teams |\n| --------- | ---------- |\n| 0.893 | 299|\n| 0.892 | 89 |\n| 0.809 | 78 |\n\n## Final thoughts\n\nWishing the medal winners a hearty congratulations and best of luck to one and all! Happy learning and best regards for the future Kaggle competitions and otherwise!\n\nRegards,\nRavi R",
      "votes": null
    },
    {
      "id": "3218207",
      "postDate": "06/06/2025 01:20:46",
      "content": "<p>Good analysis! It's so sad to make it as my first competition.lol🤣</p>",
      "rawMarkdown": "Good analysis! It's so sad to make it as my first competition.lol🤣",
      "votes": null
    },
    {
      "id": "3218211",
      "postDate": "06/06/2025 01:39:52",
      "content": "<p>Thank you for sharing the leaderboard shake-up statistics—really appreciate your effort and contribution to the community!</p>",
      "rawMarkdown": "Thank you for sharing the leaderboard shake-up statistics—really appreciate your effort and contribution to the community!",
      "votes": null
    },
    {
      "id": "3218219",
      "postDate": "06/06/2025 01:52:32",
      "content": "<p>Kind of weird — no wonder Dieter says there's always sadness in this competition.<br>\nBut there are still some clues, right? Top 1 is still Top 1.<br>\nRespect to all of you guys. 🫡</p>",
      "rawMarkdown": "Kind of weird — no wonder Dieter says there's always sadness in this competition.\nBut there are still some clues, right? Top 1 is still Top 1.\nRespect to all of you guys. 🫡",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3218207,
      "author_name": "kurisew",
      "author_url": "",
      "post_date": "06/06/2025 01:20:46",
      "content": "<p>Good analysis! It's so sad to make it as my first competition.lol🤣</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3218211,
      "author_name": "potonglantie93",
      "author_url": "",
      "post_date": "06/06/2025 01:39:52",
      "content": "<p>Thank you for sharing the leaderboard shake-up statistics—really appreciate your effort and contribution to the community!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3218219,
      "author_name": "bohangcong",
      "author_url": "",
      "post_date": "06/06/2025 01:52:32",
      "content": "<p>Kind of weird — no wonder Dieter says there's always sadness in this competition.<br>\nBut there are still some clues, right? Top 1 is still Top 1.<br>\nRespect to all of you guys. 🫡</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3218194": "Hello all,\n\nWe have a moderate level of churn here as expected, given the nature and size of the dataset. Let's delve into this in detail and study the regions of interest here-\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8273630%2F01b2a152efe9a0a3e0f8fba2b4b7cc98%2FLB.png?generation=1749169944723520&alt=media)\n\n## Regions of interest\n\n|Region Label|Comments|\n| ------------ | ----------- |\n|R1 | Indicates the rough account of top 10% rankers in the competition sustaining across the leaderboards. <br> Congrats to these teams for sustaining through the churn to win a medal here! <br> We observe a lot of **local churn** here and shall delve into this element separately below |\n|R2| Indicates the teams scoring in the top 10% in the public leaderboard and falling off the perch on the private leaderboard. This pool is a bit thick, indicating some level of churn here. |\n|R3 | Congrats to these teams for beating the churn! These teams moved up the leaderboard and won a medal here on the private leaderboard from a modest public leaderboard rank! |\n|R4 | This is very similar to R2 but shakeup occurs in a region outside the medal zone|\n\n## Stability analysis of the top 10% finishers \n\n|Region Label|Last Rank|Submissions - min, mean, max| Score band | Stability analysis |\n|------ | -------- | --------- | ----------- | ---------- |\n|Gold medal   | 14 | 242, 337 , 439 | 0.93 - 0.918 | Stable teams = 10 <br> Entering from below = 4 |\n|Silver medal | 108 |1, 153.84, 428 | 0.917 - 0.894 | Stable teams = 49 <br> Entering from below = 41  <br> Entering from above = 4|\n|Bronze medal | 216 | 1, 54.64, 302 | 0.894 - 0.893| Stable teams = 21 <br> Entering from below = 78  <br> Entering from above = 9 |\n\n<br>We observe 2 important facts here-\n- A lot of the bronze medal teams have churned either way - fallen off the perch and risen into the medal zone\n- Average submissions drastically reduce across gold -> silver -> bronze regions\n\n## Other adjutant analysis \n\n|Event Label| Public LB rank|Private LB rank|\n|------ | -------- | --------- | \n| Biggest riser across leaderboards - overall | 673 | 62 |\n| Biggest riser across leaderboards - R1| 181| 47 |\n| Highest fall across leaderboards - overall| 418 | 1925 |\n| Highest fall across leaderboards - R1|65 | 156 |\n\n## Score clusters \n\nThe below table highlights the top-3 score clusters across the entire leaderboard (private leaderboard score)\n\n|Score value| Number of teams |\n| --------- | ---------- |\n| 0.893 | 299|\n| 0.892 | 89 |\n| 0.809 | 78 |\n\n## Final thoughts\n\nWishing the medal winners a hearty congratulations and best of luck to one and all! Happy learning and best regards for the future Kaggle competitions and otherwise!\n\nRegards,\nRavi R",
    "3218207": "Good analysis! It's so sad to make it as my first competition.lol🤣",
    "3218211": "Thank you for sharing the leaderboard shake-up statistics—really appreciate your effort and contribution to the community!",
    "3218219": "Kind of weird — no wonder Dieter says there's always sadness in this competition.\nBut there are still some clues, right? Top 1 is still Top 1.\nRespect to all of you guys. 🫡"
  },
  "source": "meta"
}