{
  "id": 250198,
  "title": "What is your best single model ?",
  "url": "/competitions/mlb-player-digital-engagement-forecasting/discussion/250198",
  "author_name": "",
  "post_date": "2021-07-01T15:07:33.364931300Z",
  "votes": 39,
  "comment_count": 27,
  "views": 0,
  "content": "<p>Hi kagglers, i hope you are enjoying this competition. For curiostity and in order to have a wider sight on correlation between CV and LB, i thought it is a good idea to share our best single model CV and LB.</p>\n<p><strong>For me</strong>: my best model is ANN (CV ~ 1.14x, LB ~ 1.35x) / CV method : Time Series Split over 3 periods</p>\n<p>TO BE UPDATED …</p>",
  "messages": [
    {
      "id": "1372286",
      "postDate": "07/01/2021 15:07:33",
      "content": "<p>Hi kagglers, i hope you are enjoying this competition. For curiostity and in order to have a wider sight on correlation between CV and LB, i thought it is a good idea to share our best single model CV and LB.</p>\n<p><strong>For me</strong>: my best model is ANN (CV ~ 1.14x, LB ~ 1.35x) / CV method : Time Series Split over 3 periods</p>\n<p>TO BE UPDATED …</p>",
      "rawMarkdown": "Hi kagglers, i hope you are enjoying this competition. For curiostity and in order to have a wider sight on correlation between CV and LB, i thought it is a good idea to share our best single model CV and LB.\n\n**For me**: my best model is ANN (CV ~ 1.14x, LB ~ 1.35x) / CV method : Time Series Split over 3 periods\n\nTO BE UPDATED ...",
      "votes": null
    },
    {
      "id": "1372349",
      "postDate": "07/01/2021 16:02:05",
      "content": "<p>Model: ANN<br>\nCV: 1.1345<br>\nLB: 1.3520</p>",
      "rawMarkdown": "Model: ANN\nCV: 1.1345\nLB: 1.3520",
      "votes": null
    },
    {
      "id": "1372361",
      "postDate": "07/01/2021 16:16:06",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/ulrich07\" target=\"_blank\">@ulrich07</a>,</p>\n<p>What CV scheme are you using? </p>\n<p>I have a huge CV gap using TimeSeriesSplit (LB in 1.3's is getting CV in 0.7s).</p>\n<p>Best single model LB is an ANN 1.3268. </p>",
      "rawMarkdown": "Hi @ulrich07,\n\nWhat CV scheme are you using? \n\nI have a huge CV gap using TimeSeriesSplit (LB in 1.3's is getting CV in 0.7s).\n\nBest single model LB is an ANN 1.3268.",
      "votes": null
    },
    {
      "id": "1372397",
      "postDate": "07/01/2021 16:53:50",
      "content": "<p>I think you should add the evaluation technique for calculating the CV in order to be comparable between us</p>",
      "rawMarkdown": "I think you should add the evaluation technique for calculating the CV in order to be comparable between us",
      "votes": null
    },
    {
      "id": "1372486",
      "postDate": "07/01/2021 18:41:05",
      "content": "<p>CV: April month holdout, ~0.956 and LB: 1.3807 for solo gbm model.</p>",
      "rawMarkdown": "CV: April month holdout, ~0.956 and LB: 1.3807 for solo gbm model.",
      "votes": null
    },
    {
      "id": "1372489",
      "postDate": "07/01/2021 18:42:15",
      "content": "<p>Cool results for solo model! Did you use some target lags features as best ANN kernel do?</p>",
      "rawMarkdown": "Cool results for solo model! Did you use some target lags features as best ANN kernel do?",
      "votes": null
    },
    {
      "id": "1372507",
      "postDate": "07/01/2021 19:06:45",
      "content": "<p>Yep I used some target lags and other features too. </p>",
      "rawMarkdown": "Yep I used some target lags and other features too.",
      "votes": null
    },
    {
      "id": "1372532",
      "postDate": "07/01/2021 19:27:32",
      "content": "<p>I'm using Time Series Split over 3 periods</p>",
      "rawMarkdown": "I'm using Time Series Split over 3 periods",
      "votes": null
    },
    {
      "id": "1372707",
      "postDate": "07/02/2021 00:17:29",
      "content": "<p>CV method：April month holdout<br>\nCV Score：1.3979<br>\nLB Score：1.3187</p>",
      "rawMarkdown": "CV method：April month holdout\nCV Score：1.3979\nLB Score：1.3187",
      "votes": null
    },
    {
      "id": "1372715",
      "postDate": "07/02/2021 00:27:43",
      "content": "<p>An interesting difference is in validation and lb. I seem to have a total leak in my validation. What type of machine learning model are you using for best solo model? Of course you don't have to answer if you see fit :)</p>",
      "rawMarkdown": "An interesting difference is in validation and lb. I seem to have a total leak in my validation. What type of machine learning model are you using for best solo model? Of course you don't have to answer if you see fit :)",
      "votes": null
    },
    {
      "id": "1372721",
      "postDate": "07/02/2021 00:43:41",
      "content": "<p>I'm using lightgbm only.<br>\nAnd I only use the train data that players are included in test data.</p>",
      "rawMarkdown": "I'm using lightgbm only.\nAnd I only use the train data that players are included in test data.",
      "votes": null
    },
    {
      "id": "1373098",
      "postDate": "07/02/2021 08:19:18",
      "content": "<p>I just tried an April holdout with only players included in test data. Got 1.354 validation score with 1.3392 LB. </p>",
      "rawMarkdown": "I just tried an April holdout with only players included in test data. Got 1.354 validation score with 1.3392 LB.",
      "votes": null
    },
    {
      "id": "1373126",
      "postDate": "07/02/2021 08:34:43",
      "content": "<p>Model: LightGBM<br>\nCV method: April month holdout<br>\nCV Score: 0.898<br>\nLB Score：1.322</p>\n<p>I have confirmed that the CV and LB scores are closer when training and evaluating only the players included in the test.<br>\nHowever, in my experiment, the LB score was better when all players were included. So at this time, I am using all players.</p>",
      "rawMarkdown": "Model: LightGBM\nCV method: April month holdout\nCV Score: 0.898\nLB Score：1.322\n\nI have confirmed that the CV and LB scores are closer when training and evaluating only the players included in the test.\nHowever, in my experiment, the LB score was better when all players were included. So at this time, I am using all players.",
      "votes": null
    },
    {
      "id": "1373234",
      "postDate": "07/02/2021 10:35:39",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/jacobhowardparker\" target=\"_blank\">@jacobhowardparker</a>  which test data are you talking about ? the public test data  <code>example_test.csv</code></p>",
      "rawMarkdown": "Hi @jacobhowardparker  which test data are you talking about ? the public test data  `example_test.csv`",
      "votes": null
    },
    {
      "id": "1373337",
      "postDate": "07/02/2021 12:09:27",
      "content": "<p>There is a feature(playerForTestSetAndFuturePreds) whether the player are included in test data in players.csv.<br>\nI only use the data this feature is True.</p>",
      "rawMarkdown": "There is a feature(playerForTestSetAndFuturePreds) whether the player are included in test data in players.csv.\nI only use the data this feature is True.",
      "votes": null
    },
    {
      "id": "1373341",
      "postDate": "07/02/2021 12:17:53",
      "content": "<p>Thks <a href=\"https://www.kaggle.com/xblade\" target=\"_blank\">@xblade</a> </p>",
      "rawMarkdown": "Thks @xblade",
      "votes": null
    },
    {
      "id": "1373517",
      "postDate": "07/02/2021 14:36:07",
      "content": "<p>Model: LightGBM with All Players<br>\nCV method: April month holdout<br>\nCV Score: 0.926<br>\nLB Score：1.363</p>\n<p>Model: LightGBM with only players include test set<br>\nCV method: April month holdout<br>\nCV Score: 1.507<br>\nLB Score：1.359</p>",
      "rawMarkdown": "Model: LightGBM with All Players\nCV method: April month holdout\nCV Score: 0.926\nLB Score：1.363\n\nModel: LightGBM with only players include test set\nCV method: April month holdout\nCV Score: 1.507\nLB Score：1.359",
      "votes": null
    },
    {
      "id": "1373992",
      "postDate": "07/02/2021 22:35:45",
      "content": "<p>Model: LightGBM<br>\nCV method: April month holdout<br>\nCV Score: 1.3211<br>\nLB Score：1.3407</p>",
      "rawMarkdown": "Model: LightGBM\nCV method: April month holdout\nCV Score: 1.3211\nLB Score：1.3407",
      "votes": null
    },
    {
      "id": "1374122",
      "postDate": "07/03/2021 03:38:22",
      "content": "<p>LightGBM<br>\nApril as validation<br>\nValidation : 0.689<br>\nLB : 1.3502</p>\n<p>Validation : 0.823<br>\nLB : 1.3376</p>\n<p>Validation : 0.802<br>\nLB : 1.3382</p>",
      "rawMarkdown": "LightGBM\nApril as validation\nValidation : 0.689\nLB : 1.3502\n\nValidation : 0.823\nLB : 1.3376\n\nValidation : 0.802\nLB : 1.3382",
      "votes": null
    },
    {
      "id": "1375076",
      "postDate": "07/03/2021 20:23:32",
      "content": "<p>CV method: TimeSeries holdout<br>\nCV Score: 1.6625<br>\nLB Score：1.4347</p>",
      "rawMarkdown": "CV method: TimeSeries holdout\nCV Score: 1.6625\nLB Score：1.4347",
      "votes": null
    },
    {
      "id": "1380280",
      "postDate": "07/08/2021 01:28:04",
      "content": "<p>Model: LightGBM<br>\nCV method: April as validation<br>\nCV score: 0.9397<br>\nLB: 1.3652</p>",
      "rawMarkdown": "Model: LightGBM\nCV method: April as validation\nCV score: 0.9397\nLB: 1.3652",
      "votes": null
    },
    {
      "id": "1387847",
      "postDate": "07/14/2021 13:15:48",
      "content": "<p>Model: NN<br>\nApril as validation<br>\nCV: 1.293<br>\nLB: 1.3251</p>",
      "rawMarkdown": "Model: NN\nApril as validation\nCV: 1.293\nLB: 1.3251",
      "votes": null
    },
    {
      "id": "1390662",
      "postDate": "07/17/2021 00:15:28",
      "content": "<p>LightGBM holdout</p>\n<ul>\n<li>1. Validation Score:1.1570  LB Score:1.3371</li>\n<li>2. Validation Score:1.3380 LB Score:1.3268</li>\n<li>3. Validation Score:1.4705 LB Score:1.3367</li>\n</ul>\n<p>It is the same except how to divide the training data and the verification data.</p>",
      "rawMarkdown": "LightGBM holdout\n\n- 1. Validation Score:1.1570  LB Score:1.3371\n- 2. Validation Score:1.3380 LB Score:1.3268\n- 3. Validation Score:1.4705 LB Score:1.3367\n\nIt is the same except how to divide the training data and the verification data.",
      "votes": null
    },
    {
      "id": "1392315",
      "postDate": "07/18/2021 14:39:42",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/horohoro\" target=\"_blank\">@horohoro</a>, how is your holdout made?</p>",
      "rawMarkdown": "Hi @horohoro, how is your holdout made?",
      "votes": null
    },
    {
      "id": "1393067",
      "postDate": "07/19/2021 10:53:06",
      "content": "<p>Currently, my holdout process is simple.</p>\n<ul>\n<li>Only for players with 1 players.playerForTestSetAndFuturePreds.</li>\n<li>Training data and verification data are separated by a specific date as a boundary. </li>\n<li>The 3rd is separated by 2021-04-01.</li>\n<li>The 1st and 2nd are separated by the date a while later after starting season.</li>\n</ul>",
      "rawMarkdown": "Currently, my holdout process is simple.\n- Only for players with 1 players.playerForTestSetAndFuturePreds.\n- Training data and verification data are separated by a specific date as a boundary. \n- The 3rd is separated by 2021-04-01.\n- The 1st and 2nd are separated by the date a while later after starting season.",
      "votes": null
    },
    {
      "id": "1395327",
      "postDate": "07/21/2021 06:24:06",
      "content": "<p>I see. Good job. Thanks. </p>",
      "rawMarkdown": "I see. Good job. Thanks.",
      "votes": null
    },
    {
      "id": "1396707",
      "postDate": "07/22/2021 12:15:08",
      "content": "<p>Model: NN pytorch<br>\nApril as validation<br>\nCV: 0.8240<br>\nLB: 1.3239</p>",
      "rawMarkdown": "Model: NN pytorch\nApril as validation\nCV: 0.8240\nLB: 1.3239",
      "votes": null
    },
    {
      "id": "1396917",
      "postDate": "07/22/2021 15:25:57",
      "content": "<p>Hey, <a href=\"https://www.kaggle.com/lhagiimn\" target=\"_blank\">@lhagiimn</a> wanna team up?</p>",
      "rawMarkdown": "Hey, @lhagiimn wanna team up?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1372349,
      "author_name": "lars123",
      "author_url": "",
      "post_date": "07/01/2021 16:02:05",
      "content": "<p>Model: ANN<br>\nCV: 1.1345<br>\nLB: 1.3520</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1372361,
      "author_name": "jacobhowardparker",
      "author_url": "",
      "post_date": "07/01/2021 16:16:06",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/ulrich07\" target=\"_blank\">@ulrich07</a>,</p>\n<p>What CV scheme are you using? </p>\n<p>I have a huge CV gap using TimeSeriesSplit (LB in 1.3's is getting CV in 0.7s).</p>\n<p>Best single model LB is an ANN 1.3268. </p>",
      "votes": null,
      "replies": [
        {
          "id": 1372489,
          "author_name": "sggpls",
          "author_url": "",
          "post_date": "07/01/2021 18:42:15",
          "content": "<p>Cool results for solo model! Did you use some target lags features as best ANN kernel do?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1372507,
          "author_name": "jacobhowardparker",
          "author_url": "",
          "post_date": "07/01/2021 19:06:45",
          "content": "<p>Yep I used some target lags and other features too. </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1372532,
          "author_name": "ulrich07",
          "author_url": "",
          "post_date": "07/01/2021 19:27:32",
          "content": "<p>I'm using Time Series Split over 3 periods</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1372397,
      "author_name": "enric1296",
      "author_url": "",
      "post_date": "07/01/2021 16:53:50",
      "content": "<p>I think you should add the evaluation technique for calculating the CV in order to be comparable between us</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1372486,
      "author_name": "sggpls",
      "author_url": "",
      "post_date": "07/01/2021 18:41:05",
      "content": "<p>CV: April month holdout, ~0.956 and LB: 1.3807 for solo gbm model.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1372707,
      "author_name": "xblade",
      "author_url": "",
      "post_date": "07/02/2021 00:17:29",
      "content": "<p>CV method：April month holdout<br>\nCV Score：1.3979<br>\nLB Score：1.3187</p>",
      "votes": null,
      "replies": [
        {
          "id": 1372715,
          "author_name": "sggpls",
          "author_url": "",
          "post_date": "07/02/2021 00:27:43",
          "content": "<p>An interesting difference is in validation and lb. I seem to have a total leak in my validation. What type of machine learning model are you using for best solo model? Of course you don't have to answer if you see fit :)</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1372721,
          "author_name": "xblade",
          "author_url": "",
          "post_date": "07/02/2021 00:43:41",
          "content": "<p>I'm using lightgbm only.<br>\nAnd I only use the train data that players are included in test data.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1373098,
          "author_name": "jacobhowardparker",
          "author_url": "",
          "post_date": "07/02/2021 08:19:18",
          "content": "<p>I just tried an April holdout with only players included in test data. Got 1.354 validation score with 1.3392 LB. </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1373234,
          "author_name": "ulrich07",
          "author_url": "",
          "post_date": "07/02/2021 10:35:39",
          "content": "<p>Hi <a href=\"https://www.kaggle.com/jacobhowardparker\" target=\"_blank\">@jacobhowardparker</a>  which test data are you talking about ? the public test data  <code>example_test.csv</code></p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1373337,
          "author_name": "xblade",
          "author_url": "",
          "post_date": "07/02/2021 12:09:27",
          "content": "<p>There is a feature(playerForTestSetAndFuturePreds) whether the player are included in test data in players.csv.<br>\nI only use the data this feature is True.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1373341,
          "author_name": "ulrich07",
          "author_url": "",
          "post_date": "07/02/2021 12:17:53",
          "content": "<p>Thks <a href=\"https://www.kaggle.com/xblade\" target=\"_blank\">@xblade</a> </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1373126,
      "author_name": "nomorevotch",
      "author_url": "",
      "post_date": "07/02/2021 08:34:43",
      "content": "<p>Model: LightGBM<br>\nCV method: April month holdout<br>\nCV Score: 0.898<br>\nLB Score：1.322</p>\n<p>I have confirmed that the CV and LB scores are closer when training and evaluating only the players included in the test.<br>\nHowever, in my experiment, the LB score was better when all players were included. So at this time, I am using all players.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1373517,
      "author_name": "joshspchang",
      "author_url": "",
      "post_date": "07/02/2021 14:36:07",
      "content": "<p>Model: LightGBM with All Players<br>\nCV method: April month holdout<br>\nCV Score: 0.926<br>\nLB Score：1.363</p>\n<p>Model: LightGBM with only players include test set<br>\nCV method: April month holdout<br>\nCV Score: 1.507<br>\nLB Score：1.359</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1373992,
      "author_name": "ramikhreas",
      "author_url": "",
      "post_date": "07/02/2021 22:35:45",
      "content": "<p>Model: LightGBM<br>\nCV method: April month holdout<br>\nCV Score: 1.3211<br>\nLB Score：1.3407</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1374122,
      "author_name": "assign",
      "author_url": "",
      "post_date": "07/03/2021 03:38:22",
      "content": "<p>LightGBM<br>\nApril as validation<br>\nValidation : 0.689<br>\nLB : 1.3502</p>\n<p>Validation : 0.823<br>\nLB : 1.3376</p>\n<p>Validation : 0.802<br>\nLB : 1.3382</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1375076,
      "author_name": "enric1296",
      "author_url": "",
      "post_date": "07/03/2021 20:23:32",
      "content": "<p>CV method: TimeSeries holdout<br>\nCV Score: 1.6625<br>\nLB Score：1.4347</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1380280,
      "author_name": "leewook",
      "author_url": "",
      "post_date": "07/08/2021 01:28:04",
      "content": "<p>Model: LightGBM<br>\nCV method: April as validation<br>\nCV score: 0.9397<br>\nLB: 1.3652</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1387847,
      "author_name": "marktenenholtz",
      "author_url": "",
      "post_date": "07/14/2021 13:15:48",
      "content": "<p>Model: NN<br>\nApril as validation<br>\nCV: 1.293<br>\nLB: 1.3251</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1390662,
      "author_name": "horohoro",
      "author_url": "",
      "post_date": "07/17/2021 00:15:28",
      "content": "<p>LightGBM holdout</p>\n<ul>\n<li>1. Validation Score:1.1570  LB Score:1.3371</li>\n<li>2. Validation Score:1.3380 LB Score:1.3268</li>\n<li>3. Validation Score:1.4705 LB Score:1.3367</li>\n</ul>\n<p>It is the same except how to divide the training data and the verification data.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1392315,
          "author_name": "zacchaeus",
          "author_url": "",
          "post_date": "07/18/2021 14:39:42",
          "content": "<p>Hi <a href=\"https://www.kaggle.com/horohoro\" target=\"_blank\">@horohoro</a>, how is your holdout made?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1393067,
          "author_name": "horohoro",
          "author_url": "",
          "post_date": "07/19/2021 10:53:06",
          "content": "<p>Currently, my holdout process is simple.</p>\n<ul>\n<li>Only for players with 1 players.playerForTestSetAndFuturePreds.</li>\n<li>Training data and verification data are separated by a specific date as a boundary. </li>\n<li>The 3rd is separated by 2021-04-01.</li>\n<li>The 1st and 2nd are separated by the date a while later after starting season.</li>\n</ul>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1395327,
          "author_name": "zacchaeus",
          "author_url": "",
          "post_date": "07/21/2021 06:24:06",
          "content": "<p>I see. Good job. Thanks. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1396707,
      "author_name": "lhagiimn",
      "author_url": "",
      "post_date": "07/22/2021 12:15:08",
      "content": "<p>Model: NN pytorch<br>\nApril as validation<br>\nCV: 0.8240<br>\nLB: 1.3239</p>",
      "votes": null,
      "replies": [
        {
          "id": 1396917,
          "author_name": "zacchaeus",
          "author_url": "",
          "post_date": "07/22/2021 15:25:57",
          "content": "<p>Hey, <a href=\"https://www.kaggle.com/lhagiimn\" target=\"_blank\">@lhagiimn</a> wanna team up?</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1372286": "Hi kagglers, i hope you are enjoying this competition. For curiostity and in order to have a wider sight on correlation between CV and LB, i thought it is a good idea to share our best single model CV and LB.\n\n**For me**: my best model is ANN (CV ~ 1.14x, LB ~ 1.35x) / CV method : Time Series Split over 3 periods\n\nTO BE UPDATED ...",
    "1372349": "Model: ANN\nCV: 1.1345\nLB: 1.3520",
    "1372361": "Hi @ulrich07,\n\nWhat CV scheme are you using? \n\nI have a huge CV gap using TimeSeriesSplit (LB in 1.3's is getting CV in 0.7s).\n\nBest single model LB is an ANN 1.3268.",
    "1372397": "I think you should add the evaluation technique for calculating the CV in order to be comparable between us",
    "1372486": "CV: April month holdout, ~0.956 and LB: 1.3807 for solo gbm model.",
    "1372489": "Cool results for solo model! Did you use some target lags features as best ANN kernel do?",
    "1372507": "Yep I used some target lags and other features too.",
    "1372532": "I'm using Time Series Split over 3 periods",
    "1372707": "CV method：April month holdout\nCV Score：1.3979\nLB Score：1.3187",
    "1372715": "An interesting difference is in validation and lb. I seem to have a total leak in my validation. What type of machine learning model are you using for best solo model? Of course you don't have to answer if you see fit :)",
    "1372721": "I'm using lightgbm only.\nAnd I only use the train data that players are included in test data.",
    "1373098": "I just tried an April holdout with only players included in test data. Got 1.354 validation score with 1.3392 LB.",
    "1373126": "Model: LightGBM\nCV method: April month holdout\nCV Score: 0.898\nLB Score：1.322\n\nI have confirmed that the CV and LB scores are closer when training and evaluating only the players included in the test.\nHowever, in my experiment, the LB score was better when all players were included. So at this time, I am using all players.",
    "1373234": "Hi @jacobhowardparker  which test data are you talking about ? the public test data  `example_test.csv`",
    "1373337": "There is a feature(playerForTestSetAndFuturePreds) whether the player are included in test data in players.csv.\nI only use the data this feature is True.",
    "1373341": "Thks @xblade",
    "1373517": "Model: LightGBM with All Players\nCV method: April month holdout\nCV Score: 0.926\nLB Score：1.363\n\nModel: LightGBM with only players include test set\nCV method: April month holdout\nCV Score: 1.507\nLB Score：1.359",
    "1373992": "Model: LightGBM\nCV method: April month holdout\nCV Score: 1.3211\nLB Score：1.3407",
    "1374122": "LightGBM\nApril as validation\nValidation : 0.689\nLB : 1.3502\n\nValidation : 0.823\nLB : 1.3376\n\nValidation : 0.802\nLB : 1.3382",
    "1375076": "CV method: TimeSeries holdout\nCV Score: 1.6625\nLB Score：1.4347",
    "1380280": "Model: LightGBM\nCV method: April as validation\nCV score: 0.9397\nLB: 1.3652",
    "1387847": "Model: NN\nApril as validation\nCV: 1.293\nLB: 1.3251",
    "1390662": "LightGBM holdout\n\n- 1. Validation Score:1.1570  LB Score:1.3371\n- 2. Validation Score:1.3380 LB Score:1.3268\n- 3. Validation Score:1.4705 LB Score:1.3367\n\nIt is the same except how to divide the training data and the verification data.",
    "1392315": "Hi @horohoro, how is your holdout made?",
    "1393067": "Currently, my holdout process is simple.\n- Only for players with 1 players.playerForTestSetAndFuturePreds.\n- Training data and verification data are separated by a specific date as a boundary. \n- The 3rd is separated by 2021-04-01.\n- The 1st and 2nd are separated by the date a while later after starting season.",
    "1395327": "I see. Good job. Thanks.",
    "1396707": "Model: NN pytorch\nApril as validation\nCV: 0.8240\nLB: 1.3239",
    "1396917": "Hey, @lhagiimn wanna team up?"
  },
  "source": "meta"
}