{ "id": 304735, "title": "Forecast the different medal zones.", "url": "/competitions/g-research-crypto-forecasting/discussion/304735", "author_name": "Kirderf", "post_date": "2022-02-02T09:46:27.295000", "votes": 23, "comment_count": 34, "views": 0, "content": "

Work done, submission selection completed. Now just follow the exciting coming three month re-runs 🙂

\n

Based on your validation and testing, what do you think will be the thresholds for the different zones, starting with bronze level, silver level, gold level and the top winner's score?

\n

My guess: Bronze 0.01, Silver 0.02, Gold 0.033, Winner 0.038

\n

Updated with suggested levels in average posted in this topic, excluding extreme values:

\n

Winner 0.049
\nGold 0.035
\nSilver 0.022
\nBronze 0.014

", "messages": [ { "id": 1672779, "postDate": "2022-02-02T09:46:27.297Z", "content": "

Work done, submission selection completed. Now just follow the exciting coming three month re-runs 🙂

\n

Based on your validation and testing, what do you think will be the thresholds for the different zones, starting with bronze level, silver level, gold level and the top winner's score?

\n

My guess: Bronze 0.01, Silver 0.02, Gold 0.033, Winner 0.038

\n

Updated with suggested levels in average posted in this topic, excluding extreme values:

\n

Winner 0.049
\nGold 0.035
\nSilver 0.022
\nBronze 0.014

", "rawMarkdown": "Work done, submission selection completed. Now just follow the exciting coming three month re-runs 🙂\n\nBased on your validation and testing, what do you think will be the thresholds for the different zones, starting with bronze level, silver level, gold level and the top winner's score?\n\nMy guess: Bronze 0.01, Silver 0.02, Gold 0.033, Winner 0.038\n\nUpdated with suggested levels in average posted in this topic, excluding extreme values: \n\nWinner 0.049\nGold 0.035\nSilver 0.022\nBronze 0.014", "votes": 23 }, { "id": 1673705, "postDate": "2022-02-02T22:39:10.483Z", "content": "

Good luck to everyone!! It will be very exciting to watch the results coming in. We might see a >0.06 winner?! 🔥

", "rawMarkdown": "Good luck to everyone!! It will be very exciting to watch the results coming in. We might see a >0.06 winner?! 🔥", "votes": 5, "replies": [ { "id": 1673748, "postDate": "2022-02-02T23:47:11.127Z", "content": "

And thank you for a good and interesting challenge and competition. 👍
\nYes that would be great! Looking forward to reading that write-up 😊

", "rawMarkdown": "And thank you for a good and interesting challenge and competition. 👍\nYes that would be great! Looking forward to reading that write-up 😊", "votes": 1 } ] }, { "id": 1674360, "postDate": "2022-02-03T12:32:30.690Z", "content": "

Winner: .055
\nGold: . .055 - .040
\nSilver: .040 - .022 (I will score .022-.023 so humbly did put myself into silver region :-)
\nBronz: .0220 - .0145

\n

You should thinkof a prize for the best guesser..

", "rawMarkdown": "Winner: .055\nGold: . .055 - .040\nSilver: .040 - .022 (I will score .022-.023 so humbly did put myself into silver region :-)\nBronz: .0220 - .0145\n\nYou should thinkof a prize for the best guesser..", "votes": 3 }, { "id": 1673190, "postDate": "2022-02-02T15:08:29.133Z", "content": "

Agree with your guesses, most people here who think they have done well will likely end up nearer to < 0.00 in the next 3 months IMO..

", "rawMarkdown": "Agree with your guesses, most people here who think they have done well will likely end up nearer to < 0.00 in the next 3 months IMO..", "votes": 3, "replies": [ { "id": 1674196, "postDate": "2022-02-03T09:46:38.673Z", "content": "

😄 yeah my only chance of a medal in this comp is that most people above me overfitted to the public LB

", "rawMarkdown": "😄 yeah my only chance of a medal in this comp is that most people above me overfitted to the public LB", "votes": 1 } ] }, { "id": 1672861, "postDate": "2022-02-02T10:53:07.680Z", "content": "

I wouldn't make any bet here!
\nI was surprised that just predicting the negative of the re-computed target for (t-16) was getting a score of 0.0292 on the LB, and 0.0288 on the latest supplemental data.
\nI've had a hard time trying to get \"significantly\" better than this simple predictor (but I have no experience with forecasting…).
\nAlso, I found it interesting to plot the evolution of the score of models as the horizon increases by 16 hours increments.
\nFor this simple predictor, this plot looks like this for the LB test period (top) and on the latest supplemental data (bottom):
\nhttps://postimg.cc/CnjdtDpf/68b82a2f
\nOne can see that the score can vary significantly over the horizon (data for the beginning of the horizon is not that relevant as the scoring period is small).

\n

I just keep thinking that we just bought ourselves some lottery tickets and that future market conditions will significantly impact the performance of our models.
\nGood luck to all the competitors (crossing fingers :-) )

", "rawMarkdown": "I wouldn't make any bet here!\nI was surprised that just predicting the negative of the re-computed target for (t-16) was getting a score of 0.0292 on the LB, and 0.0288 on the latest supplemental data.\nI've had a hard time trying to get \"significantly\" better than this simple predictor (but I have no experience with forecasting...).\nAlso, I found it interesting to plot the evolution of the score of models as the horizon increases by 16 hours increments.\nFor this simple predictor, this plot looks like this for the LB test period (top) and on the latest supplemental data (bottom):\nhttps://postimg.cc/CnjdtDpf/68b82a2f\nOne can see that the score can vary significantly over the horizon (data for the beginning of the horizon is not that relevant as the scoring period is small).\n\nI just keep thinking that we just bought ourselves some lottery tickets and that future market conditions will significantly impact the performance of our models.\nGood luck to all the competitors (crossing fingers :-) )", "votes": 3, "replies": [ { "id": 1672870, "postDate": "2022-02-02T11:04:35.760Z", "content": "

Yes luck will be a great and the most valuable indicator I guess, anyway for me and my solution :) Maybe just a simple random forest would had been best in the end ;)

", "rawMarkdown": "Yes luck will be a great and the most valuable indicator I guess, anyway for me and my solution :) Maybe just a simple random forest would had been best in the end ;)" }, { "id": 1673196, "postDate": "2022-02-02T15:13:12.950Z", "content": "

Your link is not working for me but yeah I tried to account for changing market conditions, I trained each ensemble model on different market regime and tried to dynamically weight it for the predictions, my ideal would have been retrain the models periodically by keeping track of all new test data from API but alas would not pass the time constraints..

", "rawMarkdown": "Your link is not working for me but yeah I tried to account for changing market conditions, I trained each ensemble model on different market regime and tried to dynamically weight it for the predictions, my ideal would have been retrain the models periodically by keeping track of all new test data from API but alas would not pass the time constraints.." }, { "id": 1673676, "postDate": "2022-02-02T21:57:51.220Z", "rawMarkdown": "", "isDeleted": true }, { "id": 1673710, "postDate": "2022-02-02T22:49:59.553Z", "content": "

At each time step, one can recompute the target (the 15 minute residualized returns) that corresponds to the time step 16 minutes before (using the algorithm described in this discussion: https://www.kaggle.com/c/g-research-crypto-forecasting/discussion/286778
\nThis means, that one has to maintain a shifting window of the past 3750 minutes to perform this calculation.
\nWhat I meant is that using the negative of this value for the prediction at time t, one gets a score of 0.0292 on the LB.
\nFor the plots that I posted, I took a 16 hours granularity (this value is \"arbitrary\" and is introducing some confusion, but one should not pay to much attention to it. I should have chosen 24 hours instead, but 16 hours was a good trade-off in my experiments!)
\nActually, I've just noticed this other post which describes the same idea: https://www.kaggle.com/c/g-research-crypto-forecasting/discussion/304891
\nHope this answers your question.

", "rawMarkdown": "At each time step, one can recompute the target (the 15 minute residualized returns) that corresponds to the time step 16 minutes before (using the algorithm described in this discussion: https://www.kaggle.com/c/g-research-crypto-forecasting/discussion/286778\nThis means, that one has to maintain a shifting window of the past 3750 minutes to perform this calculation.\nWhat I meant is that using the negative of this value for the prediction at time t, one gets a score of 0.0292 on the LB.\nFor the plots that I posted, I took a 16 hours granularity (this value is \"arbitrary\" and is introducing some confusion, but one should not pay to much attention to it. I should have chosen 24 hours instead, but 16 hours was a good trade-off in my experiments!)\nActually, I've just noticed this other post which describes the same idea: https://www.kaggle.com/c/g-research-crypto-forecasting/discussion/304891\nHope this answers your question." }, { "id": 1673719, "postDate": "2022-02-02T22:54:25.600Z", "content": "

Okay makes sense, I actually had this as a feature myself but not for such a long period and did not get 0.0292 by that alone.

", "rawMarkdown": "Okay makes sense, I actually had this as a feature myself but not for such a long period and did not get 0.0292 by that alone." }, { "id": 1674082, "postDate": "2022-02-03T08:06:42.463Z", "content": "

Your \"minustarget\" experiment is really interesting. For your rolling stats to calculate beta ( for this specific examle ), did you select the 3750 window, starting from t-16, since they are the only available ones?

", "rawMarkdown": "Your \"minustarget\" experiment is really interesting. For your rolling stats to calculate beta ( for this specific examle ), did you select the 3750 window, starting from t-16, since they are the only available ones?" }, { "id": 1674105, "postDate": "2022-02-03T08:30:52.713Z", "content": "

Good remark, you're right. Technically, the total buffer size covers the previous: (3750 + 16) minutes, to be able to use the calculated target at (t-16) as a predictor at time t.

", "rawMarkdown": "Good remark, you're right. Technically, the total buffer size covers the previous: (3750 + 16) minutes, to be able to use the calculated target at (t-16) as a predictor at time t." } ] }, { "id": 1673688, "postDate": "2022-02-02T22:05:40.663Z", "content": "

My predictions:
\nBronze: 0.001±1
\nSilver: 0.005±1
\nGold: 0.01±1
\nWinner: 0.015±1

\n

Unfortunately, this can destroy the leaderboard:
\n\"\"
\nExpected position in the leaderboard = 0.5 * random() ± 0.4 * (Elon Musk tweets about bitcoin) + 0.1 * skill

", "rawMarkdown": "My predictions:\nBronze: 0.001±1\nSilver: 0.005±1\nGold: 0.01±1\nWinner: 0.015±1\n\nUnfortunately, this can destroy the leaderboard:\n![](https://inteng-storage.s3.amazonaws.com/img/iea/BxG21M1DO9/sizes/dogecoin-value-jumps-and-dives-on-meme-inspired-dogeday_md.jpg)\nExpected position in the leaderboard = 0.5 * random() ± 0.4 * (Elon Musk tweets about bitcoin) + 0.1 * skill", "votes": 2 }, { "id": 1674935, "postDate": "2022-02-03T22:04:40.037Z", "content": "

Winner: .048
\nGold: .036
\nSilver: .028
\nBronze: .018

", "rawMarkdown": "Winner: .048\nGold: .036\nSilver: .028\nBronze: .018", "votes": 1 }, { "id": 1673679, "postDate": "2022-02-02T22:00:41.597Z", "content": "

Winner 0.045+
\nGold 0.04
\nSilver 0.029
\nBronze 0.02

", "rawMarkdown": "Winner 0.045+\nGold 0.04\nSilver 0.029\nBronze 0.02", "votes": 1 }, { "id": 1673195, "postDate": "2022-02-02T15:12:05.410Z", "content": "

Bronze: 0.01
\nSilver: 0.015
\nGold: 0.025
\nWinner maybe in 0.04+

", "rawMarkdown": "Bronze: 0.01 \nSilver: 0.015\nGold: 0.025\nWinner maybe in 0.04+", "votes": 1 }, { "id": 1673537, "postDate": "2022-02-02T19:57:09.693Z", "content": "

I would say:
\nBronze: 0.005
\nSilver: 0.01
\nGold: 0.03

", "rawMarkdown": "I would say:\nBronze: 0.005\nSilver: 0.01\nGold: 0.03", "votes": 2, "replies": [ { "id": 1677087, "postDate": "2022-02-05T14:16:56.797Z", "content": "

Same here.

", "rawMarkdown": "Same here." }, { "id": 1677199, "postDate": "2022-02-05T15:19:32.057Z", "content": "

so you're saying I have a chance

", "rawMarkdown": "so you're saying I have a chance" } ] }, { "id": 1673157, "postDate": "2022-02-02T14:41:29.673Z", "content": "

I think the winner will definitely be greater than .05. Most likely they will get a score between .06-.065.

", "rawMarkdown": "I think the winner will definitely be greater than .05. Most likely they will get a score between .06-.065. ", "votes": 1 }, { "id": 1673092, "postDate": "2022-02-02T14:02:56.133Z", "content": "

I guess winner should be 0.04~0.05.

", "rawMarkdown": "I guess winner should be 0.04~0.05.", "votes": 1 }, { "id": 1693122, "postDate": "2022-02-16T12:58:40.797Z", "content": "

I don't care much of score. I just want myself in rank zone.

", "rawMarkdown": "I don't care much of score. I just want myself in rank zone." }, { "id": 1678922, "postDate": "2022-02-06T21:11:57.563Z", "content": "

Private LB is always a rollarcoaster!

", "rawMarkdown": "Private LB is always a rollarcoaster!", "replies": [ { "id": 1681596, "postDate": "2022-02-08T15:17:18.810Z", "content": "

it will be retsaocrallor 📉

", "rawMarkdown": "it will be retsaocrallor 📉" } ] }, { "id": 1673247, "postDate": "2022-02-02T15:53:05.843Z", "content": "

Given an infinite number of monkeys and an infinite number of submissions…..maybe 0.01

", "rawMarkdown": "Given an infinite number of monkeys and an infinite number of submissions.....maybe 0.01" }, { "id": 1672953, "postDate": "2022-02-02T12:05:42.193Z", "content": "

winner 0.0000 😯

", "rawMarkdown": "winner 0.0000 😯", "replies": [ { "id": 1672967, "postDate": "2022-02-02T12:15:28.310Z", "content": "

Haha, oboy, a complete LB shake. Yes why not, isn't easy.

", "rawMarkdown": "Haha, oboy, a complete LB shake. Yes why not, isn't easy." }, { "id": 1674514, "postDate": "2022-02-03T15:07:04.423Z", "content": "

I think > 0. These are 15minute returns and there are no transaction costs imposed. As such, you could certainly get > 0 (even though this might not translate to a profitable signal in the real world after taking into account market impact).

", "rawMarkdown": "I think > 0. These are 15minute returns and there are no transaction costs imposed. As such, you could certainly get > 0 (even though this might not translate to a profitable signal in the real world after taking into account market impact).", "votes": 3 } ] }, { "id": 1672929, "postDate": "2022-02-02T11:44:27.390Z", "content": "

I hope you're right ha

", "rawMarkdown": "I hope you're right ha ", "replies": [ { "id": 1672939, "postDate": "2022-02-02T11:58:35.177Z", "content": "

Yes maybe just wishful thinking, would be nice to listen to others guesses :)

", "rawMarkdown": "Yes maybe just wishful thinking, would be nice to listen to others guesses :)" } ] }, { "id": 1683165, "postDate": "2022-02-09T16:02:02.180Z", "rawMarkdown": "", "votes": 1, "isDeleted": true, "replies": [ { "id": 1684617, "postDate": "2022-02-10T15:53:56.263Z", "content": "

You mean 0.027 or 0.27? 0.27 sounds high.

", "rawMarkdown": "You mean 0.027 or 0.27? 0.27 sounds high.", "votes": 1 } ] }, { "id": 1673855, "postDate": "2022-02-03T03:14:34.983Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 1673705, "author_name": "Carlos Stein N Brito", "author_url": "", "post_date": "2022-02-02T22:39:10.483000", "content": "

Good luck to everyone!! It will be very exciting to watch the results coming in. We might see a >0.06 winner?! 🔥

", "votes": 5, "replies": [ { "id": 1673748, "author_name": "Kirderf", "author_url": "", "post_date": "2022-02-02T23:47:11.127000", "content": "

And thank you for a good and interesting challenge and competition. 👍
\nYes that would be great! Looking forward to reading that write-up 😊

", "votes": 1, "replies": [] } ] }, { "id": 1674360, "author_name": "GUNER", "author_url": "", "post_date": "2022-02-03T12:32:30.690000", "content": "

Winner: .055
\nGold: . .055 - .040
\nSilver: .040 - .022 (I will score .022-.023 so humbly did put myself into silver region :-)
\nBronz: .0220 - .0145

\n

You should thinkof a prize for the best guesser..

", "votes": 3, "replies": [] }, { "id": 1673190, "author_name": "JM", "author_url": "", "post_date": "2022-02-02T15:08:29.133000", "content": "

Agree with your guesses, most people here who think they have done well will likely end up nearer to < 0.00 in the next 3 months IMO..

", "votes": 3, "replies": [ { "id": 1674196, "author_name": "Marco Gorelli", "author_url": "", "post_date": "2022-02-03T09:46:38.673000", "content": "

😄 yeah my only chance of a medal in this comp is that most people above me overfitted to the public LB

", "votes": 1, "replies": [] } ] }, { "id": 1672861, "author_name": "Hugues", "author_url": "", "post_date": "2022-02-02T10:53:07.680000", "content": "

I wouldn't make any bet here!
\nI was surprised that just predicting the negative of the re-computed target for (t-16) was getting a score of 0.0292 on the LB, and 0.0288 on the latest supplemental data.
\nI've had a hard time trying to get \"significantly\" better than this simple predictor (but I have no experience with forecasting…).
\nAlso, I found it interesting to plot the evolution of the score of models as the horizon increases by 16 hours increments.
\nFor this simple predictor, this plot looks like this for the LB test period (top) and on the latest supplemental data (bottom):
\nhttps://postimg.cc/CnjdtDpf/68b82a2f
\nOne can see that the score can vary significantly over the horizon (data for the beginning of the horizon is not that relevant as the scoring period is small).

\n

I just keep thinking that we just bought ourselves some lottery tickets and that future market conditions will significantly impact the performance of our models.
\nGood luck to all the competitors (crossing fingers :-) )

", "votes": 3, "replies": [ { "id": 1672870, "author_name": "Kirderf", "author_url": "", "post_date": "2022-02-02T11:04:35.760000", "content": "

Yes luck will be a great and the most valuable indicator I guess, anyway for me and my solution :) Maybe just a simple random forest would had been best in the end ;)

", "votes": 0, "replies": [] }, { "id": 1673196, "author_name": "JM", "author_url": "", "post_date": "2022-02-02T15:13:12.950000", "content": "

Your link is not working for me but yeah I tried to account for changing market conditions, I trained each ensemble model on different market regime and tried to dynamically weight it for the predictions, my ideal would have been retrain the models periodically by keeping track of all new test data from API but alas would not pass the time constraints..

", "votes": 0, "replies": [] }, { "id": 1673676, "author_name": "", "author_url": "", "post_date": "2022-02-02T21:57:51.220000", "content": "", "votes": 0, "replies": [] }, { "id": 1673710, "author_name": "Hugues", "author_url": "", "post_date": "2022-02-02T22:49:59.553000", "content": "

At each time step, one can recompute the target (the 15 minute residualized returns) that corresponds to the time step 16 minutes before (using the algorithm described in this discussion: https://www.kaggle.com/c/g-research-crypto-forecasting/discussion/286778
\nThis means, that one has to maintain a shifting window of the past 3750 minutes to perform this calculation.
\nWhat I meant is that using the negative of this value for the prediction at time t, one gets a score of 0.0292 on the LB.
\nFor the plots that I posted, I took a 16 hours granularity (this value is \"arbitrary\" and is introducing some confusion, but one should not pay to much attention to it. I should have chosen 24 hours instead, but 16 hours was a good trade-off in my experiments!)
\nActually, I've just noticed this other post which describes the same idea: https://www.kaggle.com/c/g-research-crypto-forecasting/discussion/304891
\nHope this answers your question.

", "votes": 0, "replies": [] }, { "id": 1673719, "author_name": "JM", "author_url": "", "post_date": "2022-02-02T22:54:25.600000", "content": "

Okay makes sense, I actually had this as a feature myself but not for such a long period and did not get 0.0292 by that alone.

", "votes": 0, "replies": [] }, { "id": 1674082, "author_name": "GUNER", "author_url": "", "post_date": "2022-02-03T08:06:42.463000", "content": "

Your \"minustarget\" experiment is really interesting. For your rolling stats to calculate beta ( for this specific examle ), did you select the 3750 window, starting from t-16, since they are the only available ones?

", "votes": 0, "replies": [] }, { "id": 1674105, "author_name": "Hugues", "author_url": "", "post_date": "2022-02-03T08:30:52.713000", "content": "

Good remark, you're right. Technically, the total buffer size covers the previous: (3750 + 16) minutes, to be able to use the calculated target at (t-16) as a predictor at time t.

", "votes": 0, "replies": [] } ] }, { "id": 1673688, "author_name": "", "author_url": "", "post_date": "2022-02-02T22:05:40.663000", "content": "

My predictions:
\nBronze: 0.001±1
\nSilver: 0.005±1
\nGold: 0.01±1
\nWinner: 0.015±1

\n

Unfortunately, this can destroy the leaderboard:
\n\"\"
\nExpected position in the leaderboard = 0.5 * random() ± 0.4 * (Elon Musk tweets about bitcoin) + 0.1 * skill

", "votes": 2, "replies": [] }, { "id": 1674935, "author_name": "Student", "author_url": "", "post_date": "2022-02-03T22:04:40.037000", "content": "

Winner: .048
\nGold: .036
\nSilver: .028
\nBronze: .018

", "votes": 1, "replies": [] }, { "id": 1673679, "author_name": "BiteMe", "author_url": "", "post_date": "2022-02-02T22:00:41.597000", "content": "

Winner 0.045+
\nGold 0.04
\nSilver 0.029
\nBronze 0.02

", "votes": 1, "replies": [] }, { "id": 1673195, "author_name": "cymchen", "author_url": "", "post_date": "2022-02-02T15:12:05.410000", "content": "

Bronze: 0.01
\nSilver: 0.015
\nGold: 0.025
\nWinner maybe in 0.04+

", "votes": 1, "replies": [] }, { "id": 1673537, "author_name": "Man of the year", "author_url": "", "post_date": "2022-02-02T19:57:09.693000", "content": "

I would say:
\nBronze: 0.005
\nSilver: 0.01
\nGold: 0.03

", "votes": 2, "replies": [ { "id": 1677087, "author_name": "Hantian Zheng", "author_url": "", "post_date": "2022-02-05T14:16:56.797000", "content": "

Same here.

", "votes": 0, "replies": [] }, { "id": 1677199, "author_name": "Marco Gorelli", "author_url": "", "post_date": "2022-02-05T15:19:32.057000", "content": "

so you're saying I have a chance

", "votes": 0, "replies": [] } ] }, { "id": 1673157, "author_name": "Kyle Peters", "author_url": "", "post_date": "2022-02-02T14:41:29.673000", "content": "

I think the winner will definitely be greater than .05. Most likely they will get a score between .06-.065.

", "votes": 1, "replies": [] }, { "id": 1673092, "author_name": "Yirun Zhang", "author_url": "", "post_date": "2022-02-02T14:02:56.133000", "content": "

I guess winner should be 0.04~0.05.

", "votes": 1, "replies": [] }, { "id": 1693122, "author_name": "Aditya Sharma", "author_url": "", "post_date": "2022-02-16T12:58:40.797000", "content": "

I don't care much of score. I just want myself in rank zone.

", "votes": 0, "replies": [] }, { "id": 1678922, "author_name": "N. Peker Çelik", "author_url": "", "post_date": "2022-02-06T21:11:57.563000", "content": "

Private LB is always a rollarcoaster!

", "votes": 0, "replies": [ { "id": 1681596, "author_name": "yuanzhe zhou", "author_url": "", "post_date": "2022-02-08T15:17:18.810000", "content": "

it will be retsaocrallor 📉

", "votes": 0, "replies": [] } ] }, { "id": 1673247, "author_name": "Ricardo Colomer", "author_url": "", "post_date": "2022-02-02T15:53:05.843000", "content": "

Given an infinite number of monkeys and an infinite number of submissions…..maybe 0.01

", "votes": 0, "replies": [] }, { "id": 1672953, "author_name": "yuanzhe zhou", "author_url": "", "post_date": "2022-02-02T12:05:42.193000", "content": "

winner 0.0000 😯

", "votes": 0, "replies": [ { "id": 1672967, "author_name": "Kirderf", "author_url": "", "post_date": "2022-02-02T12:15:28.310000", "content": "

Haha, oboy, a complete LB shake. Yes why not, isn't easy.

", "votes": 0, "replies": [] }, { "id": 1674514, "author_name": "marketneutral", "author_url": "", "post_date": "2022-02-03T15:07:04.423000", "content": "

I think > 0. These are 15minute returns and there are no transaction costs imposed. As such, you could certainly get > 0 (even though this might not translate to a profitable signal in the real world after taking into account market impact).

", "votes": 3, "replies": [] } ] }, { "id": 1672929, "author_name": "Maximiliano Diaz Battan", "author_url": "", "post_date": "2022-02-02T11:44:27.390000", "content": "

I hope you're right ha

", "votes": 0, "replies": [ { "id": 1672939, "author_name": "Kirderf", "author_url": "", "post_date": "2022-02-02T11:58:35.177000", "content": "

Yes maybe just wishful thinking, would be nice to listen to others guesses :)

", "votes": 0, "replies": [] } ] }, { "id": 1683165, "author_name": "", "author_url": "", "post_date": "2022-02-09T16:02:02.180000", "content": "", "votes": 1, "replies": [ { "id": 1684617, "author_name": "Kirderf", "author_url": "", "post_date": "2022-02-10T15:53:56.263000", "content": "

You mean 0.027 or 0.27? 0.27 sounds high.

", "votes": 1, "replies": [] } ] }, { "id": 1673855, "author_name": "", "author_url": "", "post_date": "2022-02-03T03:14:34.983000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1672779": "Work done, submission selection completed. Now just follow the exciting coming three month re-runs 🙂\n\nBased on your validation and testing, what do you think will be the thresholds for the different zones, starting with bronze level, silver level, gold level and the top winner's score?\n\nMy guess: Bronze 0.01, Silver 0.02, Gold 0.033, Winner 0.038\n\nUpdated with suggested levels in average posted in this topic, excluding extreme values: \n\nWinner 0.049\nGold 0.035\nSilver 0.022\nBronze 0.014", "1673705": "Good luck to everyone!! It will be very exciting to watch the results coming in. We might see a >0.06 winner?! 🔥", "1674360": "Winner: .055\nGold: . .055 - .040\nSilver: .040 - .022 (I will score .022-.023 so humbly did put myself into silver region :-)\nBronz: .0220 - .0145\n\nYou should thinkof a prize for the best guesser..", "1673190": "Agree with your guesses, most people here who think they have done well will likely end up nearer to < 0.00 in the next 3 months IMO..", "1672861": "I wouldn't make any bet here!\nI was surprised that just predicting the negative of the re-computed target for (t-16) was getting a score of 0.0292 on the LB, and 0.0288 on the latest supplemental data.\nI've had a hard time trying to get \"significantly\" better than this simple predictor (but I have no experience with forecasting...).\nAlso, I found it interesting to plot the evolution of the score of models as the horizon increases by 16 hours increments.\nFor this simple predictor, this plot looks like this for the LB test period (top) and on the latest supplemental data (bottom):\nhttps://postimg.cc/CnjdtDpf/68b82a2f\nOne can see that the score can vary significantly over the horizon (data for the beginning of the horizon is not that relevant as the scoring period is small).\n\nI just keep thinking that we just bought ourselves some lottery tickets and that future market conditions will significantly impact the performance of our models.\nGood luck to all the competitors (crossing fingers :-) )", "1673688": "My predictions:\nBronze: 0.001±1\nSilver: 0.005±1\nGold: 0.01±1\nWinner: 0.015±1\n\nUnfortunately, this can destroy the leaderboard:\n![](https://inteng-storage.s3.amazonaws.com/img/iea/BxG21M1DO9/sizes/dogecoin-value-jumps-and-dives-on-meme-inspired-dogeday_md.jpg)\nExpected position in the leaderboard = 0.5 * random() ± 0.4 * (Elon Musk tweets about bitcoin) + 0.1 * skill", "1674935": "Winner: .048\nGold: .036\nSilver: .028\nBronze: .018", "1673679": "Winner 0.045+\nGold 0.04\nSilver 0.029\nBronze 0.02", "1673195": "Bronze: 0.01 \nSilver: 0.015\nGold: 0.025\nWinner maybe in 0.04+", "1673537": "I would say:\nBronze: 0.005\nSilver: 0.01\nGold: 0.03", "1673157": "I think the winner will definitely be greater than .05. Most likely they will get a score between .06-.065. ", "1673092": "I guess winner should be 0.04~0.05.", "1693122": "I don't care much of score. I just want myself in rank zone.", "1678922": "Private LB is always a rollarcoaster!", "1673247": "Given an infinite number of monkeys and an infinite number of submissions.....maybe 0.01", "1672953": "winner 0.0000 😯", "1672929": "I hope you're right ha ", "1683165": "", "1673855": "" } }