I am very excited that kaggle finally hosted a crypto ML comp.
\nI've been trading crypto currency for almost a year, and I made huge profits and I also lose a lot.
\nSo I consider myself as a Semi-Professional Trader.
[1]Crypto Market is a highly Periodic Market，it has bull run and bear run. Markets will have totally different trends between bull and bear, so the data from bull market means nothing in bear market.
\n[2]I found that the data includes vwap, this is a very useful indicator, but what really useful is the \"vwap slope\"，vwap slope is computed by two adjacent data(vwap slope = (Day2 vwap - Day1 vwap)/(day2-day1)) , bigger slope is a sign of an incoming upward(downward) trend.
\n[3]Volumes make sense on bigger timeframe, at least Daily timeframe.
\n[4]Most of the time, the markets hang, so at least 30% of data is nonsense.you should put your eyes on those bigger trends, find their patterns.
\n[5]Support and resistence line really matters in Crypto markets, the bigger timeframe you trade the more important it is. So I suggest you should go to tradingview to find some key Sup/Res lines during the time.
\n[6]Another useful indicator is Moving Avervage/Exponatial Moving Average, you can do some research.
Some doubts.
\nAs you can see, the market has bull run and bear run, which means if the crypto markets is bullish until 2022 may 3, then the model which uses bull market data will win a lot. And those bear market data are just noises.the market is actually unpredictable, so your model is just repeating the history, if 2022 may 3th is a bull market, and your model luckly learns a lot of bull market patterns, then your model surely wins a lot.
\nSo if someone uses 3 accounts, one account uses bull market, one uses bear market and final one uses all data, then he/she will have huge advantage.
So it's just like tossing a coin, it's not ML.
", "messages": [ { "id": 1597656, "postDate": "2021-11-27T18:45:47.233Z", "content": "I am very excited that kaggle finally hosted a crypto ML comp.
\nI've been trading crypto currency for almost a year, and I made huge profits and I also lose a lot.
\nSo I consider myself as a Semi-Professional Trader.
[1]Crypto Market is a highly Periodic Market，it has bull run and bear run. Markets will have totally different trends between bull and bear, so the data from bull market means nothing in bear market.
\n[2]I found that the data includes vwap, this is a very useful indicator, but what really useful is the \"vwap slope\"，vwap slope is computed by two adjacent data(vwap slope = (Day2 vwap - Day1 vwap)/(day2-day1)) , bigger slope is a sign of an incoming upward(downward) trend.
\n[3]Volumes make sense on bigger timeframe, at least Daily timeframe.
\n[4]Most of the time, the markets hang, so at least 30% of data is nonsense.you should put your eyes on those bigger trends, find their patterns.
\n[5]Support and resistence line really matters in Crypto markets, the bigger timeframe you trade the more important it is. So I suggest you should go to tradingview to find some key Sup/Res lines during the time.
\n[6]Another useful indicator is Moving Avervage/Exponatial Moving Average, you can do some research.
Some doubts.
\nAs you can see, the market has bull run and bear run, which means if the crypto markets is bullish until 2022 may 3, then the model which uses bull market data will win a lot. And those bear market data are just noises.the market is actually unpredictable, so your model is just repeating the history, if 2022 may 3th is a bull market, and your model luckly learns a lot of bull market patterns, then your model surely wins a lot.
\nSo if someone uses 3 accounts, one account uses bull market, one uses bear market and final one uses all data, then he/she will have huge advantage.
So it's just like tossing a coin, it's not ML.
", "rawMarkdown": "I am very excited that kaggle finally hosted a crypto ML comp.\nI've been trading crypto currency for almost a year, and I made huge profits and I also lose a lot.\nSo I consider myself as a Semi-Professional Trader.\n\n[1]Crypto Market is a highly Periodic Market，it has bull run and bear run. Markets will have totally different trends between bull and bear, so the data from bull market means nothing in bear market.\n[2]I found that the data includes vwap, this is a very useful indicator, but what really useful is the \"vwap slope\"，vwap slope is computed by two adjacent data(vwap slope = (Day2 vwap - Day1 vwap)/(day2-day1)) , bigger slope is a sign of an incoming upward(downward) trend.\n[3]Volumes make sense on bigger timeframe, at least Daily timeframe.\n[4]Most of the time, the markets hang, so at least 30% of data is nonsense.you should put your eyes on those bigger trends, find their patterns.\n[5]Support and resistence line really matters in Crypto markets, the bigger timeframe you trade the more important it is. So I suggest you should go to tradingview to find some key Sup/Res lines during the time.\n[6]Another useful indicator is Moving Avervage/Exponatial Moving Average, you can do some research.\n\nSome doubts.\nAs you can see, the market has bull run and bear run, which means if the crypto markets is bullish until 2022 may 3, then the model which uses bull market data will win a lot. And those bear market data are just noises.the market is actually unpredictable, so your model is just repeating the history, if 2022 may 3th is a bull market, and your model luckly learns a lot of bull market patterns, then your model surely wins a lot.\nSo if someone uses 3 accounts, one account uses bull market, one uses bear market and final one uses all data, then he/she will have huge advantage.\n\nSo it's just like tossing a coin, it's not ML.\n\n", "votes": 14 }, { "id": 1597693, "postDate": "2021-11-27T19:35:03.760Z", "content": "When you traded crypto, how long did hold for? Because in this competition we are predicting 15 minutes in advance. However, it seems like if you bought crypto and sold after 15 minutes the transaction costs and fees would be a lot more than any returns you would get.
", "rawMarkdown": "When you traded crypto, how long did hold for? Because in this competition we are predicting 15 minutes in advance. However, it seems like if you bought crypto and sold after 15 minutes the transaction costs and fees would be a lot more than any returns you would get. ", "votes": 1, "replies": [ { "id": 1597953, "postDate": "2021-11-28T04:51:11.023Z", "content": "I usually hold for at least 4hours, I am not a Ultra-short trader. 15min's trades can be affected by higher timeframe's trend, So if you want to trade in 15 mins, you should be following the higher timeframe's trend or you probably lose this trade. From a trader's view, if we are predicting 15mins, then the 1 hour and 4hour timeframe matters more than daily timeframe.
", "rawMarkdown": "I usually hold for at least 4hours, I am not a Ultra-short trader. 15min's trades can be affected by higher timeframe's trend, So if you want to trade in 15 mins, you should be following the higher timeframe's trend or you probably lose this trade. From a trader's view, if we are predicting 15mins, then the 1 hour and 4hour timeframe matters more than daily timeframe." }, { "id": 1599931, "postDate": "2021-11-30T00:34:52.410Z", "content": "If I held a crypto position for 15 minutes, I'd probably need a bucket to yack in. I've never seen anything so insanely volatile. It makes the futures markets look like the market for government bonds. Thanks for providing the heads-up advice though, I haven't much experience with crypto relative to bonds and stock.
", "rawMarkdown": "If I held a crypto position for 15 minutes, I'd probably need a bucket to yack in. I've never seen anything so insanely volatile. It makes the futures markets look like the market for government bonds. Thanks for providing the heads-up advice though, I haven't much experience with crypto relative to bonds and stock." } ] }, { "id": 1599729, "postDate": "2021-11-29T18:27:08.540Z", "content": "Interesting discussion!
", "rawMarkdown": "Interesting discussion!" }, { "id": 1599396, "postDate": "2021-11-29T12:04:13.220Z", "content": "If you have done trading for over a year, will you be willing to share your annual return rate and sharpe ratio?
", "rawMarkdown": "If you have done trading for over a year, will you be willing to share your annual return rate and sharpe ratio?" }, { "id": 1597750, "postDate": "2021-11-27T21:19:08.573Z", "content": "Curious on your set-up. Can you tell us a bit what plateform / tools you use ?
", "rawMarkdown": "Curious on your set-up. Can you tell us a bit what plateform / tools you use ?", "replies": [ { "id": 1597954, "postDate": "2021-11-28T04:52:05.157Z", "content": "You mean trading? I usually use binance or bybit to set a trade. I use tradingview to do chart analysis.
", "rawMarkdown": "You mean trading? I usually use binance or bybit to set a trade. I use tradingview to do chart analysis.", "votes": 1 } ] }, { "id": 1607643, "postDate": "2021-12-06T02:57:59.727Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 1597693, "author_name": "Kyle Peters", "author_url": "", "post_date": "2021-11-27T19:35:03.760000", "content": "When you traded crypto, how long did hold for? Because in this competition we are predicting 15 minutes in advance. However, it seems like if you bought crypto and sold after 15 minutes the transaction costs and fees would be a lot more than any returns you would get.
", "votes": 1, "replies": [ { "id": 1597953, "author_name": "BrainFuck", "author_url": "", "post_date": "2021-11-28T04:51:11.023000", "content": "I usually hold for at least 4hours, I am not a Ultra-short trader. 15min's trades can be affected by higher timeframe's trend, So if you want to trade in 15 mins, you should be following the higher timeframe's trend or you probably lose this trade. From a trader's view, if we are predicting 15mins, then the 1 hour and 4hour timeframe matters more than daily timeframe.
", "votes": 0, "replies": [] }, { "id": 1599931, "author_name": "Ryan", "author_url": "", "post_date": "2021-11-30T00:34:52.410000", "content": "If I held a crypto position for 15 minutes, I'd probably need a bucket to yack in. I've never seen anything so insanely volatile. It makes the futures markets look like the market for government bonds. Thanks for providing the heads-up advice though, I haven't much experience with crypto relative to bonds and stock.
", "votes": 0, "replies": [] } ] }, { "id": 1599729, "author_name": "Manzil Khan", "author_url": "", "post_date": "2021-11-29T18:27:08.540000", "content": "Interesting discussion!
", "votes": 0, "replies": [] }, { "id": 1599396, "author_name": "Johnp", "author_url": "", "post_date": "2021-11-29T12:04:13.220000", "content": "If you have done trading for over a year, will you be willing to share your annual return rate and sharpe ratio?
", "votes": 0, "replies": [] }, { "id": 1597750, "author_name": "Lucas Morin", "author_url": "", "post_date": "2021-11-27T21:19:08.573000", "content": "Curious on your set-up. Can you tell us a bit what plateform / tools you use ?
", "votes": 0, "replies": [ { "id": 1597954, "author_name": "BrainFuck", "author_url": "", "post_date": "2021-11-28T04:52:05.157000", "content": "You mean trading? I usually use binance or bybit to set a trade. I use tradingview to do chart analysis.
", "votes": 1, "replies": [] } ] }, { "id": 1607643, "author_name": "", "author_url": "", "post_date": "2021-12-06T02:57:59.727000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1597656": "I am very excited that kaggle finally hosted a crypto ML comp.\nI've been trading crypto currency for almost a year, and I made huge profits and I also lose a lot.\nSo I consider myself as a Semi-Professional Trader.\n\n[1]Crypto Market is a highly Periodic Market，it has bull run and bear run. Markets will have totally different trends between bull and bear, so the data from bull market means nothing in bear market.\n[2]I found that the data includes vwap, this is a very useful indicator, but what really useful is the \"vwap slope\"，vwap slope is computed by two adjacent data(vwap slope = (Day2 vwap - Day1 vwap)/(day2-day1)) , bigger slope is a sign of an incoming upward(downward) trend.\n[3]Volumes make sense on bigger timeframe, at least Daily timeframe.\n[4]Most of the time, the markets hang, so at least 30% of data is nonsense.you should put your eyes on those bigger trends, find their patterns.\n[5]Support and resistence line really matters in Crypto markets, the bigger timeframe you trade the more important it is. So I suggest you should go to tradingview to find some key Sup/Res lines during the time.\n[6]Another useful indicator is Moving Avervage/Exponatial Moving Average, you can do some research.\n\nSome doubts.\nAs you can see, the market has bull run and bear run, which means if the crypto markets is bullish until 2022 may 3, then the model which uses bull market data will win a lot. And those bear market data are just noises.the market is actually unpredictable, so your model is just repeating the history, if 2022 may 3th is a bull market, and your model luckly learns a lot of bull market patterns, then your model surely wins a lot.\nSo if someone uses 3 accounts, one account uses bull market, one uses bear market and final one uses all data, then he/she will have huge advantage.\n\nSo it's just like tossing a coin, it's not ML.\n\n", "1597693": "When you traded crypto, how long did hold for? Because in this competition we are predicting 15 minutes in advance. However, it seems like if you bought crypto and sold after 15 minutes the transaction costs and fees would be a lot more than any returns you would get. ", "1599729": "Interesting discussion!", "1599396": "If you have done trading for over a year, will you be willing to share your annual return rate and sharpe ratio?", "1597750": "Curious on your set-up. Can you tell us a bit what plateform / tools you use ?", "1607643": "" } }