{ "id": 58725, "title": "Looking for a team (0.2195 LightGBM single models)", "url": "/competitions/avito-demand-prediction/discussion/58725", "author_name": "", "post_date": "2018-06-13T03:25:06.427010500Z", "votes": 5, "comment_count": 23, "views": 0, "content": "

Hi Kagglers.
\nI am looking to join a team which has a good chance of getting gold.
\n
\nMy model is a single model of LightGBM which scores 0.2199 on the LB.
\nWith 5 seed-avg runs, it scores 0.2195, which is comparable to the sota 0.2190.
\nI guess I have exhausted the tabular features, but lacking some features on images,
\nand potentially text-features(like translation)
\n
\nSince my model is a LightGBM model, I am looking for teams with strong NN models,
\nand/or teams with strong image/text modelling but lacking tabular features.
\n
\nI live in Japan, which is JST(UTC+9), and can communicate in English/Japanese.
\nI am currently a Kaggle Novice,
\nbut I hope my past results(2gold medals) speak for my LightGBM modelling ability.
\n
\nThank you for reading, and good luck to everyone in this competition :)

", "messages": [ { "id": "342175", "postDate": "06/13/2018 03:25:06", "content": "

Hi Kagglers.
\nI am looking to join a team which has a good chance of getting gold.
\n
\nMy model is a single model of LightGBM which scores 0.2199 on the LB.
\nWith 5 seed-avg runs, it scores 0.2195, which is comparable to the sota 0.2190.
\nI guess I have exhausted the tabular features, but lacking some features on images,
\nand potentially text-features(like translation)
\n
\nSince my model is a LightGBM model, I am looking for teams with strong NN models,
\nand/or teams with strong image/text modelling but lacking tabular features.
\n
\nI live in Japan, which is JST(UTC+9), and can communicate in English/Japanese.
\nI am currently a Kaggle Novice,
\nbut I hope my past results(2gold medals) speak for my LightGBM modelling ability.
\n
\nThank you for reading, and good luck to everyone in this competition :)

", "rawMarkdown": "Hi Kagglers.
\nI am looking to join a team which has a good chance of getting gold.
\n
\nMy model is a single model of LightGBM which scores 0.2199 on the LB.
\nWith 5 seed-avg runs, it scores 0.2195, which is comparable to the sota 0.2190.
\nI guess I have exhausted the tabular features, but lacking some features on images,
\nand potentially text-features(like translation)
\n
\nSince my model is a LightGBM model, I am looking for teams with strong NN models,
\nand/or teams with strong image/text modelling but lacking tabular features.
\n
\nI live in Japan, which is JST(UTC+9), and can communicate in English/Japanese.
\nI am currently a Kaggle Novice,
\nbut I hope my past results(2gold medals) speak for my LightGBM modelling ability.
\n
\nThank you for reading, and good luck to everyone in this competition :)", "votes": null }, { "id": "342212", "postDate": "06/13/2018 05:38:42", "content": "

I have enough computer resource... and can speak Japanese.

", "rawMarkdown": "I have enough computer resource... and can speak Japanese.", "votes": null }, { "id": "342213", "postDate": "06/13/2018 05:42:12", "content": "

We(onodera and me) can manage a lot of computing resource, and variational data for this competition. \nIf you consider merge group, we will spend data and machine resources.

", "rawMarkdown": "We(onodera and me) can manage a lot of computing resource, and variational data for this competition. \nIf you consider merge group, we will spend data and machine resources.", "votes": null }, { "id": "342226", "postDate": "06/13/2018 06:06:56", "content": "

I have neural nework models(0.2206) but I did not create good light gbm.(0.2221)\nIf you merge me, we will get the gold.

\n\n

note:I can speak Japanese.

", "rawMarkdown": "I have neural nework models(0.2206) but I did not create good light gbm.(0.2221)\nIf you merge me, we will get the gold.\n\nnote:I can speak Japanese.", "votes": null }, { "id": "342322", "postDate": "06/13/2018 10:23:29", "content": "

I also have GPU resources. I think that I can extract features from the image somewhat. Can I join the team too?

", "rawMarkdown": "I also have GPU resources. I think that I can extract features from the image somewhat. Can I join the team too?", "votes": null }, { "id": "342393", "postDate": "06/13/2018 12:35:15", "content": "

Hi nardtree, you are my Idol.
\nI think I have enough resource, since I am always ready to spend a few thousand dollars on GCP.
\nMaybe next time when I am a Master :D

", "rawMarkdown": "Hi nardtree, you are my Idol.
\nI think I have enough resource, since I am always ready to spend a few thousand dollars on GCP.
\nMaybe next time when I am a Master :D", "votes": null }, { "id": "342394", "postDate": "06/13/2018 12:36:10", "content": "

Hi Keiku-san.
\nThank you for you offer.
\nHowever, I think I have enough resource, since I am always ready to spend a few thousand dollars on GCP.
\nHope to see you more often in future competitions :)

", "rawMarkdown": "Hi Keiku-san.
\nThank you for you offer.
\nHowever, I think I have enough resource, since I am always ready to spend a few thousand dollars on GCP.
\nHope to see you more often in future competitions :)", "votes": null }, { "id": "342395", "postDate": "06/13/2018 12:37:45", "content": "

Hi ONODERA,
\nI am sure you are very good at Japanese, but I need an actual model to advance in the competition :D

", "rawMarkdown": "Hi ONODERA,
\nI am sure you are very good at Japanese, but I need an actual model to advance in the competition :D", "votes": null }, { "id": "342399", "postDate": "06/13/2018 12:41:12", "content": "

Hi tereka,
\nThank you for offering an actual model :)
\nAs I have PMed you on twitter,
\nI am considering all offers until at least tomorrow at the earliest, so thank you for your patience.

", "rawMarkdown": "Hi tereka,
\nThank you for offering an actual model :)
\nAs I have PMed you on twitter,
\nI am considering all offers until at least tomorrow at the earliest, so thank you for your patience.", "votes": null }, { "id": "342405", "postDate": "06/13/2018 12:49:37", "content": "

Hello,pocket .You can join our team if you want.Now we have strong nn's for text+image ,but dont have strong numerical features.We can perfectly complement each other

", "rawMarkdown": "Hello,pocket .You can join our team if you want.Now we have strong nn's for text+image ,but dont have strong numerical features.We can perfectly complement each other", "votes": null }, { "id": "342766", "postDate": "06/14/2018 04:16:08", "content": "

Are you really spending thousands of dollars for competitions? Wow. Is that the norm for high rankers?

", "rawMarkdown": "Are you really spending thousands of dollars for competitions? Wow. Is that the norm for high rankers?", "votes": null }, { "id": "342767", "postDate": "06/14/2018 04:19:58", "content": "

@Lem Lordje Ko: I don't know what the norm is, but it is certainly possible to get a gold medal (or even a #1 ranking) without spending that much money.

", "rawMarkdown": "Lem Lordje Ko: I don't know what the norm is, but it is certainly possible to get a gold medal (or even a #1 ranking) without spending that much money.", "votes": null }, { "id": "342769", "postDate": "06/14/2018 04:21:43", "content": "

Actually, I spent 2k dollars for the last Talking Data competition.

", "rawMarkdown": "Actually, I spent 2k dollars for the last Talking Data competition.", "votes": null }, { "id": "342806", "postDate": "06/14/2018 06:34:02", "content": "

o.O

", "rawMarkdown": "o.O", "votes": null }, { "id": "342858", "postDate": "06/14/2018 08:31:39", "content": "

Hello,

\n\n

I have a NN language model to extract features from the description and the title. I have not spent time on building a structured model as my key focus was to improve my NLP skills. If you are interested in learning more drop me a note.

\n\n

Thanks

", "rawMarkdown": "Hello,\n\nI have a NN language model to extract features from the description and the title. I have not spent time on building a structured model as my key focus was to improve my NLP skills. If you are interested in learning more drop me a note.\n\nThanks", "votes": null }, { "id": "342859", "postDate": "06/14/2018 08:34:04", "content": "

Wow .. assembling a desktop might be more economical ... in 5K dollars you can build a really good machine.

", "rawMarkdown": "Wow .. assembling a desktop might be more economical ... in 5K dollars you can build a really good machine.", "votes": null }, { "id": "342867", "postDate": "06/14/2018 08:58:47", "content": "

I don’t think so. What if it has been broken? Appropriate memory and cpu for the motherboard? We have a lot of things to consider to set up those kind of stuff.

\n\n

If you don’t have enough resources and want me to help, please summon me ;)

", "rawMarkdown": "I don’t think so. What if it has been broken? Appropriate memory and cpu for the motherboard? We have a lot of things to consider to set up those kind of stuff.\n\nIf you don’t have enough resources and want me to help, please summon me ;)", "votes": null }, { "id": "342893", "postDate": "06/14/2018 10:45:06", "content": "
\n

I don’t think so. What if it has been broken? Appropriate memory and cpu for the motherboard? We have a lot of things to consider to set up those kind of stuff.

\n
\n\n

Yes, assembling one's own machine means involves taking responsibility of maintenance, I guess that's why cloud services are costly.

\n\n
\n

If you don’t have enough resources and want me to help, please summon me ;)

\n
\n\n

Will remember to reach out to you in future when I am low on resources :)

\n\n

Also, I assembled my own machine, so if you ever plan on doing that you can reach out to me :)

", "rawMarkdown": "> I don’t think so. What if it has been broken? Appropriate memory and cpu for the motherboard? We have a lot of things to consider to set up those kind of stuff.\n\nYes, assembling one's own machine means involves taking responsibility of maintenance, I guess that's why cloud services are costly.\n\n> If you don’t have enough resources and want me to help, please summon me ;)\n\nWill remember to reach out to you in future when I am low on resources :)\n\nAlso, I assembled my own machine, so if you ever plan on doing that you can reach out to me :)", "votes": null }, { "id": "342976", "postDate": "06/14/2018 13:18:49", "content": "

Hi Arthur, thank you for your merge offer.
\nI have decided to merge with tereka, since he has a good NN model already,
\nand I really want to learn a lot from him to open up my game.
\nMaybe we can merge later on in the competition, when we are both close to gold :)
\nGood luck in this competition!

", "rawMarkdown": "Hi Arthur, thank you for your merge offer.
\nI have decided to merge with tereka, since he has a good NN model already,
\nand I really want to learn a lot from him to open up my game.
\nMaybe we can merge later on in the competition, when we are both close to gold :)
\nGood luck in this competition!
", "votes": null }, { "id": "342977", "postDate": "06/14/2018 13:22:23", "content": "

Hi Guillaume,
\nThank you for your offer.
\nHowever, it is hard for me to determine how good your features are, without a model on the LB.
\nI have decided to merge with tereka for now.
\nGood luck to you in this competition :)

", "rawMarkdown": "Hi Guillaume,
\nThank you for your offer.
\nHowever, it is hard for me to determine how good your features are, without a model on the LB.
\nI have decided to merge with tereka for now.
\nGood luck to you in this competition :)", "votes": null }, { "id": "342985", "postDate": "06/14/2018 13:34:36", "content": "

Hi all. \nI just wanted to comment that spending a few thousand dollars is certainly not the norm
\n (atleast not in tabular data competitions), nor is it necessary.
\nMy first gold medal was won by my local machine(16GB RAM).
\nThe next one was I used GCP, which cost me a few hundred dollars.

\n\n

I am just stating that I am ready to pull the trigger whenever necessary.
\nEveryone is in a different situation, so I will not advice anyone to do so.

\n\n

I am really LOVING machine learning at the moment, so if I can reduce the learning curve just by spending money, I am ready to do so.
\nIt is also worth to note that I am a professional software engineer, and I spend a few hundred hours per competition, which is potentially worth a few thousand dollars in terms of opportunity cost anyways....

", "rawMarkdown": "Hi all. \nI just wanted to comment that spending a few thousand dollars is certainly not the norm
\n (atleast not in tabular data competitions), nor is it necessary.
\nMy first gold medal was won by my local machine(16GB RAM).
\nThe next one was I used GCP, which cost me a few hundred dollars.
\n\nI am just stating that I am ready to pull the trigger whenever necessary.
\nEveryone is in a different situation, so I will not advice anyone to do so.
\n\nI am really LOVING machine learning at the moment, so if I can reduce the learning curve just by spending money, I am ready to do so.
\nIt is also worth to note that I am a professional software engineer, and I spend a few hundred hours per competition, which is potentially worth a few thousand dollars in terms of opportunity cost anyways....
", "votes": null }, { "id": "343083", "postDate": "06/14/2018 17:08:12", "content": "

oh makes sense. thanks for the insight!

", "rawMarkdown": "oh makes sense. thanks for the insight!", "votes": null }, { "id": "343729", "postDate": "06/16/2018 00:07:38", "content": "

@onodera, dear friend, can I summon you in the future? ^_^

", "rawMarkdown": "onodera, dear friend, can I summon you in the future? ^_^", "votes": null }, { "id": "343770", "postDate": "06/16/2018 03:19:06", "content": "

Yes of course!!!

", "rawMarkdown": "Yes of course!!!", "votes": null } ], "comments": [ { "id": 342212, "author_name": "onodera", "author_url": "", "post_date": "06/13/2018 05:38:42", "content": "

I have enough computer resource... and can speak Japanese.

", "votes": null, "replies": [ { "id": 342395, "author_name": "pocketsuteado", "author_url": "", "post_date": "06/13/2018 12:37:45", "content": "

Hi ONODERA,
\nI am sure you are very good at Japanese, but I need an actual model to advance in the competition :D

", "votes": null, "replies": [] } ] }, { "id": 342213, "author_name": "nardtree", "author_url": "", "post_date": "06/13/2018 05:42:12", "content": "

We(onodera and me) can manage a lot of computing resource, and variational data for this competition. \nIf you consider merge group, we will spend data and machine resources.

", "votes": null, "replies": [ { "id": 342393, "author_name": "pocketsuteado", "author_url": "", "post_date": "06/13/2018 12:35:15", "content": "

Hi nardtree, you are my Idol.
\nI think I have enough resource, since I am always ready to spend a few thousand dollars on GCP.
\nMaybe next time when I am a Master :D

", "votes": null, "replies": [] } ] }, { "id": 342226, "author_name": "tereka", "author_url": "", "post_date": "06/13/2018 06:06:56", "content": "

I have neural nework models(0.2206) but I did not create good light gbm.(0.2221)\nIf you merge me, we will get the gold.

\n\n

note:I can speak Japanese.

", "votes": null, "replies": [ { "id": 342399, "author_name": "pocketsuteado", "author_url": "", "post_date": "06/13/2018 12:41:12", "content": "

Hi tereka,
\nThank you for offering an actual model :)
\nAs I have PMed you on twitter,
\nI am considering all offers until at least tomorrow at the earliest, so thank you for your patience.

", "votes": null, "replies": [] } ] }, { "id": 342322, "author_name": "keiku322", "author_url": "", "post_date": "06/13/2018 10:23:29", "content": "

I also have GPU resources. I think that I can extract features from the image somewhat. Can I join the team too?

", "votes": null, "replies": [ { "id": 342394, "author_name": "pocketsuteado", "author_url": "", "post_date": "06/13/2018 12:36:10", "content": "

Hi Keiku-san.
\nThank you for you offer.
\nHowever, I think I have enough resource, since I am always ready to spend a few thousand dollars on GCP.
\nHope to see you more often in future competitions :)

", "votes": null, "replies": [] }, { "id": 342766, "author_name": "lemonkoala", "author_url": "", "post_date": "06/14/2018 04:16:08", "content": "

Are you really spending thousands of dollars for competitions? Wow. Is that the norm for high rankers?

", "votes": null, "replies": [] }, { "id": 342767, "author_name": "peterhurford", "author_url": "", "post_date": "06/14/2018 04:19:58", "content": "

@Lem Lordje Ko: I don't know what the norm is, but it is certainly possible to get a gold medal (or even a #1 ranking) without spending that much money.

", "votes": null, "replies": [] }, { "id": 342769, "author_name": "onodera", "author_url": "", "post_date": "06/14/2018 04:21:43", "content": "

Actually, I spent 2k dollars for the last Talking Data competition.

", "votes": null, "replies": [] }, { "id": 342806, "author_name": "christofhenkel", "author_url": "", "post_date": "06/14/2018 06:34:02", "content": "

o.O

", "votes": null, "replies": [] }, { "id": 342859, "author_name": "tezdhar", "author_url": "", "post_date": "06/14/2018 08:34:04", "content": "

Wow .. assembling a desktop might be more economical ... in 5K dollars you can build a really good machine.

", "votes": null, "replies": [] }, { "id": 342867, "author_name": "onodera", "author_url": "", "post_date": "06/14/2018 08:58:47", "content": "

I don’t think so. What if it has been broken? Appropriate memory and cpu for the motherboard? We have a lot of things to consider to set up those kind of stuff.

\n\n

If you don’t have enough resources and want me to help, please summon me ;)

", "votes": null, "replies": [] }, { "id": 342893, "author_name": "tezdhar", "author_url": "", "post_date": "06/14/2018 10:45:06", "content": "
\n

I don’t think so. What if it has been broken? Appropriate memory and cpu for the motherboard? We have a lot of things to consider to set up those kind of stuff.

\n
\n\n

Yes, assembling one's own machine means involves taking responsibility of maintenance, I guess that's why cloud services are costly.

\n\n
\n

If you don’t have enough resources and want me to help, please summon me ;)

\n
\n\n

Will remember to reach out to you in future when I am low on resources :)

\n\n

Also, I assembled my own machine, so if you ever plan on doing that you can reach out to me :)

", "votes": null, "replies": [] }, { "id": 342985, "author_name": "pocketsuteado", "author_url": "", "post_date": "06/14/2018 13:34:36", "content": "

Hi all. \nI just wanted to comment that spending a few thousand dollars is certainly not the norm
\n (atleast not in tabular data competitions), nor is it necessary.
\nMy first gold medal was won by my local machine(16GB RAM).
\nThe next one was I used GCP, which cost me a few hundred dollars.

\n\n

I am just stating that I am ready to pull the trigger whenever necessary.
\nEveryone is in a different situation, so I will not advice anyone to do so.

\n\n

I am really LOVING machine learning at the moment, so if I can reduce the learning curve just by spending money, I am ready to do so.
\nIt is also worth to note that I am a professional software engineer, and I spend a few hundred hours per competition, which is potentially worth a few thousand dollars in terms of opportunity cost anyways....

", "votes": null, "replies": [] }, { "id": 343083, "author_name": "lemonkoala", "author_url": "", "post_date": "06/14/2018 17:08:12", "content": "

oh makes sense. thanks for the insight!

", "votes": null, "replies": [] }, { "id": 343729, "author_name": "yifanxie", "author_url": "", "post_date": "06/16/2018 00:07:38", "content": "

@onodera, dear friend, can I summon you in the future? ^_^

", "votes": null, "replies": [] }, { "id": 343770, "author_name": "onodera", "author_url": "", "post_date": "06/16/2018 03:19:06", "content": "

Yes of course!!!

", "votes": null, "replies": [] } ] }, { "id": 342405, "author_name": "kingarthur7", "author_url": "", "post_date": "06/13/2018 12:49:37", "content": "

Hello,pocket .You can join our team if you want.Now we have strong nn's for text+image ,but dont have strong numerical features.We can perfectly complement each other

", "votes": null, "replies": [ { "id": 342976, "author_name": "pocketsuteado", "author_url": "", "post_date": "06/14/2018 13:18:49", "content": "

Hi Arthur, thank you for your merge offer.
\nI have decided to merge with tereka, since he has a good NN model already,
\nand I really want to learn a lot from him to open up my game.
\nMaybe we can merge later on in the competition, when we are both close to gold :)
\nGood luck in this competition!

", "votes": null, "replies": [] } ] }, { "id": 342858, "author_name": "z7700903", "author_url": "", "post_date": "06/14/2018 08:31:39", "content": "

Hello,

\n\n

I have a NN language model to extract features from the description and the title. I have not spent time on building a structured model as my key focus was to improve my NLP skills. If you are interested in learning more drop me a note.

\n\n

Thanks

", "votes": null, "replies": [ { "id": 342977, "author_name": "pocketsuteado", "author_url": "", "post_date": "06/14/2018 13:22:23", "content": "

Hi Guillaume,
\nThank you for your offer.
\nHowever, it is hard for me to determine how good your features are, without a model on the LB.
\nI have decided to merge with tereka for now.
\nGood luck to you in this competition :)

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "342175": "Hi Kagglers.
\nI am looking to join a team which has a good chance of getting gold.
\n
\nMy model is a single model of LightGBM which scores 0.2199 on the LB.
\nWith 5 seed-avg runs, it scores 0.2195, which is comparable to the sota 0.2190.
\nI guess I have exhausted the tabular features, but lacking some features on images,
\nand potentially text-features(like translation)
\n
\nSince my model is a LightGBM model, I am looking for teams with strong NN models,
\nand/or teams with strong image/text modelling but lacking tabular features.
\n
\nI live in Japan, which is JST(UTC+9), and can communicate in English/Japanese.
\nI am currently a Kaggle Novice,
\nbut I hope my past results(2gold medals) speak for my LightGBM modelling ability.
\n
\nThank you for reading, and good luck to everyone in this competition :)", "342212": "I have enough computer resource... and can speak Japanese.", "342213": "We(onodera and me) can manage a lot of computing resource, and variational data for this competition. \nIf you consider merge group, we will spend data and machine resources.", "342226": "I have neural nework models(0.2206) but I did not create good light gbm.(0.2221)\nIf you merge me, we will get the gold.\n\nnote:I can speak Japanese.", "342322": "I also have GPU resources. I think that I can extract features from the image somewhat. Can I join the team too?", "342393": "Hi nardtree, you are my Idol.
\nI think I have enough resource, since I am always ready to spend a few thousand dollars on GCP.
\nMaybe next time when I am a Master :D", "342394": "Hi Keiku-san.
\nThank you for you offer.
\nHowever, I think I have enough resource, since I am always ready to spend a few thousand dollars on GCP.
\nHope to see you more often in future competitions :)", "342395": "Hi ONODERA,
\nI am sure you are very good at Japanese, but I need an actual model to advance in the competition :D", "342399": "Hi tereka,
\nThank you for offering an actual model :)
\nAs I have PMed you on twitter,
\nI am considering all offers until at least tomorrow at the earliest, so thank you for your patience.", "342405": "Hello,pocket .You can join our team if you want.Now we have strong nn's for text+image ,but dont have strong numerical features.We can perfectly complement each other", "342766": "Are you really spending thousands of dollars for competitions? Wow. Is that the norm for high rankers?", "342767": "Lem Lordje Ko: I don't know what the norm is, but it is certainly possible to get a gold medal (or even a #1 ranking) without spending that much money.", "342769": "Actually, I spent 2k dollars for the last Talking Data competition.", "342806": "o.O", "342858": "Hello,\n\nI have a NN language model to extract features from the description and the title. I have not spent time on building a structured model as my key focus was to improve my NLP skills. If you are interested in learning more drop me a note.\n\nThanks", "342859": "Wow .. assembling a desktop might be more economical ... in 5K dollars you can build a really good machine.", "342867": "I don’t think so. What if it has been broken? Appropriate memory and cpu for the motherboard? We have a lot of things to consider to set up those kind of stuff.\n\nIf you don’t have enough resources and want me to help, please summon me ;)", "342893": "> I don’t think so. What if it has been broken? Appropriate memory and cpu for the motherboard? We have a lot of things to consider to set up those kind of stuff.\n\nYes, assembling one's own machine means involves taking responsibility of maintenance, I guess that's why cloud services are costly.\n\n> If you don’t have enough resources and want me to help, please summon me ;)\n\nWill remember to reach out to you in future when I am low on resources :)\n\nAlso, I assembled my own machine, so if you ever plan on doing that you can reach out to me :)", "342976": "Hi Arthur, thank you for your merge offer.
\nI have decided to merge with tereka, since he has a good NN model already,
\nand I really want to learn a lot from him to open up my game.
\nMaybe we can merge later on in the competition, when we are both close to gold :)
\nGood luck in this competition!
", "342977": "Hi Guillaume,
\nThank you for your offer.
\nHowever, it is hard for me to determine how good your features are, without a model on the LB.
\nI have decided to merge with tereka for now.
\nGood luck to you in this competition :)", "342985": "Hi all. \nI just wanted to comment that spending a few thousand dollars is certainly not the norm
\n (atleast not in tabular data competitions), nor is it necessary.
\nMy first gold medal was won by my local machine(16GB RAM).
\nThe next one was I used GCP, which cost me a few hundred dollars.
\n\nI am just stating that I am ready to pull the trigger whenever necessary.
\nEveryone is in a different situation, so I will not advice anyone to do so.
\n\nI am really LOVING machine learning at the moment, so if I can reduce the learning curve just by spending money, I am ready to do so.
\nIt is also worth to note that I am a professional software engineer, and I spend a few hundred hours per competition, which is potentially worth a few thousand dollars in terms of opportunity cost anyways....
", "343083": "oh makes sense. thanks for the insight!", "343729": "onodera, dear friend, can I summon you in the future? ^_^", "343770": "Yes of course!!!" }, "source": "meta" }