4 ms·

Hey folks, I am the back-end/infra dev on this project, AMA. It's built with a combination of Elixir/Phoenix, Rails, and React (+Native). If you'd like to work

by hakunin 10y ago

Hey folks, I am the back-end/infra dev on this project, AMA. It's built with a combination of Elixir/Phoenix, Rails, and React (+Native). If you'd like to work with me on stuff like this feel free to apply by shooting us an email with some info about yourself. You can find the contact at http://www.crossfield.com http://www.crossfield.com

rbobby 10y ago
Pretty cool idea and the implementation seems on target too. Just a quick note on one possible way to adjust votes: weighting by success. A vote from a fan that consistently votes for a successful play (defined as 0 or more yards gained?) should count for more. A vote from a fan with a bad record should count for less (though maybe just keep it at 1 as the minimum and favor success more). This might also help with gaming the system by opponents (especially if non-successful votes move you down faster than successful votes move you up). This could also lead to fan promotion opportunities... "fan of the game" or "best 5 fans of the game"... "fan of the season"... etc.
- SomeStupidPoint 10y ago
  You should go from 1 to -1; being consistently wrong is useful information, being randomly right or wrong is not.
  - munin 10y ago
    you probably want this? https://en.wikipedia.org/wiki/Brier_score https://en.wikipedia.org/wiki/Brier_score
- Someone 10y ago
  Good idea, but I see a fee problems: - it is hard to define "successful play". "defined as 0 or more yards gained" doesn't cut it, as it will reward low-risk plays, and those sometimes are very bad. For example, with two seconds on the clock, 80 yards to go, and 5 points behind, gaining 1 or even 20 yards buys you nothing. - I can still see adversaries game the system. Say I'm a fan of club C. Then, my strategy would be to gain reputation when this club plays anything but me, and then to use the hard-won reputation for a few crucial plays when this team plays my team. I won't get become the best fan, but if I have enough friends who do this, too, this will still make an impact. - majority voting has two problems. Firstly, I think it will lead to dull plays; secondly, I think there will be a problem keeping fans who have minority opinions interested, and I fear that will be the majority of fans. You could prevent that by increasing a voter's weight whenever their choice isn't the winning one, and resetting it to the 'real' value when it is. That allows both the genius, the maverick, and the fool with minority opinions to occasionally have their choice picked.
cagenut 10y ago
would greatly appreciate as much detail as you're willing to type about how you're hosting and deploying your elixir/phoenix codebase
- hakunin 10y ago
  We went with good old Heroku and Redis for the real-time Elixir component. It isn't a full-blown OTP setup by far, but we just needed stable performance with many websocket connections, for which it didn't even break a sweat. We have set up OTP clusters for other projects however, but that's a story for another day.
f_allwein 10y ago
There's an interesting question by user IanDrake further down on this page on how the voting works. Can you address that?

by hakunin 10y ago

rbobby 10y ago

Pretty cool idea and the implementation seems on target too. Just a quick note on one possible way to adjust votes: weighting by success. A vote from a fan that consistently votes for a successful play (defined as 0 or more yards gained?) should count for more. A vote from a fan with a bad record should count for less (though maybe just keep it at 1 as the minimum and favor success more). This might also help with gaming the system by opponents (especially if non-successful votes move you down faster than successful votes move you up). This could also lead to fan promotion opportunities... "fan of the game" or "best 5 fans of the game"... "fan of the season"... etc.

SomeStupidPoint 10y ago
You should go from 1 to -1; being consistently wrong is useful information, being randomly right or wrong is not.
- munin 10y ago
  you probably want this? https://en.wikipedia.org/wiki/Brier_score https://en.wikipedia.org/wiki/Brier_score
Someone 10y ago
Good idea, but I see a fee problems: - it is hard to define "successful play". "defined as 0 or more yards gained" doesn't cut it, as it will reward low-risk plays, and those sometimes are very bad. For example, with two seconds on the clock, 80 yards to go, and 5 points behind, gaining 1 or even 20 yards buys you nothing. - I can still see adversaries game the system. Say I'm a fan of club C. Then, my strategy would be to gain reputation when this club plays anything but me, and then to use the hard-won reputation for a few crucial plays when this team plays my team. I won't get become the best fan, but if I have enough friends who do this, too, this will still make an impact. - majority voting has two problems. Firstly, I think it will lead to dull plays; secondly, I think there will be a problem keeping fans who have minority opinions interested, and I fear that will be the majority of fans. You could prevent that by increasing a voter's weight whenever their choice isn't the winning one, and resetting it to the 'real' value when it is. That allows both the genius, the maverick, and the fool with minority opinions to occasionally have their choice picked.

cagenut 10y ago

would greatly appreciate as much detail as you're willing to type about how you're hosting and deploying your elixir/phoenix codebase

hakunin 10y ago
We went with good old Heroku and Redis for the real-time Elixir component. It isn't a full-blown OTP setup by far, but we just needed stable performance with many websocket connections, for which it didn't even break a sweat. We have set up OTP clusters for other projects however, but that's a story for another day.

f_allwein 10y ago

There's an interesting question by user IanDrake further down on this page on how the voting works. Can you address that?