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
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.
- 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.
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.
This is a really cool idea. I wonder if the play calling is a pure democracy or if they have an algo to make good play callers have more sway.
They could do something like, for every play I vote for that wins, the results of that play add to the weight of my vote. So, for a sack, interception, etc, I lose weight, but I gain weight for every play with positive yardage.
Obviously that's a bit naïve, but you get the idea.
However weights are assigned, they could carry over from game to game and lead to the best play callers in the crowd having the most sway. It would also solve the problem of non-fans trying to sabotage the game.
Having an algorithm that increases your ability to call plays by giving your votes more weight based on number of correct plays would be really interesting. Make it biased so that you need to maintain a significant accuracy to get a vote above 1 but generally manage the voting so that the shrewd plays get made more often... that'd be really neat.
Good point. Instead of just achieving "engagement" from the fans, they could attempt to crowdsource the best calls.
They could also extend the tracking / rating of play calls to fans at home. Their votes wouldn't have influence like fans at the game, but award / subtract points just the same if their calls succeed. Then when they do attend games, their rating already exists.
At home vs at game could be determined by having purchased a ticket.
They could even use that to monetize people who want to vote but don't want to go to the game (and even monetize away games). Purchase of an at-game ticket comes with a free voters pass. Or can can buy a voter pass on a game-by-game basis or a season pass.
In the 1980s QUBE cable had a similar system where people at home could vote on plays in a football game. I have a vague memory of doing this once at a friend's house ... they had this new thing called cable TV.
"QUBE tried a wide array of program formats for two-way participation, including games, talent contests, astrology shows, exercise classes, interactive games, football games in which viewers called the plays, town meetings and public hearings involving Federal agencies."
Norm Macdonald is apparently a part-owner of this team and was even doing some play-by-play / color commentary from the booth, to the great dismay of fans of his podcast.
I like the idea but couldn't this somewhat easily be 'hacked' with a botnet type of thing allowing one crafty person to basically take over and play-call circumventing the crowd?
Taking it further couldn't this person make poor calls for the opposing team, aka punting on first down... seems like they gloss over the challenge of preventing this type of exploit but maybe it's not as easy as it seems?
The threat of a botnet is something any service that depends on popular vote has to overcome; it'll take engineering and operations effort, but it's not a showstopper.
As for the hostile vote exploit, remember that a play is "bad" mostly because it doesn't work well against the opponent's strategy, not because it's inherently bad (if it were, it wouldn't be listed as an option by the team). Making this exploit work requires you to know the results of your team's vote ahead of time. My guess is that it doesn't reveal vote results until after the play, but I haven't used the app so I wouldn't know.
The article mentions that the vote results are only revealed to the coach, who speaks to the QB over the headset to tell him what play he should run. As you surmised, the vote results are revealed to the app users after the play is run. This is also to prevent the opposing defensive coordinator from logging into the app, seeing the results and countering the play.
