The Harvard Network Accessible Dartboard

These guys hacked a dartboard to serve scores over a wireless network. There is an OpenGL client that grabs the scores, runs the games, stores the results in a database, and suggests moves based on player's past performance. On top of all this, the client looks exactly like the dartboard, so it can be projected over the real thing.

Comments

Two things we can notice here:

[ksp]

1)
"Pedro V. Sander" desperately needs to get a life according to the stats!

2)
You can be no good at darts as long as you play alone, it is easy to cheat since the PC never notices when you miss the board altogether.

BTW: Pedro, I am not suggesting that there is a link between 1) and 2).

Security concerns, anti-cheat mechanisms

[Xerion]

Seems like every great game will eventually be ruined by cheaters, see Starcraft, CounterStrike, ARC... I wonder if this wireless dartboard can detect cheaters (e.g stand in front of the board and stab it point blank, in order to up your stats). Hope the AI is smart enough to /ban such users. And maybe literally "kick" them by ways of a 10 ton hydrolic metal boot nailed behind the player's spot.

Re:I want one of those!

[rm-r]

This could easily be done on a monochromed and thresholded (thresheld?) image to give white balls on a black background and then using Hough transforms to give the ball positions. The area detected could then be sampled from the original image to obtain the ball colour. This is cool! If only I had a pool table so I could do this ;-(

Re:I want one of those!

[nat5an]

Actually, given that you know the positions of the balls at the beginning of the game (assuming that you specify what game you're playing), it's pretty easy to track them if you know where they start. This can even be done fairly well with people, who are much more complex in their movement patterns than billard balls.

In fact, you could do LOTS of useful things with a system like this. Combine it with two other cameras that get side views of the table, and the computer can get the angle and location of your cue and predict what will happen if you hit the cue ball at this angle with varying levels of force.

The only thing you'd have a problem with is predicting what would be affected by varying levels of chalk on the tip of the cue, but if you played consistently, the system could be trained to predict very well.

Hey, this would be a good topic for my thesis, and a lot more fun than some other topics :-D

Also some game theory on the matter...

[mcarbone]

We also applied some game theory to the game 301 and to a game we invented called 30-Block, which you can read about here. We can solve 301 fairly easily, but 30-Block turns out to be quite intractable.

The more interesting part of this paper discusses probability models we use to predict where players will hit based on where they aim. It's interesting: if you are a perfect player, you have the highest expected value when aiming for Triple-20 (obviously), but the worse you get, the best place to aim in the boards spirals inward until it gets to double bulls-eye (which minimizes how often you miss the board).