Othello is a particularly tricky game to model, since there isn't a particular orientation to the board, nor is there a "direction" that players "move".

This is similar to the way tic-tac-toe works ... it doesn't matter how you orient the board, so you want to make sure these two boards are both considered "equal"...

 x | o |         |   |
---+---+---   ---+---+---   
 o | x |       o | x |
---+---+---   ---+---+---   
   |   |       x | o |

...and if you are keeping track of the game state, and potential game state, you want to model the board insome way that considers those teh same "state".

Personally, I'd start with a simpiler game (backgammon is good, because it is very linear) but no matter what game you decide to use, make sure you search for existing documentation on game modeling, strategies, and board "ranking".

If the main thing you are intested in developing is a learning algorithm, then I would first spend some time looking for an existing framework/API for building an Othello playing bot. (many College AI classes have frameworks like this provided by the instructor which include all of the game modeling you need ... you just impliment some "player" interface.) Then you can focus on what you are really interested in, and not spend a lot of time worrying about modelling the game itself.

UPDATE: You should definitley consider using Games::Goban for representing boards/pieces, and you probably ought to read up on the SGF Format mentioned in the "See Also" section.