I am currently experimenting with killer-moves, splitting of the evaluation function (opening, middle and end-game), aspiration windows and all kinds of small tricks. For now, I'm leaving transposition tables for later. The idea is to optimise this part of the engine so that the impact of transposition tables will be maximum.
I'm running tournaments on my computer and I wrote a little Java application to upload the file to my FTP account every now and again. The current tourney is between 9 engines, 15 minutes a side and 2 rounds. For up-to-date, semi-real time updates, visit the test page.