We actually have a standard client - server architecture, not a peer-to-peer one. Game clients connect to game servers, and game server (optionally) report games to the configured master server (and could report statistics, too). Now there are safeguards against cheating clients in place, and some for servers reporting false games to the master server, too. However, assuming all you control is the master server, there is no easy way to prevent the servers to report fake statistics, even more so with an open source, game.
The way other multi-player games solve this issue is by assigning a unique ID to every game installation - whether client or (possibly dedicated) server -, and make users create accounts to login to a central server before being able to join an Internet server. The master server monitors the clients' logins and match the game statistics reported by the client with those reported by the servers. If they don't match, the client is considered to be cheating. I assume they also use a software to detect irregularities (such as a player who previously played very bad becoming very good at once) and blacklist based on this, too.
Maintaining a central ladder is therefore combined with a high maintenance overhead, and unless we have volunteers available who are willing to do daily manual moderation on this _and_ we have people who are able and willing to code and tweak/maintain the cheat detection algorithms, this may turn out to be not really something we want to do.
But then, it's always easy to shut some service down if we realize we cannot maintain it. So, once we have clearly defined goals and a well considered plan for the implementation I'm all for it.