Great article!

As for the DMV, I concur with LP. The sys.dm_os_performance_counters view is constantly incrementing various counters of performance that SQL Server reports to the O/S. As written, after the first deadlock occurs the proc will signal an e-mail every time the job runs. If you capture the results at the last run of the sproc, then compare the current run against those values you'll have an accurate measure of deadlocks in the interval between job runs. The you can signal an e-mail only when there's an increase in the counter since the last run.

As for the volatility of the values in the DMV, they are all reset to 0 with a reboot or an instance restart.

Thanks for the article, and I hope this helps!

Carter