The new ART runtime is still experimental. It will make your phone less stable, and some apps just won't work at all. That's why the option to switch from Dalvik to ART is in the developer options: it's only there to let developers and system integrators test their stuff and help Google fix problems.
For this reason, it's a little pointless to measure the performance of ART right now, since bugs in it might cause artificially low or high performance, even in situations where the bug doesn't occur or isn't apparent. Fixing a bug might reduce performance (increasing battery or memory use, or decreasing speed) because extra checks or a slower algorithm are needed to avoid the bug; or, it might increase performance by making some computation unnecessary. Also, further work on improving the performance of ART will certainly be done before it's turned on for everybody.
Combining the effects of fixing all bugs, and all the performance improvements that are made, it's very hard to predict how the finished product will perform. The performance could completely change before ART is ready for mainstream use.