"without any click made by me and even without refreshing the page or the adsense unit."

thats strange.. I dont think that its feasible to run an AJAX script directly within an ad.. or can it be done ??

It's certainly possible – it's all an included javascript block / iframe.

Is there a problem with tracking mouse-overs?

Thanks JohnMu – actaully no problem tracking mouse-overs.. and its good data for UX'ers and Human Interaction studies..

<< Is there a problem with tracking mouse-overs? >>

Only if it's described as cost-per-*click*, and instead the advertisers are paying for cost-per-"hover".

> Is there a problem with tracking mouse-overs?

I don't see any problem...
I guess Google needs any kind of data to analyze click-rates, click-fraud, and so on...

If you hover over the google links on this page as well and click properties on the links there is a changing number there as well.

i think thats not the number of hovers but the number of seconds since the ad was loaded (the number gets higher, even if you're not move your mouse over the ad)

I think Henrik is right (although the number seems to be a little bit bigger than the actual no. of seconds).

nope, i must correct, its really the number of hovers on the ad – but the counting method is very broken ;)

i grab the javascript from the ad and format it, so you can better read it, http://www.digitalwar.de/google_js.html – when they would count the seconds some time/date/seconds functions/variables should exists, instead they're just incrementing 'g' on every OnMouseOver Event (see line 28). Thats btw very stupid, because it count many times when you're moving you're mouse on the ad. Google should combine this with OnMouseOut Event (OnMouseOver -> g++ -> lock g; OnMouseOut unlock g)

Could you imagine the amount of data you could get by just testing on 1% of the displayed Adsense blocks? Wow. That must be fun.

Is it possible that they are trying to detect click fraud. A few years ago there were places that would pay you to surf the web. People set up programs that ran all night and randomly would move the mouse till it got to a link. A co-worker of mine would set it up to run all night on his work machine.

If they are measuring mouse overs, it could be that they are trying to see how many times that a mouse rolls over it till it clicks on it.

Just a thought.

> Is it possible that they are trying to detect click fraud.

If I don't want their algorithms to think I'm a spammer, how many times should I mouseover before I click an ad?

Just one time ?

if CLICK DETECTED then
   if countmouseover == 0 then
   bot for click fraud detected
   else
   maybe an human
endif
endif

Bob Morton, of course this is what they are trying to track if you are moving your mouse across an ad :-P
However, the question really is if the paramter nm is actually used for this purpose...

Tracking mouse movement does seem to be a useful way to identify humans until the spammers all update their clicking software.

And what's so hard to fake that parameter and add it manually?