Welcome to WebmasterWorld Guest from 50.16.78.128

Forum Moderators: incrediBILL & martinibuster

Message Too Old, No Replies

AdSense "Experiments" "Confidence" Question

Why does loser have higher confidence rating?

   
3:28 pm on Jun 6, 2014 (gmt 0)

WebmasterWorld Senior Member ken_b is a WebmasterWorld Top Contributor of All Time 10+ Year Member Top Contributors Of The Month



I've run a few experiments now using the built-in AdSense system. The first ones worked just as one might expect. The "winner" had the highest confidence rating.

But on this last one the "loser" has a higher confidence rating? By loser I mean it is performing worse, Lower CTR, lower EPC, lower total earnings, etc.

Any guesses why that might be?

.
1:16 pm on Jun 8, 2014 (gmt 0)

WebmasterWorld Senior Member ken_b is a WebmasterWorld Top Contributor of All Time 10+ Year Member Top Contributors Of The Month



No one else has noticed this?

A glitch maybe?
2:39 pm on Jun 8, 2014 (gmt 0)

WebmasterWorld Senior Member netmeg is a WebmasterWorld Top Contributor of All Time 10+ Year Member Top Contributors Of The Month



I noticed it, I have no explanation for it.
8:20 pm on Jun 8, 2014 (gmt 0)

WebmasterWorld Senior Member Top Contributors Of The Month



Question is, "confidence in what?"

Confidence that the style performs best or confidence that the loosing ad is always going to under-perform? I'm not entirely sure either but I think confidence is related to each individual result. So, it's not a single confidence rating for the overall experiment e.g. we're confident in the results.

Winner and loser both have a result so you could be very confident that the looser actually was the worst version while only somewhat confident that the winner is definitely the best choice. It's kind of like saying, did the winner win because it's better or did the looser lose because it's worse? Which of these conclusions are we most confident in? That's my guess anyway.
10:39 pm on Jun 8, 2014 (gmt 0)

WebmasterWorld Administrator phranque is a WebmasterWorld Top Contributor of All Time 10+ Year Member Top Contributors Of The Month



perhaps it's a reference to the statistical term "Confidence Interval".
12:48 pm on Jun 9, 2014 (gmt 0)

WebmasterWorld Senior Member netmeg is a WebmasterWorld Top Contributor of All Time 10+ Year Member Top Contributors Of The Month



It is. It's a common concept in split testing, I just don't have a lot of confidence that I can explain it.

But that's why there's Wikipedia.

[en.wikipedia.org...]