Date: 10/27/25 2:02 pm From: Sean S <therefromhere168...> Subject: Re: [Maine-birds] Interesting article on Merlin
My observations have been that when it was first rolled out, the accuracy
was good although it would sometimes include non-US occurring and very
unlikely species as possible ID's. Then it got better and for a while
(sorry, I have no dates) it was working extremely well, with high
accuracy. I don't know if somewhere along the line the data set took in a
bunch of false ID's from eBird submissions by birders misidentifying
relatively common, therefore non-flagged species (so these ID's just
weren't scrutinized) but a couple of years ago its results started to get
shaky at times. This hypothesis could be completely wrong as I don't know
Merlin's inner workings. However, I can say that since it implemented
location-based ID's it's often, but not always been spotty and unreliable.
Just last week it failed to recognize a loud and obvious Carolina Wren, and
this is one of many missed and sometimes completely off-base ID's that have
been happening over the past year or more. I would also say that being a
bit too strict about individual locations (counties, etc.) has overall
undermined rather than strengthened its capabilities. To give an example,
if something rare but not impossible (say, a Lark Sparrow) showed up in my
neck of the woods, I highly doubt that it would be recognized even with a
good recording. Others may of course have completely different takes and
results. I'm still using it because its ID capability currently seems fair
to pretty good to me.
Sean Smith
On Mon, Oct 27, 2025 at 10:23 AM Jay Pitocchelli <jpitocch...> wrote:
