00:00:00
So, I mean, you could come up with a
00:00:00
model that could be 95% right, but then
00:00:03
the confidence intervals are huge and it
00:00:05
just has no meaningful interpretation.
00:00:07
To me, there's just a lot of uncertainty
00:00:09
with the draft, and so you just kind of
00:00:10
have to get used to being wrong a lot.
00:00:12
And then there's also these correlated
00:00:13
errors you mentioned. So, one of the
00:00:14
things that was the story of round two
00:00:16
and three of the draft this year was the
00:00:18
rise of the what I would call the
00:00:19
overpriced blocking tight end. You know,
00:00:21
a lot of NFL teams are following the
00:00:23
money around what is innovation in the
00:00:26
NFL, and the Rams ran a lot of three
00:00:29
tight end sets where you need a blocking
00:00:31
tight end to have the threat of the run
00:00:33
game, and teams want to be able to run
00:00:34
those. And so, they were like, "Well, if
00:00:36
that's the new innovation, we want to be
00:00:38
able to get that." It bumped up the
00:00:39
price in the draft of the blocking tight
00:00:42
ends. Seth Walder from ESPN asked a
00:00:45
bunch of team sources about a bunch of
00:00:46
players who went earlier than expected
00:00:48
and asked them, "Is that really a
00:00:49
reach?" And for the blocking tight ends,
00:00:51
he said, "Actually, I think they were
00:00:53
just a lot very undervalued by the draft
00:00:56
community this year and overvalued or
00:00:58
maybe more valued, correctly, by the
00:01:00
teams. So, you could pick these up in a
00:01:02
model, but if you just jammed everything
00:01:04
into a machine learning model, it missed
00:01:06
that all the time.
