We call emerging picks early. Here's how often the pros actually drafted them.
Live model · held-out validation
84% precision
leave-one-patch-out
68% cross-region
held-out Western leagues
3.8d lead
Measured on patches the live model (Scuttle 2.5) never trained on.
We grade it leave-one-patch-out across 6 patches, and the cross-region
figure is graded on disjoint-player Western leagues (Korean soloq forecasting the LCS/LEC meta). The
per-patch table below is the as-of replay history.
GetPrio watches high-elo solo queue and flags champions that are climbing, ideally before
they show up in a pro draft. Once a patch is done, we take the shortlist we put out early that
patch and check it against every champion picked or banned across the four major regions
(LCK, LPL, LCS, LEC) plus the international events. Every patch we've graded is here, the misses
included.
Hit rate
72%
Of the 20 picks we flag, about 14
end up picked or banned on the pro stage.
Catch rate
42%
When a champ does break out into pro play, we'd already flagged it about
2 times in 5.
Head start
4days
How far ahead of a champ's first pro game we typically call it. We see it coming.
Pooled over 12 finished patches,
new picks only, measured on our top 20 each patch.
Accuracy, patch by patch
Hit rate is how many of our calls panned out. Catch rate is how many of the
actual breakouts we'd flagged. Higher is better. Each dot is one finished patch.
Every graded patch · top 20 new picks
Patch
Hit rate
Catch rate
Days early
Flags that landed
Breakouts caught
Scored by
16.6
70%
67%
8
14/ 20
14/ 21
model
16.7
80%
52%
1
16/ 20
16/ 31
model
16.8
80%
48%
1
16/ 20
16/ 33
model
16.9
70%
54%
2
14/ 20
14/ 26
model
16.10
90%
44%
5
18/ 20
18/ 41
model
16.11
65%
54%
7
13/ 20
13/ 24
model
16.13
70%
35%
2
14/ 20
14/ 40
model
16.14
85%
59%
6
17/ 20
17/ 29
model
16.15
70%
28%
2
14/ 20
8/ 29
model
16.16
55%
26%
2
11/ 20
11/ 43
model
16.17
90%
40%
0
18/ 20
18/ 45
model
16.20
45%
24%
6
9/ 20
9/ 38
model
Called shots · what we flagged early, and what landed
The hit rate above, made concrete: the actual champions we put on our early new-pick shortlist
each patch, and which the pros went on to pick or ban.
83 of our last 120 early calls reached pro play
(69%).
The green chips are hits, each marked with how many days' head start we had. The rest we flagged, but the pros didn't take them that patch.
Our newest shortlist, with the day we flagged each pick locked in the moment we called it.
A pick stays watching until it shows up in a pro draft, then
flips to a confirmed hit with the exact head start, measured against that frozen date, so the lead
can't be moved after the fact.
61 still in play.
What counts as pro play: any champion picked or banned in the four majors (LCK, LPL, LCS, LEC)
or an international event in 2026. “New picks” drops anything that was already a standard pick
the patch before, so we're only grading fresh calls, not meta staples everyone runs anyway.
The samples are still small, so read these as a direction. They get sturdier every patch the pipeline runs.