Every slate is scored once its matches finish: for each prop a book priced, did the result land on the side of the line the projection was on. Kills O/U is the share that did. A coin gets 50%; paying the books’ margin takes roughly 53% to break even. Only slates since the 2026-08-21 rounds fix count toward the headline — earlier ones came from a different model and are listed separately below.
kills O/U
50.2%
scored slates
55
days over 50%
24 of 55
headshots O/U
50.8%
avg kill miss
-0.74
Edge size has not predicted results. Over 2,240 scored kills rows, a bigger gap between projection and line hit no more often than a small one, so nothing on this page is ordered by it. The model itself has landed on the right side of the line 50.2% of the time over 55 scored slates, against a break-even of about 53%. What did hold up in testing is the price itself: where the books disagreed, the better number hit more often than the worse one (48.3% against 44.0%) — still short of break-even, and nothing here is advice.
Kills over/under by slate
Each dot is one slate; the line is the running average of all of them so far. The table below lists every slate.
Accuracy history
Model comparison
every scored slate, by model — break-even before vig is ~53%
Model
Slates
Kills O/U
Days over 50%
HS O/U
Kill diff
(unlabelled)
12
54.7%
8
44.3%
+5.81
v2
29
53.0%
18
49.9%
+3.35
v3
55
50.2%
24
50.8%
-0.74
One row per scored slate per model.
Date
Model
Kills O/U
HS O/U
Kill diff
2026-10-18
v3
46.1%
54.5%
+1.9
2026-10-17
v3
57.8%
58.9%
-2.3
2026-10-16
v3
31.9%
39.5%
-3.9
2026-10-15
v3
61.5%
62.3%
+0.7
2026-10-14
v3
57.9%
50.0%
-0.6
2026-10-13
v3
46.4%
45.8%
-2.3
2026-10-12
v3
59.2%
62.1%
+0.6
2026-10-11
v3
58.8%
59.0%
+1.6
2026-10-09
v3
44.4%
43.3%
+2.2
2026-10-08
v3
59.4%
60.5%
-0.2
2026-10-07
v3
39.7%
61.1%
+1.8
2026-10-06
v3
62.8%
54.3%
-2.4
2026-10-05
v3
45.1%
45.3%
-0.6
2026-10-04
v3
48.2%
42.5%
-0.6
2026-10-03
v3
54.2%
50.0%
+0.4
2026-10-02
v3
46.3%
27.3%
-1.8
2026-10-01
v3
60.0%
—
+0.3
2026-09-29
v3
53.6%
36.6%
+0.1
2026-09-28
v3
44.2%
51.6%
+0.9
2026-09-27
v3
51.7%
55.1%
-0.7
2026-09-26
v3
43.7%
52.3%
+0.1
2026-09-25
v3
45.0%
32.3%
-1.9
2026-09-24
v3
52.5%
51.9%
+1.0
2026-09-23
v3
50.0%
47.1%
+1.0
2026-09-22
v3
57.1%
50.0%
-0.0
2026-09-21
v3
30.0%
30.0%
-2.7
2026-09-20
v3
40.0%
41.5%
-0.6
2026-09-18
v3
49.1%
58.5%
+0.8
2026-09-17
v3
51.4%
36.0%
-2.0
2026-09-16
v3
50.0%
60.7%
-0.9
2026-09-15
v3
69.6%
77.8%
-0.4
2026-09-14
v3
69.6%
77.8%
-0.4
2026-09-13
v3
58.0%
63.2%
-2.6
2026-09-12
v3
50.7%
50.0%
-0.8
2026-09-11
v3
70.0%
72.2%
-1.1
2026-09-10
v3
43.1%
45.1%
+0.1
2026-09-09
v3
45.5%
35.1%
-2.4
2026-09-08
v3
42.3%
51.9%
-0.9
2026-09-07
v3
45.8%
47.4%
-2.3
2026-09-06
v3
45.8%
47.4%
-2.3
2026-09-05
v3
44.1%
71.8%
-1.7
2026-09-04
v3
48.5%
52.7%
-0.6
2026-09-03
v3
42.9%
46.4%
+1.7
2026-09-02
v3
47.8%
53.2%
-1.2
2026-09-01
v3
41.1%
53.2%
-1.9
2026-08-31
v3
44.2%
53.6%
-3.3
2026-08-30
v3
50.0%
37.1%
-1.7
2026-08-29
v3
34.4%
40.0%
-0.7
2026-08-28
v3
51.4%
46.5%
+0.0
2026-08-27
v3
50.0%
45.5%
+1.6
2026-08-26
v3
51.1%
48.3%
-0.7
2026-08-24
v3
65.0%
72.7%
-2.6
2026-08-23
v3
57.6%
31.6%
-2.1
2026-08-22
v3
54.2%
52.9%
-2.7
2026-08-21
v3
41.0%
50.0%
-1.7
Below: before the 2026-08-21 rounds fix, when the model used rounds WON instead of rounds PLAYED — not comparable to the rows above.
Date
Model
Kills O/U
HS O/U
Kill diff
2026-08-20
v3
39.4%
41.7%
-7.0
2026-08-19
v3
50.8%
37.8%
+5.2
2025-08-18
v3
58.2%
50.3%
+3.5
2025-08-18
v2
58.1%
55.6%
+2.5
2025-08-16
v3
52.0%
50.5%
+1.9
2025-08-16
v2
59.1%
57.5%
+0.4
2025-08-15
v3
55.0%
63.2%
+4.7
2025-08-15
v2
75.0%
57.9%
+3.3
2025-08-13
v3
46.8%
53.8%
+5.5
2025-08-13
v2
52.8%
53.6%
+3.7
2025-08-12
v3
44.9%
53.9%
+6.3
2025-08-12
v2
42.9%
41.3%
+5.3
2025-08-11
v3
54.9%
57.1%
+5.2
2025-08-11
v2
47.9%
49.5%
+5.5
2025-08-10
v3
45.9%
51.9%
+3.0
2025-08-10
v2
38.0%
41.5%
+4.3
2025-08-09
v3
50.3%
46.8%
+1.8
2025-08-09
v2
48.9%
49.5%
+2.3
2025-08-08
v3
57.8%
46.4%
+1.8
2025-08-08
v2
57.4%
50.5%
+3.1
2025-08-07
v3
61.3%
51.0%
+1.1
2025-08-07
v2
69.6%
60.0%
+2.8
2025-08-06
v3
41.8%
51.7%
+6.5
2025-08-06
v2
41.8%
37.0%
+5.3
2025-08-05
v3
38.7%
54.1%
+8.5
2025-08-05
v2
39.9%
42.6%
+5.3
2025-08-04
v3
47.4%
53.0%
+11.6
2025-08-04
v2
42.9%
50.8%
+2.9
2025-08-03
v3
61.3%
53.3%
+10.8
2025-08-03
v2
57.1%
50.0%
+2.2
2025-08-02
v3
55.9%
52.6%
+11.0
2025-08-02
v2
55.2%
50.5%
+2.8
2025-08-01
v3
40.7%
36.1%
+9.3
2025-08-01
v2
57.4%
44.9%
+2.8
2025-07-30
v3
36.8%
26.3%
+13.2
2025-07-30
v2
41.4%
41.4%
+6.0
2025-07-29
v3
33.8%
47.3%
+18.1
2025-07-29
v2
47.5%
51.7%
+5.1
2025-07-28
v3
54.3%
54.3%
+3.0
2025-07-28
v2
52.8%
50.6%
+1.9
2025-07-27
v3
60.0%
49.2%
+3.4
2025-07-27
v2
64.0%
53.2%
+1.3
2025-07-26
v3
55.7%
50.0%
+4.7
2025-07-26
v2
50.6%
49.4%
+3.1
2025-07-25
v3
52.1%
54.7%
+5.8
2025-07-25
v2
51.9%
51.0%
+3.7
2025-07-24
v3
61.3%
57.4%
+5.4
2025-07-24
v2
58.0%
49.0%
+2.1
2025-07-23
v3
58.8%
50.7%
+5.9
2025-07-23
v2
61.1%
51.3%
+0.7
2025-07-21
v3
41.2%
52.9%
+6.8
2025-07-21
v2
35.3%
47.1%
+6.6
2025-07-20
v3
77.1%
66.7%
+3.1
2025-07-20
v2
66.7%
70.6%
+0.4
2025-07-18
v3
48.3%
52.2%
+4.0
2025-07-18
v2
60.6%
44.4%
-0.9
2025-07-17
v2
46.9%
49.2%
+4.6
2025-07-16
v2
55.2%
46.6%
+8.0
2025-07-15
?
55.2%
46.6%
+8.0
2025-07-14
?
51.1%
55.3%
+14.2
2025-07-13
?
63.2%
57.1%
+7.7
2025-07-12
?
53.3%
46.2%
+8.9
2025-07-11
?
47.6%
44.0%
+10.3
2025-07-10
?
41.9%
48.0%
-3.7
2025-07-09
?
72.4%
44.8%
-1.4
2025-07-07
?
47.7%
53.8%
-2.2
2025-07-06
?
68.0%
52.0%
+0.1
2025-07-05
?
80.0%
20.0%
+4.7
2025-07-04
?
23.1%
30.8%
+8.7
2025-07-03
?
52.8%
33.3%
+14.3
Diagnostics
For the slate currently on the front page, not a past one.
Both sides of every prop divided by the match's projected rounds. A CS2 player lives between roughly 0.47 and 0.96 kills per round, and 0.18 and 0.47 headshots per round, so a number outside that is a broken input rather than an edge — this says which side is broken.
Kill rate0.653 KPR
Your projections imply 0.65 kills per round on average; the posted lines imply 0.65. Real CS2 sits near 0.70. The rate itself is in range, so any remaining gap is player-level, not a blanket shift.
Model out of band0 of 48
Every projection implies a kill rate a real player produces.
Lines below floor0 of 48
Every posted line implies a survivable kill rate.
Round accuracy0.976x over 1380
Across 1380 scored matches the actual round count came in at 0.976 times projected, which is close enough that the kill rates above are trustworthy.
Headshot ratio50.9%
Headshots come out at 50.9% of kills. That sits inside the 35–60% range real rosters produce, so the headshots side is not the problem.
v3 track record50.2% over 55
Across 55 scored slates, v3 has landed on the correct side of the line 50.2% of the time (24 of 55 days above even). Break-even before vig is around 53%, so this is not yet a demonstrated edge.
Chart each market is judged against its own band
Both plausible · 48Model and line both sit inside the survivable band; a gap here is a real disagreement.
Line out of range · 0The posted line implies a kill rate nobody sustains. Suspect the scrape before the market.
Model out of range · 0Your projection is outside what a CS2 player does over a series. A bug in the inputs.
Both plausible · 47Model and line both sit inside the survivable band; a gap here is a real disagreement.
Line out of range · 0The posted line implies a headshot rate nobody sustains. Suspect the scrape before the market.
Model out of range · 0Your projection is outside what a CS2 player does over a series. A bug in the inputs.
By team
How far the projections sit above the lines, in kills per round, for every team on the slate — a read on where the model runs hot or cold against the market, not a ranking of bets. Built from the whole slate; the filters above do not narrow it.