Closing Line Value Scorecard
CLV compares our entry price to the market's closing price. For pregame entries it is a conventional market-relative price measure. Most of our entries are in-play, and there the close we can capture is the last price before the market resolves — a price that already reflects most of the outcome. So the headline is the unconditioned aggregate below (how often we beat our captured close, and by how much), not any statistic conditioned on who won. Full data download, every admitted sport.
46.4% of trades beat our captured close.
Average CLV: -0.0¢ per trade.
Fewer than half of our measured fills beat the close we capture, and the average is currently negative. Per-sport tables below disclose where that aggregate comes from. CLV is one input to operational review alongside calibration, execution quality, realized P&L, and sample uncertainty.
Until this date, this page led with an outcome-conditioned CLV win-rate split. That framing is retracted. For auditability, the equivalent split on today's admitted population is 89% (n=1407) for positive CLV and 8% (n=1622) for non-positive CLV, computed on settled binary rows only (settled non-binary rows are excluded from this split); it is shown here as a diagnostic artifact, not evidence of edge.
Why: about 90% of our entries are in-play, and the close we can capture is the last quoted price before the market converges to its 0/100 resolution. That price already reflects most of the outcome — eventual winners' closes cluster near the top of the range, eventual losers' near the bottom — so "beat the close" largely restates "bought the eventual winner." The demonstration is a placebo test: replace our model with entry rules that contain no information — every entry at a flat 50c, entries shuffled across markets, sides picked by coin flip — and run them against our real closes and real outcomes. All three reproduce the same gap, and the flat-50c rule reproduces a wider one than our model does. A statistic that rewards a constant more than it rewards the forecast is measuring the close, not skill.
For in-play entries the honest instrument is a fixed-horizon markout (price movement at T+60s / T+180s after entry, before the next major game event). Our tracker records these; coverage is still thin, and we will publish that scorecard when the sample supports it. Until then, judge us on the aggregate above.
The numbers above are exchange-corrected. Our fast fill path had been recording the submitted limit rather than the executed price — 74% of MLB entries were affected, averaging 1.78c too favourable to the market and therefore making our published CLV look worse than it actually was. 43.4% of the measured population (1315 of 3030 rows) is now corrected by an exact exchange-identity match; the remainder keeps its ledger-recorded value and we do not describe those rows as exchange-verified.
Full detail, the public correction sidecar and a verification script you can run yourself are on the retraction page.
Whether our model's entry price was better than the market's closing price. Positive = we got in cheaper than where the market settled. For pregame entries this is independent of the outcome. For in-play entries — most of ours — the captured close partially encodes the outcome (see the correction above), which is why we do not condition any headline statistic on who won.
- Profit / loss (variance dominates over short windows)
- Model accuracy on a single game
- Forecasting skill, when conditioned on the outcome after the fact (the retired 91%/10% framing)
- Liquidity at our trade size
- Slippage between fair and entry
Per-sport CLV — sorted best to worst
Each row is one sport's mean CLV across the admitted live fills with a measured close. Green bars = positive and red bars = negative by this price-difference measure. Sports with under 10 measured fills are flagged as preliminary.
| Sport | Verdict | Mean CLV (all-time) | 30d | Beat close | Overall WR | n |
|---|---|---|---|---|---|---|
|
NCAAMB
unconditioned price measurement
|
preliminary
LOW N
|
+19.25c | — |
100%
4/4
|
100%
4W/0L
|
4/4 |
|
NCAAWB
unconditioned price measurement
|
preliminary
LOW N
|
+6.82c | — |
75%
3/4
|
75%
3W/1L
|
4/4 |
|
WNBA
unconditioned price measurement
|
beating close
POSITIVE OBS.
|
+3.15c | -9.39c n=18 |
45%
66/148
|
36%
54W/94L
|
148/148 |
|
LOL
unconditioned price measurement
|
beating close
POSITIVE OBS.
|
+2.85c | +22.00c n=4 |
50%
91/182
|
46%
86W/100L
|
182/186 |
|
NFL
unconditioned price measurement
|
beating close
POSITIVE OBS.
|
+2.64c | +2.64c n=14 |
43%
6/14
|
43%
6W/8L
|
14/14 |
|
WTA
unconditioned price measurement
|
beating close
POSITIVE OBS.
|
+1.96c | +3.87c n=381 |
45%
489/1083
|
42%
460W/627L
|
1083/1089 |
|
SOCCER
unconditioned price measurement
|
beating close
POSITIVE OBS.
|
+1.68c | — |
46%
40/86
|
41%
36W/51L
|
86/87 |
|
NHL
unconditioned price measurement
|
breakeven
NEAR ZERO
|
-0.47c | -10.14c n=7 |
53%
67/126
|
59%
75W/52L
|
126/128 |
|
MLB
unconditioned price measurement
|
bleeding
NEGATIVE OBS.
|
-1.07c | +5.26c n=58 |
51%
321/632
|
53%
337W/294L
|
632/632 |
|
ATP
unconditioned price measurement
|
bleeding badly
LOW OBS.
|
-3.26c | -0.14c n=48 |
39%
135/347
|
38%
134W/217L
|
347/352 |
|
CS2
unconditioned price measurement
|
bleeding badly
LOW OBS.
|
-3.44c | — |
46%
168/362
|
48%
182W/196L
|
362/378 |
|
CFB
unconditioned price measurement
|
preliminary
LOW N
|
-6.00c | -6.00c n=1 |
0%
0/1
|
0%
0W/1L
|
1/1 |
|
NBA
unconditioned price measurement
|
bleeding badly
LOW OBS.
|
-9.90c | — |
42%
17/41
|
40%
17W/26L
|
41/43 |
CLV by entry-edge bucket
For each sport with enough data, this groups mean CLV by the model-minus-entry difference recorded when the trade opened. Green and red describe the observed closing-price difference in that bucket; they do not by themselves validate the model, establish causation, or guarantee future returns.
| Edge band | n | Mean CLV | Beat % | WR |
|---|---|---|---|---|
| 8-11c | 2 | +13.50c | 100% | 100% |
| 11-15c | 1 | +26.00c | 100% | 100% |
| 15-20c | 1 | +24.00c | 100% | 100% |
| Edge band | n | Mean CLV | Beat % | WR |
|---|---|---|---|---|
| 11-15c | 1 | +32.80c | 100% | 100% |
| 15-20c | 2 | -12.25c | 50% | 50% |
| 25c+ | 1 | +19.00c | 100% | 100% |
| Edge band | n | Mean CLV | Beat % | WR |
|---|---|---|---|---|
| <5c | 8 | -1.94c | 38% | 50% |
| 5-8c | 94 | +2.06c | 44% | 38% |
| 8-11c | 22 | +9.05c | 46% | 36% |
| 11-15c | 15 | +3.01c | 40% | 27% |
| 15-20c | 3 | -1.33c | 67% | 0% |
| 20-25c | 5 | -0.24c | 60% | 40% |
| 25c+ | 1 | +49.50c | 100% | 0% |
| Edge band | n | Mean CLV | Beat % | WR |
|---|---|---|---|---|
| <5c | 12 | +12.83c | 67% | 58% |
| 5-8c | 6 | -9.25c | 17% | 33% |
| 8-11c | 65 | +2.14c | 52% | 55% |
| 11-15c | 29 | -1.73c | 52% | 45% |
| 15-20c | 26 | +14.70c | 62% | 58% |
| 20-25c | 20 | -11.83c | 20% | 20% |
| 25c+ | 24 | +7.75c | 54% | 33% |
| Edge band | n | Mean CLV | Beat % | WR |
|---|---|---|---|---|
| 8-11c | 11 | -1.73c | 36% | 36% |
| 11-15c | 3 | +18.67c | 67% | 67% |
| Edge band | n | Mean CLV | Beat % | WR |
|---|---|---|---|---|
| <5c | 474 | +0.87c | 44% | 42% |
| 5-8c | 246 | +4.28c | 46% | 41% |
| 8-11c | 146 | +0.18c | 42% | 36% |
| 11-15c | 80 | +8.32c | 59% | 55% |
| 15-20c | 61 | +2.98c | 52% | 52% |
| 20-25c | 23 | -0.80c | 44% | 39% |
| 25c+ | 53 | -3.70c | 34% | 36% |
| Edge band | n | Mean CLV | Beat % | WR |
|---|---|---|---|---|
| <5c | 1 | +28.00c | 100% | 100% |
| 5-8c | 1 | -62.50c | 0% | 0% |
| 8-11c | 11 | -3.55c | 46% | 54% |
| 11-15c | 58 | +2.11c | 47% | 43% |
| 15-20c | 10 | -3.87c | 30% | 20% |
| 20-25c | 3 | +39.20c | 100% | 67% |
| 25c+ | 2 | +8.50c | 50% | 0% |
| Edge band | n | Mean CLV | Beat % | WR |
|---|---|---|---|---|
| <5c | 15 | -4.70c | 47% | 60% |
| 5-8c | 13 | -6.69c | 38% | 38% |
| 8-11c | 35 | -5.74c | 43% | 51% |
| 11-15c | 24 | +3.88c | 71% | 67% |
| 15-20c | 19 | +12.71c | 63% | 84% |
| 20-25c | 11 | -6.56c | 46% | 54% |
| 25c+ | 9 | +4.08c | 67% | 44% |
| Edge band | n | Mean CLV | Beat % | WR |
|---|---|---|---|---|
| <5c | 274 | -3.06c | 47% | 47% |
| 5-8c | 91 | -2.17c | 62% | 70% |
| 8-11c | 48 | -8.17c | 50% | 58% |
| 11-15c | 128 | +2.68c | 52% | 57% |
| 15-20c | 47 | +2.93c | 51% | 49% |
| 20-25c | 20 | -1.40c | 40% | 40% |
| 25c+ | 24 | +12.42c | 58% | 54% |
| Edge band | n | Mean CLV | Beat % | WR |
|---|---|---|---|---|
| <5c | 49 | -10.66c | 35% | 43% |
| 5-8c | 14 | -28.24c | 0% | 7% |
| 8-11c | 94 | -4.19c | 37% | 37% |
| 11-15c | 35 | -6.42c | 37% | 31% |
| 15-20c | 94 | +1.21c | 45% | 42% |
| 20-25c | 30 | +2.61c | 40% | 27% |
| 25c+ | 31 | +6.83c | 52% | 55% |
| Edge band | n | Mean CLV | Beat % | WR |
|---|---|---|---|---|
| <5c | 89 | -5.95c | 49% | 61% |
| 5-8c | 25 | -18.52c | 36% | 40% |
| 8-11c | 33 | -1.48c | 58% | 58% |
| 11-15c | 43 | -2.90c | 46% | 51% |
| 15-20c | 57 | -4.61c | 40% | 40% |
| 20-25c | 82 | +2.45c | 48% | 38% |
| 25c+ | 33 | -0.56c | 42% | 46% |
| Edge band | n | Mean CLV | Beat % | WR |
|---|---|---|---|---|
| 8-11c | 1 | -6.00c | 0% | 0% |
| Edge band | n | Mean CLV | Beat % | WR |
|---|---|---|---|---|
| <5c | 3 | -27.40c | 0% | 0% |
| 5-8c | 9 | +4.33c | 78% | 78% |
| 8-11c | 6 | -6.00c | 50% | 50% |
| 11-15c | 4 | -33.02c | 0% | 0% |
| 15-20c | 11 | -18.22c | 27% | 18% |
| 20-25c | 7 | -3.64c | 43% | 29% |
| 25c+ | 1 | +31.50c | 100% | 100% |
Why we publish this
Realized P&L and probability calibration answer different questions, and both can be noisy over short samples. Price-based measures add another view: did the entry finish ahead of the defined market close? For pregame entries our captured CLV answers that directly; for in-play entries it only partially does (see the correction above), which is why a fixed-horizon markout is the planned execution-quality measure there.
We publish CLV by sport because aggregate results can hide sport-specific execution and timing differences. Persistent negative CLV is a warning that merits investigation; it is not a mathematical guarantee of future loss and does not identify whether the cause is model selection, timing, liquidity, or measurement. The changelog is the authoritative record of live/shadow changes.
Recorded closes and the weekly price-history backfill supply the closing-price side of this calculation; exact exchange reconciliation corrects entry prices where that evidence is available. Coverage is currently 98.6% across the admitted population. Older trades may not have CLV because the available history does not always contain a usable non-terminal close.
Citing this dataset
CC BY 4.0. Reproduce, quote, and embed freely with attribution. Programmatic access via JSON / CSV is rate-unlimited for non-abusive use.
Definition. CLV is computed as closing_price_c − entry_price_c for each trade. Closing price is the last usable non-terminal Polymarket price in the recorded history before convergence to 0/100. Entry price is the exchange-reconciled execution price where exact order evidence is available, otherwise the recorded execution value.
Coverage. Older trades may lack CLV because the close-price-tracker only stored prices in a 5–95c band; markets that resolved fast can have no non-terminal price recorded. The weekly backfill_clv.py cron pulls Polymarket price history to close that gap retroactively. Current coverage: 98.6%.
Display labels. The page labels mean CLV ≥ +1c beating close, −1c to +1c breakeven, −3c to −1c bleeding, and below −3c bleeding badly. These are disclosed display rules, not statistical proof of edge. Samples under 10 measured fills are flagged preliminary.
Operational status. CLV is one review input, not a live-trading switch by itself. Current live/shadow decisions can change faster than this scorecard; consult the changelog for the dated operational record.
Why publish CLV. Reproducible CLV requires entry timestamps, execution prices, and a closing-price archive. We publish both positive and negative measurements, the admitted population, and the correction history so readers can evaluate the limits themselves.