OQOmniQuantBeta🌏 APAC
Live
Cohort
Window
All markets · last 90 days · all horizons (default)
🌐 Every market · track record

The whole track record.
Every slice, including misses.

Every prediction — across all markets — scored on the real close-to-close move, broken down by horizon, confidence, sector, region, model and regime. Calibration, risk-adjusted returns and platform discipline included. No cherry-picking.

95% confidence interval: 43.1%–43.4% · All markets

Over these 90 days, each call's own direction happened 43.3% of the time in its market (an up call against how often prices rose, a down call against how often they fell). A forecaster with no skill would have scored about that, so our 43.2% is 0.1 points below that. When the market turns both numbers move together; the edge is what shows skill.

From 26 October 2026 OQ publishes and scores both up and down calls. Between 22 August 2026 and 26 October 2026 only up calls were published and scored, so the down calls made then are not counted; calls made before 22 August 2026 are scored in both directions, as they were published.

From 26 October 2026 OQ publishes and scores both up and down calls. Between 22 August 2026 and 26 October 2026 only up calls were published and scored, so the down calls made then are not counted; calls made before 22 August 2026 are scored in both directions, as they were published. Each call is compared with how often its own direction happened in its market over the same days, so a rise in accuracy when down calls enter a falling market reflects market direction, not skill.

By call direction. Each call is compared with how often each call's own direction happened in that market (same horizon, same days).

CallsScoredAccuracyBaseLift
UP calls353,07942.9%43.0%−0.1pp
DOWN calls70,30944.9%44.7%+0.2pp

89,460 outcomes (17.4%) moved less than the symbol's noise band and are excluded as indeterminate — not counted for or against.

43.2%directional
423,388
validated predictions

Top-conviction calls · held 5d

highest-conviction decile · 9,084 of 90,848 served calls · horizon auto-selected as the highest-accuracy slice
-0.26%
mean return per call, held to horizon
45.1%
directional accuracy of that slice
+0.8 pp
vs a 44.3% base rate

Live (served) predictions only — shadow-challenger calls, which users never saw, are excluded. No stop-loss and no price target: just the direction, held to the horizon. The base rate is how often each call's own direction happened in that market (same horizon, same days); lift is accuracy minus that rate, not minus 50%. This is a research signal, not investment advice.

The trust funnel

last 90 days · no confidence floor

Every slot the system tried, through to the cohort the headline accuracy is computed on — including the ones we declined to call. Hover a stage for its exact definition.

Attempted
3,571,292
Abstained
1,035,456
Predicted
2,535,836
Pending Validation
1,442,853
Validated
1,086,183
52.5% accurate
Net Predictions (KPI cohort)
392,792
42.7% accurate

Accuracy over time

Rolling 7-day directional accuracy against the dashed line — how often each call's own direction happened in that market on that day (an up call against how often prices rose, a down call against how often they fell). That line, not 50%, is where this book breaks even — All markets.

2026-08-18no-skill baseline (dashed)2026-11-16

Recent validation activity

most recent settle 2026-11-16

Predictions scored per day, by the day the outcome resolved — a throughput count, the honest answer to “is the scoring engine running right now?”. It carries no accuracy: every accuracy chart below is by the day each call was made, and its recent days fill in as those calls mature.

Settled (30d)
306,013
predictions scored
Most recent day
8,498
2026-11-16
Days covered
31
with settlement activity
-1,4759,29620,067
2026-10-17settled2026-11-16

Cumulative accuracy

Running accuracy across every scored prediction through each date — All markets. It converges as the sample grows: early swings shrink as evidence accumulates, so a flat right-hand tail is the sample getting large, not performance freezing. Ends short of today by design, because recent predictions are still maturing.

35.7%44.4%53.0%
2026-08-18cumulative accuracy2026-11-16

Daily accuracy vs 7-day rolling

By the day each call was made. Daily values are noisy on thin days, so the rolling line is the one to read. Recent days are still maturing (the fainter line): only their shorter-horizon calls have been scored so far, and the figure fills in as the rest settle. A day's accuracy is drawn once at least 100 of its calls are scored; until then it shows only its scored and pending counts.

21%37%54%70%
2026-08-18dailystill maturing (fills in)rolling 7dmarket base rate (rolling 7d)2026-11-15

Scored predictions per day

Calls made each day and scored so far — the sample behind each accuracy point. Recent days read low because their calls are still maturing, not because output stopped (837,542 calls in this window are still pending).

08,71417,4272026-08-18: 2,3592026-08-19: 5,4712026-08-20: 4,8472026-08-21: 4,6762026-08-22: 1,0542026-08-23: 1,1672026-08-24: 2,2912026-08-25: 2,6542026-08-26: 2,2882026-08-27: 2,0032026-08-28: 1,2322026-08-29: 1,3182026-08-30: 1,4582026-08-31: 2,5212026-09-01: 3,0842026-09-02: 3,5752026-09-03: 4,1162026-09-04: 2,4952026-09-05: 1,6722026-09-06: 1,7672026-09-07: 3,0752026-09-08: 2,3152026-09-09: 3,0802026-09-10: 2,4212026-09-11: 4,6152026-09-12: 5992026-09-13: 1,9102026-09-14: 4,5742026-09-15: 4,0722026-09-16: 9,1312026-09-17: 7,1982026-09-18: 3,8682026-09-19: 6,0062026-09-20: 4,5922026-09-21: 4,4552026-09-22: 5,1512026-09-23: 6,1762026-09-24: 4,0222026-09-25: 2,8092026-09-26: 2,4662026-09-27: 3,7532026-09-28: 6,0652026-09-29: 8,1232026-09-30: 6,6542026-10-01: 6,7992026-10-02: 4,6062026-10-03: 1,8922026-10-04: 4412026-10-05: 3,4662026-10-06: 4,3742026-10-07: 4,3632026-10-08: 4,8262026-10-09: 2,0962026-10-10: 2,4982026-10-11: 3,1822026-10-12: 4,2352026-10-13: 5,6412026-10-14: 7,9982026-10-15: 7,0032026-10-16: 4,9552026-10-17: 7,0432026-10-18: 3,3952026-10-19: 7,8262026-10-20: 7,3832026-10-21: 9,7332026-10-22: 8,5602026-10-23: 3,0832026-10-24: 3,5602026-10-25: 8,6872026-10-26: 17,4272026-10-27: 15,8612026-10-28: 13,7952026-10-29: 14,7902026-10-30: 9,2792026-10-31: 7,3052026-11-01: 8,0562026-11-02: 11,0552026-11-03: 5,1102026-11-04: 6,8122026-11-05: 7,4332026-11-06: 3,3762026-11-07: 4,0262026-11-08: 4,1772026-11-09: 2,1322026-11-10: 1,0942026-11-11: 1,9682026-11-12: 1,7082026-11-13: 1912026-11-14: 3492026-11-15: 6212026-11-16: 0
2026-08-18scored predictions (validated only)2026-11-15

By horizon

1d
51.9%45,372 calls
5d
45.3%95,777 calls
7d
43.4%85,876 calls
10d
41.9%88,587 calls
15d
39.8%63,098 calls
20d
40.2%12,997 calls
25d
38.9%8,284 calls
30d
32.5%16,365 calls
45d
39.8%6,073 calls
60d
36.1%959 calls

Bullish vs bearish

Bullish
42.9%353,079 up-calls
Bearish
44.9%70,309 down-calls

Do the probabilities mean what they say?

Brier 0.277ECE 0.179MCE 0.898n=47,7292 sparse bins hidden

The true reliability diagram: mean predicted P(up) against the realized up-rate in each probability bin. This is a different question from the confidence chart below it, which asks whether higher stated confidence means more calls come out directionally right. A model can rank direction usefully and still have probabilities that mean nothing — these two charts are how you tell those apart.

These probabilities do worse than this market's own base rate (Brier skill −0.05 ± 0.00, n 47,729): a higher stated probability did not mean a better call here.
Calibration is currently unreliable — read this chart, not the label. The snapshot tags the bias as overconfident, but that tag is only a direction test over the bin counts — it has no notion of how large the errors are, so it reads the same whether the bins miss by one point or by ninety. With ECE 0.179 and MCE 0.898 the probabilities are not usable as literal odds, whichever direction the tag names.
0%25%50%75%100%0%25%50%75%100%Bin 30-40%: said 37.4% chance of up, realized 58.0% up (95% CI 57.4-58.6%) · n=25,751 · realized − predicted +20.6ppBin 40-50%: said 44.7% chance of up, realized 63.3% up (95% CI 62.5-64.1%) · n=13,570 · realized − predicted +18.6ppBin 50-60%: said 53.0% chance of up, realized 58.9% up (95% CI 57.7-60.0%) · n=7,453 · realized − predicted +5.8ppBin 60-70%: said 63.2% chance of up, realized 59.3% up (95% CI 54.6-63.8%) · n=442 · realized − predicted −3.9ppBin 70-80%: said 75.0% chance of up, realized 34.0% up (95% CI 26.8-42.1%) · n=144 · realized − predicted −40.9ppBin 80-90%: said 85.4% chance of up, realized 36.2% up (95% CI 28.7-44.4%) · n=141 · realized − predicted −49.3ppBin 90-100%: said 93.9% chance of up, realized 33.3% up (95% CI 27.1-40.2%) · n=195 · realized − predicted −60.6pppredicted P(up)realized up-rate
Dashed diagonal is perfect calibration. Points above it mean we under-promised, below means we over-promised. Vertical bars are 95% Wilson intervals on the realized rate, and circle area is the bin's sample size — a bin of 90 predictions sitting far off the line says far less than one of 19,000.
Probability binPredicted P(up)Realized up-rate95% CInRealized − predicted
30–40%37.4%58.0%57–59%25,751+20.6pp
40–50%44.7%63.3%63–64%13,570+18.6pp
50–60%53.0%58.9%58–60%7,453+5.8pp
60–70%63.2%59.3%55–64%442−3.9pp
70–80%75.0%34.0%27–42%144−40.9pp
80–90%85.4%36.2%29–44%141−49.3pp
90–100%93.9%33.3%27–40%195−60.6pp

Does confidence mean anything?

Mean gap 15.9ppWorst band 47.7ppn=423,388

Actual directional hit-rate by the model’s own stated confidence bucket — higher confidence should mean more calls come out right. This is a different measurement from the reliability diagram above, which checks whether the calibrated probabilities match realized frequencies. Ranking well here and calibrating badly there is a real, common combination, so read them separately.

Verdict: Overconfident — across the confidence bands the stated probability runs ahead of the hit-rate we actually deliver, by 15.9 percentage points on average once each band is weighted by how many predictions are in it. It is also close to uninformative as an ordering — the delivered hit-rate barely moves across confidence bands — so do not treat stated confidence as a ranking either. We have withdrawn the high-conviction figures while the calibrator is re-fitted.Measured on the 423,388 scored predictions in the table below — 5 bands over-stated · 1 on target. Worst band: Extreme (90%+), stated 90.3% and delivered 42.6% over 13,002 calls (−47.7pp). Stated confidence is the band's mean stated confidence.
02550751000255075100Very Low (<50%): said 46%, delivered 45.4% (n=118,279)Low (50-60%): said 55%, delivered 40.9% (n=144,331)Medium (60-70%): said 64%, delivered 43.5% (n=68,826)High (70-80%): said 74%, delivered 44.1% (n=53,689)Very High (80-90%): said 85%, delivered 44.4% (n=25,261)Extreme (90%+): said 90%, delivered 42.6% (n=13,002)stated confidence %actual hit rate %
Dashed line is perfect calibration. Points below it are overconfident; circle size is the number of predictions in that bucket.
Confidence bandPredictionsWe saidWe deliveredDelivered − stated
Very Low (<50%)118,27945.8%
45.4%
−0.4pp
Low (50-60%)144,33154.8%
40.9%
−14.0pp
Medium (60-70%)68,82664.0%
43.5%
−20.5pp
High (70-80%)53,68973.7%
44.1%
−29.6pp
Very High (80-90%)25,26185.4%
44.4%
−41.0pp
Extreme (90%+)13,00290.3%
42.6%
−47.7pp

Calibration by model

lower Brier / ECE is better

The headline numbers above are the blend of every model that produced a scored call. Split out, they disagree — which is the useful part. Delivered − stated is realized accuracy minus stated confidence, the same convention as the band table above: negative means that model talks a better game than it plays.

ModelScoredAvg confidenceAccuracyDelivered − statedBrierECE
modular45,18942.4%59.9%+17.5pp0.2800.186
v182,54047.2%50.9%+3.7pp0.2360.118

Risk-adjusted returns

all markets · all horizons · last 90 days · n=265,298

Measured on the disciplined book — the signals we actually commit to after the confidence gate, not every raw signal. This is the money-weighted result.

Computed on all markets · all horizons · last 90 days, 265,298 closed trades.

Book repaired. 602,456 trades in the ledger — 3,818 re-walked against the venue their prediction actually named, and 590 excluded as unresolvable (386 on a price series frozen at a single close, 204 whose venue could not be confirmed).Trades whose candles were walked against a venue the prediction did not name have been re-walked against the venue it did name, holding entry, stop, target, horizon and direction fixed so that only the venue changed. Rows whose correct venue could not be confirmed, or whose series is frozen at a single close (a data gap rather than a tape), are excluded rather than published as flat trades.
Profit factor
0.81
gross win / loss
Trade win rate
40.4%
trades with a positive return
Expectancy
-0.46%
per trade (trade-weighted)
Kelly fraction
-0.05
implied fraction · ≤0 = no position
For analysts — Sharpe, Sortino, Calmar, information ratio, drawdown, risk/reward, VaR and CVaR
Sharpe
-0.38
risk-adj (annualized)
Sortino
-0.78
downside-only
Calmar
-2.64
turn-aware annualised return ÷ max drawdown · 88d series
Max drawdown
-4.50%
peak-to-trough

Tail risk & efficiency

How bad the bad trades get. VaR is the return the worst 1-in-20 (5%) and 1-in-100 (1%) trades beat; CVaR is the average of that tail — the number that matters when the tail actually arrives.

VaR 5%
-7.50%
worst 1-in-20 trade
CVaR 5%
-10.13%
avg of worst 5%
VaR 1%
-11.33%
worst 1-in-100 trade
CVaR 1%
-12.98%
avg of worst 1%
Information ratio
-0.38
excess return / tracking error
Risk / reward
1.19
avg win / avg loss

Returns by horizon (holding-period)

How a user actually experiences it: hold from the call to its horizon. Sharpe here is annualized the correct way — per-trade × √(252 / horizon-days), not a calendar-year compound.

This table covers all 425,719 scored signals in the window — not only the 265,298 the gated book actually traded, which the cards above describe. Their profit factor is 0.74, against 0.81 for the traded book above.

HorizonnAvg hold returnSharpe (holding-period)Trade win rateMax DD
1d28,461+0.199%0.6953.4%-7.58%
5d80,602-0.550%-0.2146.0%-33.40%
7d73,879-0.811%-0.3244.0%-47.04%
10d81,606-1.259%-0.3642.2%-59.47%
15d70,019-1.662%-0.7341.1%-67.17%
20d13,755-2.379%-0.7539.9%-74.58%
25d10,907-1.401%-0.4641.6%-70.09%
30d27,086-2.261%-0.5139.4%-82.77%
45d16,363-0.396%-0.0644.3%-73.46%
60d16,066-2.521%-0.2942.6%-61.91%
90d6,975+0.960%0.0648.2%-35.21%

Trade funnel

Generated
1,134,328
Traded
602,456
Rejected
531,872
Abstained
722,944
Why we gate. Follow every raw signal — both directions, no confidence floor, including the low-conviction tails and long-dated horizons — and the naive backtest is negative. The gate is meant to close that gap. On the window above it has not yet closed it: the gated book is also negative (profit factor 0.808). We publish that rather than the intention.

Strategy returns (gated book)

last 90 days

Cumulative return of the disciplined trade book — the same daily equity series the Sharpe, Calmar and Max-drawdown figures above are built from, not the raw-signal firehose.

Basis: all markets · all horizons · last 90 days. Compounded over 88 trading days (2026-08-19 → 2026-11-16) — which is what the annualised figure extrapolates from. Book equity: each day's return is the equal-weighted mean of that day's closed trades, compounded across 88 trading days. Expectancy is a per-TRADE mean, so the two can differ in sign when trade counts vary sharply day to day.
Total return
-3.06%
gated book · 88d series
Annualised (turn-aware)
-11.86%
annualised on portfolio turns, not calendar time: an average 13.0-day holding period redeploys capital about 28.1 times a year, and the per-trade return is compounded over those turns across 265,298 closed trades
Avg return / trade
-0.447%
per-trade mean
Avg return / day
-0.035%
per-day book mean
How these reconcile. Per-trade mean -0.447% vs per-day (book) mean -0.035%. They agree in sign here, but they are still different weightings — per-day drives the curve and Total return, per-trade drives Expectancy.

The 88 days behind the book return — each counts once, whatever its trade count

Exit dayTradesShare of tradesDay return
2026-08-194,5051.7%-0.11%
2026-08-204,1011.5%+0.07%
2026-08-214,6571.8%+0.20%
2026-08-22 · thin day880.0%+0.02%
2026-08-23 · thin day400.0%+0.00%
2026-08-243,5311.3%+0.10%
2026-08-253,1951.2%+0.04%
2026-08-262,7821.0%+0.12%
2026-08-272,5861.0%+0.09%
2026-08-283,8651.5%+0.15%
2026-08-29 · thin day360.0%+0.00%
2026-08-30 · thin day520.0%-0.00%
2026-08-313,5011.3%-0.02%
2026-09-013,8891.5%+0.02%
2026-09-023,7011.4%-0.08%
2026-09-033,0321.1%+0.10%
2026-09-043,6311.4%+0.12%
2026-09-05 · thin day570.0%+0.01%
2026-09-06 · thin day530.0%+0.01%
2026-09-071,8870.7%+0.12%
2026-09-083,3841.3%+0.02%
2026-09-094,7871.8%-0.09%
2026-09-104,3481.6%-0.21%
2026-09-117,3012.8%-0.27%
2026-09-12 · thin day640.0%-0.00%
2026-09-13 · thin day780.0%-0.00%
2026-09-145,8682.2%-0.28%
2026-09-155,8402.2%-0.41%
2026-09-165,2902.0%-0.34%
2026-09-174,4521.7%-0.04%
2026-09-188,0233.0%-0.05%
2026-09-19 · thin day860.0%+0.00%
2026-09-20 · thin day2360.1%+0.00%
2026-09-215,4112.0%+0.06%
2026-09-225,9092.2%+0.11%
2026-09-235,0021.9%-0.03%
2026-09-245,7202.2%-0.23%
2026-09-256,7942.6%-0.17%
2026-09-26 · thin day850.0%+0.00%
2026-09-27 · thin day580.0%-0.00%
2026-09-286,4472.4%-0.33%
2026-09-296,5132.5%-0.34%
2026-09-305,6662.1%-0.14%
2026-10-017,5692.9%-0.37%
2026-10-026,2632.4%-0.08%
2026-10-03 · thin day620.0%-0.00%
2026-10-04 · thin day900.0%-0.00%
2026-10-054,6981.8%+0.06%
2026-10-063,5261.3%+0.10%
2026-10-073,8131.4%-0.07%
2026-10-084,4441.7%-0.27%
2026-10-094,2991.6%-0.02%
2026-10-10 · thin day440.0%+0.00%
2026-10-11 · thin day1200.0%-0.01%
2026-10-123,1771.2%-0.06%
2026-10-133,1351.2%-0.10%
2026-10-142,5481.0%-0.04%
2026-10-154,6071.7%-0.20%
2026-10-164,0411.5%+0.02%
2026-10-17 · thin day730.0%-0.00%
2026-10-18 · thin day700.0%-0.00%
2026-10-193,2221.2%-0.12%
2026-10-202,9281.1%-0.07%
2026-10-213,0241.1%-0.06%
2026-10-224,4071.7%-0.16%
2026-10-235,1461.9%-0.26%
2026-10-24 · thin day640.0%-0.00%
2026-10-25 · thin day2170.1%-0.03%
2026-10-263,3671.3%-0.09%
2026-10-273,7351.4%+0.06%
2026-10-283,5861.4%+0.07%
2026-10-293,2771.2%-0.00%
2026-10-304,7671.8%-0.09%
2026-10-31 · thin day250.0%-0.00%
2026-11-01 · thin day100.0%+0.00%
2026-11-024,2391.6%+0.02%
2026-11-033,5811.3%+0.01%
2026-11-043,5751.3%+0.10%
2026-11-053,3021.2%+0.17%
2026-11-063,9111.5%-0.08%
2026-11-07 · thin day210.0%-0.00%
2026-11-08 · thin day1830.1%+0.00%
2026-11-093,1731.2%+0.06%
2026-11-101,7150.6%+0.07%
2026-11-11 · thin day1,3880.5%+0.05%
2026-11-12 · thin day1,3620.5%+0.03%
2026-11-131,5800.6%+0.04%
2026-11-16 · thin day3630.1%-0.00%
27 of the 88 days held only 5,025 trades (1.9% of the total), yet each counts as much as any other day in the book return. Their returns: 2026-08-22 +0.02%, 2026-08-23 +0.00%, 2026-08-29 +0.00%, 2026-08-30 -0.00%, 2026-09-05 +0.01%, 2026-09-06 +0.01%, 2026-09-12 -0.00%, 2026-09-13 -0.00%, 2026-09-19 +0.00%, 2026-09-20 +0.00%, 2026-09-26 +0.00%, 2026-09-27 -0.00%, 2026-10-03 -0.00%, 2026-10-04 -0.00%, 2026-10-10 +0.00%, 2026-10-11 -0.01%, 2026-10-17 -0.00%, 2026-10-18 -0.00%, 2026-10-24 -0.00%, 2026-10-25 -0.03%, 2026-10-31 -0.00%, 2026-11-01 +0.00%, 2026-11-07 -0.00%, 2026-11-08 +0.00%, 2026-11-11 +0.05%, 2026-11-12 +0.03%, 2026-11-16 -0.00%.
-3.9%-2.2%-0.4%1.4%
2026-08-19book equity (dashed = break-even)2026-11-16
Max drawdown
-4.50%
peak-to-trough on this curve
Calmar
-2.64
turn-aware annualised return ÷ max drawdown
Trades in book
265,298
closed trades
Trading days
88
2026-08-19 → 2026-11-16

By asset class

Stocks (Intl)
43.5%271,170 calls
Stocks (India)
40.3%72,191 calls
Stocks
45.2%71,777 calls
Crypto
48.7%3,189 calls
Commodities
40.8%3,108 calls
Forex
39.2%1,632 calls
Indices
48.9%321 calls

By region

Asia-Pacific
40.5%140,144 calls
Europe
48.6%102,728 calls
India
40.3%72,191 calls
North America
45.2%71,777 calls
Middle East & Africa
31.7%15,251 calls
Latin America
48.0%13,047 calls
Crypto
48.7%3,189 calls
Commodities
40.8%3,108 calls

By sector

Unknown
43.2%419,983 calls
Financial Services
46.1%975 calls
Industrials
40.8%517 calls
Consumer Cyclical
34.7%320 calls
Technology
49.8%309 calls
Basic Materials
40.4%285 calls
Healthcare
34.9%252 calls
Consumer Defensive
32.2%239 calls

By model

modular
42.6%328,840 calls · 77.7% of book
v18
45.3%94,372 calls · 22.3% of book
best
25.6%176 calls · 0.0% of book

Accuracy by market × horizon

7 markets · 10 horizons · dimmed under n=30

Directional accuracy for every market at every horizon, each compared with that market's own base rate at that horizon (how often each call's own direction happened there over the same window), never with a fixed 50% line. This is the joint cut, not the two marginal breakdowns above: a market can look strong overall and still trail its own base rate on the horizon you actually trade. Cohort: All markets · last 90 days.

Market1d5d7d10d15d20d25d30d45d60d
United States
47.4%
n=8,725
47.7%
n=16,009
47.3%
n=13,022
45.1%
n=15,410
41.1%
n=12,504
45.1%
n=1,521
39.9%
n=1,781
34.8%
n=1,712
39.9%
n=992
34.6%
n=101
India
58.4%
n=7,918
41.3%
n=15,631
37.8%
n=16,171
37.8%
n=15,404
38.9%
n=8,665
42.0%
n=2,526
28.5%
n=1,308
25.3%
n=4,029
38.8%
n=539
—
Europe
50.6%
n=11,112
49.3%
n=24,337
48.6%
n=19,941
47.6%
n=20,964
47.8%
n=16,048
50.9%
n=2,525
50.2%
n=1,986
46.1%
n=3,987
47.2%
n=1,530
52.0%
n=298
Asia-Pacific
52.0%
n=14,908
43.0%
n=34,035
41.8%
n=28,398
39.0%
n=28,333
34.0%
n=19,633
33.9%
n=5,070
37.5%
n=2,227
27.3%
n=4,729
37.6%
n=2,317
26.1%
n=494
Crypto
54.0%
n=472
49.1%
n=658
50.5%
n=525
44.2%
n=593
46.9%
n=520
16.7%
n=78
18.5%
n=27 · thin
62.7%
n=201
51.9%
n=104
63.6%
n=11 · thin
Forex
47.2%
n=335
39.2%
n=255
43.4%
n=332
39.1%
n=289
31.1%
n=206
34.8%
n=46
27.9%
n=43
17.5%
n=63
39.2%
n=51
16.7%
n=12 · thin
Commodities
54.1%
n=364
41.3%
n=697
40.3%
n=439
44.7%
n=640
34.9%
n=627
15.5%
n=103
40.4%
n=57
35.5%
n=124
25.0%
n=36
38.1%
n=21 · thin

Green is above that cell's own market base rate, red is below; intensity saturates at ±10 points so a thin outlier can't shout down a real one. A cell without a base rate yet is shown uncoloured, with its accuracy and n. 4 cells are dimmed for a sample under 30 — those numbers are shown as measured, but they are not yet evidence of anything. A cell with no scored predictions shows an em-dash rather than a zero.

Returns by market × horizon

10 markets · 11 horizons · dimmed under n=30

What following the calls actually returned, for every market at every horizon — the same joint cut as the accuracy grid above, in the unit that pays. These come from the trade ledger, not the scored prediction cohort, so the sample sizes are their own and will not match cell for cell. Cohort: All markets - last 90 days.

Metric
Average return per trade by market and horizon, up-calls only, All markets - last 90 days.
Market1d5d7d10d15d20d25d30d45d60d90d
United States
+0.10%
n=3,296 · 46% won
-0.12%
n=8,712 · 42% won
-0.19%
n=6,846 · 40% won
-0.43%
n=7,685 · 40% won
-1.04%
n=7,711 · 35% won
-1.87%
n=1,286 · 30% won
-1.72%
n=1,742 · 32% won
-2.11%
n=3,602 · 31% won
-0.99%
n=1,717 · 38% won
-2.11%
n=1,047 · 37% won
+4.24%
n=315 · 68% won
India
+0.39%
n=2,344 · 55% won
-0.35%
n=6,374 · 37% won
-0.60%
n=6,183 · 35% won
-0.99%
n=5,755 · 33% won
-1.84%
n=5,357 · 29% won
-2.86%
n=1,786 · 25% won
-2.38%
n=1,307 · 29% won
-2.56%
n=3,690 · 30% won
-1.07%
n=1,688 · 37% won
-1.35%
n=1,100 · 40% won
-0.52%
n=97 · 39% won
Europe
+0.14%
n=5,616 · 47% won
+0.01%
n=12,769 · 44% won
-0.08%
n=11,807 · 44% won
-0.22%
n=11,054 · 42% won
-0.46%
n=10,098 · 41% won
-1.00%
n=2,034 · 39% won
-0.82%
n=2,134 · 43% won
-0.99%
n=5,899 · 41% won
+0.08%
n=2,427 · 48% won
+0.66%
n=1,432 · 52% won
+1.53%
n=237 · 57% won
Asia-Pacific
-0.03%
n=4,865 · 40% won
-0.51%
n=19,261 · 34% won
-0.57%
n=17,026 · 36% won
-0.84%
n=15,770 · 36% won
-1.11%
n=13,065 · 34% won
-0.83%
n=4,365 · 38% won
-0.40%
n=2,809 · 42% won
-0.28%
n=9,280 · 42% won
+1.42%
n=3,531 · 52% won
+2.12%
n=3,424 · 56% won
+5.08%
n=1,132 · 65% won
Crypto
-0.44%
n=221 · 35% won
-0.26%
n=310 · 36% won
+1.09%
n=294 · 49% won
-0.17%
n=259 · 39% won
-0.56%
n=242 · 34% won
+3.76%
n=59 · 59% won
+9.86%
n=37 · 89% won
+11.70%
n=235 · 89% won
+16.21%
n=103 · 96% won
+21.02%
n=82 · 100% won
+24.67%
n=9 · 100% won · thin
Forex
-0.05%
n=207 · 49% won
-0.12%
n=101 · 37% won
-0.10%
n=173 · 46% won
+0.01%
n=145 · 60% won
-0.33%
n=99 · 47% won
-0.49%
n=31 · 48% won
-0.44%
n=60 · 48% won
-0.69%
n=102 · 41% won
-1.51%
n=47 · 32% won
-2.27%
n=12 · 8% won · thin
-0.87%
n=7 · 57% won · thin
Commodities
-0.03%
n=147 · 48% won
-0.07%
n=436 · 44% won
-0.25%
n=284 · 40% won
+0.39%
n=360 · 47% won
-0.33%
n=389 · 42% won
+0.52%
n=100 · 46% won
+3.43%
n=66 · 68% won
+4.42%
n=232 · 65% won
+6.39%
n=95 · 75% won
+7.70%
n=92 · 76% won
+13.68%
n=11 · 91% won · thin
INDX
+1.08%
n=10 · 80% won · thin
+0.00%
n=43 · 60% won
+0.73%
n=18 · 56% won · thin
-0.64%
n=25 · 36% won · thin
-0.07%
n=28 · 61% won · thin
+2.86%
n=1 · 100% won · thin
+2.33%
n=7 · 86% won · thin
+0.54%
n=18 · 56% won · thin
-0.34%
n=2 · 50% won · thin
+2.18%
n=2 · 50% won · thin
+2.25%
n=4 · 75% won · thin
Latin America
+0.15%
n=982 · 41% won
+0.23%
n=1,660 · 42% won
+0.29%
n=1,883 · 44% won
+0.59%
n=1,686 · 46% won
+0.66%
n=1,592 · 46% won
+2.56%
n=424 · 58% won
+3.52%
n=386 · 62% won
+3.32%
n=1,026 · 60% won
+2.94%
n=346 · 62% won
+4.05%
n=282 · 64% won
+2.38%
n=45 · 56% won
Middle East & Africa
-1.39%
n=7 · 0% won · thin
-0.43%
n=531 · 36% won
-1.29%
n=2,678 · 29% won
-1.26%
n=2,683 · 32% won
-1.93%
n=1,932 · 28% won
-2.56%
n=609 · 28% won
-2.63%
n=460 · 29% won
-2.78%
n=840 · 31% won
-3.74%
n=208 · 29% won
+0.44%
n=153 · 46% won
+3.98%
n=5 · 60% won · thin

Green is a profit, red is a loss — the pivot is zero, not the 50% of the accuracy grid, because this is a return and the null result is “made nothing”. Intensity saturates at ±2% per trade, so a thin outlier can't shout down a real one. 16 cells are dimmed for a sample under 30 — those numbers are shown as measured, but they are not yet evidence of anything, and the annualized view magnifies them hardest. A market/horizon pair with no trades shows an em-dash rather than a zero.

What “annualized” means here. The average trade in a cell, compounded at that cell's own observed holding period — (1 + avg return)periods / avg hold days − 1. Periods is 252 trading sessions for equities and commodities and 365 days for crypto and forex, because a share cannot be traded on a weekend and a coin can. It assumes the capital is redeployed continuously into trades exactly like the ones measured. That is an extrapolation, not a realised return: nobody earned these figures, and a cell with a short hold and a thin sample can print a number in the hundreds of percent from a handful of trades. Each cell's tooltip states how many round trips a year its figure assumes — read that before believing the headline. A 1d trade is an intraday round trip: bought at the open, sold before the close. It is stored with the same entry and exit date, and is counted as a one-day hold — which is why the 1d column carries the largest annualized numbers on the grid, off some of the smallest per-trade averages.
Returns and risk figures come from the trading book, which is long-only: they include long (up) calls only. Short trades are excluded: OQ's trading book is long-only, so this grid shows returns from up (long) calls only. Down calls are published and scored for accuracy but are not traded. Neutral-direction trades belong to neither side and are excluded as well — 0 trades on that basis.
Cohort and basis. Realised returns from the oq_trades_v1 simulated ledger, scoped by exit_at inside the window; market bucket read from the trade's own exchange where the row carries one, otherwise inferred from the symbol via core.exchange_taxonomy.market_for_symbol; annualized = ((1 + mean(pnl_pct)/100) ** (periods / mean holding days) - 1), where periods is 252 trading sessions for equities and commodities and 365 calendar days for crypto and FX. A 1d trade is an intraday round trip (entry_at == exit_at) and is counted as a one-day hold. On multi-day rows the mean return covers EVERY trade in the cell while the holding period is measured only on those that ran past the first bar, so the two rest on different subsets — compare n with n_with_hold per cell. Ledger: 403,006 trades over the requested 90-day window. On the multi-day rows, 41,867 trades closed inside their first session — a stop or target hit on day one. Their returns are in every average here, but they carry no holding period, so the annualized figure compounds a full-sample return over a holding period measured only on the trades that ran longer. Compare a cell’s n with its holding sample in the tooltip to see how far apart those two are. Cross-venue repair: separately, 3,818 trades were priced against a venue the prediction did not name — one bare ticker collapsing several venues and currencies into one record — and have been re-walked against the venue it did name. A further 590 could not be resolved and are excluded. These cells are computed on the repaired book, so they differ from figures published before that correction.

Net edge vs benchmark

Portfolio return minus what the benchmark did over the same window, per market, each against its own named local index. Positive means the calls beat simply holding that index; negative means they did not, and it is printed here either way.

Measuring net edge against each market's benchmark…

Performance by market regime

computed 2026-11-16

Not just accuracy per regime — the return, Sharpe and win rate of the calls made in each market state. This is where you find out whether an edge is real or is one regime carrying the whole record.

Tagged predictions
199,825
carry an ex-ante regime tag
Total in window
200,000
predictions considered
Tag coverage
99.9%
of the window is tagged
Regimes observed
4
volatility x trend

Volatility x trend (ex-ante tags)

The market state recorded at the moment each call was made — knowable in advance, so these numbers are the honest ones.

RegimenAccuracyAvg returnSharpeTrade win rateCoverage
high trending84,062
46.9%
-0.89%-1.0646.9%42.0%
medium trending76,612
50.8%
+0.09%0.1650.8%38.3%
medium mean reverting27,917
51.5%
+0.17%0.5251.5%14.0%
high mean reverting10,971
48.8%
-0.78%-1.2948.8%5.5%

Realized move size (post-hoc diagnostic)

Not a regime, and not a performance claim. These rows bucket the predictions with no ex-ante tag by how large the move turned out to be — which conditions on the outcome and is not knowable when the call is placed. A high accuracy in the “small” bucket is expected and means nothing tradeable: a small realized move is easy to call directionally after the fact, and a near-0% average return on it is arithmetic, not a red flag. Read this as a diagnostic of where errors concentrate, never as an achievable edge.
Move sizenAccuracyAvg returnSharpeTrade win rateCoverage
small73
53.4%
+0.11%0.8453.4%0.0%
mid57
57.9%
+0.49%1.8757.9%0.0%
large45
40.0%
-2.58%-4.0240.0%0.0%

Discipline & coverage

Most attempts never become a call at all — the gate abstains upstream, before a prediction is issued, and those abstentions are counted against us in the two figures on the right. What remains is scored in full. Committed calls here are the published cohort, which applies no confidence floor — every scored prediction is included.

Committed calls
423,388
Abstained
722,944
23.3% of attempts
Accuracy (committed)
43.2%
Coverage rate
36.9%
of all attempts
Acc if abstain = wrong
16.0%
every abstention counted as a miss

Abstention by horizon

platform-wide 23.3%

Share of attempted calls the engine declined to publish, per horizon. Higher means more was withheld.

HorizonAttemptedAbstainedAbstention rate
1d251,121128,400
51.1%
5d271,60250,423
18.6%
7d266,32833,300
12.5%
10d263,54438,882
14.8%
15d311,83658,943
18.9%
20d259,31851,457
19.8%
25d254,78747,243
18.5%
30d351,390106,678
30.4%
45d290,63770,289
24.2%
60d292,38971,439
24.4%
90d284,07265,890
23.2%

The intelligence stack

last promotion 2026-07-06

Which models are serving the predictions scored above, and the rules that govern replacing them.

Active models
1
serving live predictions
Auto-promotion
enabled
challenger can replace champion
Modular combo
active
per-symbol model blending
Swarm consensus
60%
agreement needed to commit

Currently serving

Cross-reference these against the “By model” breakdown above — a version with a small share of the book is a challenger being measured, not the champion.

DAILY_LIGHT_v20260622_1414_90d_AU

Top forecasters

No per-agent forecast records scored yet. This leaderboard fills once the swarm attributes settled outcomes to individual agents — until then there is nothing here to rank, and we would rather show that than an empty table.

Cohort definition & audit trail

canonical_trust_strict

Exactly which predictions the numbers above are computed over, and what was filtered out to get there. Published so the accuracy figure can be checked rather than taken on faith.

Strict directional accuracy — a prediction is scored correct only when the symbol moved beyond its noise band and the model called that direction. Band-indeterminate outcomes (see indeterminate.band_description for the exact rule in force) are excluded from both the numerator and the denominator — not counted for or against. Cohort also excludes archived, abstained, out-of-universe, and shadow-challenger predictions. No confidence floor is applied: every scored prediction is in the cohort, including the ones the model was least sure of. From 26 October 2026 OQ publishes and scores both up and down calls. Between 22 August 2026 and 26 October 2026 only up calls were published and scored, so the down calls made then are not counted; calls made before 22 August 2026 are scored in both directions, as they were published.

This window spans a change to the published cohort on 2026-10-09: it mixes calls published under the old rule and the new one, so a move in the figure can be the population changing, not the model. Calls below the former 0.52 confidence cut are now published and scored. Before this date about 31% of batch calls were withheld by that cut; the confidence tiers (TRADE_CANDIDATE / WATCH / BELOW_WATCH) are now display labels only.

Cohort size
1,261,755
predictions in scope
Snapshot age
2.0h
stale past 6h
Validator last run
2026-11-16
outcome scoring pass
Facet health
degraded
2 failed
Some facets failed to build: walk-forward, daily by maturation. Sections backed by those are hidden rather than shown with partial numbers.

Cohort invariants

Every prediction counted above satisfies all of these. Each one removes a population that would otherwise flatter the number.

  • 1archived != True (exclude pre-2026-05-07 broken-pipeline cohort)
  • 2abstain != True (exclude model-declined predictions)
  • 3in_universe != False (exclude out-of-universe symbols)
  • 4is_neutral != True (exclude neutral predictions from directional accuracy)
  • 5is_shadow != True (exclude shadow-challenger predictions — the ~75% not shown to users)
  • 6confidence >= 0 (no confidence floor — every scored prediction is included, including the lowest-confidence ones)
  • 7prediction_type not in ['causal', 'crowd', 'sentiment'] and source not in ['predict-any', 'realtime'] (published model calls only — not non-model readings, not on-demand runs)
  • 8direction as published: UP or DOWN, except DOWN calls made 2026-08-22 to 2026-10-26 (never published, never scored)

Filters applied

The query this snapshot was built with. null means unfiltered.

days back
90
exchange filter
—
model version
—
horizons
—
direction
—

Excluded from the cohort

Printed exactly as the backend reports it — where a count is not separately tracked it says so rather than showing a zero.

archived count excluded
see #202 archive script — 11,153 docs archived 2026-05-16
abstained count
722,944
out of universe count
not separately surfaced
neutral count
89,460
low confidence count
not separately surfaced

Top-conviction ledger

All markets · last 90 days

The highest-conviction decile of settled calls, ranked on the calibrated probability — not the stated confidence. Held to the horizon with no stop and no target, which is why these rows reconcile with the hero number above and not with the traded book.

Sign in to inspect the raw ledger

This is the conviction slice itself. These are the individual predictions behind every aggregate number above — one row per call, with the outcome it actually settled to. The aggregates stay public; the row-level record is for signed-in members.

Sign in

Audit ledger

All markets · last 90 days · scored rows only

Every scored prediction in the canonical cohort, newest settlement first. This is the same set of rows the headline accuracy is computed over — the six invariants noted beneath the table are applied identically — so this is the panel to audit against. The recent-settled panel is a wider, non-canonical cohort and will not tie out.

Sign in to inspect the raw ledger

This is the canonical audit trail. These are the individual predictions behind every aggregate number above — one row per call, with the outcome it actually settled to. The aggregates stay public; the row-level record is for signed-in members.

Sign in

Recent settled predictions

All markets · last 100 returned

The latest calls that have already settled — pending ones are pulled from the same response and filtered out here, so this is scored rows only. This cohort is not the canonical one. It excludes archived and shadow-challenger rows only — it does not apply the six canonical invariants the headline accuracy above is computed on, so tallying these rows will not tie out to that number. The audit ledger panel is the cohort that does. Every market, because the scope above is set to Global.

Sign in to inspect the raw ledger

This is the recent settled record. These are the individual predictions behind every aggregate number above — one row per call, with the outcome it actually settled to. The aggregates stay public; the row-level record is for signed-in members.

Sign in

How we measure

Directional

Right if the price moves the way we said by the horizon. Unambiguous, checkable.

Close-to-close

Official closes at the horizon date — not intraday ticks we could pick to flatter.

No look-ahead

Locked before the outcome window opens. Out-of-sample, every time.

Against the base rate

Each call is compared with how often its own direction happened in its market and horizon — not a fixed 50%.

The record is the proof. Now see the picks.

Today’s high-conviction calls for Asia-Pacific.

View picks