ioxer
Log inStart free
TRACK RECORD · LIVE VALIDATION

The proof is the live forward record.

A backtest can be curve-fit; a live, point-in-time record can’t. So the real proof of IOX is the forward track record — every daily score saved before the outcome is known, then checked against what actually happened. It is early, and we say so plainly: full performance metrics publish here only once the sample is statistically significant.

01

Where we are right now

Live forward recording opened 29 May 2026. Every day the engine scores its ~50-coin universe and stores each score point-in-time — so nothing here is fitted after the fact. Outcomes fill in as each horizon matures:

11 factors tested · 1 survived~50 liquid coins80% conformal intervalsout-of-sample · FDR · survivorship-aware
1-dayLIVE

Recording forward 1d outcomes since 29 May 2026.

7-dayLIVE

Recording forward 7d outcomes since 29 May 2026 — the primary IOX horizon.

30-dayLIVE

Recording forward 30d outcomes since 29 May 2026; first cohorts maturing now. A longer sanity check — past the signal’s strongest 1–14d range.

90-dayRECORDING

Scored point-in-time since 29 May 2026; first 90d outcomes mature ~27 Aug 2026. Long-run context, not where the edge is expected.

LIVE SO FAR · EARLY

Loading the live record…

RAW RECORD · POINT-IN-TIME

The raw record — verify it yourself

Every daily score is saved before the outcome is known. Below is the most recent day whose 7-day outcomes have fully matured (2026-08-27) — and you can download the entire record as CSV and check every row yourself.

98 recorded score dates4794 matured 1-day outcomes4495 matured 7-day outcomes10 excluded (data anomalies)Download full record (CSV) ↓
AssetIOXP·7dFundingEntry1d7dStatus
SAND10055.2%-0.116%$0.0419-1.8%-6.6%matured
JST9850.9%-0.016%$0.101-3.4%+6.8%matured
ATOM9650.3%-0.001%$1.5+3.9%-3.1%matured
ZIL9450.2%+0.003%$0.00274-2.0%-4.1%matured
XTZ9250.1%+0.004%$0.226+1.7%-3.4%matured
ENJ9050.1%+0.005%$0.0269-2.3%-8.4%matured
AXS8850.1%+0.005%$0.940-0.2%-2.3%matured
TRX8650.0%+0.006%$0.336+0.3%-3.5%matured
FIL8450.0%+0.007%$0.704-0.6%+13.7%matured
ADA8249.9%+0.008%$0.210+1.3%-4.1%matured
HBAR8049.9%+0.009%$0.0780+0.8%-4.3%matured
DOT7849.9%+0.009%$0.855+2.7%+0.8%matured
BNB7649.9%+0.009%$704.41+1.1%-2.2%matured
AVAX7349.9%+0.009%$7.37+2.0%-2.7%matured
INJ7149.9%+0.009%$5.45+0.4%-12.7%matured
SOL6949.9%+0.009%$99.72+9.1%+0.3%matured
FET6749.9%+0.009%$0.164-0.4%-7.3%matured
NEAR6549.9%+0.010%$1.88+1.9%-0.4%matured
ETH6349.9%+0.010%$2,493.38+0.5%-4.2%matured
XLM6149.8%+0.010%$0.183+1.9%-4.2%matured
CRV5949.8%+0.010%$0.313+4.2%+14.9%matured
ALGO5749.8%+0.010%$0.0891+2.1%+4.5%matured
ZEC5549.8%+0.010%$812.89-0.6%-0.0%matured
YFI5349.8%+0.010%$2,312.64+1.1%-5.0%matured
VET5149.8%+0.010%$0.00635+14.1%+2.6%matured
UNI4949.8%+0.010%$4.37+7.0%+33.5%matured
RUNE4749.8%+0.010%$0.477+6.7%-1.1%matured
THETA4549.8%+0.010%$0.176+1.4%-5.0%matured
GRT4349.8%+0.010%$0.0175+0.8%-7.3%matured
SUSHI4149.8%+0.010%$0.199-0.9%-3.1%matured
XRP3949.8%+0.010%$1.4+3.3%-3.7%matured
NEO3749.8%+0.010%$2.28-3.0%-12.0%matured
LTC3549.8%+0.010%$50.1-0.9%-0.8%matured
KSM3349.8%+0.010%$3.51+1.2%+2.3%matured
IOTA3149.8%+0.010%$0.0421-0.0%-6.8%matured
ICP2949.8%+0.010%$2.38+3.0%+2.6%matured
SNX2749.8%+0.010%$0.227-0.3%-8.2%matured
ETC2449.8%+0.010%$7.79+0.2%-6.8%matured
EOS2249.8%+0.010%$0.0773-0.2%-1.1%matured
EGLD2049.8%+0.010%$3.4+3.0%+52.6%matured
DOGE1849.8%+0.010%$0.0866+2.4%-5.8%matured
MANA1649.8%+0.010%$0.0743+1.5%-1.2%matured
DASH1449.8%+0.010%$38.58+2.5%+10.3%matured
COMP1249.8%+0.010%$19.5-0.7%-1.8%matured
CHZ1049.8%+0.010%$0.0141+1.7%-5.2%matured
LINK849.8%+0.010%$11.52+3.0%-3.9%matured
BAT649.8%+0.010%$0.0675+2.6%+0.2%matured
AAVE449.8%+0.010%$125.98+1.7%+1.0%matured
BCH249.8%+0.010%$266.47+0.7%-8.6%matured
XMR049.3%+0.022%$437.83+4.6%+13.0%matured

Raw stored returns (absolute, not excess-vs-field). Model: iox-v1 · S=-z(funding_7d) · 1 factor (out-of-sample-tested). Early sample — not yet statistically significant; the full CSV is the complete record, including any gaps. Excluded rows are outcomes we quarantine as data anomalies — a coin’s price feed broke (e.g. a token redenomination), so its realized return is meaningless; the raw value stays in the CSV (flagged) but is kept out of every metric. One convention note, in the open: each row stores the point-in-time entry price used when the score was saved, and every return from 16 Jun 2026 onward reproduces exactly as (the day-N opening price ÷ that entry price − 1). Cohorts on or before 15 Jun 2026 were filled with an earlier exit convention (the price at fill time, not the day’s open), so some of those early rows won’t reproduce to the cent — the differences are a fraction of a percent, with no look-ahead and no change to any direction.

02

What publishes here — once it’s significant

When the live sample is large enough to mean something (months, not weeks — non-overlapping windows make honest significance slow), this page will show the real numbers, against a clear benchmark:

  • Calibration — does “P = 54%” actually land near 54%? (reliability curve)
  • Hit-rate by IOX bucket — do higher scores outperform more often?
  • Average forward return by bucket — and the spread top-vs-bottom, after the costs a real trade pays.
  • Interval coverage — does the 80% uncertainty band actually contain ~80% of outcomes?

Until then: this is early, directional data — not a performance claim, and not investment advice. Survivorship-corrected backtests informed the design, but a backtest is a hypothesis, not a result.

The bar is fixed in advance. “Significant” here is not “whenever a number first crosses a line” — that is how honest-looking results get manufactured (check every day, publish the moment it dips). The threshold was written into the code before the live sample matured, so the goalposts can’t move: 26 non-overlapping 7-day windows (~6 months of live data), a mean rank-IC whose 95% interval excludes zero at |t| ≥ 2, still holding after delisted coins are added back, after real trading costs, and in both calm and stormy regimes. We publish the verdict when that pre-set bar is reached — whatever it says, pass or fail.

03

What already failed — shown, not buried

The honesty is the product. Candidate factors that didn’t survive the same gauntlet funding crowding passed were dropped — and we show them:

1· Out-of-sample IC2· Survives FDR3· Survivorship-robust4· Tradeable + adds over funding

Market-level candidates — stablecoin “dry powder”, options skew / DVOL, futures term-structure basis, implied-minus-realised vol — were also tested, but each is one number for the whole market (or covers only 1–2 coins), so it cannot RANK a 50-coin cross-section. Their honest home is a regime/context readout, never a per-coin score.