Experimental data research

Forecast ledger: track record and calibration

A timestamped record of what the model said, what horizon it named and whether a real archived snapshot exists for the outcome date.

Experimental section. Every page publishes its method, coverage and limitations; history accumulates only from real timestamped snapshots and is never back-filled.

Response generated: 2026-07-12 20:26 UTCHistory accumulating: 1 archived observation · 2026-07-12 to 2026-07-12

Updates: daily archive snapshot after the nightly data pipeline (~03:30 UTC); on weekdays an extra market reading at 15:30 UTC. Live tables refresh continuously.

Model signals today

BUY · 52NEUTRAL · 70SELL · 141

Mean model confidence by horizon

1 day
83%
7 days
70%
30 days
51%
1 year
23%

Confidence honestly decays with the horizon — a volatility-based estimate a year out cannot be as confident as tomorrow’s.

Research snapshot

Key measurements

Min. evaluations for summary
5
D1 exact evaluations
0
D7 exact evaluations
0
D30 exact evaluations
0
Y1 exact evaluations
0

Observed records

50 records shown.

“Actual” and “Direction hit” fill in as due dates arrive and archive snapshots for those dates exist — that is the point of the ledger: outcomes cannot be backfilled or retouched.

First evaluations as the archive matures: 1 day — 2026-07-13 · 7 days — 2026-07-19 · 30 days — 2026-08-11.

Date Ticker Signal Price Target 7d Target 30d Confidence 30d Actual 30d Direction hit
2026-07-120GSELL$0.18447$0.17812$0.1655540.0%
2026-07-122ZNEUTRAL$0.07373$0.07626$0.0812951.0%
2026-07-12AAVEBUY$98.23$101.612$108.30340.0%
2026-07-12ACESELL$0.0734$0.07236$0.070341.0%
2026-07-12ADASELL$0.16429$0.16551$0.1679348.0%
2026-07-12AERONEUTRAL$0.51799$0.51785$0.5175724.0%
2026-07-12AIXBTSELL$0.0192$0.01919$0.0191752.0%
2026-07-12ALGOSELL$0.08411$0.08305$0.0809658.0%
2026-07-12ALTNEUTRAL$0.00654$0.00677$0.0072155.0%
2026-07-12ANIMESELL$0.00281$0.00286$0.0029653.0%
2026-07-12APEBUY$0.15576$0.16112$0.1717341.0%
2026-07-12APEXSELL$0.27324$0.26703$0.2547550.0%
2026-07-12APTSELL$0.6199$0.62485$0.6346564.0%
2026-07-12ARSELL$1.981$1.971$1.95154.0%
2026-07-12ARBNEUTRAL$0.09381$0.09704$0.1034352.0%
2026-07-12ASTERSELL$0.62595$0.62087$0.6108359.0%
2026-07-12ATOMSELL$1.588$1.601$1.62967.0%
2026-07-12AVAXSELL$6.456$6.316$6.03856.0%
2026-07-12AVNTSELL$0.09706$0.09787$0.0994954.0%
2026-07-12AXSSELL$0.99703$0.98832$0.9710735.0%
2026-07-12AZTECNEUTRAL$0.01531$0.01584$0.0168847.0%
2026-07-12BABYSELL$0.01304$0.01271$0.0120523.0%
2026-07-12BANANANEUTRAL$3.467$3.587$3.82348.0%
2026-07-12BCHBUY$246.15$254.624$271.39258.0%
2026-07-12BERASELL$0.19234$0.18572$0.1726252.0%
2026-07-12BIGTIMESELL$0.00716$0.00704$0.0068163.0%
2026-07-12BIOSELL$0.02854$0.02799$0.0269143.0%
2026-07-12BLURNEUTRAL$0.01704$0.01763$0.0187920.0%
2026-07-12BNBSELL$579.29$588.667$607.22273.0%
2026-07-12BOMESELL$0.00041$0.00041$0.000457.0%
2026-07-12BRETTSELL$0.00535$0.00532$0.0052736.0%
2026-07-12BSVSELL$13.469$13.235$12.77251.0%
2026-07-12BTCNEUTRAL$64,206$65,381.02$67,705.9673.0%
2026-07-12CAKENEUTRAL$1.414$1.439$1.48867.0%
2026-07-12CASHCATNEUTRAL$0.16425$0.16425$0.1642590.0%
2026-07-12CCSELL$0.13579$0.13414$0.1308868.0%
2026-07-12CELONEUTRAL$0.07069$0.07312$0.0779434.0%
2026-07-12CFXNEUTRAL$0.04457$0.04474$0.0450861.0%
2026-07-12CHIPNEUTRAL$0.03273$0.03366$0.035540.0%
2026-07-12COMPNEUTRAL$16.967$17.372$18.17465.0%
2026-07-12CRVSELL$0.20776$0.2088$0.2108557.0%
2026-07-12DASHSELL$34.885$35.03$35.31753.0%
2026-07-12DOGESELL$0.07364$0.07343$0.0730269.0%
2026-07-12DOTSELL$0.84347$0.84475$0.8472863.0%
2026-07-12DYDXSELL$0.13322$0.1352$0.1391120.0%
2026-07-12DYMSELL$0.01562$0.01571$0.0158952.0%
2026-07-12EIGENNEUTRAL$0.2338$0.24185$0.2577820.0%
2026-07-12ENANEUTRAL$0.08105$0.08305$0.0870243.0%
2026-07-12ENSSELL$4.138$4.125$4.159.0%
2026-07-12ETCSELL$6.922$6.859$6.73669.0%

Sources and coverage

Primary market data: Hyperliquid public API. Historical comparisons use only archived Spottrench snapshots that actually exist.

Calculation method

Forecasts are frozen at generation time. An outcome is scored only when the required due-date snapshot exists; missing outcomes stay null and do not enter hit-rate or calibration metrics.

Snapshot method

Each horizon is scored separately on exact due-date snapshots: direction, normalized target MAE versus a no-change baseline, daily-observed target hit and confidence calibration.

Frequently asked questions

What does this ledger record?

Every day it freezes the model’s verdict, price and algorithmic targets for each market. When the due date arrives, the actual archived price is placed next to the frozen target — direction accuracy and error are computed only from those real pairs.

Can past forecasts be edited?

No. A forecast is written once, on the day it was made, and outcomes come only from snapshots that were actually saved later. If a snapshot is missing, the outcome stays empty rather than being reconstructed.

What does “confidence” mean?

A deterministic 0–100 estimate derived from recent volatility, not a neural network’s opinion. It honestly decays with the horizon: a one-year target can never be as certain as tomorrow’s.

When will accuracy numbers appear?

Directional accuracy and error versus a no-change baseline are published per horizon once at least five exact evaluations exist — after about a week for 1-day and 7-day horizons, and after a month for 30-day ones.

Known limitations

  • A short ledger is descriptive, not proof of predictive skill. Listing changes and overlapping horizons can also distort small samples.
  • Missing due-date snapshots remain null; they are never interpolated or backfilled.
  • Target hit means seen in a saved daily snapshot. A miss is reported only for a complete daily window; this does not claim an intraday high/low was or was not touched.
  • Summary rates are withheld until at least 5 exact evaluations exist for that horizon; y1 can remain accumulating much longer.
  • Early samples are exploratory and are not evidence of future accuracy.

Experimental statistics are not a recommendation or financial advice. Correlation does not establish causation, tracked wallets are only a sample, and past relationships may disappear.