Market

Prices refresh every 3 seconds. Screening and evaluations, every 15.

System

Live feeds
8
of 25 collected
Freshest bar
2m
ago
Decisions / hour
164
87 passed every gate
Signal chain
Intact
1385 verified

Serving crypto_direction_h15 v202608131548. Decisions are counted from rows the predictor wrote — a process that is running but deciding nothing shows as zero here rather than as healthy.

Performance

Signals sent
1385
1223 resolved
Hit rate
53.6%
needs 54.05%
Won / lost
655 / 568
settled at the expiry bar
Expectancy
-0.009
per unit staked

Live feeds

25 instruments
InstrumentLast1hHighLowBarsAge
BTCUSDT crypto78,760.59-0.21%78,760.5978,760.591k2m
AVAXUSDT crypto7.5360+0.12%7.53607.53601k2m
DOGEUSDT crypto0.09002+0.28%0.090020.090021k2m
ADAUSDT crypto0.2221+0.51%0.22210.22211k2m
XRPUSDT crypto1.4853+0.39%1.48531.48531k2m
BNBUSDT crypto706.0800+0.63%706.0800705.70001k2m
SOLUSDT crypto100.1300+1.88%100.130099.73001k2m
ETHUSDT crypto2,478.95-0.26%2,478.952,478.111k2m
GBP/USD (ECN) forex1.3639+0.04%1.36391.36391k33m
EUR/JPY (ECN) forex185.6620+0.05%185.6840185.65801k33m
EUR/USD (ECN) forex1.1669+0.04%1.16691.16691k33m
NZD/USD (ECN) forex0.5967+0.10%0.59680.59671k33m
USD/CAD (ECN) forex1.3842+0.00%1.38421.38421k33m
USD/JPY (ECN) forex159.1050+0.01%159.1240159.09901k33m
GBP/JPY (ECN) forex217.0100+0.05%217.0220217.01001k37m
USD/CHF (ECN) forex0.8023-0.00%0.80230.80221k37m
AUD/USD (ECN) forex0.7154+0.06%0.71540.71541k37m
XAU/USD (broker feed) metal4,415.45+0.23%4,415.594,414.950292h
GBP/USD OTC forex1.3533+0.10%1.35391.35320292h
GBP/USD (broker feed) forex1.3490-0.02%1.34911.34900292h
AUD/CAD OTC forex1.0435-0.08%1.04371.04300292h
EUR/USD (broker feed) forex1.1523-0.01%1.15231.15230292h
EUR/USD OTC forex1.1682-0.04%1.16831.16810292h
BTC/USD (broker feed) crypto63,374.4-0.21%63,394.2463,372.20292h
XAU/USD OTC metal4,359.61+0.01%4,359.614,359.590292h
How we know  — the measurements behind every claim here

Not established. One result clears on paper; none survive being counted honestly.

Best hit rate across 16 evaluations was 51.11%, against a 54.05% breakeven at an 85% payout. Below that line every trade loses money on average.

None of the 6 evaluations measured against a one-line rule beat it — always-up, momentum, or bet-against-the-last-bar, on the same rows. Whatever the models found is reproduced by betting against the previous bar, which is the bid-ask bounce rather than the market, and it is not capturable: that reversion sits inside the spread you cross to open the position.

1 of 16 clears breakeven on the raw trade count, and 0 still does once overlapping trades are counted once. A 15-minute call entered at 10:00 and another at 10:01 share fourteen of their fifteen minutes; they are not two independent pieces of evidence.

Gradient boosting on 40 indicator features

InstrumentExpiryMeasured onHit rate95% lowNeedsOne-line ruleBeats itSamples
EUR/USD (ECN)1mfresh51.11%54.05%51.15%no101k
EUR/USD (ECN)1mall bars51.96%54.05%148k
EUR/USD (ECN)5mfresh50.14%54.05%50.52%no105k
EUR/USD (ECN)5mall bars50.77%54.05%154k
EUR/USD (ECN)15mfresh50.14%54.05%50.45%no106k
EUR/USD (ECN)15mall bars50.91%54.05%156k
GBP/USD (ECN)1mfresh50.88%54.05%51.12%no85k
GBP/USD (ECN)1mall bars51.52%54.05%155k
GBP/USD (ECN)5mfresh50.36%54.05%50.64%no88k
GBP/USD (ECN)5mall bars50.86%54.05%161k
GBP/USD (ECN)15mfresh50.16%54.05%50.33%no88k
GBP/USD (ECN)15mall bars50.32%54.05%163k

Sequence model (TCN + transformer)

InstrumentExpiryMeasured onHit rate95% lowNeedsOne-line ruleBeats itSamples
pooled1mall bars51.15%54.05%55k
pooled5mall bars51.19%54.05%119k
pooled15mall bars51.42%54.05%137k

Model A + Model B meta-labelling gate

InstrumentExpiryMeasured onHit rate95% lowNeedsOne-line ruleBeats itSamples
pooledpooledall bars56.64%52.52%54.05%3k → 564

fresh counts only bars whose recent quotes were moving; all bars includes stale ones, where a quote catching up looks like a prediction. pooled means the run mixed expiries or instruments and cannot honestly be split. Where samples read 3k → 564, the second number is how many independent observations remain after overlap — the confidence bound uses that one.

Volta · 1.1M bars across 25 instruments · signals are hash-chained and cannot be edited after the fact · not financial advice