Are Drishti's AI crypto trading signals accurate?

Drishti publishes AI-generated crypto perpetual-futures trading signals for BTC, ETH and other assets about every 15 minutes on Delta Exchange India, and tracks every outcome publicly — wins and losses, never deleted. As of 2026-07-16: a 50% win rate and 1.28 average reward-to-risk (R) across 580 signals (191 wins, 188 losses, 4 open), with a Sharpe ratio of 1.98.

Will Bitcoin or Ethereum go up tomorrow?

Drishti does not predict a single price — it issues directional signals (long or short) with a specific entry, stop-loss and take-profit, then publishes whether each one won or lost. The current live calls for BTC, ETH and other perpetual futures are shown below and updated continuously.

Latest AI trading signals — BTC, ETH & crypto perps

SOLUSD SHORT — entry 76.11, TP 74.10, SL 77.55 · open
XAUTUSD SHORT — entry 4,034, TP 4,006, SL 4,058 · open
PAXGUSD SHORT — entry 4,022, TP 3,992, SL 4,042 · open
SLVONUSD SHORT — entry 51.88, TP 51.15, SL 52.55 · open
SLVONUSD SHORT — entry 51.55, TP 50.35, SL 52.45 · lost
XRPUSD LONG — entry 1.12, TP 1.14, SL 1.10 · lost
DOGEUSD LONG — entry 0.0747, TP 0.0760, SL 0.0736 · lost
BTCUSD LONG — entry 65,377, TP 66,470, SL 64,550 · lost

How Drishti works and how the track record is measured

← Back to signals

How Drishti works

Every 15 minutes, around the clock, Drishti asks a frontier AI model for one trade idea on each of eleven crypto & metal perpetuals — then tracks every call honestly to its outcome. Here is the whole journey, from raw market data to a scored signal.

ModelClaude Opus 4.8 Prompt— Data feeds— Assets— Cadenceevery 15 min Win rate—

1 The 15-minute loop

One cycle, four moves — then it repeats, forever. Tap any step to see what happens inside it.

Under the hood this is three phases on a server — track the open signals, decide new ones, publish the result — running about 96 times a day.

2 What we look at

Before each decision the AI is handed 144 engineered features per asset, grouped into the eight families below and pulled live from thirteen independent data sources. In plain terms:

Price & momentum

Is the trend up, down, or sideways — and how strong? Read across four time-frames at once (5m to 4h).

EMA trend stack
RSI, ATR, ADX strength
multi-timeframe alignment

Delta · Binance

Crowd positioning

Are traders leaning too far long or short? Funding rates and open interest reveal where the crowd is — and where it might get squeezed.

funding rate + 30d z-score
open interest & 24h change
top-trader long/short %

Binance · CoinGlass · Coinalyze

Options & fear

What is the market paying to hedge, and how greedy or fearful is sentiment right now?

25-delta skew & implied vol (BTC/ETH)
liquidation pressure
Fear & Greed index

Deribit · Alternative.me

Big money

Are the large spot Bitcoin and Ether ETFs taking money in or letting it out this week?

net ETF inflow 1d / 5d
flow velocity

Farside (BTC/ETH)

Macro backdrop

The weather outside crypto — the dollar, stocks, rates and gold — plus how much capital is sitting in stablecoins waiting.

DXY, S&P 500, VIX, US 10Y, gold
BTC dominance
stablecoin supply trend

yfinance · CoinGecko · DefiLlama

How far it usually moves

A read on each asset's typical daily range, so a target is set to a distance price can realistically cover within a day.

7d / 30d average daily range
busy-day range
cross-asset correlation

Delta · derived

Cross-venue & flow

Do the other big venues agree? Funding, basis and order-flow read across exchanges catch a divergence before price reacts.

spot–perp basis & Coinbase premium
Hyperliquid & Bybit cross-venue funding
spot-vs-perp CVD & global long/short
open-interest concentration by exchange

CoinGlass · Hyperliquid · Bybit

Volatility & extremes

How fast is it moving versus its own history — and is price stretched to an extreme that tends to snap back?

IV term structure + implied-vs-realized premium
realized-vol percentiles (7d / 30d)
distance from recent highs & lows
mean-reversion z-score

Deribit · derived

144 features8 families · 13 data sources

Claude Opus 4.8temperature 0 · reasoning first · fixed schema

DrishtiDecisiondirection · levels · expectation

Signalpublic card + Slack alert

If a feed is down it simply shows n/a — one missing source never stops the cycle.

3 How to read a signal

Every signal is one structured answer the AI fills in. Hover or tap a label to see what each part means.

BTCUSD long confidence 72 bull_trend

Entry$104,200

Stop-loss$102,600

Take-profit$107,400

R : R1.3

Target+2.0%

Regime: 1h and 4h EMAs stacked bullish, ADX rising — a real trend, not chop.
Thesis: funding is only mildly positive and ETF flows turned net-positive, room to run up.
Pre-mortem: a stop below the 102.6k swing low invalidates the idea cleanly.

signals in the 1.0–1.5 R:R band have run ≈54% historical win rate

In plain English The trend is up and the crowd is not over-leaned, so this bets on a push higher, with a clear exit if it breaks down.

4 Honesty & track record

A forecast no one keeps score of is just noise. Every Drishti signal is tracked to its outcome — and the scoring leans against us, never for us.

— resolved win rate

—signals tracked

—won

—lost

—expired

A clean denominator

Win = take-profit hit first. Loss = stop hit first. If neither happens within 48 hours it expires and is excluded — expiries never pad the win rate either way.

Pessimistic on ties

If a single minute-bar touches both the stop and the target, we record it as a loss. On price gaps we fill at the worse price. We would rather understate results than flatter them.

Calibration, graded

The AI states its own odds (take-profit / stop / expire). When the signal closes, those odds are scored against what actually happened — so we can tell whether its confidence is real or just talk.

Illustrative only — not financial advice. Drishti generates and tracks signals to study how AI reasons about markets. Nothing here is a recommendation to trade.