Hybrid Indices and HFV: Deriv’s New Synthetics for Algorithmic Scalping (2026)

Deriv’s Synthetic Indices catalog has more than doubled in 12 months. In 2026 it gained the new Deriv Hybrid Indices (Crash/Boom + Volatility), High Frequency Volatility (HFV) with 2 ticks per second, and the new Crash/Boom 50 and 150. For anyone trading with EAs and AI, that means new battlegrounds — and new opportunities.

Why synthetic indices matter for AI-powered EAs

Synthetic Indices are a Deriv-exclusive product: algorithmically generated indices that don’t depend on real markets. Advantages for algorithmic trading:

• 24/7, no breaks: the market never closes, ideal for continuous backtesting
• Engineered volatility: each index has a defined statistical profile (Vol 75 = 75% annualized volatility)
• Zero news risk: no NFP, FOMC, or politician’s tweet moves the price
• Infinite liquidity: Deriv is the counterparty, so there’s no slippage from a thin order book
• Faithful backtesting: the same engine runs live and in the tester (rare among CFDs)

1. Deriv Hybrid Indices — the new format

Launched in 2026, it combines two classic synthetic behaviors:

• Crash/Boom behavior: structured directional movement followed by sudden spikes/drops
• Volatility behavior: continuous, random fluctuation

The result: the asset trends for extended periods, then enters an unstable/choppy phase, and finally produces a spike event (up or down) with variable timing.

Annualized volatility: roughly 20% (lower than traditional Crash/Boom, which runs 40-60%).

Ideal for: EAs that combine trend detection + unstable-regime recognition + anti-spike filtering. Long-sequence LSTM/Transformer models (50+ candles) shine here.

2. High Frequency Volatility (HFV) — pure scalping

Added in April 2026, HFV Indices have one crucial difference from traditional Volatility Indices: they generate 2 ticks per second (vs. 1 tick every 2 seconds on traditional Vol indices). That’s 4x more ticks per unit of time.

For AI-powered EAs: HFV indices are built for high-frequency algorithmic systems. With Build 5572 + CUDA, a small 1D CNN or LSTM model can run inference in <1ms, matching the pace of new ticks.

3. Crash/Boom 50 and 150 — rounding out the family

Added to the catalog in May 2026. The numbers represent the expected average spike frequency (in ticks):

• Crash 50 / Boom 50: a spike roughly every ~50 ticks — more frequent, more aggressive
• Crash 150 / Boom 150: a spike roughly every ~150 ticks — rarer, more predictable

Together with the classic Crash/Boom 300, 500, and 1000, you now have 5 spike-frequency levels to choose from. For EAs running specific strategies (anti-spike, ride-the-spike), that allows for fine calibration.

Crash/Boom 150 has already logged over $10 billion in volume since launch, showing strong retail demand.

Full Deriv Synthetic Indices catalog (updated May 2026)

Strategy: an HFV 75 scalper EA with a 1D CNN

A practical combination of the 2026 additions:

//+------------------------------------------------------------------+
//| HFV75 Scalper - 1D CNN ONNX + Build 5572 CUDA                   |
//+------------------------------------------------------------------+
#property strict
#resource "\\Files\\cnn1d_hfv75.onnx" as uchar Model[]

long handle = INVALID_HANDLE;
datetime last_signal = 0;

int OnInit()
{
    // Build 5572+: GPU CUDA + granular logging
    handle = OnnxCreateFromBuffer(Model, ONNX_GPU_DEVICE_N | ONNX_LOGLEVEL_WARNING);
    if(handle == INVALID_HANDLE) return INIT_FAILED;

    const long in_shape[]  = {1, 30, 4};  // 30 ticks, OHLC
    const long out_shape[] = {1, 2};      // BUY/SELL probabilities
    OnnxSetInputShape(handle, 0, in_shape);
    OnnxSetOutputShape(handle, 0, out_shape);

    EventSetTimer(1);  // 1 second - HFV generates 2 ticks/s
    return INIT_SUCCEEDED;
}

void OnTimer()
{
    if(TimeCurrent() - last_signal < 5) return;  // 5s cooldown

    matrix input(30, 4);
    for(int i = 0; i < 30; i++) {
        input[i][0] = (float)iOpen(_Symbol, PERIOD_M1, i);
        input[i][1] = (float)iHigh(_Symbol, PERIOD_M1, i);
        input[i][2] = (float)iLow(_Symbol, PERIOD_M1, i);
        input[i][3] = (float)iClose(_Symbol, PERIOD_M1, i);
    }

    vector output(2);
    if(!OnnxRun(handle, ONNX_DEFAULT, input, output)) return;

    // High threshold for scalping (rare but reliable signals)
    if(output[0] > 0.75) { /* BUY signal */ last_signal = TimeCurrent(); }
    else if(output[1] > 0.75) { /* SELL signal */ last_signal = TimeCurrent(); }
}

Risks specific to synthetics

• They’re not real markets: there’s no economy, flow, or structure behind them — just a cryptographically secure RNG
• Perfect backtest ≠ perfect live: you still deal with spread, latency, and risk management
• Trading addiction: 24/7 availability + easy access = real psychological risk. Set daily limits
• Spike risk on Crash/Boom: a wide stop loss is mandatory — a 30% spike in a single tick is normal
• Single-broker concentration: Synthetic Indices only exist on Deriv. Diversify your exposure