A systematic NQ futures strategy that reads live order flow inside every volume bar and only trades when institutional conviction is overwhelming. Three locks. One edge.
Every candlestick is a summary. Green means up, red means down. That's all most traders see. But inside every bar is a forensic record — every buyer, every seller, every aggressive order at every price level. The candlestick is the cover. DeltaEdge reads the pages.
DeltaEdge requires three independent confirmations before entry. No single signal is enough. All three must open simultaneously.
While other strategies look at the outside of a candle, DeltaEdge looks inside. Using volumetric bars, the strategy reads bid/ask volume at every price level within each 1,000-tick bar.
The four highlighted rows show aggressive buying at 6.3x, 6.0x, 3.4x, and 3.0x the selling volume.
This isn't retail noise. Large participants are reaching up and taking liquidity at every price level. They're not waiting. They're pushing.
The Point of Control sits at 25068 — the price with the heaviest volume — confirming buyers are holding gains at elevated prices.
Backtested on NQ 100 futures across January 2025–February 2026. One contract. Fixed 22-tick target and 140-tick stop on 1,000-tick volumetric bars. Contract scaling to 10 is on the development roadmap — these results reflect the conservative single-contract baseline.
ORB breakouts, SuperTrend, and volumetric imbalances are established market mechanics. The rules weren't mined from historical trades — they were built from known order flow behavior, then validated against the data. Hypothesis first, confirmation second.
Three entry conditions, one fixed target, one fixed stop. There's almost nothing to overfit. Compare that to strategies with 15 tuned parameters — every added knob is a degree of freedom that molds itself to historical noise.
14 months of data spanning trending runs, choppy consolidation, FOMC volatility, earnings seasons, and varying VIX environments. A curve-fit strategy works in one regime and breaks in others. This held across all of them.
If the backtest were overfit, the forward test would degrade significantly. Instead it's tracking nearly identically — 89% backtest, 90% forward. That convergence is exactly what a non-overfit strategy looks like.
Live simulation since February 17, 2026. Real data, real fills, real slippage. Every trade logged with 56 data points.
Two tracks are running in parallel. The research track forward tests the full 22-tick target / 140-tick stop configuration — identical to the backtest — collecting the complete 56-column dataset without any modifications. This is the scientific baseline. No changes until the data earns them.
The funded track launched January 24, 2026 on an Apex 50K Performance Account with risk parameters adjusted for prop firm compliance — tighter 100-tick stop, maximum 8 trades per day. Same entry logic. Same three locks. Just a narrower risk envelope to meet funded account rules.
Full backtest parameters. Unchanged. Every trade feeds the 56-column dataset that powers the V2 and V3 roadmap. This data is never compromised.
Same entry logic. Tighter risk envelope. 8 consecutive wins in a single session is common — and the strategy remains consistency rule safe because no single trade represents an outsized portion of daily P&L.
DeltaEdge works. The research is about making it prop firm compliant and optimally efficient.
Losses cluster after extended winning streaks. If exhaustion is measurable — declining delta, weakening stacks — we skip the trade that stops out.
67% of winners show MAE under 20 ticks. Losers blow through 100+ immediately. A tighter stop could cut losses without flipping winners.
Every trade logs delta ratio, stack count, POC position, cumulative delta. Which variables correlate with wins? The 56-column dataset will reveal it.
When delta thins and imbalance quality degrades but price still rises — that's exhaustion. If it has a measurable signature, it becomes a filter.
Every trade generates a fingerprint across 56 dimensions. Trade mechanics, order flow, volumetrics, context, and excursion — all logged automatically.
Baseline performance first. Hypotheses from observed patterns. Data collection without changes. Statistical validation. Then — with evidence — implementation.
No premature optimization. No curve fitting. The strategy earns improvements through statistical proof, not storytelling.
Look — I'm no revolutionary quant. I'm just using good filtering. But the 56-column data engine isn't just logging trades. It's building the dataset that makes each version smarter than the last. Here's the plan, the ceiling, and why we'll never pretend otherwise.
You can't get to 100%. Here's the mechanical reason: your entry is based on information at the moment of entry. But the market is moved by information that arrives after your entry — a sudden iceberg order, a headline, a large player deciding to unwind. No amount of historical pattern matching predicts a random exogenous event.
The 5–15% of trades that lose aren't pattern failures. They're genuinely unpredictable events. No model fixes that.
Every filter you add fits your historical data better. V2 with 2 filters is safe. V3 with a model trained on 56 features and 3,000 trades is borderline. V4 retraining weekly on 50 trades risks learning noise — patterns that existed by coincidence but don't repeat. Win rate looks amazing in backtest, falls apart live. That's the overfitting trap, and we refuse to fall into it.
The practical ceiling for a strategy like this is probably 90–93%. We're already close. Beyond that, every marginal percent costs more complexity, more fragility, and more risk. We'd rather run a robust 90% forever than chase a fragile 96% that breaks.