> ## Documentation Index > Fetch the complete documentation index at: https://docs.askfutures.com/llms.txt > Use this file to discover all available pages before exploring further. # Spread z-score, COT and correlation > Build an advanced ES trade gated by an ES-NQ spread z-score, weekly Commitment-of-Traders positioning, and a rolling correlation filter. This is the deep end. You'll trade **ES** — but only when the *relationship* between ES and NQ is stretched, the big commercial players agree, and the two markets are still moving together. You fade a statistical stretch in that relationship, and take the trade in one liquid leg. It pulls together four advanced ideas at once. If z-scores or Commitment-of-Traders data are new to you, read [signals & indicators](/concepts/signals-indicators-series) and the [trend-pullback tutorial](/tutorials/trend-pullback-adaptive-risk) first. **You'll learn:** using the **ES-NQ spread** as a signal, a **20-bar z-score** reversion trigger, a **weekly COT** positioning bias, a **30-bar rolling correlation** gate, trading **ES outright**, and exits at the mean or a dollar stop. **Time:** about fifteen minutes. ## The idea, in plain English Four conditions working together to time an ES trade: * **The instrument** — you trade **ES outright**, one market. * **The signal** — the **ES minus NQ spread**, tracked as a single series so you can measure how stretched the relationship is. * **The trigger** — when the spread's **20-bar z-score climbs above +2**, ES is unusually rich versus NQ, so **short ES**, betting the stretch snaps back. * **The bias** — only when **commercials are net short ES** in the weekly [Commitment-of-Traders](/concepts/where-data-comes-from) report (the hedgers' positioning agrees with the fade). * **The gate** — only when the two markets' **30-bar rolling correlation is above 0.8** (they're moving together tightly enough for the spread signal to mean something). * **The exits** — **cover when the z-score returns to 0** **or** on a **\$1,000 stop**. **Why trade ES alone, not the two-leg spread?** A spread built from two *different* markets (ES and NQ) can't be traded as a single instrument here — the engine trades **same-root** calendar spreads only (e.g. two CL contract months). So you use the ES-NQ spread purely as a **signal** and express the trade in one liquid leg, ES. You still fade the same stretch — you just place one order instead of two. ## Step 1 — Describe it Type it as one connected idea. AskFutures handles the spread construction, the z-score window, the weekly COT lookup, and the correlation gate from this single description: > Trade ES. When the ES-NQ spread's 20-bar z-score exceeds +2 and commercials are > net short ES in the weekly COT, short ES; cover when the z-score returns to 0 or > on a \$1,000 stop. Only when the 30-bar rolling correlation between ES and NQ is > above 0.8. A spread-signal reversion behaves differently from a single-market breakout — there's no opening-range or intraday-session breakout to lean on. State your exits explicitly (mean and dollar stop, as above) so AskFutures doesn't fall back to a default it has to guess at. ## Step 2 — Read the strategy card There's a lot packed into this one. Walk the card top to bottom and confirm each piece. One market is **traded** — **ES**. NQ is only *read* (never traded) to build the ES-NQ spread signal and the correlation gate, and ES COT is read for the positioning bias. See [futures & symbols](/concepts/futures-and-symbols). `ES - NQ`, built as a single series the z-score is measured on. This is the **signal** that times your ES trade — not an instrument you trade directly. **Short ES when the spread's 20-bar z-score is above +2** — two standard deviations rich versus its own recent average. **Commercials net short ES** in the weekly report. COT is a `1w` source, so the weekly positioning is held constant across the bars inside that week. See [where data comes from](/concepts/where-data-comes-from). **30-bar rolling correlation between ES and NQ above 0.8.** When the two decouple, the spread stops mean-reverting reliably — so the gate stands the strategy down. **Cover ES at the mean** (z-score returns toward 0) **or** a **\$1,000 stop** — whichever comes first. See [risk & trade management](/concepts/risk-and-trade-management). The **Strategy Flow** chart makes the gating order clear: ```mermaid theme={null} flowchart TD COR{"ES-NQ 30-bar
correlation > 0.8?"} -->|no| W["Stand aside"] COR -->|yes| Z{"Spread 20-bar
z-score > +2?"} Z -->|no| W Z -->|yes| COT{"Commercials
net short ES?"} COT -->|no| W COT -->|yes| SH["Short ES"] SH --> X{"First to hit?"} X -->|"z-score to mean"| M["Cover at mean"] X -->|"-$1,000"| ST["Stop"] ``` COT data is **weekly and reported with a lag** — the published positioning reflects a snapshot from earlier in the week, not the live tape. Treat it as a slow-moving bias, not a precise timing tool. Check the **Assumptions** notes on the card for exactly how the weekly series is aligned to your bars. ## Step 3 — Backtest and read the results > Backtest it. With four conditions stacked — and one of them a slow weekly COT bias — expect a **small number of trades**. That's expected for a selective reversion strategy; the filters are doing exactly what you asked. Backtest results are **hypothetical and simulated**. No real trades were placed, so live outcomes can differ. Reported P\&L is net of modeled slippage (default **1 tick per round-turn trade**) and commission (default **\$2.50/side** for full-size contracts like ES). Past performance — actual or simulated — does not guarantee future results. Always test before you trade. What to check, in order: Four filters can leave you with very few — or zero — trades. If it's too sparse to judge, relax one gate (correlation above 0.7, or z-score above 1.5) and re-run. Mostly `cover at mean` exits is the shape you're hoping for — stretches that snapped back. A run of `stop` exits suggests ES kept running rich versus NQ, often a sign the correlation gate let through a regime that was actually breaking down. Compare against a version with the COT filter removed (next step) to see whether the positioning bias actually earned its place. ## Step 4 — Iterate Change one condition at a time and [compare versions](/guides/version-and-compare) to isolate what each filter contributes. > Remove the COT filter and backtest again. > Loosen the correlation gate to 0.7. > Use a z-score threshold of +1.5 instead of +2. > Also take the long side: buy ES when the z-score drops below -2. > Optimize the z-score threshold and the correlation gate. Handing the z-score level and correlation threshold to the [optimizer](/guides/optimize-a-strategy) is the natural way to map out which combinations would have held up — without running each one by hand. This example is **illustrative, not a recommendation**. It exists to exercise an ES-NQ spread signal, a z-score trigger, a weekly COT bias, a rolling-correlation gate, and dollar-stop exits all at once — not because this exact combination is profitable. Use it to learn the mechanics, then build and test your own ideas. ## Next steps How COT, correlation, and continuous-contract data are sourced and aligned. Z-score, rolling correlation, and the spread builder explained. Run the strategy with and without each filter, side by side. Why even a complex strategy produces the same numbers every time.