Systematic Futures Execution — Lead Strategy: Caliber Patterns
Built on evidence, not intuition.
Caliber Patterns is Caliber's lead strategy for NQ and MNQ futures — classical chart-pattern detection, session-aware risk, and a development process where nothing gets promoted to live without a logged reference case behind it. Caliber's original structural-level engine (v1.7.1/v1.8) continues running alongside it.
NET P&L +$16,940PROFIT FACTOR 1.88MAX DRAWDOWN 17.8%WINDOW 90 DAYS
System
How Caliber thinks
Three principles govern every decision the engine makes — none of them optional, none of them adjustable mid-session.
Structural levels
Every entry is anchored to a real price level — prior day, week, or month highs and lows, Asian session extremes — not an indicator lagging behind price.
Session discipline
London, New York, and Asian sessions are treated as genuinely different markets — each with its own blackout windows, risk posture, and entry rules.
Evidence-gated
No feature ships without a logged reference case first. Backtest evidence precedes every version. Forward-test evidence precedes every live decision.
Performance — Caliber Patterns (Lead Strategy)
The numbers, as they actually are
Rendered in Caliber's own live dashboard format — the same readout used to monitor every session in production. This is the base MAIN line: the confirmed, live-tested Caliber Patterns product, not an experimental variant.
Backtest Complete90 Days · MNQ · Caliber Patterns MAIN
Metric
Value
Detail
Total P&L
+$16,940.00
—
Profit Factor
1.88
—
Max Drawdown
—
17.8%
Backtested performance on historical data — not a projection of future results. Simulated and live execution can diverge; Caliber's own development process exists specifically to close that gap. Full variant comparison, including win rate and trade count, is in the Caliber Patterns section below.
Also Running — Caliber v1.7.1/v1.8
Caliber's original structural-level engine continues running in production alongside the new lead strategy. Its own schedule-effect finding, shown below, is specific to v1.7.1 — not Caliber Patterns.
Full Trading Week (incl. Friday)
Total P&L+$13,743.00
Win Rate67.31%
Profit Factor2.005
Max Drawdown$3,588.50
Tue–Thu Only (practiced schedule)
Total P&L+$14,853.00
Win Rate73.17%
Profit Factor2.582
Max Drawdown$3,348.50
Methodology
Nothing ships on a hunch
The process a feature has to survive before it ever reaches a live account.
01
Reference case logged first
Every proposed change starts as a documented, observed instance — not a theory about what might work.
02
Backtest across the full window
Historical evidence is gathered before a single line of live logic changes, with walk-forward validation against unseen data.
03
Forward-tested before trusted
A change earns its place in production only after proving out in real, forward-moving market conditions.
04
Version-locked, not patched live
Complete builds ship at defined milestones. No incremental edits to a script that's actively trading.
On honesty about performance
Backtested and simulated results are a starting point, not a guarantee. Every strategy's internal record and its real, broker-confirmed outcome are two different things — and the gap between them is treated as seriously as any other risk in the system.
Caliber's own monitoring exists to catch that gap in real time, not just to report a return figure and move on.
Caliber Patterns — Lead Strategy
The primary engine now, built the same way
A standalone classical chart-pattern recognition system — evidence-gated independently of Caliber's original structural-level engine, and now the strategy Caliber leads with. v1.7.1/v1.8 continues running alongside it, not replaced by it.
Nine classical formations
Head & Shoulders and Inverse H&S, Double Top and Double Bottom, Rising and Falling Wedges, Ascending and Descending Channels, and Symmetrical Triangles — detected from swing-pivot structure, not a lagging indicator.
Kept deliberately separate
Its own codebase, its own version line, its own evidence trail. Nothing here touches Caliber's live structural-level files — a failed idea stays contained, a working one doesn't inherit token debt it didn't earn.
Indicator first, bot second
Shipped as pure pattern detection before a single order was ever placed. Only after the read was trusted did it become a single-entry bot with a fixed bracket — and now a P1/P2/runner fork under separate test.
Backtested on historical data, $0 commission/slippage modeled — see the note below before reading this as a forward-looking number.
On honesty about this number
This is the strongest of five variants tested on the same 90-day window — and it's still an experimental trial, not the base "Caliber Patterns Bot" product. It adds a higher-timeframe directional bias filter (a 4H EMA read) on top of the same pattern-detection engine: longs only fire above it, shorts only below. The base MAIN line, running the confirmed asymmetric long/short bracket without this filter, returned +$16,940 (PF 1.88, 17.8% max drawdown) over the identical window — a real, live result, just a smaller one.
Worth knowing before reading too much into the number above: one of the other four variants tested on this same window (a P1/P2/Runner partial-exit fork) showed a max drawdown that took simulated equity negative against its $10,000 starting balance, driven entirely by its long side. Five different exit and filter architectures, tested on identical data, producing results this far apart is itself the evidence for why nothing here gets promoted to the base bot without its own forward-test track record — a strong backtest number is a candidate, not a conclusion.
