PMTS Year-End 2025: 53.60% Return in 6 Months, AI Pattern Recognition, and What We Learned

Six months ago, we wrote that institutional capital was deployed and the system was “performing.” Today, we can be specific.

PMTS delivered a 53.60% return over the second half of 2025.

Verified. Auditable. Every trade visible on the public dashboard. No cherry-picking. No hypothetical backtests. Real capital, real broker, real results.

This post breaks down what happened, what the AI improvements contributed, and — because honesty is a core value — what went wrong along the way.

The Numbers — In Context

Before we celebrate, let us put 53.60% in proper context:

• Time period: July 1 – December 28, 2025 (approximately 6 months)
• Instrument: XAUUSD (Gold) exclusively
• Broker: MultiBank Group (regulated, 11+ licenses)
• Maximum drawdown during the period: ~9.7%
• Sharpe ratio: Above 4.0 (annualized)
• Win rate: Above 80%
• Total trades: 2,100+ across all accounts

These results were achieved during a period of extraordinary gold market activity. Geopolitical tensions, central bank gold purchases reaching record levels, and persistent inflation uncertainty created ideal conditions for a trend-following system with Smart Money detection. We do not expect 53% every six months. What we expect is that the system performs well across different market regimes — and 2025 was a powerful validation of the architecture.

Past performance does not guarantee future results.

AI Pattern Recognition — The Breakthrough of 2025

The single biggest technical advancement of the second half of 2025 was the integration of AI-powered pattern recognition into the trading pipeline. This goes beyond the AI Master Analyzer we built in 2023. Pattern recognition adds a visual intelligence layer that processes candlestick formations, chart structures, and price action patterns the way an experienced chartist would — but across 4 timeframes simultaneously and without fatigue.

How It Works

The system now identifies and classifies 23 distinct candlestick and chart patterns:

• Single-candle reversals: Hammer, inverted hammer, shooting star, doji (standard, dragonfly, gravestone, long-legged), spinning top, hanging man
• Multi-candle reversals: Engulfing (bullish/bearish), morning star, evening star, three white soldiers, three black crows, harami, piercing line, dark cloud cover
• Continuation patterns: Rising/falling three methods, tasuki gap, side-by-side white lines
• Structure patterns: Double top/bottom, head and shoulders, ascending/descending triangle, wedge, flag, pennant

Each pattern is detected algorithmically using precise mathematical definitions (not approximate visual matching), then assigned a reliability score based on historical accuracy for XAUUSD specifically. The AI Master weighs pattern signals alongside the existing technical, fundamental, and sentiment data.

Impact on Performance

In A/B testing during September-October 2025, the pattern recognition module improved entry timing by an average of 12 points (gold pips). It also reduced false signals by approximately 18% — particularly during range-bound markets where the previous system would sometimes misidentify consolidation as the beginning of a trend.

Real-Time News Adaptation — No More 15-Minute Delays

One of the criticisms we made of our own system in 2023 was latency during high-impact events. The 15-minute refresh cycle meant that by the time the AI processed a surprise Fed statement or an unexpected jobs report, the market had already moved 30-50 points.

In Q3 2025, we deployed the Smart Refresh 2.0 system that fundamentally changed how PMTS responds to breaking news:

1. Pre-event positioning (T-30 minutes): The system queries the economic calendar via Finnhub API. 30 minutes before any high-impact event (FOMC, NFP, CPI, ECB, BOJ), it triggers a full analysis refresh and adjusts position sizing and stop distances
2. Real-time event detection (T+0): When the event fires, our NLP engine processes the first headlines within seconds. If sentiment diverges significantly from consensus expectations, an immediate signal refresh is triggered
3. Post-event confirmation (T+15 minutes): A follow-up analysis runs 15 minutes after the event to assess whether the initial market reaction is holding or reversing — critical for avoiding fakeouts
4. Volatility-triggered refresh: Independent of scheduled events, any price movement exceeding 0.5% since the last analysis triggers an immediate recalculation

The result: during the November 2025 Fed meeting (which surprised markets with hawkish language), PMTS detected the bearish shift within 90 seconds of the statement release, adjusted its signal from BUY to NEUTRAL, and avoided a 40-point drop that would have hit the previous stop-loss level.

Intermarket Intelligence — Reading the Full Picture

Gold does not trade in isolation. In H2 2025, we significantly expanded the intermarket analysis capabilities:

• US 10-Year Treasury Yield (US10Y): Real-time tracking via Yahoo Finance. Rising yields = bearish for gold (opportunity cost of holding non-yielding assets). The AI now weights this as the single strongest correlation signal
• Real Yields (US10Y minus inflation expectations): The most predictive single indicator for gold direction. When real yields fall, gold rises — and vice versa. Our system now calculates and tracks this in real time
• DXY (US Dollar Index): Inverse correlation monitored continuously. Dollar strength = gold weakness
• VIX (CBOE Volatility Index): Fear gauge. VIX above 25 = safe-haven demand for gold
• Brent Crude Oil: Inflation proxy. Rising oil prices support gold as an inflation hedge
• S&P 500: Risk-on/risk-off indicator. Equity selloffs drive capital to gold

These six intermarket indicators are now combined into a Composite Intermarket Score ranging from -100 (extremely bearish for gold) to +100 (extremely bullish). The AI Master uses this score as a confirmation layer — it will not initiate a BUY signal when the intermarket score is below -30, regardless of what technical indicators suggest.

Session-Specific Forecasting

Another addition in H2 2025: the AI now generates individual forecasts for each trading session:

• Asian Session (01:00-09:00 CET): Typically lower volume, narrower ranges. The AI tracks Shanghai Gold Exchange premiums and PBOC activity for Asian-specific catalysts
• European Session (09:00-14:30 CET): London Fix, ECB decisions, European macro data. The kill zone (08:00-11:00) is where the day’s range is typically established
• US Session (14:30-22:00 CET): Highest liquidity, US macro releases, COMEX close. The kill zone (14:00-17:00) produces the strongest moves

Each session forecast includes direction, probability, expected price range, key catalysts, and recommended entry windows. This information is now visible on the AI Analysis page for all users.

The Months That Tested Us

Not every month was positive. Transparency demands we discuss the difficult periods:

August 2025: The Consolidation Trap

Gold entered a 3-week consolidation range between $2,480 and $2,520. The system generated multiple false breakout signals, leading to a series of small losses. The pattern recognition module was not yet fully deployed, and the old system misidentified range compression as trend initiation.

Lesson: This directly accelerated the deployment of pattern recognition, which now identifies consolidation patterns (triangles, rectangles, wedges) and suppresses trend-following signals during range-bound conditions.

October 2025: Flash Crash Recovery

A flash crash driven by cascading stop-losses on October 14th dropped gold 80 points in 4 minutes. Our trailing stop on one position was hit 20 points below the calculated level due to slippage — a harsh reminder that backtested stop-losses and real stop-losses are not the same thing.

Lesson: We adjusted the ATR multiplier for stop-loss calculation from 1.5x to 1.8x during high-volatility regimes (VIX above 25), providing additional buffer for slippage events.

The Expert Advisor — PMTS Signal Trader

In Q4 2025, we developed a new Expert Advisor: the PMTS Signal Trader. Unlike the original AI Trader (which reads signals and executes), the Signal Trader introduces a fundamentally different approach to losing positions:

It never closes in losses.

Instead of closing a position when the signal reverses, the Signal Trader opens a hedging counter-position at 50% of the original size. It then manages the pair — the original position and the hedge — until one or both can be resolved profitably. This approach reduces the psychological impact of realizing losses and allows time for the original thesis to play out.

The Signal Trader also incorporates all the new data streams: session forecasts, intermarket composite score, H4 candle prediction, and a 10-layer close validation system (up from the original 7 layers) that includes intermarket alignment and session confirmation.

Platform Growth

Some numbers from the second half of 2025:

• Active trading accounts: Growing from single digits to double digits across multiple countries
• Languages served: 7 (English, Spanish, Arabic, German, Portuguese, Hindi, Chinese)
• Blog launched: You are reading it — 6 posts documenting the complete PMTS journey
• XAUUSD Newsletter subscribers: Growing organically through the analysis page
• Accounting system: Full UAE-compliant module with IFRS 9 investment tracking, VAT calculations, and share links for external accountants

What 2026 Will Bring

Our priorities for the year ahead:

1. Scaling capital under management — With a verified 6-month track record, we are in a stronger position to onboard institutional capital
2. Multi-instrument expansion — XAUUSD remains the core, but we are testing the system on EURUSD, GBPJPY, and crude oil with promising early results
3. AI next generation — Evaluating Claude (Anthropic) and Gemini (Google) alongside GPT-4o for potential multi-model consensus — the same principle we apply to trading bots, applied to AI models
4. Regulatory expansion — Exploring additional regulatory frameworks to serve clients in the EU and GCC region
5. Independent performance audit — Commissioning a third-party verification of all trading results

A Decade in One Paragraph

In 2015, one developer hypothesized that algorithms could read institutional order flow. In 2017, we proved confluence works. In 2019, we went modular. In 2020, risk management saved us. In 2022, we built infrastructure and hired a team. In 2023, AI gave us contextual reasoning. In 2024, we opened the platform to investors. In 2025, institutional capital arrived, and the system delivered 53.60% in six months.

We do not know what the next decade holds. But we know the foundation is solid, the team is exceptional, and the results are real.

Past performance does not guarantee future results. Trading involves substantial risk of loss. Only invest capital you can afford to lose.

— Lorenzo Ballanti, Founder & CEO, Elysium Media FZCO, Dubai Silicon Oasis, UAE

All performance data referenced in this post is available for verification on the PMTS Live Dashboard. We encourage due diligence.