Titan Analytics Engine:
Pipmaster France’s proprietary analytics infrastructure
Most Expert Advisors are developed almost exclusively based on backtests. At Pipmaster, we have chosen a complementary approach: observing, measuring, and using artificial intelligence to analyze the actual behavior of Titan Breakout under real-world market conditions.
The goal is not to present a “magical artificial intelligence” capable of predicting the market. Artificial intelligence is used as an analytical tool capable of leveraging data from the trading robot, technical logs, rejected signals, executed trades, and market context.
A comprehensive evaluation of each H1 signal
At the close of each H1 candle, Titan Breakout does not simply stop at the first unmet condition. All of the strategy’s filters are systematically evaluated: candlestick structure, EMA trend, ADX strength, ADX slope, Donchian breakout, Kill Zone, structural context, and imbalance zones.
Each filter is scored individually (as valid or invalid), and the complete results are stored in a database. This comprehensive matrix provides precise information on which filters blocked each signal, under what market conditions, and how frequently.
Simulate the rejected signals to understand their true value
When a signal is rejected by a filter, a fundamental question remains unanswered: Would that trade have been profitable if it had been executed? The Ghost Trades mechanism answers this question by simulating, in the backend, each rejected signal using the same management rules as for actual trades.
The same stop-loss, the same take-profit, the same exit rules. Titan Breakout did not place the order, but the backend simulates the result using real H1 data and records the outcome: TP reached, SL hit, or trade expired.
Ghost Trade - Simulation backend
│
├── Signal rejeté par un ou plusieurs filtres
├── Backend simule le trade sur données H1 OHLC réelles
├── Même SL / TP / Règles de gestion que les trades réels
└── Résultat enregistré : TP, SL, expiration, PnL simulé (R)
│
▼
Comparaison Ghost vs Réels
│
├── Si ghost WR > réel WR → Filtre potentiellement trop restrictif
├── Si ghost WR ≈ réel WR → Filtre pertinent
└── Si ghost WR < réel WR → Filtre protège efficacement
Track what happens during and after each trade
For each live trade, the system continuously measures two key price movements over the life of the position: The MFE (Maximum Favorable Excursion – how far the price moved in the desired direction before closing) and the MAE (Maximum Adverse Excursion – how far the price moved against the position before closing).
These two metrics, expressed in R (risk multiples), help us understand the true profile of each trade: Did a trade that closed in profit nearly hit the stop-loss before reversing? Was a losing trade profitable at +2R at some point?
Post-close analysis goes a step further: Once the trade is closed, the backend continues to monitor the price over the next 24 H1 candles. The goal is to determine whether the original take-profit level would have been reached after the trade closed and, if so, over how many candles.
This information makes it possible to estimate the potential return that could have been achieved with a different management approach, and to identify recurring early exits.
How data is collected and analyzed
At the close of each H1 candle, Titan Breakout evaluates all market conditions and records a comprehensive set of results: Candlestick structure, EMA trend, ADX strength, Donchian breakout, Kill Zone, structural context, and areas of imbalance.
This data is sent in real time to a MySQL database hosted on OVH infrastructure. Automated scripts then take over: simulating ghost trades, calculating MFE/MAE, performing post-close analysis, and synchronizing external market data.
[H1 Candle Close]
│
▼
Titan Breakout - MT5
EvaluateNewBarSignal() - Comprehensive evaluation of all filters
│
├── Score: Body_ok, ema_trend, adx_ok, adx_slope, breakout
├── Score: KillZoneOK(), StructuralContextOK(), FVG
└── Records: Signal_generated + full score for each filter
│
▼
Secure HTTP transmission to the Pipmaster backend
│
├── ea_signals.php → H1 signals + full filter matrix
├── ea_logs.php → Technical logs
├── ea_trades.php → Open/Close
└── ea_snapshots.php → Daily snapshot
│
▼
OVH MySQL database
│
├── ea_signals → H1 signal matrix
├── ea_ghost_trades → Rejected signal simulation
├── ea_trades → Actual trades + MFE/MAE + post-close
├── ea_market_data → H1/D1 Twelve Data
└── ea_logs / ea_snapshots
MT5 logs are enhanced with data from Twelve Data
To avoid limiting the analysis to MT5 logs alone, the system also retrieves external market data via the Twelve Data API.
An automated script identifies the active patterns recently observed by the robot, then retrieves the corresponding H1 and D1 candlesticks. This data helps put Titan Breakout’s decisions into their real-world context: trend, volatility, price momentum, and market structure.
market_data_sync.php - OVH Cron
│
├── Reads the active symbols from the last 7 days
├── Queries the Twelve Data API
├── Retrieves H1 and D1 data
└── Inserts the data into ea_market_data
Data used:
├── H1: Approximately 168 candles - 7 days
└── D1: Approximately 30 candles - 30 days
A dedicated internal API: Direct access to the database for artificial intelligence
All collected data is accessible in real time via a secure internal API. This API allows you to query the OVH MySQL database directly, without an intermediary and without relying on files generated at fixed intervals. Each query returns the most recent data available.
Two dedicated Artificial Intelligence sessions (an ANALYSIS session and a DESIGN session) access this data on demand. Each time a session is opened, a compact snapshot of the system’s state is automatically loaded: current assumptions, recent decisions, deployed changes, and AI parameters.
ea_query.php — Direct MySQL API (OVH)
│
├── GET ?section=signals → H1 Signal Matrix (all scored filters)
├── GET ?section=ghost → Ghost Trades (simulations of rejected signals)
├── GET ?section=trades → Actual trades + MFE/MAE + post-close
├── GET ?section=market → H1 / D1 market data
├── GET ?section=state → Compact snapshot (AI session startup)
└── POST ?section=decisions → Decisions, assumptions, modifications
│
▼
Dedicated Artificial Intelligence sessions
│
├── ANALYSIS session → Behavioral analysis, data interpretation
└── DESIGN session → Technical modifications + database writing
Artificial intelligence as an analyst, not as a trader
Artificial intelligence does not play a role in executing trades. It does not replace strategy, does not adjust positions in real time, and does not claim to predict the future.
His role is different: analyzing data, comparing contexts, identifying recurring patterns, pinpointing the causes of rejection, studying drawdown periods, and helping to develop recommendations for improvement based on quantified observations.
An internal infrastructure that directly benefits users
Although this analytical infrastructure remains internal to Pipmaster, it directly benefits Titan Breakout users. The areas for improvement identified through real-world data allow us to update the EA on a regular basis, in a more structured and objective manner.
The goal is not to make random changes to the EA or to react emotionally to a few losing trades. The goal is to identify recurring trends, understand the system’s potential weaknesses, and gradually improve its robustness based on quantitative evidence.
Developing an EA as a living system
Serious algorithmic trading isn't about looking for a miracle robot. It involves continuously maintaining a consistent, measurable system that can be objectively analyzed.
Titan Breakout follows this approach: a structured strategy, strict risk management, continuous monitoring, and a commitment to constant improvement 🚀