Financial Technology / Algorithmic Trading

Paksa Hunter Pro: An MT5 Trading Agent Built to Work Live, Not Only in Backtests

Most trading robots look good in a backtest and fail quietly on a live account. We built Paksa Hunter Pro, a MetaTrader 5 trading agent for gold (XAUUSD) and other markets, that reads six timeframes as one market, explains every decision, handles real broker conditions and gives traders a full control dashboard inside MT5.

Paksa IT Solutions (in-house product)
4 weeks (Sep - Oct 2026)
October 2026
MQL5 MetaTrader 5 Algorithmic Trading XAUUSD Risk Management Trading Dashboard
6
Timeframes in one decision
86
Dashboard settings
15
Releases in 4 weeks
Paksa Hunter Pro MT5 trading agent: chart analysis

Problem Statement

The Business Challenge

Traders who buy an Expert Advisor (EA) for MetaTrader 5 run into the same problems again and again. The robot trades in the backtest but stays silent on a live account. It behaves differently on another broker. It forgets its state after a restart or a move to a VPS. When it does not trade, nobody can tell why.

Our own first versions of Paksa Hunter Pro had these problems too. Separate engines read the market and disagreed, so the robot missed clear entries. Broker rules that never appear in a backtest, such as freeze levels, volume limits and requotes, refused orders without a clear reason. The dashboard said RUNNING while a protection lock was blocking every trade.

The goal was a trading agent a trader can trust and understand: one decision for the whole market, strict risk control, correct behaviour on real brokers, and an honest explanation of everything it does.

Our Approach

The Solution

We rebuilt the product around one agent. It reads M1, M5, M15, H1, H4 and D1 as one market and makes one decision. We measured thousands of entry signals on two years of gold data: waiting for pullbacks into a zone had no edge, while strong trends confirmed by H4 and D1 kept going. So the agent trades momentum: it enters on the next M5 push of a strong H1 trend and never waits for hours at an old level. Every trade has a stop-loss and a target at the broker from the first second.

Around the agent we built a chain of protections that every order must pass: capital allocation chosen by the trader, risk per trade, daily, weekly and maximum drawdown locks, an open-risk budget, news and market-shock filters, and prop-firm rules.

We then made the agent survive the real world. It reads broker rules (filling mode, stop and freeze levels, volume limits, time zone), retries requotes, rebuilds its state from the account history after a restart or a VPS move, and detects a second copy trading the same account.

Finally we gave traders a full control dashboard inside MT5, with 86 settings in 9 sections and pages for trades, orders, performance and history. The agent writes its reasoning on the chart and in the dashboard.

One agent, one decision

Paksa Hunter Pro reads six timeframes of the same market (M1, M5, M15, H1, H4 and D1) and makes one decision from them. H1 sets the direction, H4 and D1 must agree, M5 times the entry and ATR measures every distance, so the same rules work on gold, indices and currencies. On every candle the agent writes what it sees and what it is waiting for, on the chart and in its dashboard.

Paksa Hunter Pro dashboard in MetaTrader 5 explaining why the agent is waiting: H1 trend up, price in the buy zone, waiting for M5 to turn up
Live GOLD chart: the dashboard says what the agent is doing and exactly why it is not trading yet (account values blurred).

The problems we solved, with evidence

Every fix below started from evidence, not from a guess: a live log, a Strategy Tester report or an MQL5 Market validation report. Where a broker condition could be reproduced in the tester, we reproduced it before fixing it.

ProblemEvidenceWhat we changed
Waiting at old zones for a pullback lost money, and the agent sat idle for hours while price ranSignal study: every entry signal on GOLD 2025 and 2026 real ticks (4,459 signals), measured in RMomentum entry: only strong H1 trends confirmed by H4 and D1, on the next M5 push; designed on 2025, confirmed on unseen 2026 (v2.00)
Separate engines disagreed on direction, so the robot missed clear entriesLive logs: one module saw a buy, another a sell, a third filtered both outOne agent reads M1 to D1 and makes every decision (v1.112)
After price bounced from the entry zone, the robot kept waiting at a level price had leftLive GOLD log, 19 Oct 2026: the bounce was seen at 16:09 and refused 9 times by an over-strict target ruleA used zone is cancelled and the bounce’s order block becomes the next entry area; the target is the high or low the trend is heading back to (v1.115, refined in v1.118)
A chart forecast drew a straight line to the target from every new candleScreenshot belowTwo conditional scenarios that start at the level that must trade first and run at the pace the market really moved (v1.114)
Stop-loss changes sent while the broker had frozen the positionMQL5 Market validation: 21 refused modifications on EURUSDNo stop change is sent while the broker would refuse it (v1.116)
Orders larger than the broker’s per-symbol volume limitMQL5 Market validation: 21 refused orders on GBPUSD; reproduced on a test symbol with a 0.2-lot limitOrder size is trimmed to the limit: 11 refused orders before, none after (v1.117)
A single requote silently cancelled a valid entryCode audit: a requote was treated as a final rejectionRequotes and price changes are retried after a pause (v1.118)
Loss limits and trade plans forgotten after a restart or a move to a VPSCode audit of what survives a restart and what a MetaTrader VPS receivesLoss baselines, trade plans and missing stop-losses are rebuilt from the account history (v1.118)
Two copies of the EA trading the same accountSeen on a client PC: the same account open in two MT5 terminalsA second copy on the same PC refuses to start; a copy on another PC or VPS is detected and the trader alerted (v1.118)
The dashboard said RUNNING while a loss lock or news pause blocked every tradeDashboard compared with the entry checksThe header shows the real pause, for example “DAILY LOSS LOCK – NO NEW TRADES” (v1.118)
Dashboard panels mixed after a timeframe change; the timeframe list lay over the trading buttonsScreenshot belowLists open away from the trading buttons on the top layer, leftover panels are cleared, the chart scrolls freely (v1.118)
Long backtests stopped part-way and produced an empty reportStrategy Tester: the test log reached about 1 GBStatus logging in the tester cut to changes only; every report is checked for a complete run before its numbers are used (v1.118)
Earlier build: a forecast line drawn straight from the current candle to a far target
Before: the first forecast ran straight from the current candle to a target hours away. Replaced by conditional scenarios at a realistic pace.
Earlier build: settings rows showing through the overview page after a timeframe change
Before: after a timeframe change the settings panel showed through the overview page. Fixed in v1.118.

Built for real brokers, not only for the tester

  • Reads the broker’s filling mode, minimum stop distance, freeze level and per-symbol volume limit before every order.
  • Measures the broker’s time zone automatically, so trading hours, the Friday cutoff and the weekend close are correct on any broker.
  • Retries requotes, resizes take-profit steps after a partial fill, and never sends an order at a price that moved too far.
  • One trade per setup, checked against the broker’s own history, so a restart can never open the same trade twice.
  • Stop-loss and take-profit sit at the broker on every trade, so they work even when the computer is off.

A complete control dashboard inside MetaTrader 5

Most Expert Advisors on the market are set up only through the MT5 input window and show little or nothing while they run. Paksa Hunter Pro comes with its own dashboard, so the trader controls and understands the agent without leaving the chart.

Paksa Hunter Pro settings panel with sections for capital, protection, news, session, prop firm, management, chart, signals and assistant
86 settings in 9 sections, changed from the chart and reviewed before they apply.
  • Controls: BUY and SELL tickets with a risk preview, break-even for all trades, RUN / STOP and KILL ALL.
  • Pages: overview, open positions, orders, performance, execution details and trade history.
  • Settings in 9 sections: capital, protection, news, trading session, prop-firm rules, trade management, chart drawing, signals and assistant.
  • Explanations: what the agent is doing, why it is not trading right now, and the current setup with its zone, entry, stop and target.
  • Assistant: a daily briefing, news heads-ups, an explanation of each trade and its result, and optional phone notifications.
  • Navigation: switch symbol and timeframe from the panel, edit stop-loss and take-profit of open trades.

How we test

  • More than 190 complete MetaTrader 5 Strategy Tester runs across 15 releases, on over 20 instruments: gold, silver, oil, stock indices and major currency pairs.
  • Real-tick data for every result we rely on, with the strategy designed on one period and confirmed on a later period it has not seen.
  • Broker conditions reproduced on test symbols (freeze levels, volume limits) before a fix is accepted.
  • Each new build compared trade by trade with the previous one, so a fix cannot change trading by accident.
  • MQL5 Market automatic validation used as an outside check of broker handling.

What traders get

  • Understanding: no black box. Every decision and every reason for not trading is shown in plain words.
  • Protection first: a stop-loss on every trade, no grid, no martingale, no averaging into losses, and trading stops on its own at the limits the trader sets.
  • Control: capital and risk chosen by the trader, changed from the chart, with prop-firm rules built in.
  • Reliability: the same behaviour on different brokers, after restarts and on a VPS.
  • Analysis on the chart: market structure, trendlines with channels, retracement levels, entry zones and conditional scenarios with targets.

Trading leveraged products such as gold and currencies carries a high risk of loss. Paksa Hunter Pro does not guarantee profits, and test results do not predict future results.

Outcomes

Results & Impact

Paksa Hunter Pro went from separate engines to one trading agent in 15 releases over 4 weeks, with more than 190 MetaTrader 5 Strategy Tester runs on over 20 instruments, most of them on real ticks.

Every problem in this case study was found with evidence (live logs, tester reports or marketplace validation reports), fixed, and checked again. Where a broker condition can be reproduced, we reproduced it first: on a test symbol with a 0.2-lot volume limit, the old build had 11 orders refused, the new build none.

The trader now sees exactly what the agent is doing and why it is not trading, and controls everything from one dashboard.

We publish no profit figures. Real-tick backtests decide what ships, strategy research continues, and the risk controls are built so that a losing period stays small.

Work With Us

Ready to Start Your Transformation?

Let's discuss how Paksa IT Solutions can deliver similar results for your organization.