A DCA signal portfolio on gold, designed to control risk and target sustainable growth β automated operation, tested in real conditions with positive backtest results.
For experienced tradersMost of the time, market is ranging.
About 3/4 of the time price oscillates within range instead of trending β that's the environment where average price (DCA) works best.
But old-style DCA is a double-edged sword: unlimited order stacking, arbitrary step and lot β that's gambling. Eats well when market is calm, then blow up the whole account in just one reversal.
β Transform from luck into a system with high control & monitoring β predictable and clear profit.
Smart DCA is a system disciplined price averaging on gold: enter orders on grid, time by market state, and always put risk control first. The biggest difference is every order series cycle has a fixed-capital SL level, and capital is spread across many symbols offset in rhythm instead of concentrated in one place.
Currently specialized for GOLD β platform designed to expand to many other products in the future.
~40,000 lines of code, 109 dedicated modules + multi-axis indicator system to identify ranging zones; each series has a fixed-capital SL.
Market analysis across many axes (mode, volatility, price zoneβ¦) and through many layers, using professional algorithms (Hidden Markov Model/HMM, machine learning, forward testing); 146,160 backtest scenarios run monthly; re-simulate the most dangerous phases to extract safe & effective parameters.
10 offset symbols + daily monitoring; running live for many months, fills match simulation closely β not just backtest on paper.
Result: most of the configuration space yields positive results β meaning results do not depend on hitting one lucky parameter set.
95% is coverage statistic during test period β not win probability, not forecast, not promise. The stored column for fixed set group is in-sample (parameters chosen on that very period).
Does not run a fixed configuration: at each moment, system selects appropriate parameter set, knows how far the risk goes and what the profit expectation is per market state β to balance efficiency and safety.
Each order series cycle has a fixed-capital SL level β know the boundary upfront, not "lose to who knows where".
Many symbols use different criteria so SL days don't overlap β reduce risk concentration at one moment.
Loss ceiling declared upfront per series and per EA Β· indicator filter only has permission to block new series opens Β· news filter & spread filter Β· lot ceiling, order count, and tier count.
Runs by fixed rules + daily monitoring; operating live, real fills match simulation closely.
Portfolio comprises many symbols divided by configuration generation method (ranking Β· adaptive formula Β· optimization), each symbol spans a scale from conservative to growth; allocation between symbols is operator-configurable. All symbols observe the same risk ceiling.
Select configuration by ranking criteria with ceiling; ranges from conservative (SKY/SUPER) to growth (MOON/MAX). GOLDSKY is the "anchor" with almost no SL; GOLDMAX is the riskiest configuration, running live.
Deep research group: each symbol is one independent machine learning axis (volatility Β· amplitude Β· cycle), built on international standard machine learning algorithms combined with orchestrating AI layer. Advisor adjusts configuration per market state; freeze thresholds & anti-overfit constraints at each step.
Two symbols special of the system. GOLDAI recognizes market context each day through multi-layer cascade then picks the optimal configuration for that exact context. GOLDAI2 is the upgraded version β adds guard advisor layer independent monitoring, intervenes defensively on "wrong stance" days.
| Symbol | Stored Β· before 01/06 | Live Β· from 01/06 | ||||
|---|---|---|---|---|---|---|
| Profit | SL | Drawdown | Profit | SL | Drawdown | |
| Ranking group β fixed set | ||||||
| GOLDSKY | +$54k | 0 | $9.2k | +$3.2k | 0 | $2.4k |
| GOLDSUPER | +$75k | 0 | $12k | β$2.3k | 1 | $10.2k |
| GOLDMOON | +$131k | 1 | $12.6k | +$19.2k | 1 | $11.1k |
| GOLDMAX | +$166k | 1 | $18k | +$24.1k | 1 | $10.0k |
| Formula group β adaptive Β· Machine Learning + AI | ||||||
| GOLDSMART | +$57.1k | 1 | $12.1k | +$19.0k | 0 | $5.4k |
| GOLDAIA | +$60.4k | 0 | $7.1k | +$19.0k | 0 | $5.4k |
| GOLDAIB | +$66.3k | 0 | $6.2k | +$16.1k | 0 | $5.4k |
| GOLDAIC | +$56.4k | 0 | $9.0k | +$19.8k | 0 | $5.4k |
| Optimization group | ||||||
| GOLDAI | β | β | β | +$7.1k | 1 | $10.0k |
| GOLDAI2 | β | β | β | +$11.2k | 0 | $6.0k |
Measurement notes: column before 01/06 of fixed set (SKY/SUPER/MOON/MAX) is backtest throughout ~10 months and is in-sample β parameters chosen on that very period, so not read as forward test; formula set (SMART/AIA/B/C) is causal forward testing 28/01β30/05 (~4 months, contains the biggest shock). Two periods longβshort differ so don't directly compare prior column magnitude β read by signal (all positive) and depth of loss (quanh 1 SL). GOLDAI & GOLDAI2 (optimization group) leave column empty before 01/06 because context-aware configuration selection only runs when accumulated data series is sufficient β counted only from when the system has enough live trading data.
Through both harsh and recent live trading periods, most codes positive with only 0β1 stop-loss hits, and stop-loss days fall on different days β exactly as designed for phase offset. GOLDSUPER negative after 01/06 is intentional: its role is phase offset to reduce total portfolio risk.
Each code is designed and tested on nominal operating unit $10,000 = 1 unit max-loss (1 SL). This is software configuration parameter, not recommended capital: each code runs independently at exactly one unit, and multiple codes can run simultaneously. When running simultaneously, OSS research records reference multiplier among configurations β group multiplier 2 achieves higher efficiency ratio (profit/drawdown) during testing so reference weight is double group multiplier 1.
| Group | Symbol | Reference multiplier | Nominal unit |
|---|---|---|---|
| Rank | GOLDSKY | 2 | $20k |
| GOLDMOON | 2 | $20k | |
| GOLDSUPER | 1 | $10k | |
| GOLDMAX | 1 | $10k | |
| Formula | GOLDAIB | 2 | $20k |
| GOLDSMART | 1 | $10k | |
| GOLDAIA | 1 | $10k | |
| GOLDAIC | 1 | $10k | |
| Optimization | GOLDAI2 | 2 | $20k |
| GOLDAI | 1 | $10k | |
| Total when running full 10 codes | 14 | 14 units Β· $140k | |
How to read the table: multiplier 2 (SKY Β· MOON Β· AIB Β· AI2) is the group achieving higher efficiency ratio per OSS data. Column nominal unit is the configuration scale when running full 10 codes per reference multiplier β not mandatory: running 1 code or any subset is fine, as long as you maintain correct multiplier ratio among running codes. Absolute scale is your decision and your configuration.
Software is designed and tested with 2β3Γ safety margin against nominal scale for each series: minimum 2Γ is the technical requirement for the grid to reach intended depth without hitting margin call mid-way; 3Γ is the margin where current parameter set is tested in most stable state. Amount above nominal scale is technical buffer of grid mechanism, not investment reserve. Running below 2Γ margin is outside test range and not supported. Harshest period in data already occurred and was controlled within this margin.
Because margin is 2β3Γ, drawdown calculated on total account balance is always much smaller than same number calculated on nominal scale β in exchange, return on total balance is lower.
Account type does not change how calculation works β nominal scale, stop-loss level, and 2β3Γ test margin all calculate the same way on cent account or standard account. The difference is the same nominal scale corresponds to much smaller actual amount on cent account, so easier to maintain correct design ratio and correct test margin without needing large balance. Therefore encouraged to run on cent account, especially when running multiple codes simultaneously.
We are ready to share mechanism details, structure of each code, and capital allocation suited to your goal.
Website: tradingauto.org
This document is informational, not investment advice or guarantee. Gold investing has risk and you may lose part or all capital; past results do not guarantee future. Figures are test results in limited period, calculated on nominal capital, not long-term representative.
Service nature: software rental. TRADINGAUTO provides neutral technical tools (research, test, automate) for customers to rent and self-operate on their own account β not capital solicitation, delegated management, asset management or investment solicitation/recruitment. Provider holds no money and does not make trading decisions for customers.
No intervention in use purpose & source. What customers rent and use for, which platform/broker they run on, account source is customer's right & responsibility β provider does not designate, does not control, does not intervene, not tied to any broker and receives no commission through this channel.
Customers should demo test and research thoroughly mechanism, parameters, risk before going live; using means you understand and accept your risk. Software provided as-is: customers bear responsibility for setup, configuration, capital allocation, and trading results. Provider is not liable for losses or incidents during use (broker execution, fees, slippage, connection loss, third-party errorsβ¦).
Use lawfully: customers commit not to use software for unlawful purpose and comply with local laws; any violation is customer's responsibility, unrelated to provider.
Intellectual property: software, algorithm, parameters, and brand belong to TRADINGAUTO; customers only get limited-duration usage rights. Strictly forbidden: copying, account sharing, re-renting, redistribution, reverse engineering, or modification without written consent.