How the strategy works
ST Phoenix trades SuperTrend direction changes on a Mean Renko chart. Each brick is a fixed amount of price movement, so quiet markets produce no noise and busy sessions produce plenty of signals.
SuperTrend entries
A trade opens when a fast SuperTrend flips direction on a closed brick.
Colour exit
A side in profit is banked as soon as a brick of the opposite colour closes. This is where most profit comes from.
Martingale grid
If price runs against a position, the EA adds larger trades at widening intervals to pull the average price back.
Turbo Snowball
When one side is up by a small target, it pays off the other side's worst loser. In v3.5 it can pay off part of a loser, shrinking the grid step by step.
Vault floor
Part of each gain is locked in as an equity floor. If equity touches it, everything closes and the gain is protected.
Phoenix lives
After a floor hit the EA restarts from the new balance, a limited number of times. Its state survives terminal restarts and server reboots.
Compounding
The first lot grows with equity, so profits are reinvested automatically.
The backtest station
To test properly, we built an unattended testing rig next to the live MT4 terminal, so hundreds of runs could happen without touching the trading account.
Hidden tester
A cloned MT4 runs on an invisible display in its own Wine prefix with trading disabled, driven by scripts that set up, start and save each test.
10 years of data
EURUSD one-minute history from 2016 onwards, downloaded through MT4's History Center and checked against broker prices.
Renko on demand
A script rebuilds the Mean Renko chart offline at any brick size and hands it to the Strategy Tester. No manual chart juggling.
Every tick model
All tests start from $1,000 and run a single calendar year (2026 runs to 30 October). Results and equity curves are saved automatically.
Step 1: brick size and SuperTrend
The first sweep tested 27 combinations on 2026: three brick sizes × three SuperTrend speeds × three band widths. The goal was a plateau, a region where neighbouring settings all make money, rather than one lucky best setting.
Out-of-sample check
The three best medium-brick settings were then run on 2025, a year they were never tuned on.
| SuperTrend | 2026 (tuned) | 2025 (unseen) | Verdict |
|---|---|---|---|
| Setting A (fastest) | +$5,592 | wiped | Too twitchy |
| Setting B | +$5,607 | +$5,531 (DD 54%) | Survived, barely |
| Setting C | +$5,070 | +$9,837 (DD 18%) | Chosen |
Step 2: finding a dead feature
Cutting the snowball target by 97% changed nothing, to the cent. A debug build logged every snowball check for a full year to find out why.
Lives and the vault
The best 2025 run peaked at $13.2k, then stopped trading in June: it had used all of its lives. With more lives allowed, the same setup traded the whole year and finished at +$28,033, using about eight floor hits. Each floor hit locks in gains and restarts from a higher balance.
Step 3: tuning the grid
The grid multiplier and spacing control how fast the grid stacks up against a trend. We pushed both in each direction until something broke.
| Multiplier | 2024 | 2025 | 2026 | Verdict |
|---|---|---|---|---|
| Gentlest | +$7,067 · DD 57% | – | – | Grid too weak to recover |
| Gentle | +$7,497 · DD 62% | – | – | Grid too weak to recover |
| Chosen | +$7,139 · DD 18% | +$30,172 · DD 21% | +$6,075 · DD 43% | Sweet spot |
| Steep | +$6,638 · DD 25% | +$30,622 · DD 26% | +$6,482 · DD 48% | Same money, more risk |
| Steepest | +$8,416 · DD 22% | wiped in week 1 | +$6,635 · DD 54% | Breaks |
Step 4: three risk tiers, stress-tested
The finalists were grouped by risk and re-run with a much wider spread, about five times the normal test spread, to mimic real spread and slippage.
Show the numbers as a table
Equity curves
The Strategy Tester's own balance graphs for each tier and year, at the 1-pip stress spread.
Curve showcase
Fifteen standout runs, grouped by risk. Each one started from $1,000 with compounding on. Runs marked stress were tested at the 1-pip spread.
What the tests can't tell you
- Renko bricks in the tester are rebuilt from one-minute data, so they may be slightly smoother than live bricks.
- Real execution adds variable spread, slippage and swaps beyond the 1-pip stress test.
- A martingale grid can lose the whole account in a single long trend. The vault floor limits this, but it doesn't remove it.
Next step: a live forward test of the MID setup on a small account.