Trend King EA

Autonomous 3-EMA Pullback Trading with Regime Mapping & UCB Bandit

Trend King — Auto Bandit combines a robust 3-EMA pullback baseline with regime mapping (Volatility × Trend Strength × Autocorrelation = 18 cells) and a multi-armed-bandit selector (UCB).
The EA learns online, adapts micro-variants (EMA fine-tuning, ATR SL/TP factors, pullback target, trailing mode, ADX threshold) and persists state per symbol & timeframe.

No curve-fitting. No parameter chase. Learning from real execution.

Why no backtests/optimizer?

• Online learning & path-dependence: The bandit needs real, sequential tick/trade flow with spread/slippage/margin feedback.

• Regime switching: 18 state cells + state-dependent arm selection/mutation → tester distorts intrabar breakouts & trailing.

• Multi-parameter dynamics: Arms mutate adaptively; classic WFO/optimizers don’t represent this behavior.

Bottom line: The only meaningful “optimization” is a demo warm-up. Then go live.

What the EA actually does

• Baseline: 3-EMA trend + pullback (Fast/Mid/Slow) with AO baselines (“Variant B”).

• Regime mapping: ATR% (vol), ADX (trend), ACF(1) → 18 cells.

• Bandit arms (per cell):

  • EMA fine scaling (± around baseline),

  • ATR-SL× / ATR-TP×,

  • Pullback target (Fast vs. Mid),

  • Trailing mode (ATR vs. Donchian),

  • ADX threshold factor.
    Selection via UCB; weakest arms get mutated.

• Risk logic: Risk% per trade (lots from stop distance), gross-risk cap across open positions per symbol/TF, spread guard, configurable positions per side.

• Persistence: Learning per symbol/TF with file storage (see paths).

Inputs – keep it intentionally lean

Leave AutoMode = true. The bandit handles the fine details. Day-to-day you only set:

• AutoMode = true (autonomous mode)

• RiskPerTradePct – risk per trade (%)

• GrossRiskCapPct – equity-based gross risk cap

• MaxOpenPositions – max concurrent positions (per symbol/TF)

• MaxSpreadPoints – spread guard (optional; 0 = off)

• Slippage , MagicOverride – if needed

Everything else can stay default. (ADX/ATR/Donchian etc. are auto-tuned via AO baselines + bandit.)

Install & start (practical workflow)

1) Install

• Copy TrendKing_AutoBandit.mq4 to ...\MQL4\Experts\ .

• Restart MT4 or refresh the Navigator.

2) Attach to chart

• Open the symbol/TF, attach the EA.

• Set only the minimal inputs above (especially RiskPerTradePct ).

• Important: Do not use the MT4 Strategy Tester/Optimizer.

3) Demo warm-up (the only optimization)

• Let the EA run for weeks or until you have enough trades and a stable, positive equity/R curve.

• During warm-up the EA learns per regime cell, mutates weak arms, adjusts EMA/ATR/trailing, and continually chooses the strongest micro-variant.

4) Go live (state transfer)

• Once demo performance is stable/positive: copy the state files (see below) and attach the EA with the same settings on your live account.

• Done – no classic optimization required.

File paths & transfer (Demo → Live)

The EA stores learning state in MT4 Common Files (shared by terminals under the same Windows user).

• Filename (per symbol/TF):
  tkbandit_<SYMBOL>_<TF>_v3601.dat
  Example: tkbandit_EURUSD_60_v3601.dat

• Primary folder (with subfolder; Windows):
  C:\Users\<YourName>\AppData\Roaming\MetaQuotes\Terminal\Common\Files\BanditTK\

• Fallback (if no subfolder exists):
  ...\Terminal\Common\Files\ (files in root)

Copying:

• From (demo PC/user): ...\Common\Files\BanditTK\

• To (live PC/user): ...\Common\Files\BanditTK\

• Close both MT4s, copy, then start the live MT4.

• Use the same symbol & timeframe on the live chart—otherwise a different file will load.

If demo & live run under the same Windows user, both see the same Common folder—often no copy is needed.

“Can I still backtest?”

You can, but results are not reliable (relaxed filters, different spread/slippage, no real interaction). Only the real demo performance matters.

FAQ (short & honest)

How do I “optimize” the EA?
You don’t. Warm-up on demo until stable positive. Then copy state and go live.

How many charts?
Multiple symbols/TFs are fine—each chart learns independently (own file). Start with 3–6 pairs/TFs, expand as you wish.

May I tweak (e.g., ADX on/off)?
You can, but in AutoMode the bandit auto-tunes fine settings. Recommendation: keep defaults, just control risk/guards.

How do I reset learning?
Delete the relevant tkbandit_*.dat for that symbol/TF (with MT4 closed). On next start it reseeds.

When is warm-up “enough”?
Practically: sufficient trades plus a smoothly positive curve on demo over your intended holding horizon (days/weeks). Stability beats any single number.

Checklist (TL;DR)

• ✅ Put EA in ...\MQL4\Experts\

• ✅ Attach to chart, AutoMode=true, set RiskPerTradePct

• ✅ No tester/optimizer

• ✅ Demo warm-up until stable positive

• ✅ Copy files:
  ...\Common\Files\BanditTK\tkbandit_<SYMBOL>_<TF>_v3601.dat
  from Demo → Live (same Common folder may mean nothing to do)

• ✅ Start live — no further optimization

Notes & practical tips

• Have hundreds of bars per TF available.

• Spread/slippage/margin matter—use a stable VPS/server.

• Consider MaxSpreadPoints (e.g., 25–35 on 5-digit FX) to filter news spikes.

• Use GrossRiskCapPct as a second safety net (e.g., 1.5–2.0%).

Risk notice

Trading involves risk. Use only capital you can afford to lose. Past performance is not indicative of future results.

Version / files

• EA: Trend King — Auto Bandit v3.6

• Bandit state version: v3601 (embedded in filenames)