Kobold EA MT5
- Experts
-
Tomasz Wojciech Forszpaniak
I started trading at 14 because I thought I had found an easy way to make money. Now 4 years later at 18, I’m an algorithmic trader obsessed with finding out whether an idea actually works. - Versão: 1.12
- Ativações: 10
KOBOLD | JPN225 M30
No Grid · No Martingale · No Averaging · Stop Loss on Every Trade · No Overnight Positions · Source code included
Three strategies. One Nikkei chart. Every one of them tested until it either broke or held.
In German folklore, a kobold is a small spirit that works in the mines while everyone else sleeps. Miners even named the metal cobalt after it. KOBOLD does the same job on your chart: three independent strategies on the Nikkei 225, working quietly through the European session.
Buy the Lifetime license and you get the complete, commented source code. Send me a private message here on MQL5 after purchase. It isn't shared to prove KOBOLD is legitimate; the tests do that. It's shared so you can study the three edges, challenge them and build on them. Source code is not included with rentals.
Backtest results
JPN225 M30, 2021 01 01 to YTD. $10,000 starting balance, fixed lot of 15, so there is no compounding.
- Net profit: +$35,283
- Profit factor: 1.63
- Win rate: 59.1% (773 of 1,307)
- Trades: 1,307, all long
- Max drawdown: 17.7%
- Recovery factor: 11.49
- Average win / average loss: $117.99 / $104.72
Robustness: every strategy passed my full test suite
A backtest only shows that a strategy made money in the past. The real question is whether that was a real edge, luck, or settings tuned until the curve looked good. Before a strategy goes into KOBOLD, it has to survive tests built to break it:
- Costs and timing: doubled fees, extra slippage, and every entry delayed by one bar
- Luck: significance on both the data used to build it and data it never saw, beating 200 random-entry strategies, the Harvey-Liu haircut for how many ideas were tested, and Monte Carlo resampling
- Overfitting: held-out performance, walk-forward periods, probability of backtest overfitting, and a look-ahead audit
- Fragility: nearby versions of every setting, broad parameter changes, randomly removing 20% of trades, and removing the 10 best trades
All three strategies passed every test. Most ideas my research produces don't.
The three strategies
- Edge 1 compares the Nikkei with GBPJPY and buys when the index sits high in its recent range and momentum hasn't broken down.
- Edge 2 buys calm, steady markets with no recent drawdown, as long as the move isn't overextended.
- Edge 3 buys quiet, sideways conditions close to the 60-bar average, when no opening-range breakout has happened.
Each has its own stop loss (2 to 3 ATR), take profit (4.5 ATR).
Protection, built in
- Max Daily Loss % (default 4%): when KOBOLD's result for the day, closed and open trades together, reaches this share of your equity, it closes its positions and stops until the next day.
- Min Margin Level % (default 300%): it won't open a trade that would push your margin level below this.
- One switch turns Protection on or off, and you set both levels yourself.
On top of that, built in limits stop any single trade from risking more than 5% of equity. The strategy's own stop loss is never moved.
When it trades
Monday to Friday. Entries between 11:30 and 16:30 server time, and everything is closed by 17:00. On the common GMT+2 / GMT+3 broker server, that's 04:30 to 09:30 New York time. Nothing is held overnight, so there is no swap or prop violation.
Setup in 3 steps
- Open a JPN225 chart on M30 (your broker may call it JP225 or NIKKEI225)
- Attach KOBOLD
- Set your Trade Size and max dd limit
Edge 1 also reads GBPJPY. KOBOLD finds it automatically and adds it to Market Watch. If your broker uses an unusual name, type it into the GBPJPY Symbol Name input.
Lifetime license: full source code
Buy the Lifetime license and you get the complete, commented source code. Send me a private message here on MQL5 after purchase. It isn't shared to prove KOBOLD is legitimate; the tests do that. It's shared so you can study the three edges, challenge them and build on them. Source code is not included with rentals.
Good to know
- Long only, so it sits out falling markets.
- Edge 1 needs GBPJPY at your broker. Without it, Edges 2 and 3 still trade normally.
- Losing streaks happen: up to 12 losses in a row in the backtest.
- On a hedging account each strategy can hold its own trade. On a netting account KOBOLD holds one trade at a time.
Requirements
- JPN225 on M30, with GBPJPY available at your broker
- Broker server time GMT+2 / GMT+3 recommended
- VPS recommended
Trading involves risk. Backtest and robustness results do not guarantee future performance. Only trade with money you can afford to lose.
