Performance · Backtested, not a live track record
The numbers we actually got.
No theoretical projections. Every figure on this page comes from the same backtesting engine the live bots run on: the same code, the same data, the same parameters. The drawdowns are published alongside the returns because they are the more useful number.
- AlphaTrading, annualised
- +23.1%
- AlphaTrading, avg max drawdown
- −37.5%
- ATSB, annualised
- +15.8%
- ATSB, avg max drawdown
- −44.2%
- How to read them
- In-sample: an upper bound
How we test
Backtesting held to a standard.
Not all backtests are equal. Ours are run with strict rules to prevent the most common forms of overfitting and look-ahead bias.
Real historical OHLC
Daily open, high, low and close fetched via the MarketStack API. No synthetic data, and no adjusted-price corrections that would add look-ahead bias.
Five full years
Each parameter set is tested against the whole five-year dataset, not a chosen date range. The result is the average, not the best-case year.
A full parameter grid
Every dimension is varied systematically: 384 combinations for Turtle, 528 for ATSB. The winner is the one that survives across the whole sample.
Results
Best configuration, per bot.
The winning parameter set across each bot's full test period. The live bots run these exact parameters.
- Annualised return, 5-year backtest average
- +23.1% / yr
- Total return over the window
- +182%
- Average maximum drawdown
- −37.5%
- Best / worst ticker
- +1,046% / −34%
- Risk per trade, hard cap
- 2% of equity
- Combinations tested
- 384
The average is carried by a single ticker that returned +1,046%. Its weakest lost 34%.
- Annualised return, 5-year backtest average
- +15.8% / yr
- Total return over the window
- +108%
- Average maximum drawdown
- −44.2%
- Profit factor
- 1.80
- Profitable instruments
- All tested
- Combinations tested
- 528
Entry at a score of 4 or more; hard stop 3% from entry. A profit factor of 1.80 means $1.80 earned for every $1.00 lost.
How to read these numbers
Both figures are backtested, not a live track record. Each is the best-performing configuration out of a full parameter sweep, 384 combinations for AlphaTrading and 528 for ATSB, measured over roughly five years, about 1,260 trading days, on the universe each bot trades today.
Because the configuration was chosen after seeing those results, the returns are in-sample and should be read as an upper bound. Live performance is normally lower. We are working through out-of-sample validation and will publish those results alongside these when they are ready.
A 37.5% peak-to-trough decline is a normal part of trend following. Anyone who would abandon the system there should not subscribe to it.
Parameter comparison
What was tested, and what won.
The ranges we searched, and the values that came out ahead. The winners are the parameters the live bots run today.
| Parameter | Range tested | AlphaTrading | ATSB |
|---|---|---|---|
| Fast period | 5, 10, 15, 20 days | 20-day channel | 5-day |
| Slow period | 20, 30, 50, 55 days | n/a, a channel rather than an average | 20-day |
| Stop size | 3%, 5%, 7%, 10% / 1 to 3 × ATR | 1.5 × ATR | 3% |
| Entry threshold | Score 2, 3, 4 / breakout channels | 20-day breakout | Score ≥ 4 |
| Pyramid / sizing | 1 to 4 units / 50 to 100% of size | Up to 4 units, 2% risk each | Score-weighted, 75 to 100% |
| Trailing stop | On / off | On, 10-day channel | On, MA5 pullback |
-
—
What this data is, and isn'tSimulated, not live
Everything on this page is simulated historical performance, not live trading results. Backtests assume perfect execution at closing prices, no slippage and no market impact. In real trading all three affect returns.
The purpose of this data is to show the research methodology, not to promise returns. Sheylar Global runs these bots because the evidence supports it, not because any result is guaranteed. For a full account of the risks in following them, read the risk and transparency page before subscribing.
Run the backtests yourself.
Both simulators are free. Test every parameter combination, the same ones we ran, and see exactly how each strategy behaves on any instrument.