Moving-average cross lost money.

After costs on BTCUSDT 15m: −62.35% in-sample, −18.46% out-of-sample.

Equity, day by day

Equity curve of Moving-average cross on BTCUSDT In-sample, 2025-11-18 to 2026-07-31: equity went from 10,000 to 3,764.73. Out-of-sample, 2026-07-31 to 2026-11-18: from 10,000 to 8,154.34. The lowest daily point was 3,764.73, the highest 10,388.56. 02,5005,0007,50010,000 Equity In-sample Out-of-sample 2026-012026-042026-072026-10
Each sample starts again from 10,000. The value axis starts at 0.

After fees and slippage

Total return, in-sample

−62.35%

Clear loss: t is −2.95 over 572 trades.

Total return, out-of-sample

−18.46%

Cannot be told apart from zero: t is −0.68 over 263 trades.

Sample Trades Win rate Average R t-statistic Total return Max drawdown
In-sample 2025-11-18 to 2026-07-30 572 20.80% −0.31 −2.95 −62.35% 63.33%
Out-of-sample 2026-07-31 to 2026-11-18 263 19.01% −0.05 −0.68 −18.46% 31.18%

Computed by perpmath 0.1.0 on 2026-11-18 11:40 UTC. A backtest is arithmetic on past prices, not a forecast.

Win rate, drawn to scale

In-sample: 20.80%, 119 wins in 572 trades Full width: all 572 trades
Out-of-sample: 19.01%, 50 wins in 263 trades Full width: all 263 trades

A win is a trade with a profit after costs.

The averages

Fast is the simple average of the last 20 closes. Slow, of the last 50.

The cross

Fast crosses above slow: close any short, go long. Fast crosses below slow: close any long, go short.

The stop

Long: the signal bar’s close minus 2 × ATR. Short: the close plus the same. ATR is averaged over 14 bars.

No target

No profit target. A position ends at the opposite cross or the stop. After a stop it waits, flat, for the next cross.

Fixed parameters

Fast average, in bars fast_period
20
Slow average, in bars slow_period
50
ATR length, in bars atr_period
14
Stop distance, in ATR stop_atr_mult
2

Set before the first run on real data. Not adjusted since.

Costs

Fee
0.055% per side
Slippage
0.02% per fill, against you
Position size
100% of equity
Starting capital
10,000

Fee: Bybit’s base taker rate, so check the current table. Slippage: an assumption, not a measurement.

Fills

Decision
Bar close
Fill
Next bar open
Stop and target in one bar
Stop first
Look-ahead check
Passed, 9 shortened runs

What is not modelled

  • Funding payments.
  • Leverage, margin and liquidation.
  • Maker fees, partial fills, latency, order-book depth and market impact.
  • The order of events inside a bar, beyond stop-first.

Data

Market
BTCUSDT, 15m bars, Bybit public data
First bar
2025-11-18 11:30 UTC
Last bar
2026-11-18 11:15 UTC
Bars
35,040, 0 missing
Out-of-sample from
2026-07-31 00:00 UTC
Data file SHA-256
ac4acc429d7f33dd10e480195bce85f5639714b5d46869499b4275666f9fa037
Result file SHA-256
13c4163eb9c20bf25c917704ed3e251fd13585384a79fd676aed0924e59af84b

Run it yourself

perpmath download BTCUSDT 15m --start 2025-11-18T11:30:00Z --end 2026-11-18T11:30:00Z
perpmath backtest ma_cross --symbol BTCUSDT --interval 15m --config configs/ma_cross.json --out results/ma-cross-btcusdt.json