5. Backtest in C++

examples/cpp/tutorial/first_mm_backtest.cpp backtests first_mm for 60 s on a synthetic market and checks the result; like the quick start, it needs no registration.

The configuration

auto cfg = bt::BacktestConfig::single_instrument("BTCUSDT", 0.01_px, 0.00001_qty);
cfg.duration = seconds(60);            // of simulated time
cfg.set_seed(7);                       // the synthetic market and the latency model
cfg.generator.limit_rate_per_s = 400;  // a busier market than the defaults
cfg.generator.market_rate_per_s = 30;
cfg.generator.market_qty_median_lots = 1500;
cfg.transport.fees = sim::FeeModel::from_bps(10.0, 10.0);  // Binance spot VIP 0: 0.1 % both sides
cfg.params = {{"edge_bps", "0.002"}, {"max_position", "0.004"}, {"report_ms", "0"}};

The synthetic market's spread is one tick (0.01 USDT) at 60,000 USDT, so the edge is 0.002 bps: 0.012 USDT, about one tick. Pages 8 and 9 use 5 bps. cfg.params takes the same strings as a configuration file.

Run and check

const bt::BacktestResult first = bt::run_backtest<tutorial::FirstMM>(cfg);
const bt::BacktestResult second = bt::run_backtest<tutorial::FirstMM>(cfg);
std::fputs(first.summary_table().c_str(), stdout);

int failures = 0;
const auto expect = [&failures](bool ok, const char* what) {
  std::printf("%s: %s\n", ok ? "ok    " : "FAILED", what);
  failures += ok ? 0 : 1;
};
expect(first.metrics.fills > 0, "the strategy traded");
// Largest |position| at the end of a 1 s bar, base units.
expect(first.metrics.inventory_max <= 0.004, "the position stayed within max_position");
expect(first.outbound_sha256 == second.outbound_sha256, "two runs sent the same orders");
return failures == 0 ? 0 : 1;

run_backtest returns fills, orders, an equity curve and summary metrics.

"$BIN"/first_mm_backtest

backtest first_mm  seed=7  md_events=18478  steps=20023  wall=0.01s
  net pnl                        -27.0083
  realized / unrealized / fees   0.0067 / 0.0000 / 27.0150
  fills (maker / taker)          467 (467 / 0)
  orders / cancels / replaces    762 / 316 / 0
  inventory mean / |mean| / max  -0.00025 / 0.00172 / 0.00387
  outbound messages / sha256     1078 / f29af3d82986f7e18291c51a7d46503cf3097cfff09a22f2068e1f60c29f5deb
...
where the PnL came from (quote currency, 27014.99 traded notional)
  gross spread capture                 0.0046  +0.002 bps of 20974.79, over the 366 of 467 fills that had a venue mid
  mid drift after the fills            0.0021  adverse selection + the open inventory, marked at the final mid
  fees paid                          -27.0150
  = net                              -27.0083
...
ok    : the strategy traded
ok    : the position stayed within max_position
ok    : two runs sent the same orders

Amounts are in USDT, the position in BTC. FirstMM captures 0.002 bps of the traded notional gross and pays 10 bps in maker fees. Read the decomposition and the markouts, not the net PnL (Backtesting).

Next: 6. Register it