Skip to content

Quick start

Requires a FastMM build (Install). The code is in examples/quickstart/; CI builds and runs it (ctest examples.quickstart).

1. The strategy

MyMM quotes one bid and one ask half_spread_bps away from the mid on every book update:

#include "fastmm/strategy.hpp"

using namespace fastmm;

struct MyParams {
  FASTMM_PARAMS(MyParams)
  FASTMM_PARAM_BPS(half_spread_bps, 0.005_bps, 0_bps, 1000_bps, "half spread around the mid, bps")
  FASTMM_PARAM(Qty, quote_qty, 0.001_qty, 0_qty, 1000_qty, "quantity per side, base units")
};

struct MyMM : StrategyBase<MyParams> {
  static constexpr std::string_view name() noexcept { return "my_mm"; }

  void on_book(auto& ctx, InstrumentId id, const auto& book) noexcept {
    if (!book.is_valid()) return ctx.pull_quotes(id);  // empty or crossed
    const Instrument& inst = ctx.instrument(id);
    const Price half = book.mid() * params().half_spread_bps;
    DesiredQuotes q;
    q.bid(inst.round_price(book.mid() - half, Side::Buy), inst.round_qty(params().quote_qty));
    q.ask(inst.round_price(book.mid() + half, Side::Sell), inst.round_qty(params().quote_qty));
    ctx.set_quotes(id, q);  // diffed against resting orders
  }
};
  • FASTMM_PARAM_BPS and FASTMM_PARAM declare typed parameters with a default, a range and a description. Ratio (basis points) and Qty are 64-bit fixed-point values.
  • on_book is a hook: the engine calls it after each book update. Other hooks cover trades, fills, timers and connection changes (Strategy API).
  • set_quotes sets the desired quotes; the engine sends new, cancel and replace messages for the difference.

2. The backtest

#include "my_mm.hpp"

#include "fastmm/backtest/backtest_runner.hpp"

#include <cstdio>

int main(int argc, char** argv) {
  auto cfg = bt::BacktestConfig::single_instrument("BTCUSDT", 0.01_px, 0.00001_qty);
  cfg.generator.market_qty_median_lots = 1500;  // synthetic market: larger taker orders
  cfg.transport.fees = sim::FeeModel::from_bps(10.0, 10.0);  // Binance spot VIP 0: 0.1 % both sides
  auto source = bt::open_data(argc > 1 ? argv[1] : "synthetic");  // .fmj, .csv or synthetic
  const bt::BacktestResult result = bt::run_backtest<MyMM>(cfg, source.get());
  std::fputs(result.summary_table().c_str(), stdout);

  const char* out = argc > 2 ? argv[2] : "runs/quickstart";  // equity, fills, orders, summary
  if (!result.write_all(out)) {
    std::fprintf(stderr, "cannot write the results to %s\n", out);
    return 1;
  }
  std::printf("\nresults in %s/ -- for the HTML report: fastmm report %s\n", out, out);
}

run_backtest<MyMM> builds the engine with your strategy, a simulated venue with a matching engine and latency, and runs it for 60 s of simulated time.

3. Build and run

In the FastMM build tree:

cmake --build --preset release --target my_mm_backtest
./build/release/examples/quickstart/my_mm_backtest   # writes runs/quickstart/
python3 tools/report.py runs/quickstart              # writes runs/quickstart/report.html

Open runs/quickstart/report.html: one self-contained page with the equity curve, the inventory, where the PnL came from, the markouts and the fill quality (Run report). python3 tools/report.py needs nothing but Python 3; with the Python package installed the same command is fastmm report runs/quickstart.

In your own project, copy examples/quickstart/. Its CMakeLists.txt fetches FastMM:

cmake_minimum_required(VERSION 3.25)
project(my_mm LANGUAGES CXX)
set(CMAKE_CXX_STANDARD 20)

if(NOT TARGET fastmm::backtest)
  include(FetchContent)
  set(FASTMM_BUILD_NET OFF CACHE BOOL "backtests need no networking")
  FetchContent_Declare(fastmm GIT_REPOSITORY https://github.com/ziyangliu-666/FastMM
    GIT_TAG v0.6.0 GIT_SHALLOW TRUE)
  FetchContent_MakeAvailable(fastmm)
endif()

add_executable(my_mm_backtest main.cpp)
target_link_libraries(my_mm_backtest PRIVATE fastmm::backtest)

cmake -S examples/quickstart -B build/quickstart -G Ninja -DCMAKE_BUILD_TYPE=Release
cmake --build build/quickstart && ./build/quickstart/my_mm_backtest

Add -DFETCHCONTENT_SOURCE_DIR_FASTMM=$PWD to the first command to build against this checkout instead of downloading FastMM.

4. Read the result

The report is the whole run on one page; the summary table it printed is the same numbers:

backtest my_mm  seed=1  md_events=7129  steps=7694  wall=0.01s
  net pnl                        -2.1594
  realized / unrealized / fees   0.0011 / 0.0001 / 2.1606
  fills (maker / taker)          37 (37 / 0)
  orders / cancels / replaces    283 / 245 / 0
  inventory mean / |mean| / max  ...
  outbound messages / sha256     ... / ...

where the PnL came from (quote currency, 2160.60 traded notional)
  gross spread capture                 0.0005  +0.004 bps of 1320.60, over the 23 of 37 fills that had a venue mid
  fees paid                           -2.1606
  = net                               -2.1594
  ...
  • Money is in the quote currency (USDT), quantities in the base currency (BTC).
  • fills (maker / taker): quotes are post-only, so taker is 0.
  • The run charges Binance spot VIP 0 fees, 10 bps each side, so the spread this strategy captures does not cover them (Backtesting).
  • outbound messages / sha256: a hash of the order messages sent; it is identical on every run (Determinism).

The report adds what a table cannot show: when the losses arrived, how much inventory the strategy carried while it earned them, and whether the mid kept moving against the fills afterwards.

Next