Slipstream

Not for professional use. Slipstream was built as a study project to learn about basic stock market functionality and to learn Go. It is not audited, not battle-tested, and its fill model is a deliberate simplification of how real markets work. Do not use it to make trading decisions or to manage real money.

A backtesting engine for equity strategies, written in Go. It replays historical 1-minute bid/ask/last quotes through a simulated venue with order latency, cash constraints, and stale-data handling, then reports what the portfolio did.

AI Notice

The production code was not written with AI. However, AI was used to write the tests and to help find some bugs.

Known simplifications

There are gaps between this and a real simulator, so it's worth being explicit:

• No order book or depth. Orders match against a single top-of-book quote and always fill in full — no partial fills, no queue position, no market impact.
• Bid/ask are estimated from OHLC candles, not observed. Using the bar's high as the ask and low as the bid overstates the spread badly.
• Time-in-force is stored on the order but not enforced; nothing expires.
• No shorting, margin, fees, or slippage beyond the spread.
• No corporate action handling beyond whatever the source data already adjusted for.

Running it

go run ./cmd/backtest

Data files are not in the repo. The engine expects one quote file per ticker at {basePath}/{quoteFilePattern}, with {ticker} substituted in, and each line formatted as:

2026-04-27 04:00:00, 269.212800, 270.300000, 269.740000

that is: timestamp, bid, ask, last. If you only have OHLC candles (timestamp,open,high,low,close,volume — the FirstRate Data format), cmd/genbidask will synthesize quote files from them; the ticker and paths are constants at the top of its main.

Configuration

Settings live in config.json, layered over the defaults in src/config/config.go. A missing file is fine — the defaults run. A malformed one is an error.

{
  "latency": {
    "baseOrderDelayTicks": 2,
    "orderDelayJitterTicks": 3,
    "rngSeed": 1
  },
  "engine": {
    "renderThrottleMs": 0,
    "idlePollMs": 1
  },
  "data": {
    "tickers": ["AAPL"],
    "basePath": "./_data/firstrate",
    "quoteFilePattern": "stock_update_month_1min_quote/{ticker}_month_1min_quote.txt",
    "metricsOutputPath": "./_output/metrics.txt",
    "metricsAppend": true,
    "maxQuoteStalenessMinutes": 60
  },
  "portfolio": {
    "startingCash": 100000
  }
}

renderThrottleMs is the one worth calling out: at 0 (the default) the engine runs flat out and prints only the final summary. Set it to something watchable (say 100) to get the live NAV/positions view, which redraws and pauses every tick.

The run summary is printed to the console and written to metricsOutputPath.

Tests

go test ./...