Skip to content
AMSTOS

Research

In build

A real research environment, not a returns chart.

Read-only. Never trades. Free tier at launch.

  • No backtests without fees.
  • No peeking at future data.
  • No results tuned to one lucky setting.

Just research you can trust.

Backtesting

Built so results survive contact with reality.

  • Real-world costs

    • Transaction costs
    • Slippage
    • Liquidity constraints
    • Historical data by period and asset
  • Bias controls

    • Look-ahead controls
    • Survivorship-bias controls
    • Out-of-sample testing
    • Walk-forward testing
  • Robustness

    • Parameter sensitivity
    • Monte Carlo simulation
    • Robustness reports
    • Parameter sweeps

Methodology

How we keep backtests honest.

  1. 01

    Train and test on different data

    Walk-forward and out-of-sample tests judge a strategy on data it never saw.

  2. 02

    Charge every trade its real cost

    Fees, slippage and liquidity limits apply to every simulated fill.

  3. 03

    Stress the result, not just the average

    Parameter sensitivity and Monte Carlo runs show whether a result holds across markets and periods.

24/7 algorithm execution

Approved strategies run on cloud infrastructure, never a laptop.

Planned
  • Strategy container

    Packages the approved algorithm with its dependencies and config.

  • Execution worker

    Runs the strategy on schedule against its market connection.

  • Risk controls

    Max position, loss and exposure limits.

  • Monitoring

    Tracks health, trades, latency and errors.

  • Alerting

    Notifies you of failures, trades and risk breaches.

  • Audit log

    Records every decision, order and system event.

  • Kill switch

    Stop an algorithm immediately.

  • Failover

    Restarts workloads after infrastructure failures.

Hosting is priced per algorithm, separately from any plan.

FAQ

Questions, answered.

Why do most backtests look better than reality?

They skip fees and slippage, peek at future data or ignore delisted assets. Amstos controls for each.

What is walk-forward testing?

Training a strategy on one period, then testing it on a later period it has never seen.

Can I backtest my own strategy?

Yes. Run it against the periods and assets you choose, with costs, slippage and bias controls applied.

Will my strategy run 24/7?

Planned. Approved strategies will run on cloud infrastructure, priced per algorithm and separate from any plan.

How much history is there?

Up to 8 years of historical crypto data, by tier: 1 year on Free, 4 on Pro, 8 on Intelligence.

Test before you trust.

Early access opens soon.

  • Read-only API keys
  • Never executes trades
  • Free tier at launch