Desktop research tool for crypto strategies

Prove the strategy before you trust it.

Crypto Proving downloads real Binance candles to your PC, lets you write buy and sell rules without code, tries hundreds of thousands of variations in seconds, and then checks whether the winner survives a year it has never seen.

The optimizer at 54.5 percent of 390,625 backtests, showing speed, busy workers and a live top 20 table
The optimizer mid-run: 91,634 backtests per second on 12 workers, with the best results listed as they are found.

How it works

Five steps from raw candles to a result you can defend.

01 · DATA

A year of candles, stored on your own PC.

Pick a coin, a candle size and a period, then download once from Binance. Every later test reads from disk, so nothing is fetched twice. No account and no API key needed.

Market Data page with coin picker, candle size buttons, period chips and a table of stored candles

02 · STRATEGY

Buy and sell rules, without writing code.

Build each condition from dropdowns: an indicator, an operator such as “is below” or “crosses above”, and a number. The page translates your rules into plain English as you go, and templates like Sharp dip buyer give you a starting point.

Strategy Builder with entry rules, exit rules, stop loss and take profit, and a plain English summary

03 · BACKTEST

Replay history, candle by candle.

See ending capital, drawdown, win rate and profit factor next to an equity curve drawn against simply holding the coin. The trade list tells you why each trade was entered and why it was closed.

Backtest result with summary cards, an equity curve against buy and hold, and a trade list

04 · OPTIMIZE

Stop guessing “RSI 30 or 25?”

Give each number a range and the optimizer tests every combination across all your CPU threads. Before anything runs you see the exact number of backtests and the memory it will take. Afterwards, sort by any column and filter out results you would not trust.

Sortable results table with parameters, ending capital, return, drawdown, trades, win rate and profit factor

05 · PROVE

The best of 390,625 tries is partly luck.

The only fair test is a period the optimizer never saw. The out-of-sample check runs the winning settings on an earlier year and puts both results side by side. If the return turns negative, the settings were fitted to noise, and you know before any money is involved.

Out-of-sample check comparing the optimization period with an earlier, unseen year

Principles

Built so you cannot fool yourself by accident.

No peeking at the future

A rule is checked when a candle closes. The trade happens at the next candle’s open.

Costs are on by default

Every backtest pays a 0.1 % fee and 0.05 % slippage unless you change them.

Every run is reproducible

Same strategy, same data, same result. Each run is saved so you can open it again.

Nothing is quietly dropped

Results hidden by a filter are marked rejected, never deleted.

What it does not do.

  • It never places trades. Crypto Proving is a research tool and has no connection to your exchange account.
  • It keeps the model simple. Long positions only, one trade at a time, all capital per trade, no leverage.
  • It does not predict. A setting that passed one unseen year is a good sign, not a guarantee.