QubitQuant Terminal Beta

Is your strategy real, or just lucky?

Ask a question about a market. An agent researches it, backtests it with costs, then tries to break the result.

Join the beta list Watch a real run

Follow one question to its verdict.

Ask
Research
Data
Code
Backtest
Validate

Start with a question

Type what you want to find out, or drop in a paper or a data file.

ResearchDataStrategyCodeBacktestValidate

What do you want to find out?

Does volatility targeting add value on gold?

PaperarXiv 2603.01298, read in full

Volatility-Managed PortfoliosAlan Moreira, Tyler Muir. SSRN 2659431, abstract only

GOLD 1h124,057 hourly bars, 2006 to 2026. Bundled sample

gold_vol_target.pyWrote strategies/gold_vol_target.py

Backtest result GOLD 1hrunning

Net profit: $0
Sharpe: 0.87
Max drawdown: 50.7%
Trades: 0

Not yet: do not trust this result

Deflated Sharpe ratio: Charged for all 13 variants tried on this strategy so far, the probability the Sharpe is above zero is 0.0%.

29 pass4 warn3 fail10 info3 skip

Failed: Deflated Sharpe ratioHaircut Sharpe and multiple testingRandom-entry benchmark
Cautions: Data qualityProbability of backtest overfitting (CSCV)Sub-period consistencyAlpha versus buy-and-hold

An idea, a question about a market, or a paper you want to test… Paper or data

A recreation of one real session on 4 October 2026, on the bundled GOLD 1h sample data. See the original screenshot.

See every check from this run

Try enough ideas and one of them will look brilliant.

Every line is a strategy with no edge. Add more, and the best one still looks good.

Simulated random walks, 504 daily steps, zero expected return. Not market data.

Variants tried13

Best one made: +56.9%
Its Sharpe: 1.54
Ended up: 7 of 13

The terminal counts every variant you try and charges the result for all of them.

It runs on your machine.: Your strategy files and your data stay on your disk.
It is in beta.: There is no public download yet. Join the list and you get one email when there is.
It does not trade.: There is no broker connection, and it gives no signals or promise of returns.

Questions people ask first.

What data does it use?

Ten sample datasets ship with it, and you can import your own CSV, Parquet, Excel and JSON files. The examples on this site use the bundled gold sample.

What is Kepler?

Our hosted finance agent. It reads your data library and runs backtests for you. You can also use an agent already installed on your machine. In both cases the verdict is computed from the checks.

Will it tell me my strategy is good?

Only if the checks say so. The run on this page beat buy-and-hold in the backtest and still got “do not trust”.

Who builds this?

Two engineers. We are not a fund or a signal service.