QUÉ BOLÁASERE AI · EXPERIMENTEspañol ↗Try a rule.
See what happened.
Choose an asset, strategy, and position-sizing method. Compare the historical simulation with holding the same asset; the original SMA strategy remains available.
New: comparison curves, notebook and free PDF reports ↗The original comparison is preserved: long when SMA 10 exceeds SMA 30; cash otherwise. Fixed size uses 100% with no costs. Martingale or Antimartingale use the new next-open model.
New simulations: 10% of balance per entry. Martingale doubles after a loss; Antimartingale after a win; both capped at 10% → 20% → 40%, resetting to 10% when the streak breaks. Long only, no leverage; exit after at most 10 candles. Fee 0.1% and slippage 0.1% each side. Indices are price references; futures exclude contract multipliers and expiry mechanics. The original hold benchmark uses 100% with no costs and is shown separately. An additional reference holds 10% of initial capital from the first eligible open to the final close with the same costs. It does not match Martingale/Antimartingale’s changing exposure. Increasing size does not create an edge or guarantee recovery of losses.
Source and lookback depend on the instrument: Coinbase provides roughly up to 12 days (1h), 75 days (6h), or 300 days (1d); Yahoo Finance roughly 3 months at hourly intervals and 2 years daily. Yahoo data may be delayed.
This page shows up to 100 results from this browser or the current session, capped at 256 KB. The newest results are retained. Private server storage is not a personal history recoverable from this page; its status is reported separately.
Recover older history · review and original backup
Download the unchanged original, up to 4 MB, even if it cannot be validated. Recovery validates every record and prepares up to 100 recent results within 256 KB. Local history is replaced only after your confirmation; opening this page does not migrate it automatically.
No experiments are archived yet. The first one you run will start the history.
Strategy builder
Visual A/B rules, chronological validation, sensitivity and final holdout.
Build rules. Compare outcomes.
Compare two rules without coding. Three chronological windows evaluate stability; the final 20% of the evaluation sample remains reserved until candidates are locked. No winner is selected automatically.
History format and assumptions
time,closedAt,open,high,low,close,volume · Unix timestamps in seconds. 120–2,000 rows, up to 512 KB; no duplicate, future or overlapping rows. CSV must use one timeframe and one market. The public source returns up to 300 bars; long periods may not be covered. Continuous futures without rollovers or margin modeling.
Normalized capital 100; long or cash, no leverage. 61 warmup bars; each window starts in cash and is liquidated at its last close with costs. Drawdown is measured at closes; no taxes, dividends or interest.
Candidate A
Enter when average 5 exceeds 20; exit when it no longer does.
Loss/profit exits are decided at the close and executed at the next open; these are not intrabar stops. 0 disables that limit.
Candidate B
Enter above the prior 20-bar high; exit below the 10-bar low.
Loss/profit exits are decided at the close and executed at the next open; these are not intrabar stops. 0 disables that limit.
Saved versions · 0/20
Save the comparison after reviewing the reserve. Only configuration and summary are retained, not the full history.
Hypothetical historical simulations, with no probability of future profitability. A short or repeatedly inspected sample does not validate a strategy.