Track Record Methodology

What “simulated execution” means

No real order reaches any market, and the $1,000 per sequence is simulated capital. No real funds are at risk at any point. Two environments produce the fills, and each sequence on /performance is labelled with the one that measured it:

• Alpaca paper. US stocks, and crypto launched before September 27, 2026. Fills are simulated by Alpaca's paper trading environment against real market data.
• Candle simulation. Crypto launched from September 27, 2026. The ladder is computed exactly as for Alpaca paper, then advanced by replaying public one-minute market candles (high, low, open, close). This covers every crypto instrument in the catalogue, where Alpaca paper covered only the ones Alpaca lists.

The two are reported as separate populations and never added into one average: they are different measurements of the same rule, and a blended figure would describe neither.

Real execution would differ from these simulated results. Among other things, live orders face trading fees, slippage, partial fills, queue position for limit orders, and the actual liquidity available at a quoted price. Paper fills do not reproduce these frictions faithfully, so simulated results do not account for real-world fill quality.

What is included, what is excluded

Inclusion. Every sequence the engine launches appears on /performance, chronologically and whatever the outcome: in progress, completed and canceled alike. There is no filtering on results.

One documented exclusion. A small number of early sequences were affected by a pricing-precision limitation that has since been corrected. Those sequences are flagged and excluded from the public figures; nothing is deleted, every flagged row keeps its full order history, and the corresponding logs are available upon request.

No selected start date. The auto-launch rule entered production in June 2026 and has run continuously since. The reporting window starts when the protocol started, not at a date chosen for its results; the /performance page derives its “live since” date directly from the first auto-launched sequence on record.

How the headline figures are computed

Closed within 14 days. Of the sequences launched at least 14 days ago, the share that completed within 14 days of launch. A sequence younger than 14 days is not counted either way; one still open after 14 days counts against the figure.

Projections. The profit realised by closed sequences since the protocol started, divided by the average capital in play over the same period (each sequence's budget weighted by the time it stayed open, sequences still open included, with no profit counted for them yet), then annualised and compounded. It is a return per dollar while in play: capital kept in reserve between launches is not counted. A high-level range, not a forecast.

Limitations

The track record is published for transparency, and that transparency includes its known biases:

• It is a simulation. Paper fills or candle replay, no live fees or slippage. Live results would differ, and the direction of the difference is not guaranteed to be favorable.
• The history is short. The protocol has only run since June 2026. Aggregate figures computed on a small number of closed sequences can move materially as new sequences close.
• The catalogue evolves. The set of instruments eligible for auto-launch changes over time as coverage grows, so different periods of the record sample different universes.
• The early crypto record is a narrow slice. Until September 27, 2026, crypto ran on the Alpaca paper account, so an instrument could only enter the record if Alpaca listed it: 31 of the 247 crypto instruments in the catalogue. Read the Alpaca paper crypto figures as describing the instruments Alpaca happens to carry, not the catalogue. This bias is why crypto moved to the candle simulation, which covers the whole catalogue. US stocks are almost unaffected: 148 of 149 are listed on Alpaca.
• Candle replay is not an order book. A candle records that the price traded at a level; it does not record that an order of a given size would have been executed there. Two rules make the simulation round against itself rather than in its own favour. A price level counts as reached only when a candle strictly crosses it, never on an exact touch, because an order resting at the extreme of a minute is the one least likely to have been executed. And a minute in which a buy is triggered never also closes the sequence: a candle records that the price reached both its high and its low, not in which order, so closing in that same minute would take a gain out of a sequence of events nobody can verify.
• The thresholds are ours. The 4-and-4 rule is a design choice made by Tradingale. The record describes the model at these settings only; different settings would produce a different record.

Data access

Aggregated figures and the full per-sequence table are published on /performance. Detailed transaction logs for any sequence, including the flagged defective rows, are available upon request: support@tradingale.com.