REPLYOPSAI / METHODOLOGY

Research should show
the inconvenient results.

ReplyOpsAI is an independent automated-trading lab. We build systems, run forward PAPER experiments and preserve favorable and unfavorable outcomes under explicit version labels.

FORWARD TESTINGVERSIONED SYSTEMSSANITIZED METRICSFAILURES PRESERVED
01

What we publish

Experiment configuration, system versions, closed and open trade counts, realized PAPER PnL, failures and material strategy changes. The evidence mode is shown next to the result rather than buried in a disclaimer.

02

What we do not claim

A small winning sample does not establish a durable edge. A profitable backtest does not establish forward profitability. PAPER execution does not establish live performance. Each stage answers a narrower question.

BACKTESTHistorical simulation→PAPERForward simulation→LIVEReal capital
03

Infrastructure isolation

Trading systems and the public website run on separate infrastructure. Exchange credentials and bot-control interfaces are not exposed to the website. A sanitized metrics feed is the bridge between the research environment and public pages.

TRADING VPSmodels · execution · credentials
→
SANITIZED METRICSaggregate results only
→
PUBLIC SITEread-only research pages
04

Version history is evidence

When a strategy or model changes materially, the previous result remains attached to the previous version. We do not merge a redesign into the old record and present the combined history as one unchanged system.

05

What our metrics mean

METRICINTERPRETATION
Realized PnLClosed-trade PAPER outcome for the displayed version; open PnL is separate.
ExpectancyAverage realized result per closed trade in the recorded sample.
Profit factorGross winning PnL divided by gross losing PnL; unstable or undefined when the sample has no losses.
DrawdownPeak-to-trough decline on the recorded closed-PnL path, not a promise about future risk.
Win rateShare of winning closed trades; it must be read together with payoff size and sample count.
06

How to read a small sample

Early results are operational evidence, not statistical proof. A handful of trades can reveal broken exits, excessive overlap or infrastructure problems, but it cannot establish stable profitability. We therefore show trade counts next to performance and keep collecting forward observations.

READ NEXT

Methodology in practice.

CURRENT RESEARCH

See the methodology applied to real experiments.