Alphanume Research Track Record, Methods, and Strategy Results

A research archive is evidence of process. It is not automatically an audited investment track record, and the distinction should stay visible.

Readers evaluating Alphanume Research usually want to know two things: whether the work is serious, and whether the published strategies held up. Those questions deserve more than a highlight reel.

This page explains how to read our public work, which evidence is available, and what cannot be concluded from it. It is current as of 2026-09-04 and should be updated when the archive, methodology, or any formally reported result changes.

What the public record includes

The Alphanume Research archive contains dated articles on quantitative strategies, options, event-driven trades, market mechanics, alternative data, and research systems. Individual posts may include a hypothesis, source data, implementation steps, backtests, failure analysis, or a reconstruction of a documented market event.

The archive is useful because publication dates establish when an analysis appeared publicly. A reader can inspect what we chose to study, how we framed uncertainty, and whether later work acknowledges failure. The archive does not by itself establish that every illustrated trade was executed with capital.

Our public About page describes the audience and editorial scope. It should be read as a statement of what we aim to publish, not as third-party verification of skill or performance.

What we mean by track record

Track record can refer to at least four different things. Mixing them produces false confidence.

• Publication record: dated evidence that a hypothesis, test, or conclusion was made public.
• Backtest record: hypothetical results from a rules-based historical simulation.
• Paper portfolio: forward-tracked signals without verified brokerage execution.
• Live audited record: actual positions and returns reconciled to accounts under a defined calculation standard.

Alphanume Research primarily publishes research studies and strategy experiments. Unless a post explicitly supplies stronger documentation, its results belong in the publication or backtest categories. They should not be combined into a single live, audited composite.

This conservative classification protects the reader. It also makes the work easier to improve, because every claim is judged against the evidence it actually has.

The method we want readers to audit

1. Mechanism first. State why an effect might exist and who bears the economic cost.
2. Point-in-time data. Use only information available at the simulated decision timestamp, including the correct filing and revision history.
3. Frozen rules. Define the universe, signal, entry, exit, sizing, and portfolio constraints before reading the final result.
4. Realistic execution. Model spreads, fees, borrow, delays, capacity, and unfilled orders at the instrument level.
5. Honest comparison. Use a benchmark or control group that shares the important risk and event exposures.
6. Robustness. Vary economically meaningful assumptions and reserve later data for confirmation.
7. Failure reporting. Preserve dead trades, rejected hypotheses, and conditions under which the effect disappears.

Our article on quant research at industrial scale explains the systems side of this approach. The practical standard is also summarized in How to Read a Backtest Honestly.

How to read the strategy studies

You Can Still Find Mispriced Options is best evaluated as an options-research case study. Ask how mispricing is defined, when the inputs become known, whether contracts were executable, and how selection affects the sample. The interesting claim is not that every anomaly becomes a trade. It is that a narrow structural question can be tested.

Most Quant Newsletters Would Never Post This is the clearest dated results summary in the public archive. It reports one year of strategy and dataset outcomes as of August 28, 2026, including drawdowns, sample limits, and weak sleeves. These remain publisher-reported research results, not an audited composite or proof of future performance.

The archive also contains experiments that are negative, conditional, or operationally difficult. Those outcomes matter. A research publication that only preserves winning charts teaches readers to select stories, not build systems.

A result also has a lifecycle. A hypothesis can move from exploratory to specified, from specified to retested, and from retested to retired as costs or market structure change. Publication dates should remain fixed while status notes carry their own dates. Quietly rewriting an old thesis destroys the evidence that a public archive is supposed to preserve.

A results ledger that avoids category errors

study: [canonical article URL]
publication_date: [public timestamp]
evidence_class: publication/backtest/paper/live-audited
hypothesis_frozen_before_test: yes/no/unclear
data_vintage_recorded: yes/no
universe_and_delistings: stated/partial/missing
execution_model: stated/partial/missing
reported_result: [metric with period]
drawdown_or_tail: [metric or not reported]
dead_trades_preserved: yes/no/unclear
independent_replication: [citation or none]
current_status: active/retested/retired/educational

This ledger is more useful than one performance number. It prevents a historical simulation from being presented beside a live result without a label. It also makes missing information visible.

Mutable figures should carry an as-of date, test window, sample size, cost model, and revision note. If a result cannot be tied back to a frozen article or dataset, do not promote it into a track-record summary.

Aggregation is especially risky. Combining unrelated option, equity, crypto, and event studies into one average return hides capital overlap, leverage, correlation, and periods when no trade was possible. A composite is meaningful only after allocation rules and cash treatment are fixed. Until then, keep study-level evidence separate.

What readers should not infer

• A published backtest is not evidence that Alphanume or a subscriber received the simulated fills.
• A profitable case study is not a forecast of future returns.
• Several related studies are not independent observations of skill.
• A mechanism that appears credible can still be too small, crowded, costly, or fragile to trade.
• An archive maintained by the publisher is first-party evidence unless independently replicated or audited.

Readers should also watch for survivorship at the article level. A publication can leave old posts online and still choose only successful tests for a landing page. That is why we prefer linking the complete archive and describing the evidence class of each highlighted study.

How Alphanume data fits into the process

Research content and research infrastructure are related, though they are not the same product. Alphanume provides market and event data intended to help researchers build point-in-time studies. The documentation explains available datasets and interfaces.

Using our data does not validate a strategy, and publishing a strategy does not validate every dataset. A serious workflow checks coverage, timestamps, adjustments, revisions, and missingness independently. Our post on point-in-time data explains why an apparently small timestamp error can reverse an event study.

Reproducibility has degrees. A reader may be able to reproduce the direction of a result without matching every observation, or reproduce a dataset while reaching a different portfolio outcome. Those disagreements should be logged with code version, data vintage, and decision rules. A clean discrepancy is more valuable than a forced match.

Commercial disclosure is simple: Alphanume publishes both this page and Alphanume Research. Treat product links and editorial descriptions as first-party information. Verify material claims against the underlying data and your own execution environment.

A due-diligence standard for following the work

Pick one study that overlaps your mandate and try to break it. Recreate the eligible universe, reconstruct information availability, use a worse fill model, and test a later period. Record discrepancies instead of smoothing them away.

1. Save the canonical article and publication timestamp.
2. Write the claimed mechanism in one sentence.
3. List every field needed to reproduce the decision.
4. Mark fields that were revised or unavailable in real time.
5. Rebuild the baseline before adding filters.
6. Stress costs, delays, capacity, and one plausible control group.
7. Classify the outcome as reproduced, directionally similar, not reproduced, or not testable.

That exercise is the most honest way to evaluate our work. Browse the public archive, select a mechanism you understand, and subscribe only if the process helps you make better research decisions. The goal is not agreement. It is a record clear enough to challenge.