Xuer Capital

Strategy validation and research systems

Find out whether a strategy does what its backtest claims.

We re-build the strategy from its rules, run it on data it could actually have seen, charge it the costs you would pay at your capital, and test whether the result survives the number of things that were tried. Or we build you a system that can ask that question of every idea you have.

Two services

Strategy validation

An independent answer to one question: does this strategy do what its backtest says, on data it could have known, at your capital, after counting what was tried?

You send

  • The rules, precise enough to re-implement without asking.
  • The claimed results and where they came from.
  • The capital it would run at, and how many variants were tried.

You receive

  • A written report in five fixed parts: replication, costs at your capital, luck, forward test, verdict.
  • A verdict in plain words: replicates and survives; replicates but does not survive; does not replicate; or could not be assessed, with the reason.
  • The rules frozen with a fingerprint, re-checked on new data at stated dates, so the strategy keeps being tested after the report.

Limits

  • Daily and monthly rules. No intraday rules, no options.
  • No advice on whether to trade it. You get what was found; you decide.

Research-system development

A system of your own that can test that question properly, for every idea, without a human remembering the rules each time.

You send

  • The questions you want answered and the markets you trade.
  • The data you hold or pay for.
  • Where it has to run and who will use it.

You receive

  • A point-in-time data store: each value used only once it was public.
  • A backtester that fills at realistic times and charges fees, spreads and minimums at your capital.
  • The statistical checks from our audits built in, and frozen forward tests with paper tracking.
  • Code, tests and documentation you own. Exact deliverables are fixed in the written scope.

Limits

  • Every build is scoped before work starts; nothing is sold as a ready-made product.
  • We build the means of testing strategies, not the strategies.

What exists and what is scoped

The validation work runs on a research system we built and use for our own trading research. Some of what a client may want is already in it; the rest is work we would scope with you before starting.

Works today

  • Daily and monthly rules on US-listed stocks and ETFs, and on Bitcoin.
  • Point-in-time inputs: prices at the close, US macro data as first released, company filings by filing date.
  • Decisions on a completed bar, fills at the next open, the client's own fill assumption reported beside it.
  • Broker fees with per-order minimums, half the spread, fund expenses; cash earning the 3-month T-bill.
  • Drawdown on month-end marks and on daily closes.
  • Luck controls: Reality Check, deflated Sharpe at several trial counts, overfitting probability across a family, and a paired test of Sharpe differences.
  • Frozen forward tests with fixed checkpoints, and paper tracking of frozen rules on live prices.

Scoped per client

  • Data we do not hold: your own feeds, other exchanges, other countries.
  • Futures, other venues and asset classes beyond those listed.
  • Building on your infrastructure, or connecting to your broker and data vendors.
  • Your universe, your benchmarks, your reporting format.
  • Ongoing forward-test reporting beyond the checkpoints in a report.

Evidence: two published strategies, audited

To show the work, we audited two well-known public strategies the same way we would audit yours. US-listed ETF data through August 2026. Each verdict is stated against what the author actually claimed, not against an imaginary standard.

Faber — Global Tactical Asset Allocation

A Quantitative Approach to Tactical Asset Allocation (2007, updated 2013). Five asset classes, each held only above its 10-month moving average.

The claim

Equity-like returns with bond-like volatility and drawdowns. Claimed 10.48% a year for 1973–2012 with a worst drawdown of −9.5%.

What we found

  • It replicates: yearly returns for 2008–2012 within two points of the paper.
  • Since publication (2013 to August 2026): 6.0% a year against 6.9% for holding the same five assets, with a worst loss of −8.5% on daily closes against −28.2%.
  • The paper's −9.5% is read off month-end values. On daily closes the worst loss inside the paper's own out-of-sample years was −14.9% (2011).
  • The higher Sharpe ratio is about one standard error above holding's (p = 0.31 since publication): in the claimed direction, not distinguishable from zero.
  • At $10,000 after real costs, about 4.6% a year from 2007 against 5.4% for holding.

Verdict

Replicates and delivers the drawdown control it claims. It provides no return edge over holding, and the paper never claimed one.

Antonacci — Global Equities Momentum

Dual Momentum Investing (2014). Holds US or non-US stocks, whichever is stronger, when US stocks beat T-bills; otherwise bonds.

The claim

17.4% a year for 1974–2013 with a worst drawdown of −22.7%, against 8.9% for world equities.

What we found

  • The 40-year headline cannot be checked on ETF data, which starts in 2008. The mechanism can: it sidestepped 2008 (+5.0% against −30.4% for the S&P 500).
  • Since the book (2014 to August 2026): 8.84% a year, against 7.86% for its three assets held equally, 10.59% for world equities and 13.76% for the S&P 500.
  • In March 2020 it lost −33.7% on daily closes, exactly what the S&P 500 lost. The month-end figure shows −19.4%.
  • Its edge over its own three assets is not distinguishable from luck at any trial count, including the book's own four lookbacks.

Verdict

Replicates in character. The headline cannot be verified on post-2008 data, and the edge over holding its own assets is not distinguishable from luck.

Found along the way: the fill assumption is worth half a point a year

Both papers assume each trade fills at the close of the day the signal fires. Filling at the next morning’s open instead cost Faber’s rule 0.49 points a year (about 2.4 standard errors) and Antonacci’s 0.33 (too noisy to call). Holding Faber’s five assets showed no such gap. A month-end rule’s backtest can earn something a realistic order does not, so every audit we run reports both.

Full reports, with methods, data defects found and the frozen forward tests, are available on request. Independent analysis; not affiliated with or endorsed by the authors. Past results, real or simulated, do not predict future ones.

How to hire us

  1. Step 1

    Send the enquiry

    Use one of the two buttons below. Each opens an email listing exactly what the work needs.

  2. Step 2

    Get a written scope

    We reply with what can and cannot be tested, what is missing, the deliverables, the timeline and the price.

  3. Step 3

    Receive the work

    A report in five fixed parts, or a system you own, with the frozen forward test or handover the scope names.

Strategy validation

Opens an email that asks for the rules, the claimed results, the data, your capital and how many variants were tried. Send what you have; we will say what is missing.

Start a validation enquiry

Research system

Opens an email that asks for the question the system must answer, the markets, the data you hold, where it should run and your timeline.

Start a build enquiry

Email client not set up? Write to inquiries@xuercapital.com.