Skip to content
Pohuant

Research log

Before evaluating each hypothesis we wrote down the question, the data and the decision rule, with a verifiable date in Git. Then we evaluated it once. We publish every result, including the ones that did not work.

6 studies: 2 rejected and 4 inconclusive. None showed that these factors anticipate the direction of prices.

  1. fx_cross_sectional_reversal:v1

    Inconclusive

    Reversal of relative currency strength

    Question
    Do the currencies that rose most against the others tend to reverse over the next 20 days?
    Data
    Fed H.10 rates, 28 crosses, 1999–2014 (held-out sample, evaluated once).
    Result
    Rank correlation (IC) +0.022 · Newey-West t +1.03 · 63% positive years
    What changed
    No production change. Prospective readings will come in 2027 and 2028.

    Registered in commit f9ac29a

  2. fx_carry_spot:v1

    Inconclusive

    Interest-rate differential (carry) and exchange-rate direction

    Question
    Does the rate differential across the 8 currencies indicate the direction of the spot exchange rate over 20 days?
    Data
    Fed H.10 rates and 3-month interbank rates, 2002–2026.
    Result
    Rank correlation (IC) +0.032 · Newey-West t +1.59 · 76% positive years
    What changed
    No production change. The relationship has been close to zero since 2010.

    Registered in commit d754eca

  3. reer_valuation_review:v1

    Rejected

    Valuation by real effective exchange rate (REER)

    Question
    Does a currency that is expensive in real terms against its 10-year trend tend to weaken over the next 60 days?
    Data
    BIS REER indices and Fed H.10 rates, 2004–2014.
    Result
    Rank correlation (IC) -0.046 · Newey-West t -0.93 · 45% positive years
    What changed
    Valuation no longer counts as a validated factor.

    Registered in commit 6273036

  4. csfx_momentum_review:v1

    Inconclusive

    Momentum of relative currency strength

    Question
    Do strong currencies stay strong over the next 20 days?
    Data
    Fed H.10 rates, seven currencies, 1985–1998.
    Result
    Rank correlation (IC) +0.027 · Newey-West t +1.12
    What changed
    Relative strength became experimental.

    Registered in commit 811a745

  5. cftc_flows_review:v1

    Rejected

    CFTC positioning and currency direction

    Question
    Does CFTC futures positioning indicate the direction of currencies over 20 days?
    Data
    CFTC TFF reports and Fed H.10 rates, 2010–2014.
    Result
    Rank correlation (IC) -0.018 · Newey-West t -0.50
    What changed
    Positioning no longer counts as a validated factor. With this result no factor was left for a directional score.

    Registered in commit 811a745

  6. energy_inventory_spot:v1

    Inconclusive

    EIA inventories and WTI crude direction

    Question
    Do US inventories against their seasonal norm indicate the direction of WTI over 20 days?
    Data
    EIA weekly inventories and Cushing spot price, 1987–2014.
    Result
    Rank correlation (IC) +0.002 · Newey-West t +0.04 · 46% positive years
    What changed
    It stays experimental. The effect seen since 2015 did not exist before; it will be measured again in 2027 and 2028.

    Registered in commit c0ccf89

IC is the Spearman correlation between the factor and the subsequent price move: 0 means no relationship. A t below 2 in absolute value cannot be told apart from chance.