In search of alpha.
We survey the ground for alpha, one hypothesis at a time. Most of it is barren. Here is the evidence.
Every hypothesis is written down before anyone looks at the data, with its pass and fail criteria fixed in advance. Then it is tested with real costs, out of sample and against chance. Here we publish the ground that turned out barren: where we dug, why it looked promising, what we found and what we learned.
up session down session surveyed windowillustrative, not market data
The survey so far
76
hypotheses surveyed
51
found barren
25
still under evaluation
67%
barren share
Barren ground gets a full field report. Ground still under evaluation is only a number until it is settled.The ledger →
First case
H-042 · event / calendar
There is no bounce after earnings. The loser keeps falling
Barren because: net return of zero in one variant and of the opposite sign in the other
publishes September 23, 2026 up session down session what we expected surveyed windowillustrative, not market data
Coming up
One field report a week. Each one leaves quarantine thirty days after its verdict.
H-044
Short-term momentum in large caps does not exist, and the last door closed
Barren because: negative net return, with no selection structure even before costs
publishes September 23, 2026H-059
The shorting premium is real, and you still cannot collect it
Barren because: significance (t 0.69) and a ranking indistinguishable from chance
publishes September 25, 2026How we judge a strategy
1Pre-registrationThe rule and its pass/fail criteria are written before anyone looks at the data.2Real costsCommissions, slippage, funding and borrow, and a pass at double the friction.3Out of sampleThe effect must show up in the stretch the original author could not see.4Against chanceShuffled versions of the same rule that keep the market drift and destroy the selection.5MultiplicityEvery attempt is counted, by family of mechanism, and the threshold tightens with it.6SurvivorshipUniverses rebuilt date by date with delisted names, or a 0.6 discount on the Sharpe.7No collidersEvery condition in a rule must be measurable before entry and explain itself.8One lookNon-renewable tests run once, with a date and guards against spending them twice.9Nothing deletedBarren ground keeps its full record: the numbers, the criterion and what would bring us back.
One field report a week, in your inbox
Where we dug, why it looked promising, what we found and what we learned. No signals, no promises.
Subscribe