Track record

The desk's public register: every closed thesis, published unchanged after its close, and the metrics that measure the edge behind them. The page needs no account.

• To judge the desk on its results before subscribing. Hit rate, realised moves and edge against comparable instruments sit above the theses that produced them.
• To recompute a metric. The list under the band is complete and unfiltered: every figure shown can be rebuilt from the cards.
• To read a closed thesis in full. Clicking a card opens the detail — chart, memo, evidence ledger, log, closure note.

Two zones. At the top, the metrics band — three blocks side by side: All, then Markets and Trades. Below it, the list of closed theses, grouped into Markets then Trades, most recent first. The cards are the ones from the Theses view; a closed position is frozen and no longer moves. Clicking one opens the detail page.

What is measured

A thesis is measured from its entry to its close. The entry and close prices are frozen at the open and close timestamps; the distance between the two, in the direction of the call, sets the verdict.

What happens after the close does not enter the metrics. A Market thesis closed on convergence is measured on that convergence; the Polymarket question's resolution, often weeks later, does not change it.

The All block

The hit rate covers the theses closed to date. The 95% CI gives the range the true rate sits in given how many calls stand behind it: the shorter the register, the wider the range. It tightens with every close.

The All block shows no median move: a percentage of a price and a point of probability do not average together. That metric exists per call type only.

The Markets and Trades blocks

One block per call type: Markets for bets anchored to a Polymarket question, Trades for directional calls on an asset. The counter beside each title gives the number of closed positions of that type.

Beat peers

The metric isolates the desk's edge. It compares the realised move with what comparable instruments did over the same holding period: the other Polymarket markets priced at the same level at the moment of entry for a Market thesis, the asset's own history for a Trade thesis. The value shown is the share of those comparables the call came out ahead of.

A call can win simply because the instrument was moving: a Polymarket longshot decays over time, an asset in trend keeps running. Comparing against instruments that started at the same level over the same span removes the part of the move that was available to anyone, and leaves only what came from the desk's edge.

A call picked at random would land at 50% — a coin flip. Everything that holds durably above it measures the desk's work, not luck.

A low value on a won thesis is therefore not a contradiction: the call paid, and so did the comparables.

The metrics line under each card

The cards

Identical to those in the Theses view: header, question, entry → close, sparkline over the holding period, closure note, and the count of material at the foot. Each element is documented on Theses.

• The full register, open positions included: Theses.
• On a closed Market thesis, the Polymarket flow for that question: Markets.
• On a closed Trade thesis, the asset's current reading factor by factor: Trend matrix.

• Only closed positions appear here. An open thesis shows up in no part of the page — not the list, not the counters, not the metrics — so the band says nothing about the current book. Following the running book as it moves — entry, target, today's state, the log as it is written — takes a subscription.
• The figures are not returns. They are moves in the underlying instrument, with no position size and no leverage, and nothing is compounded from one thesis to the next. Hawkline gives you research-backed trade ideas and convictions, not a portfolio to copy.
• Moves are gross. Spread, financing and execution are not deducted; they depend on your broker.
• The number of calls governs how the rate reads. On a short register, a call type can post a good rate without yet being distinguishable from a random draw.