Vomit Elegy started this week. It is a C#/.NET tool that reads Kalshi’s public market data, builds portfolios from payoff constraints the exchange itself declares, and logs only the opportunities it can prove.
the idea
Most arbitrage scanners answer “are the prices wrong?” Vomit Elegy asks a narrower question: “is there a portfolio whose minimum payout is guaranteed in every admissible state, at executable depth, after fees?” If the answer needs a hand-wave, it does not get logged.
The research write-up in docs/research/niche-analysis.md is blunt about why:
published full-universe recon found that apparent edges were almost always fee
illusions, shallow depth, or capital lockup. A scanner that emits those is noise.
the first two adapters
- MECNET — in a mutually exclusive event, at most one market settles YES.
Buying NO on
Nselected markets pays at leastN - 1dollars per basket. The evaluator picks the most profitable fee-positive subset at displayed depth. - DIRECNET — threshold implication. If
value > 60impliesvalue > 50, buying NO on the stronger claim and YES on the weaker one guarantees at least one payout.
The scan runs in two stages: screen the whole logical graph from fixed-point snapshots, then verify the strongest events against fresh order books before logging a signal.
what it is not
Deliberately not a trading bot. There is no order placement, no keys, no
execution path. The vendored KalshiSharp SDK sits in vendor/ reserved for a
later phase that may never come. What exists today is recon: an append-only
JSONL log of opportunities, near misses, scan coverage, and rejection counts —
plus a local read-only dashboard and a recorder that snapshots books for
replay.
Kalshi’s historical API has markets, trades, and candlesticks but not order-book depth, so backtests are explicitly theoretical quotes. The README says so, which felt worth repeating here.
The bet is that the interesting artifact is not the trades — it is the constraint-health dataset: where incoherence appears, how long it persists, and why apparent opportunities fail.