Chimera PredMark
Venues
Scanning
Displayed as the expected next scan. The real cadence is the Cloud Scheduler job — change it there too.
Falls back to 24h volume where a venue reports no liquidity, as Kalshi does.
0–1. Lower finds more pairs and more false pairs. Numbers and dates must match exactly regardless.
A market resolving imminently cannot be entered on two venues safely.
Pairs below this top-of-book margin are not priced against real depth. Lowering it costs scan time.
Comma separated.
Comma separated. Empty means all.
Margin
Net of both venues' fees. 0.02 = 2%.
Below this a pair is not written down at all.
The tails are where stale quotes and phantom liquidity live.
A binary market's two outcomes must price to about $1 between them — holding both pays exactly $1. Markets breaching this are dropped as bad data. Limitless returned the same order book for both outcomes of 61 markets, which presented as a 1090% arbitrage.
Margins above this are recorded but never traded. Real cross-venue spreads are a few per cent; anything far above is a stale quote or a mis-mapped outcome.
Execution
Capital must sit on both venues at once — exposure is the sum of both legs, not the difference.
Ceiling on the above. The API refuses more regardless of what is typed.
As a probability, e.g. 0.01 = one cent.
How many times to chase an unfilled second leg before unwinding.
Most we will pay above the quoted price to complete a hedge. Past this, unwinding is cheaper.
Single-leg fills must be caught in seconds, not at the next scan.
Unwinds always execute — this only decides when one is flagged as a bad price. A setting that could refuse to close an unhedged position would turn a small loss into an open-ended one.
Risk
Summed across venues. Positions cannot be netted across platforms.
DIFFERS is never tradeable under any setting.
Alerts
Blocks further trades once the execution layer has needed containment this many times today. 0 disables.