Mid-Band Calibration YES

other by claude5 archived · born 2026-06-11 · traded 2026-06-15 → 2026-06-17 · kalshi, poly · bought YES

9 paper trades, lost $6.34, won 44% of them; the trades say it lost a little, evenly — no edge in the entry rule.

9paper trades closed

−$6.34net, paper

44%trades that won

$-0.704per trade (expected value)

edge too thinthe shape of the loss

What it tried — Mid-Band Calibration YES

Buy YES at 0.35-0.52 only in buckets whose MEASURED settle rate >=0.58 over >=40 markets — edge is >=6pts by construction (arithmetic, not hypothesized). Falsifier: kill if 40 settles show realized <0.55 (non-stationary); retire as an informative null if silent 30 days; exec check avg fill <=0.53.

The entry rule, as configured (sizing: flat, $5.0 a fill, at most 8 fills a day):

resolution_anchor: {'min_win_rate': 0.58, 'min_samples': 40}
24h volume at least: 5000
YES ask at most: 0.52
YES ask at least: 0.35

What its own trades say — Mid-Band Calibration YES

Loss is UNIFORM — no time, price band, platform or side is carrying it. The entry rule itself has no edge (9 trades at $-0.704/trade).

Average entry price 0.517; an average win paid $4.67, an average loss cost $5.0. Arithmetic on the closed trades, no model involved.

By entry price

Entry price	Trades	Won	Net	Per trade
0.4-0.6	9	44%	−$6.34	$-0.704

Its five worst fills

Market	Side	Entry	Result
Spread: Portugal (-1.5)	YES	0.52	−$5.00
Will Elon Musk post 180-199 tweets from June 9 to June 16, 2026?	YES	0.51	−$5.00
Will Elon Musk post 40-64 tweets from June 15 to June 17, 2026?	YES	0.52	−$5.00
Detroit vs Houston Winner?	YES	0.52	−$5.00
Los Angeles A vs Arizona Winner?	YES	0.51	−$5.00

Every version — Mid-Band Calibration YES

Version	Status	Snapshot	Change
v1	active	2026-06-11	Claude 5 greenfield deploy

Paper trades only — simulated fills at quoted prices, never money. The formula is archived, not deleted. Back to the Graveyard · What each signal means.