Calibration Ledger

Performance

Realized P&L, calibration backtest, dual-track scoring outcomes per event. Every binary event scored against both tracks regardless of which we ran — the only honest way to grade the referee.

Net Total P&L
+A$14,322
since 2026-05-19
Realized Cumulative
+A$11,808
5 closed positions
Unrealized Open
+A$2,514
DELL + AVGO
Return on Capital
+14.3%
A$100k base · 9 trading days

§01Realized P&L by event

AUD · per closed position
+334
DE
+1,282
MRVL T1
+2,529
ROST
−531
AVGO*
+424
ZS
+2,229
MRVL T2
+134
CRM
+2,407
MDB
+2,200
ORCL

* AVGO 5/22 stopped at $411 (−A$531); re-entered 5/22 PM at $416 (still open as carry, separate position). MRVL counted twice: T1 trim 5/20 + T2 residual 5/27.

Closed positions9
Winners8
Losers1 (AVGO stop)
Win rate89%
Avg winner+A$1,567
Avg loser−A$531
Win/loss ratio2.95
By primary trackSAMPLE
POP-SELL wins7 / 7
POP-SELL losses0
HOLD-THRU wins1 / 1
HOLD-THRU losses0
STOP-OUT1 (AVGO 5/22)
Track-pick accuracyn=4 · 4/4
Sharpe-class9 DAYS
Period return+14.3%
Largest drawdown−A$531 (0.5%)
Max consecutive losses1
Capital deployed~38–99% (variable)
Buffer breach0
Concentration flags3 (logged)

§02Calibration backtest · 20-event target

n = 4 events so far

The Fade-Risk referee gets graded on every binary event regardless of which track we ran. A directionally correct call (referee chose POP-SELL and the print faded, or chose HOLD-THRU and the print held) is a WIN. The goal is to hit n=20 with ≥ 75% referee accuracy before v3.

#01
DE
2026-05-21
PredictedFade-Risk ~8 · POP-SELL primary · pre-print exit @ $570.82
RealizedBEAT +14.1% EPS · spike to 576 → round-tripped to 540.76. Hard fade.
P&L+A$334 (vs hold-to-now −A$926; better by ~A$1,260)
POP-SELL WIN HOLD-THRU LOSS Fade-Risk 8 correctly predicted full round-trip. Referee directionally correct.
#02
ROST
2026-05-21
PredictedHold-thru conv 7.86 · Fade-Risk ~5 · SPLIT call
RealizedBEAT-and-RAISE Q1 · spike $237.88 (+9.5%) → settled $231.75 (+6.7%). Not full round-trip.
P&L+A$2,529 (POP-SELL run, both tracks would have won)
POP-SELL WIN HOLD-THRU WIN Fade-Risk 5 correctly predicted SPLIT behavior. Contrast with DE FR8 hard-fade — referee discriminates.
#03
ZS
2026-05-26
PredictedPop-sell conv 8.4 · Fade-Risk 8 · beat-and-light-guide pattern flagged
RealizedBeat headline + Q4 guide LIGHT → −20.8% AH to $146.24
P&L+A$424 exit pre-print (vs hold-to-now ~−A$1,400; better by ~A$1,800)
POP-SELL WIN HOLD-THRU LOSS Confirms 3/3 premium-multiple BEAT_AND_GUIDE_FADE pattern (DE/MDB/ZS). Caps hold-through conviction at 5.6 for class.
#04
MRVL · CRM
2026-05-27
PredictedPOP-SELL conv 8.5 / 7.0 · Fade-Risk 8 · stacked binary day
RealizedBoth exited into pre-print strength · MRVL $207.41, CRM $180.00
P&L+A$2,229 (MRVL T2 residual) · +A$134 (CRM) = +A$2,363 combined
POP-SELL WIN Both vindicated. CRM had BofA downgrade headline already — Fade-Risk 8 correctly predicted exit-into-strength was optimal.
Calibration · n = 4 · referee accuracy

4 / 4 directionally correct on Fade-Risk referee track-pick. Three HIGH (≥7) → POP-SELL ran and won. One MID (~5) → SPLIT called, both tracks won. Zero referee misses in the sample. Confidence in v2.1 Phase 1 thresholds: PROVISIONAL PASS, gated on reaching n=20.

Next calibration events: DELL 5/28 (FR 8/10, POP-SELL) · AVGO 6/03 (FR 3/10, HOLD-THRU) · CRWD 6/03 · ADBE 6/11.

§03All closed positions

since 2026-05-19
TkrShEntryExit Days heldTrackP&L AUDOutcome
DE30562.85570.82~3POP-SELL+334.3Pre-print exit vindicated
MRVL (T1)40168.53191.46~2TP1 trim+1,282.3Scale-out into strength
ROST91211.12231.00~3POP-SELL+2,529.1Beat-and-raise, AH spike sold
AVGO (1st)76427.50411.00~2STOP−531.2Stop-out · re-entered later
ZS48176.05~182.00~5POP-SELL+424.1Pre-print exit · vindicated
MRVL (T2)41168.53207.41~7POP-SELL+2,229Residual exit first 1–3 min
CRM48178.07180.00~7POP-SELL+134.2Downgrade-tagged exit
MDB44323.74~360~7POP-SELL+2,407Hard rule honored: no hold-thru
ORCL83185.24~194~9HOLD+2,200Stop-trail exit on yield re-accel
Total realized +A$11,808 8W / 1L · 89% hit rate
Path to v3

Gates remaining: (1) reach n = 20 calibration events on Fade-Risk referee accuracy · (2) walk-forward Phase 10 backtest on historical universe · (3) Kelly sizing / max-loss Phase 9 integration. Provisional pass at n=4 — sample too small for production claims. The discipline is real, the sample is not yet statistical.