The Retraction Ledger · Public

Every strategy we killed —
and the bug that killed it.

Most quant shops bury their failures. We publish ours, with the exact false-positive that fooled the backtest. A verifier is only credible if it is willing to say FAIL: this ledger is the proof that, when MIZAN says a strategy is real, it means something.

40+
Strategies retracted
100%
With documented root cause
8
Distinct failure signatures
$0
Buried · everything public
"A firm that publishes its own failures is a firm whose passing claims mean something."
The retraction record is the commitment device — it cannot be faked retroactively
The Failure Signatures

Every way a backtest lies.

A fabricated track record is cheap to produce and nearly impossible to spot by eye. Each retraction below traces to one of these structural failure modes, the exact patterns MIZAN's engine now checks, independently, on every credential.

01 · Lookahead

Same-bar signal

Signal computed on a bar's own close, then booked against that same bar's return. No lag. "Buys what already went up."

02 · Lookahead

Period-boundary

Month-end values used to trade from the month's start; same-day VIX spikes used to de-risk that same day's loss.

03 · Lookahead

Volume / data leak

Full-day volume used at the open as an entry filter — information not yet knowable at decision time.

04 · Survivorship

Future-winner universe

Backtesting on today's index members back through history — buying tomorrow's winners before they were chosen.

05 · Fills

Phantom fills

Exits booked at stop prices the market never actually traded through. Free, impossible profit.

06 · Sampling

Hidden drawdown

Monthly-close sampling that masks catastrophic intra-period drawdowns at leverage.

07 · Statistics

Annualization error

Per-trade returns annualized at the wrong time base — inflating a Sharpe of 1.9 into 13.

08 · Selection

Window / cost fantasy

Edge that exists only in a favorable window, or that ignores real costs that turn it negative.

The Ledger

The graveyard, in the open.

Selected retractions. Each names what it claimed, the bug that produced the claim, and the honest result after the fix.

Crypto Squeeze family — 4h / 1h / 30m
Phantom fills
Claimed
Sharpe 8–16, Calmar 90–160, five straight positive years.
The bug
The entry trigger admitted closes below the band while the stop sat just above it, so the backtest booked instant "wins" at stop prices the market never traded through. On 79% of 4h trades. Tighter stops produced bigger phantom profits, which looked like an "optimal stop" discovery.
Honest
Caught with real money: a live bot traded it on Binance → 8 trades, 12.5% win rate, 7/8 instant stop-outs. Honest-fill Sharpe −4 to −8. Entire family killed.
US Equity Momentum family — Fund / PC / Broad
Survivorship
Claimed
Calmar 1.88, CAGR 30%, 30 of 30 positive years.
The bug
The universe held today's 322 S&P survivors back to 1997, so the engine was quietly buying future index winners before their inclusion dates. The edge was pre-inclusion buying, not stock selection.
Honest
Point-in-time membership → Calmar 0.40; modern-era (2018+) CAGR 7.7%, Sharpe 0.55. SPY-like, with worse drawdown. Whole family retired.
BTC + Alt Overlay — v1 through v19
Lookahead + hidden DD
Claimed
+7,177% CAGR, 7.3% max drawdown, walk-forward 9.98.
The bug
Regime inputs used each month's end value to trade from its start; crash months got retroactively classified as profitable shorts. Plus seven liquidation events at 6.9× leverage, entirely hidden by monthly-close sampling.
Honest
Intra-month daily mark-to-market: −86% per year, worst single day −193%. All 19 versions void.
XAU+Forex v7 · Trend v1 · S3 Hedge
Same-bar lookahead
Claimed
Calmar 5.25 (XAU), 13.9 (oil E&P), 100% positive years.
The bug
The monthly signal was computed at the month's close and booked against that same month's return — no lag. With a fast EMA equal to the current price, it was literally "did this month go up?" applied to that month. The celebrated "EMA(1/3) is universal" result was the bug's signature.
Honest
With a one-period lag → Calmar 0.06, CAGR 0.7%. There was nothing to trade.
Fund v45 · PC v21 · US Broad v3 — risk overlay
Same-day lookahead
Claimed
Calmar 1.88, Sharpe 1.77 (Fund); Calmar 3.27 (PC).
The bug
The daily risk block set today's exposure from today's closes, then booked today's close-to-close return on those weights, so a VIX spike cut exposure retroactively on the very day it would have cost money. Not implementable live.
Honest
With t-1 risk signals → Calmar 0.77 (Fund) and 0.66 (PC). Both fail the hard limits.
HL Cross-Sectional Momentum v3
Annualization error
Claimed
Sharpe 13.06 — a Medallion-class number.
The bug
Per-trade returns from a 48-hour hold were annualized at the hourly rate, √(365×24). A pure units artifact — plus a lookahead coin-filter and a 2bp cost fantasy.
Honest
True Sharpe 1.89. Re-validated honestly it survives at a real ~1.5 — a genuine but ordinary edge, not a miracle.
India Gold v5
Hidden drawdown
Claimed
Calmar 4.89, max drawdown −4.3%.
The bug
Month-end equity sampling hid the intra-month drawdown at 3× leverage — the peak and the crash both happened inside a single month and the monthly series never saw them.
Honest
Daily mark-to-market true max drawdown −51% (the underlying fell 19% in two days at 3×). Fails all four hard limits.
Intraday / Day-Trade family · Liquidation-as-stop
Volume leak · cost fantasy
Claimed
High intraday Sharpe; "leverage as a free stop-loss."
The bug
Full-day volume used at the open as an entry filter (data not yet known). And at 100× leverage, fees alone are 10% of margin per round-trip — negative expectancy no position-sizing can fix.
Honest
Edge gone once volume is lagged; negative EV across all 54 configs over 968K honest trades. Retracted.
Gap Fade
Phantom fills
Claimed
59% win rate, average loss −0.010%, reward/risk 220×, walk-forward 6.8.
The bug
The average loss is the stop distance. The backtest assumed a 0.010% stop always filled at the stop price. It cannot: a stop that tight is inside the spread. Live paper on 2026-06-05 stopped out all five positions intraday. Same signature as the phantom-fill family — the profit was in the fill assumption, not the signal.
Honest
Expectancy is unknown and the 220× reward/risk is an artifact. Never traded with real money. Retired permanently 2026-07-10: daemon stopped, launch agent removed, code archived. A KeepAlive flag had silently restarted it after an earlier halt — it ran one more paper session before we caught it.

This is the product. The discipline that kills these is the same engine that credentials the survivors.