Track record.

Real fills from broker accounts · honest backtests · nothing hidden

A note on scope and honesty — updated Aug 19, 2026

Quanta Privé trades only personal capital. We do not accept, solicit, or manage external funds — the principal is not licensed to do so and does not intend to be. This page exists to make the practice visible, not to raise money.

Every P&L number is paper-account, from real broker fills (Alpaca & IBKR paper). No live-money numbers exist yet — the pilot is deliberately capitalised on paper first. All figures below are broker-verified and pulled from the same reports engine that runs the operator’s daily Fleet Command dashboard.

Where we sit today. YTD net is +$9,178 across 337 trades. August alone has been tough — a near-flat month with the multi-day book taking real losses and Undertow near breakeven at the reduced $3.5k/trade size. We publish both sides. That is the discipline.

YTD net (paper, broker-verified)

+$9,178

337 closed trades · WR 46.6% · Jan 1 – Aug 20

MTD (August so far)

−$199

184 trades · best day +$606 · worst −$1,902

Bots live in production

Donchian pair retired 08-09 · +Bot-Donchian4 paused

Alpha-Engine research

21 / 0

campaigns run · 0 tradeable · chain dormant

What we're actually optimising for

Beat SPY on risk-adjusted return. Take profits daily / weekly / monthly — not once a decade.

SPY buy-and-hold is a formidable absolute-return machine — on paper. In practice, very few investors sit through the −55% drawdowns required to earn it, and no one who lives off the portfolio can wait ten years for the compounding to arrive. Our target is different: compound at a materially lower drawdown, and take periodic profits so gains are realised in cash, not paper.

Undertow trio — live paper fills

official Alpaca fills · period: MTD Aug

Per-bot performance — August MTD (at $3.5k/trade)

Real broker fills

Bot	Exit time	Trades	Win rate	P&L
Undertow-1	15:55 ET	46	52%	−$128
Undertow-2	14:00 ET	47	47%	−$254
Undertow-3	15:00 ET	47	53%	+$120
Combined MTD		140	51%	−$262

Context. Undertow ran at $10k/trade in July and produced +$8,683 across 134 trades. Size dropped to $3.5k/trade on Aug 3 to lengthen runway before any live-money step. August so far is near-flat — the edge is thinner at the smaller size, and market conditions have been unfavourable for event shorts.

Placebo validation

2026-07-20 audit

Event-day short performance

+0.23% / trade net

Random-day control (placebo)

−0.15% / trade net

Edge is real. Event-day short beats random-day control by +0.38% — reproduces the +0.354% event-vs-control differential from original validation. The edge is thin but survives placebo, which is what matters at this stage.

Multi-day book — August MTD

real broker fills · through Aug 20

Bot-ORB BASE

Live · tough month

First live2026-07-28

MTD trades / WR13 · 15.4%

MTD realized−$1,856

Open (unrealized)+$3,859

Open positions20 (max)

Universe300 S&P500 names

15% WR is well below target. Open book is holding significant unrealized — we’re carrying trend positions through a rough entry regime. No forced action; monitoring.

Bot-Donchian3

Live · observing

First live2026-08-04

MTD trades / WR23 · 17.4%

MTD realized−$1,866

Open (unrealized)+$141

Decision fidelity100%

DesignS&P150 · 35/20 · SPY>200d

Fidelity is clean, entries are honoured — the market is not yet rewarding the design. Not enough sample to conclude. Kept running.

Bot-Swing

Live · net positive

First live2026-07-31

MTD trades / WR6 · 33.3%

MTD realized+$122

Open (unrealized)+$502

Backtest (WF 2023-26)+$49k on $50k

Pre-registered review2026-11-10

Only positive multi-day bot MTD. Realized + open both green. Small sample; pre-registered 11-10 review governs.

Bot-Donchian4 (new · paused before live)

not trading

Fourth Donchian variant staged. Currently holds 20 open positions from setup runs but has traded 0 rounds and is realized-flat — the 🚫 flag on our internal fleet means it's paused pending a rethink after the Donchian-pair retirement. Open mark is −$856 — unrealized, no live-trading decision yet.

The point of the multi-day book: realised gains, not paper gains.

Each of the just-launched bots closes positions on rules — a trailing exit, a multi-day close, a scheduled EOD flat. That means profits are realised weekly and monthly, not "someday when we sell SPY". Combined with Undertow's daily event-shorts (which cross zero-hold at close), the fleet is designed to produce cash-in-hand P&L on a rolling basis. Live-fill numbers for these new bots will accumulate in the coming weeks and will be reflected on this page as they do.

Backtests — vs SPY on the terms that matter

annualised · with drawdown you'd actually sit through

Bot-Swing walk-forward — beats SPY on both return and drawdown

2023–2026 · $50k capital · live since 07-31

Bot-Swing (walk-forward)

~+33%/yr

+$49,036 total on $50k · worst DD −26%

SPY buy & hold (same window)

~+16%/yr

estimate · worst DD ~−14% (2023 selloff)

The story here. Bot-Swing produces roughly twice SPY's annualised return over the same 3-year window, with drawdown roughly 2× SPY's — but the top-5 trades contribute only 12% of total profit. This is broad edge from many small trades, not a few outliers. Kill switch armed at −$10k or SPY < 200d MA. Live paper since 2026-07-31.

Bot-Donchian "Quality" (retired 08-09)

withdrawn

Backtest P&L (initial)+$120,000

Sharpe (initial 5bps assumption)0.96

Sharpe (honest, survivorship-corrected)0.71

Live statusRetired forever 2026-08-09

Withdrawn: honest re-audit dropped Sharpe from 0.96 to 0.71. Combined with cost-model V2 showing real all-in RT above the surviving edge, the bot was retired forever on 08-09 via MOO winddown.

Bot-Donchian2 "Active" (retired 08-09)

withdrawn

Backtest P&L (initial)+$105,000

Sharpe (initial 5bps assumption)0.90

Sharpe (honest, survivorship-corrected)0.54

Live statusRetired forever 2026-08-09

Withdrawn: same story as the "Quality" sibling. Retired on the same day, same discipline.

Alpha-Engine research chain — 21 campaigns, 0 tradeable

CONTINUOUS mode · 2026-08-07

Campaigns run

Tradeable candidates

Chain status

Dormant

Next wake

~Nov 2026

The Alpha-Engine ran 17 sequential research campaigns without approval waits. Every one either failed the promotion gate or produced an edge too thin to survive our measured execution costs (~27.3 bps/leg). Chain is now dormant by design until deeper 2026 data matures around November. Most shops would suppress this result. We publish it because the search process, run honestly, is the credibility.

Correlated finding (08-10): the 21-campaign work was itself flagged as too shallow. A deeper ML pipeline is being built. Overfit verdicts now require a Kimi + Gemini review panel — no single-agent "this is overfit" verdict can kill a profitable result.

ORB-Multiday — conservative baseline (honest costs)

$50k · positive 3 of 5 years

OOS Total

+$7,174

2022

−$1.2k

2023

+$13.9k

2024

+$8.2k

2025

+$8.5k

2026 YTD

−$1.3k

Transparency note: An earlier iteration of this backtest quoted +$34.3k / Sharpe 2.57 — withdrawn after audit. That number was inflated ~4–5× by survivorship bias and 5bps-cost assumptions. The honest number above is what remains after applying measured 11.4bps entry / 10bps exit. Modest but positive — a baseline, not the star. The star is Bot-Swing.

How the pilot is structured

infrastructure & process

Broker footprint

IBKR paper gateways4 (Bot-C fleet + Donchian + ORB + Swing)

Alpaca paper accounts3 (Undertow trio)

Cross-broker isolationPer-bot dedicated account

Live-money deploymentNot yet — deliberate

Cost discipline

Measured slippage audit1,044 fills, 2026-07-27

Open-fill bots~40bps/leg (real)

MOO-fill variant−24.5bps/leg saved

Backtest cost assumptionMatch reality, not 5bps

Governance

Config-versioning ledgerbot_epochs DB table

Config changesBetween sessions only

P&L attributionPer-epoch, never blended

Kill switchesPer-bot + portfolio-level

What we retired — and why

since 2026-05-28

Bot	Fate	Reason
Bot-Donchian & Bot-Donchian2	Retired forever 2026-08-09	Honest re-audit Sharpes dropped (0.96→0.71 / 0.90→0.54); cost gap
Bot-Bounce	Retired 2026-08-03	Live-realized was soft — MOO winddown executed clean
Bot-DWide	Decommissioned FOREVER 2026-07-18	"Who buys falling knives?" — liquidated at −$2,710
Bot-C H2 / H3	Retired to replay-only 2026-07-10	Underperformed the H1/H4 variants; freed accounts
Bot-C H5 v1 / v2	Killed 2026-07-15	Bug test cleared bot; operator decision to retire
Bot-ORB2	Paused, then repurposed	Never traded; slot given to Bot-Swing 2026-07-31

Retiring underperforming strategies quickly — and publishing that we did — is the discipline. A track record without failed experiments is not a real track record.

Quanta Privé

A private quantitative trading practice. Numbers on this page will update as the pilot progresses to live capital. Discipline before edge. Edge before scale.

Nothing on this page constitutes investment advice or an offer to buy or sell any security.