Harvests the drift
After a company beats earnings, its stock tends to keep drifting for ~60 trading days. Helm enters the session after a positive surprise across a fixed 20-name large-cap universe and rides that window.
Post-earnings drift, run as a system · invite-only
Steer your capital with a system, not a guess.
Helm harvests post-earnings-announcement drift across 20 large-cap names — and shows you every gate, sizing step and broker order before anything is placed. You hold the helm; the strategy does the rowing.
20
large-cap universe
~60
session drift hold
100%
decisions shown
crisis periods
backtest — not a forecast
What Helm does
A research-grade strategy you can actually steer.
Sizes & de-risks automatically
Risk-parity sleeves and a 10% volatility-target overlay set each position — blended with a mean-reversion sleeve. No name runs away from you; exits fire on time or at a −15% hard stop.
Total transparency
See the gate-by-gate decision graph, the live drift timeline, the exact sizing chain, the current model state and the broker order plan — before anything is placed. Nothing is a black box.
How Helm compares
Where a rules-based PEAD strategy sits among well-known benchmarks, ranked by Sharpe (return per unit of risk). Periods, fees and methods differ — context, not a ranking.
| Strategy | CAGR | Sharpe | Sortino | Max DD | Period · basis |
|---|---|---|---|---|---|
| Renaissance Medallion1 | 39% | 2.0 | — | minimal | 1988–2018 net · reported |
| Helm v4.2your strategy | 11.8% | 1.36 | 1.94 | −11.0% | 2001–2026 · backtest |
| Berkshire / Buffett2 | 20% | 0.79 | — | −44% | 1976–2017 · AQR study |
| S&P 500 buy & hold4 | 10.2% | 0.61 | 0.86 | −55% | 2001–2026 · benchmark |
| 60 / 40 portfolio | 6.5% | 0.60 | 0.82 | −30% | 2001–2026 · blend |
| Nasdaq-100 (QQQ) | 10.5% | 0.55 | 0.72 | −53% | 2001–2026 · index |
| Avg equity hedge fund3 | 6–8% | 0.50 | — | — | industry estimate |
| US Aggregate bonds | 3.5% | 0.40 | 0.55 | −17% | 2001–2026 · index |
1 Medallion (Renaissance Technologies): ~66% gross / ~39% net per year, 1988–2018; reported no losing year, so realized drawdown was minimal. Closed to outside capital. 2 Berkshire Hathaway Sharpe 0.79 vs market 0.49 — Frazzini, Kabiller & Pedersen, Buffett's Alpha (1976–2017). 3 Rough industry estimate; varies widely by fund and period. 4 S&P 500 figures are Helm's own backtested benchmark over the same 2001–2026 window as Helm (a like-for-like comparison; CAGR 10.2%, Sharpe 0.61, Sortino 0.86, max DD −55.2%). Official index returns over other periods and dividend treatments differ — see S&P Dow Jones Indices.
Index, blend and bond figures are approximate over 2001–2026 and vary by data source. Other funds' figures are reported or third-party estimates over different periods, fees and methodologies — not directly comparable. Helm figures are modeled / backtested and not a forecast.
How it works
One position, four disciplined steps.
-
01
Earnings beat
A name in the universe posts a positive earnings surprise.
-
02
Enter +1 session
Go long at the next close — sized by the risk-parity & vol-target rules.
-
03
Ride the ~60-day drift
Hold through the post-earnings drift window, tracked session by session.
-
04
Exit on time / −15% stop
Close at ~60 sessions, or earlier if the −15% hard stop is hit.
Sleeves. PEAD is the executable long book that runs on your broker today. RSI2 mean-reversion and a regime filter are model-only for now — visible in research, not yet placed live. The production model is un-hedged: a parabolic short hedge was dropped under research review (its edge proved to be a look-ahead artifact).
Trade it your way
We find the trades. You choose how to place them.
Helm's edge is the decision logic — the gated, risk-sized, time-boxed post-earnings book. That signal is broker-agnostic: follow it by hand in any account, or connect a supported broker and let Helm keep your holdings reconciled to the target — with every order shown to you before it's placed.
Manual — advisory
Helm hands you the daily action list: which names, what size, when to enter and exit, and the exact order plan. You place the orders yourself in any brokerage. Nothing moves without you — and it works everywhere.
Trading 212 — auto-pie
Connect your Trading 212 Invest / ISA key and Helm reconciles your holdings to the target book each day, placing only the orders needed to close the gap. Long book only — T212's API can't short, so the hedge stays advisory.
Interactive Brokers — full book
The power tier: the full v4.2 book, global markets and professional execution over IBKR's API. Scaffolded today — the natural home for the strategy in full.
Brokers
Manual · any broker
Trading 212 · live
Interactive Brokers · coming
Alpaca · exploring
Saxo · Schwab · Tradier · researching
Your keys stay yours. Broker API keys are encrypted at rest and never echoed back; Helm previews every order before placing it, and currently trades demo / paper accounts only.
Transparency
Nothing is a black box.
Every position carries its full reasoning — the gates it passed, the model state that sized it, and where it sits in the drift calendar. This is what you see in the app, on every name.
Why this decision
AAPL OPEN LONG
earnings 2026-04-30·+3.5% surprise·day 28 / 60
Positive surprise → long
+3.5%
Liquidity ≥ $1M/day
pass
In the executable target book
yes
Drift window · 60 sessions
32d left · mid
OPEN LONG
10.0%
The same gate chain renders for every name — pass, fail or hold — so you always know why.
Model state
VOL-TARGET 100%
PEAD/RSI2 45%/55%
REGIME bull
PER NAME 4.5%
GROSS 45% · cash 55%
The exact sizing chain that turns a signal into a target weight.
Drift timeline
entry → forecast close · 60 sessions
today
AprMayJunJulAug
NKE
+1.7%11d
AAPL
+5.3%32d
AMZN
−11.1%31d
XOM
−3.7%32d
earning
losing
forecast
Built like research
A lab notebook, not a tip service.
Every rule earns its place: hypothesis → test → paired-bootstrap validation, logged with a dated verdict. The strategy's simulated track record, version lineage and experiment log all live in the app — newest first, nothing hidden.
hypothesis
test
validation
Sortino ratio
MODELED
1.9
Return per unit of downside risk on the v4.2 model, 2001–2026 backtest.
11.8%
CAGR · modeled
−11.0%
max drawdown
Modeled / backtested figures — not a forecast and not a promise of future results.
Performance
Backtest · 2001–2026
Helm v4.2 vs the S&P 500.
A $10k investment, 2001–2026, in the v4.2 model versus a buy-and-hold S&P 500 baseline — risk-adjusted, on a linear scale.
Final value
$154,539
S&P 500 · $106,367
CAGR
11.8%
S&P 500 · 10.2%
Sortino
1.94
S&P 500 · 0.86
Sharpe
1.36
S&P 500 · 0.61
Max drawdown
−11.0%
S&P 500 · −55%
Calmar
1.07
S&P 500 · 0.18
Growth of $10k + drawdown · 2001–2026 · daily data
Helm v4.2
S&P 500
Real backtest on daily data, 2001–2026 (Helm v4.2 model (un-hedged) vs SPY total return). Lower panel = drawdown; per-crisis depths annotated (Helm teal / S&P grey). Figures are modeled / backtested, not a forecast.
The science
Grounded in four decades of peer-reviewed research.
Helm's rules aren't invented — every sleeve implements a documented, replicated market anomaly. The core references behind the strategy:
Overreaction & mean-reversion
-
1985Journal of FinanceDe Bondt & Thaler — Does the stock market overreact?
Founding evidence that prices over- and under-react — the behavioral basis for drift.
1985-debondt-thaler-overreact.pdf -
1988J. Financial EconomicsPoterba & Summers — Mean reversion in stock prices
Long-horizon reversion — the basis for Helm's RSI2 mean-reversion sleeve.
1988-poterba-summers-mean-reversion.pdf
Post-earnings-announcement drift
-
1989J. Accounting ResearchBernard & Thomas — PEAD: delayed response or risk premium?
The canonical demonstration of post-earnings drift — Helm's core signal.
1989-bernard-thomas-pead-delayed-response.pdf -
1990J. Accounting & EconomicsBernard & Thomas — Prices don't fully reflect future earnings
Why the drift persists — slow incorporation of the earnings signal.
1990-bernard-thomas-prices-not-reflect-earnings.pdf -
2006J. Accounting ResearchLivnat & Mendenhall — PEAD: analyst vs time-series surprise
How to measure the surprise that triggers an entry — directly informs Helm's gate.
2006-livnat-mendenhall-pead-analyst-vs-timeseries.pdf -
2021J. Behavioral & Exp. FinanceFink — A review of the post-earnings-announcement drift
Modern survey confirming the anomaly survives across decades and markets.
2021-fink-pead-review.pdf
Momentum & modern evidence
-
1990Journal of FinanceJegadeesh — Predictable behavior of security returns
Short-horizon return predictability that underpins systematic timing.
1990-jegadeesh-predictable-behavior.pdf -
1996Journal of FinanceChan, Jegadeesh & Lakonishok — Momentum strategies
Links momentum to earnings news — the bridge between drift and price trend.
1996-chan-jegadeesh-lakonishok-momentum.pdf
References are provided for context and do not constitute an endorsement of Helm by their authors. Academic findings describe historical samples; they do not predict future results.