Kristjan Kullamägi — The Real Method (April 2026)

The three setups

All three share the same universe filter and risk model. The differentiator is the entry signal and timeframe.

1. Breakout (long)

Universe (daily, end-of-prior-day):

Stock had a 30–100% move over the prior 1-3 months — established leader status
Has formed a 2–8 week tight base on rising 10/20 EMA
Base range ≤ 25% (high vs low of base)
Average daily $-volume ≥ $20M, ADR ≥ 4-5%

Entry: first 5-min bar that closes above prior-day high on volume ≥ 1.3× 20-day average. Stop: session low at entry. Manage: protect breakeven after day 3, trail on 10-EMA after day 5. Risk: 0.5–1% / trade.

2. Episodic Pivot (long)

Universe (intraday, on the open):

Stock gaps up ≥ 10% at the open
First-15-min volume ≥ 20-day average daily volume (massive day-1 demand)
Above all daily MAs (50/150/200) — institutional context
Not already extended over the prior 3 months (max +40%)

Entry: first 5-min bar after the OR window that holds the gap (close > open AND close ≥ session open). Stop: low of entry bar OR session low — capped at 1.5× ATR(14). Manage: 3R take-profit OR EOD flatten in the intraday version; daily version trails on 20-EMA for weeks.

3. Parabolic Short (short)

Universe (daily, end-of-prior-day):

Stock up ≥ 50–100% in a few days/weeks (large cap) or ≥ 300–1000% (small cap)
Closed up ≥ 3 of the last 5 days
Today is day 3+ of the run — never short on day 1 or 2 (secondary push risk)
Avoid: stocks with credible fundamental catalyst (earnings beat, FDA approval, real M&A) — short the hype, not the news

Three intraday triggers (use first that fires after 09:35 ET):

ORL break-down — 5-min bar closes below the low of the 09:30–09:35 5-min range
First red 5-min — stock opens up and grinds; first 5-min that closes red is the trigger
VWAP rejection — after an initial drop, price rallies back to session VWAP, prints a red 5-min, fails to reclaim

Stop: high of day at entry. Target: 10-day SMA (where these stocks usually bounce). Cover partial every ~10% decline. R:R: 1:1 to 1:3 — accuracy over magnitude.

Hard requirements: short locate / borrow available (90% are HTB — need IBKR or Centerpoint, not Alpaca's easy-to-borrow list).

Universe — 35-name seed list

Names he or his copy-traders have publicly traded on stream/X 2023–2025. This is a seed watchlist — actual entries each day come from the scanner.

SMCI, NVDA, MSTR, MSTZ, DJT, COIN, MARA, RIOT, CLSK, AVGO, ANET, APP, PLTR, AI, SOUN, IONQ, RGTI, QBTS, TSLA, RKLB, ASTS, HIMS, CVNA, AFRM, SOFI, NET, CRWD, SMR, OKLO, VST, CEG, SEDG, ENPH, AMD, ARM, DELL

Documented 2024–2025 trades

SMCI (Feb 2024 parabolic), NVDA (Mar 2024 cooling), MSTR (Nov 2024 — entered via MSTZ for size), DJT (multiple post-IPO short cycles), ABAT (small-cap supernova short).

Polygon Stocks Starter viability

Verdict: 95-100%. Kullamägi is purely US equities. Polygon gives 1-min bars + trades/quotes for everything he trades, 15-year history for backtests, snapshot endpoint for premarket gap scanning, and full options chain for his EP-via-calls and parabolic-short-via-puts variants. The only gap is short-locate availability, which Polygon doesn't provide — IBKR's reqMarketRule handles this at runtime.

What we changed and why (April 2026 rebuild)

The first round of backtests on the synthesized multi-trader bot lost money on every Kullamägi module. After deep-reading the synthesis corpus from research/qullamaggie-corpus/ and re-auditing every line of the implementation, the dominant cause was the universe, not the rules. We rebuilt around six changes:

Dynamic point-in-time leader universe. Strategies no longer hardcode 25-40 mega-caps. Each Monday the runner ranks all ~5,300 US common stocks by simultaneous 1m/3m/6m ROC, applies ADR ≥ 4, $5M dollar volume, and above 50-SMA filters, excludes biotech (the corpus's "halts bypass stops" rule), and passes the top 200-300 to the strategy as the week's scan set. This is the single biggest fix.
Pure-trail exits by default. Three independent corpus studies (HackerTrader's 64% CAGR backtest, Whos_Agent's E0 study, mxteo's analysis) found the published "sell 1/3 at day 3-5" rule HURTS returns versus pure trail. Calendar partials are now opt-in per-strategy yaml.
Friday no-entry, earnings blackout, theme cap. Three discipline rules applied across all swings: no new entries Fridays (Whos_Agent moderator validation); skip if any earnings within 5 trading days (except for EP, which IS the earnings trade); cap concurrent positions per theme bucket at 2-3 (avoid the "all biotech blew up" pattern).
Strict bull regime gate. Replaced the old "10/20-SMA cross + both rising" with HackerTrader's mechanical "SPY > 50-SMA AND 50-SMA rising" — matches his 64% CAGR study and is more permissive on small pullbacks.
Cap-aware parabolic short. The yaml carries both thresholds (large-cap 50%, small-cap 300%) and the strategy picks based on market cap from the cached ticker payload. Earlier version applied 50% to everyone, missing legitimate small-cap supernovas and false-firing on shallow large-cap rallies.
Borrow-cost accounting on shorts. ~3.6%/yr daily borrow fee subtracted from realized P&L on the parabolic short — closes the live-vs-backtest gap that made earlier short backtests look better than they would in production.

Result on a 3-year window with the rebuilt system + Breitstein's prior-bar ratchet trail wired into the swing runner:

Parabolic short swing — PF 1.19 IS / 1.48 OOS, 50.3% win rate, 157 trades. The clear winner. OOS BETTER than IS (no overfit signature).
Episodic Pivot swing — PF 0.77 IS / 1.59 OOS, 54% win rate, 22 trades. Promising but sample small (EP is intrinsically rare — 3-4 weeks per quarter).
Breakout swing — PF 0.54 IS / 1.56 OOS, 41% win rate, 58 trades. The Breitstein trail lifted win rate from 21% to 41%.
Hougaard ORB intraday — PF 0.61 IS / 0.73 OOS, 1305 trades. Still loses. US-ETF index ORBs are structurally harder than DAX/Dow futures where his rules originate.

See the trades browser for every single signal with full P&L and exit detail.

Kristjan Kullamägi — the actual method.