61 TESTED · 15 GO · 22 CAUTION · 24 LOTTO

Trade like a quant.

We put 61 options and stock strategies through twelve years of real market data, charged them realistic trading costs, and graded every one on what it returned. All of it is published, including the strategies that lost money.

12 yrs
of option chains
197M
option-day bars
547
tickers scanned
268
written findings
Strategy verdict registryas of 2026-08-11
15 GO22 CAUTION24 LOTTO
GO
Covered straddle (full)
n=3,207,534
≈42%
LOTTO
Iron condor
−13.8%/trade
GO
Wheel (cash-secured puts)
n=1,926,136
13-30%
LOTTO
Pre-earnings iron butterfly
−86%/event
GO
Covered call
n=1,712,985
≈7-21%
LOTTO
Long butterfly
−95%/trade
12 yrs
of option chains
2014-06-02 → 2026-06-10
1.37 TB
historical archive
full chains, trades, IV, greeks
197M
option-day bars
the rows every backtest ran over
12.2M
distinct contracts
every strike and expiry, not a sample
547
tickers scanned
the live research universe
34.8M
live snapshots
22.4 GB across 86 tables

The historical archive and the live snapshot store are separate systems. Every verdict below was computed on the archive — the live data doesn't get a vote on what worked. Row counts measured 2026-08-09.

THE STRATEGY VERDICT REGISTRY · AS OF 2026-08-11

Every strategy we could test, graded the same way.

Each one ran over twelve years of real option chains, paid realistic costs to get in and out, and got graded once. There are no second runs for strategies that came back ugly.

152224
GO
CAUTION
LOTTO

24 of 61 · more than half of everything we tested lost money once real costs were counted

GO
15

Made money after realistic trading costs, with the worst case capped. This is the short list we'd run ourselves.

CAUTION
22

Works, but only under specific conditions. Read the caveat before putting size on it.

LOTTO
24

Lost money once real costs were counted, or we couldn't show it works. We publish these anyway, and there are a lot of them.

The full registry lives in the app: every strategy, its numbers, and the research behind each verdict.

268 WRITTEN FINDINGS · 717 BRIEFS · 2026-05-15 2026-08-10

Every verdict has the research behind it — including the ones that killed our own ideas.

Each one is dated and written out: what we tested, what came back, and what would change our mind. Plenty of them overturn something we'd said the week before. Here are the eight most recent.

2026-08-09Selling calls further from the money helps a little, but not enough to rescue the strategyREFUTES
2026-08-09An earnings pattern held up, but not for the reason we thought it didCORRECTS
2026-08-08Our best index strategy failed its own test, so we downgraded it the same dayDOWNGRADE
2026-08-08An entry-timing signal vanished once we included companies that went bustREFUTES
2026-08-07A screen we built to find bargains turned out to just be finding penny stocksREFUTES
2026-08-06Our stock-picking edge disappeared under stricter statisticsREVERSES
2026-08-05Our own past results were flattered by leaving out companies that went underCORRECTS
2026-08-05Being choosier didn't rescue it, and we had to correct one of our own earlier claimsCORRECTS

8 of 268. The rest ship inside the app.

HOW WE DECIDE

We write the pass mark before we run the test.

Run a backtest a hundred different ways and one version will always look good. So the rule for what counts as a pass gets written down and committed before the test runs, and the verdict is whatever the rule says. We don't get to change our minds after seeing the answer.

In August, one of our own top-rated strategies went through this. It passed three of the four checks, failed the last one, and got downgraded the same day.

Did it make money overall?Yes
Does it survive dropping any single year?Yes
Does it hold up across separate time periods?Yes
Is the edge bigger than the noise?No

We also push a strategy we know is bad through the same four checks. If it doesn't fail, the test isn't measuring anything.

DEEP RESEARCH · ON DEMAND

A full research note on any name, in about two minutes.

Point it at a ticker and five specialists work through it in order, each starting from what the one before found. What comes back reads like a sell-side note: a rating, a price target, and both sides of the argument.

Reports get saved with their date, so you can look back at what was said before a move instead of after it.

01
Market analyst

Price action, trend, support and resistance, where it sits in its own range

02
News and catalysts

What's been reported, what's scheduled, and what could move it next

03
Options and sentiment

What the options market is pricing in, and how positioning is leaning

04
Fundamentals

Earnings, margins, balance sheet, valuation against its own history and its peers

05
Portfolio manager

Has to reconcile all four into one call — including a pre-mortem on what would make it wrong

And then we grade it.

Every report is stamped with the price when it was written and the target it named. At one week, one month, three months and six months it gets scored against what the stock actually did — and against simply owning the index instead. A rating nobody marks afterwards is just an opinion.

NIGHTLY ANALYST · 20:00 EVERY DAY

Every name you follow gets re-read overnight.

Each evening the whole universe runs through the same three-step argument, including the quiet names and last week's winners.

01

The bull case

One analyst builds the strongest argument to buy it

02

The bear case

Another builds the strongest argument against — independently

03

The judge

A third reads both and has to commit to a call, with its reasoning written down

The split is deliberate. Ask a single model for its view and it tends to talk itself into whatever it said first. Two models arguing opposite sides, with a third judging, turns out to be much better at killing weak cases.

547
names in the universe
20:00
runs every evening
333
covered on the last run
28k
reads logged to date
AI PORTFOLIO TOURNAMENT · LIVE SINCE 2026-06-16

36 AI portfolio managers, a million paper dollars each, all scored against SPY.

One class gets an identical prompt and governor, so the model is the only variable: Haiku against Sonnet against Opus against GPT-5, Gemini, Grok, DeepSeek, Qwen and GPT-OSS. The other class gets mandates instead — covered premium, wheel, momentum, value, concentrated, generalist. Every fill is charged five basis points against the seat, on both sides.

36
model seats
$1M
paper capital each
6,978
executed trades
1,968
logged decisions
5 bps
friction per side
SPY
benchmark
Ranked over 20 common marks from 2026-07-1415 of 32 beat SPY
#
Seat
Excess
1
Opus · Covered-Premium
claude-opus-4-8
+6.96%
2
Grok 3 (xAI)
grok/grok-3
+5.52%
3
Qwen 2.5 14B · Covered-Premium (local, free)
ollama/qwen2.5:14b
+4.78%
THE RESULT SO FAR

The mandate beat the model.

Split the ranked seats by what they were told to do: the covered-premium and wheel mandates, the same families our backtests grade GO, pull away from everything else.

MandateBeat SPYMedian
Covered / wheel7/9+3.21%
Everything else8/23-0.89%

Nine covered seats over nineteen marks is a small sample, so treat this as an early read rather than a conclusion.

Marked to Aug 10, 2026 close, on paper capital.

Most of these seats may well end up losing to SPY. Finding that out is the point, and they sit on the same scoreboard we hold our own strategies to.

OPTIONS FLOW

See the big trades as they print, and which side was in a hurry.

Every large options trade, sized and sorted, with the side that initiated it worked out. That last part matters most: a headline saying “$103M of bullish calls” means the opposite if those calls were being sold rather than bought.

Sweeps and blocks

Orders split across exchanges to fill fast, and single large prints

Who initiated

Whether the trade hit the ask or the bid — buyer urgency or seller urgency

Premium spent

Dollars committed, not just contract counts

Repeat interest

When the same strike and expiry keeps getting hit across days

Read by a model

A written take on what the day's flow in a name actually adds up to

Flow tells you where money moved. Whether that structure has ever been worth trading is a question for the verdict board.

OPTIONS LAB

Build the trade, then check it against twelve years of history.

Six tools for deciding whether a specific trade is worth doing before you place it, on live quotes instead of a textbook example.

Build a Spread

Put together any combination of calls and puts and see the profit and loss at expiry, where you break even, the odds of finishing profitable, and how the position behaves as the stock moves.

Profit Band

For covered strategies, the exact price range where you still make money at expiry, and where you stop.

Premium Richness

Whether options on a name are expensive relative to how much the stock has been moving. Rich premium is the one condition our research found reliably matters.

Options Activity

Screen the whole universe by volume, open interest and how expensive options are versus their own history.

Open Interest

The largest established positions across the universe, ranked by dollars at risk rather than contract count.

Live Activity

A written synthesis of the day's flow as it develops.

DEALER POSITIONING

The one part of the market that isn't an opinion.

When you buy an option, a market maker sold it to you, and now they have to hedge it by trading the stock. They don't get a choice about that, which makes it one of the few predictable flows in the market. These boards show where the hedging is concentrated and which way it pushes.

Gamma exposure

How much stock dealers have to buy or sell for each move in price, strike by strike. Where it piles up, moves get damped; where it flips negative, moves get amplified.

The flip level

The price where dealer hedging stops cushioning the market and starts accelerating it. Worth knowing before you size a directional trade.

Max pain

The strike where the most options expire worthless — and the level price often drifts toward into expiry.

Short-dated tide

Same-day and weekly options now drive a large share of daily volume. This tracks how much of it there is and which direction it leans.

Underneath sits the slower positioning data — futures positioning, institutional holdings, short interest, retail sentiment surveys and congressional filings.

YOUR BOOK, AUDITED NIGHTLY

Connect your broker and the same standard applies to your own positions.

SnapTrade and IBKR sync in your real positions and fills. From there it's desk work, and the book gets re-read in writing every day.

Broker syncSnapTrade + IBKR — positions and fills, not screenshots
Portfolio greeksdelta, gamma, vega, theta netted across the book
Risk dashboardconcentration, tail exposure, margin equivalence
Pin riskthe strikes your expiry is drifting into
Auto-rollroll candidates surfaced while there's still time to act
P&L attributionwhat moved it — spot, vol, time, or your own fills
Confluence scanwhen several independent boards point at one name
Position watchper-position triggers, not a generic price alert
Daily portfolio auditthe whole book re-read every day, in writing
BOOK GREEKSEXAMPLE
Delta
+1,284
net long the underlying
Gamma
−46.2
short gamma from written options
Vega
−2,410
gains if volatility falls
Theta
+512
collected per day

An illustrative covered-premium book: long the shares, short the options, collecting decay. Connect a broker and these fill in from your own positions.

NEWS + SOCIAL ALERTS

A hundred alerts a day land on the rail. Your phone hears about almost none of them.

News, filings, social and flow all come through one rail, and most of it doesn't deserve an interruption. A push has to clear every bar below; the rest waits until you go looking.

THE RAIL · EVERYTHING LANDS HERE
Market news
Insider Form 4
Congressional trades
Truth Social
X accounts
Unusual flow
IV spikes
Macro prints
…100+ a day
THE GATE
severity≥ HIGH
flow premium≥ $10,000,000
push ceiling6 / hour

All three must pass. Everything else stays on the rail.

WHAT REACHES YOUR PHONE
High severity only

A macro print that moves the tape. Eight figures of one-sided flow. An IV spike on a name you hold.

PushoverntfyWeb push (PWA)

If your phone buzzes, something cleared a high bar. The rest sits on the rail until you have a reason to look.

HOW THIS GETS BUILT

Built in the open, with the receipts committed.

Themis started on 2026-04-14 as an options scanner. The first written finding landed a month later, and they haven't stopped since.

There's no marketing team here; the page is written from the research files themselves. That's why the findings section leads with the studies that overturned our own conclusions.

2026-04-14
Repo opened
2026-05-15
First written finding
2026-08-10
Most recent finding
268
Findings to date
30 SCHEDULED JOBS · UNATTENDED
Intradayoptions flow, tide, news alerts, catalyst sweep
16:35–17:30snapshots, covered cache, IV rank, open interest, portfolio audit
20:00nightly multi-model analyst across the universe
06:15–06:45WAL checkpoint, earnings, sentiment, AI portfolio rebalance
WeeklyCOT, positioning, peer baselines, insider Form 4

Figures on this page were exported from the research machine on Aug 11, 2026. They refresh on a schedule, not on request — which is why nothing here can spin or fail to load.

PRICING

Three plans, all built on the same evidence.

Start free for 7 days and cancel any time. Every tier sees the same verdict registry; the higher ones just point more of the machine at your own book.

Beginner

Beginner.

$9.99/mo

The daily picks, in plain English, with the verdict and the backtested receipts attached to every one.

  • Daily picks · verdict, win rate, sample size, worst case
  • Plain-English walkthroughs of every structure
  • Cash-secured puts, covered calls, wheel, spreads
  • The full verdict registry — including what we grade LOTTO
  • Paper-trade journal
  • Morning brief
Start 7-day trial →
Most chosen
Pro

Pro.

$29.99/mo

The whole cockpit. Options flow, dealer gamma, the nightly AI analyst, and your own book audited daily.

  • Everything in Beginner
  • Options flow, open interest, dealer gamma and max pain
  • Nightly AI analyst · bull, bear and judge on every name
  • Deep Research on demand · 5 sequential analysts per ticker
  • Broker sync · SnapTrade and IBKR, with daily book audit
  • Multi-leg builder · verticals, calendars, condors, jade lizards
  • News, social and flow alerts with push
Start 7-day trial →
Quant

Quant.

$99.99/mo

The evidence layer underneath. Every finding, every ledger, and the pre-registration protocol behind each verdict.

  • Everything in Pro
  • The full written findings corpus, dated and searchable
  • Pre-registered robustness batteries · bootstrap CI, leave-one-year-out, walk-forward folds
  • Per-strategy ledgers with the return basis on every figure
  • The 36-seat AI portfolio tournament, seat by seat
  • Regime cuts · VIX tiers, IV/RV, term structure
  • Direct line to the person who built it
Start 7-day trial →

Start with the evidence.

Look up the strategy you're already running and see how it held up over twelve years of real data. Every verdict comes with the research behind it, failures included.