How it works
QuantMaven sits between your thesis and your buy button. It reads what you wrote, checks the rules you set for yourself, and gives you a 1-3 second pause before you click buy. It is not an adviser. It does not pick stocks. It is the friction that turns good intentions into actual discipline.
1. The permit cycle
Every trade goes through the same four steps. The whole flow takes 30-60 seconds, most of it spent describing your thesis.
You enter real numbers — entry, stop, shares (or strategy + strike for options). Risk is derived live from those values, not from a slider you guess. The live current price + day change show up under the asset so you can anchor entry / stop to reality. For options, you pick a strike from the real chain with mid bid/ask, IV, delta, and open interest visible. Direction is inferred (stop below entry on a stock = long; long_put = short, etc.) — no LONG/SHORT toggle to remember.
The rule engine evaluates 28 possible gates — sizing, cooldowns, market regime, news, earnings windows, IV rank, time-of-day, you name it. The fastest applicable verdict wins. The verdict view shows an aggregate Risk Score (0–100) PLUS every rule that was checked with its own state chip + confidence bar — so you see the picture, not just the headline.
Three options. Every one of them writes a trade row — including the one where you do nothing. That's deliberate: respecting a rule has to be as durable as breaking one.
When you close the position, open it in /journal (it lands on the OPEN filter by default — your unclosed trades surface first) and tap CLOSE TRADE. Enter the exit price (auto-fills from the live quote for stocks). The system computes real pnl_usd from your entry, exit, and size. Outcome (win / loss / scratch) is derived from the sign — no self-reporting. Then it asks if you followed your plan and how you felt. That data drives your discipline trend, the AI weekly note, and the P&L-by-rule-adherence chart on /insights.
2. Every page in the app
Six surfaces. Each does one thing well. None of them places trades — that's between you and your broker.
Everything starts here. Pick an asset (live price shows below it with a session badge — LIVE / AFTER HOURS / PRE-MARKET / CLOSED). Toggle stock or option. Enter the actual trade details (entry/stop/shares for stocks; strategy + chain-picked strike for options) — risk is derived live and direction is inferred. Submit, get a verdict + per-rule confidence breakdown, choose an action. Active mode chips at the top let you switch rule sets in one tap. The desktop sidebar shows your active mode, today's discipline ring, the last verdict, and the market-context panel (VIX, session state, earnings guard).
A single 0-100 number you can move. Green ≥70, amber 40-69, red <40. The 7-day sparkline shows trend. The 3-stat grid (rules followed, override rate, blocks respected) tells you what specifically you're doing well or poorly. The AI Weekly Note pulls one observation in 25 words or less.
Lands on the OPEN view by default — your unclosed positions surface first. Three filters compose: STATUS (Open / Closed / All) · VERDICT (Pass / Warn / Block) · ACTION (Placed / Passed on it / Overrode). Each card shows verdict + action chips + mode badge + realized $ for closed trades. Tap to open the detail page: original thesis, coach note, trade-shape panel (entry/stop/shares for stocks; strategy/strikes/expiry/contracts for options), and either a CLOSE TRADE button (open trades) or a Realized P&L panel (closed).
Seven intent cards: My active rules · Today's discipline · Recent trades · Override history · This week · P&L breakdown · Concepts. Tap one, get a direct answer. The P&L breakdown is the headline: realized $ split across (stock vs option) and (rules-followed vs warned-placed vs overrode). That last one is the receipt for whether your discipline pays. No chat, no LLM filler — every number is computed server-side from your actual closed trades.
A free-form chat companion that reads your history. We pulled v1 because most users don't yet have enough trades for a chat AI to coach on — it just gave generic advice. Coach returns once your history can power real specificity. Until then, Insights covers the same ground deterministically.
Per-rule toggle and threshold editor for your active mode. Per-tier severity controls. The '✦ Suggest from a setup' button opens the AI Rule Suggester (next section). Changes take effect on your next permit.
3. The rules system
Rules are the gates. There are 28 of them in the engine, ranging from "max 1% risk per trade" to "block if VIX/VIX3M backwardation is steep enough to imply a volatility regime shift." You don't need to use all of them. Pick the ones that match how you trade.
Five starter presets
Pick one in onboarding and you start with 10-13 rules already configured. Each preset is grounded in published research or named educators — not arbitrary numbers.
Three tiers of severity
Every rule is tagged with a tier. Each rule set decides what happens when a tier fires.
Modes — up to 10 per user
A funded prop trader and a long-term swing investor have different rules. You can save up to 10 modes and switch between them with one tap on the mode chip strip at the top of /permit. The currently-active mode is what gates your next permit.
4. AI Rule Suggester
On /settings/rules tap "✦ Suggest from a setup." Describe what you trade in plain English, name the ticker, pick a timeframe. The AI does three things:
- 01Classifies your description into one of 8 strategy templates (Opening Range Breakout · CAN SLIM · Trend Pullback · Connors RSI(2) · Wheel · Iron Condor · Swing · Cameron Gap-and-Go).
- 02Pulls the ticker's live personality — ATR, RSI, IV rank, VIX, earnings days_until, HV30, ADX — via the same market-context engine that powers permit verdicts.
- 03Returns 4-6 RuleCards. Each card has a suggested value, the range it considered, plain-English reasoning grounded in your ticker's actual data, a named citation (tastytrade · Connors · O'Neil · Cameron · etc.), and a confidence indicator.
Every card is editable before you accept it. Skip the ones you don't want. Accept the rest and they get added to your active mode. Every session is logged for audit — the system never invents numbers, it only narrates pre-computed values.
5. Discipline score math
One number per day, 0-100. Computed atomically on the server every time you act on a verdict. The math reads from what you did, not from P&L — a discipline-respecting trader can have a bad day on the market and still see their score go up.
New users start at 50. The 7-day average drives the trend arrow. Anything ≥70 reads green on the ring; 40-69 amber; below 40 red. The score never goes negative or above 100.
6. The override gate
When the verdict is BLOCK, you can override — but only after clearing two gates. This is deliberate friction. The 20-second timer is empirically the difference between "I'm angry and want to trade" and "I've thought about it for 20 seconds and I still want to." The phrase gate forces a typed acknowledgement that muscle memory can't bypass.
- 01A 20-second drain timer counts down on screen. The override button is disabled until it hits zero.
- 02You type the exact phrase: "I am breaking my own rule." No autofill, no paste shortcuts.
- 03Both must be satisfied for the button to enable. Then it's your call.
Every override is logged. Your /insights → Override History screen shows the last 7 days with win/loss split — so you can see whether breaking your rules actually made you money or not. For most traders it doesn't. The data is on your side.
7. What this is NOT
Worth saying clearly because the market is full of "AI trading" products that blur this line.
Trading involves substantial risk of loss. Options trading is not suitable for all investors. Past performance does not guarantee future results. Consult a registered investment adviser for advice tailored to your situation.
8. FAQ
Do I have to use all 28 rules?
No. A typical user has 8-15 rules active. Presets ship 10-13. Turn off whatever doesn't fit your strategy.
Can I have different rule sets for different strategies?
Yes — up to 10 modes per user. Switch between them with one tap on the mode chip at the top of /permit.
What if the AI gets a verdict wrong?
AI is enrichment only — it never sets the verdict. The 28 rule evaluators are pure code with 95+ unit tests; the AI just adds a coaching sentence and market context. If you think a rule is calibrated wrong, edit it on /settings/rules.
Where does the live price come from?
yahoo-finance2 via our own /api/quote/[ticker] endpoint. Free, no key required. We cache 30 seconds per ticker so a typing flurry doesn't hammer them. Extended-hours sessions (POST / PRE) are surfaced — you won't see yesterday's close at 6pm.
Where does the options chain come from?
yahoo-finance2's .options() endpoint via /api/option-chain/[ticker]. Returns bid/ask/mid/IV/OI per contract; we compute delta + theta in process via Black-Scholes (5% risk-free rate, accurate to ±0.01). Free, no key. The strike picker shows ~11 strikes around ATM with quick picks for ATM, 5% OTM, and 30Δ.
Why is stop optional?
Because some traders manage exits by time, mental stop, or P&L target — not a hard stop. Forcing one is a lie. When you skip the stop, a 'Max risk $' field appears so you can express your dollar risk directly. Either path produces a valid risk_usd for the engine.
What's the per-rule confidence breakdown?
After every permit, the verdict view shows every enabled rule with a state chip (PASS / WARN / BLOCK), the engine's message, and a confidence bar. Above the list is an aggregate Risk Score (0–100). The aggregate is by construction = 100 − the verdict's underlying severity, so the headline never disagrees with the verdict state. Display only — engine still uses 'highest severity wins' so the rules you picked remain the deterministic line in the sand.
What happens if a market data feed fails?
Market-aware rules fail-PASS on null indicators. We don't silently block your trade because a data fetch errored. The verdict view shows the indicators panel as empty when this happens.
Is my thesis text private?
Yes. Thesis text is encrypted at rest (AES-256-GCM). The AI agents see the plaintext to evaluate emotional signal, but the database stores ciphertext.
What if I forget to debrief?
Trades stay open in /journal until you debrief. The discipline score only counts closed-out trades for outcome math. You can debrief days or weeks later.
Can I export my data?
Yes — /settings → Your Data → Export. GDPR-compliant. Account deletion (with confirm phrase) wipes every row via CASCADE.
When does Coach come back?
When your average user has 20+ trades. A free-form AI that has nothing to coach on just produces filler. We'd rather ship deterministic Insights now and turn Coach back on when it can actually be useful.
Ready to put your next trade through the gate?
Sign up in 30 seconds. Pick a preset. Submit your first permit. The discipline score takes care of the rest.
Get started→