LIVE FEED·EDITION 2026.05·14 CHAPTERS·8 INTERACTIVE
PAPER MODE·NO CARD
LEARN · 8 MIN · ALL LIVEEVERY CHART RESPONDS TO YOU↳ DRAG · TUNE · RUN
Trading,visualized.
Eight live demos. Fourteen chapters. Zero textbooks.
Most quant tutorials are walls of text with one screenshot. Ours is the opposite: every concept on this page comes with a chart you can drag, a slider that morphs the market, or a bot you can run in real time. Read it like a book — but every page argues with you.
Every market — every stock, every minute, every quarter — sits somewhere on this spectrum. The single most important question in quant trading is which regime are you in right now, because the answer tells you which bot family will pay off and which will burn you. Drag the knob below; watch the chart morph. The Hurst exponent — the number controlling the drag — is what our regime detector reads off real markets.
how it works →we synthesize a price path with controllable autocorrelation. Hurst > 0.5 means today's up-move makes tomorrow's up-move more likely (trending). Hurst < 0.5 means it makes tomorrow's down-move more likely (mean-reverting). The Lazybull Hurst Exponent bot reads this number off real markets.
TAKEAWAY
"The bot that wins in a trend dies in a chop. Detect first, act second."
↳Three primitives. Every product surface is a remix.
¶ 02.A
Once you see them, the whole product clicks. The dataset is the chart you're testing on. The bot is a trading rule that eats a chart and spits out a verdict. The workspace lets you stack as many bots as you want — they all see the same chart, and the output panel tallies who agreed.
01
Dataset
The chart you're testing on.
Real OHLCV from Yahoo Finance for AAPL, NVDA, BTC, NIFTY — anything Yahoo lists. Synthetic deterministic walks for symbols Yahoo doesn't have. Same input → same answer, every time.
02
Bot
A trading rule. Eats a chart, spits out a verdict.
Could be 70-year-old textbook math (Black-Scholes, RSI, SMA crossover) or a 2024 neural net trained on real markets. Either way the output is the same shape: BUY / SELL / HOLD / WARN with a confidence score.
03
Workspace
Stack as many bots as you want.
They all see the same chart. The Output panel tallies who agreed. When 5 trend bots and 3 AI bots all flash BUY at the same time, that's a signal worth noticing.
§02→03
REAL DATA · REAL MATH
§03
CHAPTER 03LIVE DEMO
One bot.One chart.Nothing fake.
↳SMA crossover. Real AMZN bars. Drag the periods.
¶ 03.A
Below is the textbook moving-average-crossover bot, running on real AMZN daily bars fetched from Yahoo on page load. Drag the period sliders — the math recomputes instantly, in your browser, with no round-trip. Every signal you see was just computed by your CPU.
↳Sharpe matters. Drawdown matters. Pure return doesn't.
¶ 04.A
Pick a bot. Pick a market scenario. Hit run. The equity curve builds bar by bar in real time, with Sharpe, max drawdown, and win rate filling in as the simulation progresses. That's when it clicks: high return ≠ good strategy. You want Sharpe — return per unit of bumpy ride — not just upside.
TERMINAL §04·INTERACTIVE
6 BOTS × 3 SCENARIOS
pick a bot
market scenario
SMA Crossover on Strong uptrendbar 2 / 160
Total return
+0.0%
Sharpe
0.00
Max DD
0.0%
Trades
0
Win rate
0%
read this →a high return doesn't mean a good bot. Look at Sharpe(return per unit of bumpy ride — >1 is good) and Max DD(the worst peak-to-trough drop the bot put you through). The same bot wins big in one regime and loses in another. That's why stacking multiple bots matters.
TAKEAWAY
"Sharpe > 1separates real edge from lucky upside. Most retail strategies don't clear it."
§04→05
STACK & VOTE
§05
CHAPTER 05WHY STACK BOTS
One bot isa guess.Six agreeingis a signal.
↳Toggle bots. Watch tier flip. Agreement is the alpha.
¶ 05.A
Watch the conviction band slide as more models fall into line. Tweak the dataset — drift, vol, seed — and see how robust the agreement is. That's how a workbench separates real edges from chart-pattern wishful thinking. The historical accuracy band of ULTRA tier consensus is 65–77% on embargoed walk-forward CV.
TERMINAL §05·INTERACTIVE
12 BOTS · 3 SCENARIOS
scenario
AMZN +14% in 30 days, low realized vol, broad market in risk-on. All trend bots fire long; mean-reversion bots quietly disagree.
Delta, Gamma, Theta, Vega, Rho. Drag the strike below — every Greek updates simultaneously. Hover any of them for a one-line plain-English explanation. By the time you've dragged the strike across the smile, you'll have the intuitions textbooks take chapters to build.
Δ
Delta
Γ
Gamma
Θ
Theta
ν
Vega
ρ
Rho
TERMINAL §06·INTERACTIVE
DRAG STRIKE · HOVER GREEK
underlying spot$100·expiry30d·iv30%
strike (drag)$100
$70 deep ITM put / OTM call$100 ATM$130 OTM put / ITM call
Δ Delta0.534
How much the option price moves when the stock moves $1.
Γ Gamma0.0462
How fast Delta changes as the stock moves.
Θ Theta-0.063
How much value the option loses each day, all else equal.
ν Vega0.114
How much the option price moves per 1% change in IV.
ρ Rho0.041
How much the option price moves per 1% change in interest rates.
teacher
Hover any Greek
Each chart shows how that Greek changes across strikes. The vertical line is your selected strike. Drag the strike slider above and watch all five update.
Price
$3.61
Δ Delta
0.534
Γ Gamma
0.0462
Θ Theta /day
-0.063
ν Vega /1%
0.114
ρ Rho /1%
0.041
§06→07
THE SMILE
§07
CHAPTER 07THE VOL SMILE
A 15%-OTM putcosts more thana 15%-OTM call.
↳Crashes happen faster than rallies. The market knows.
¶ 07.A
Black-Scholes assumes one flat volatility. Real markets don't. Out-of-the-money puts trade at higher implied vol than equidistant calls because the market knows crashes happen faster than rallies. Drag the skew and kurtosis sliders to see exactly how far reality drifts from the textbook.
TERMINAL §07·INTERACTIVE
SKEW · KURTOSIS
implied vol surface · spot $100skew 0.60 · kurt 0.50
skew (ρ)
0.60
− call-richflat+ put-rich
In real equity markets, ρ is almost always positive — when the stock drops, vol jumps. So OTM puts get bid up, OTM calls get bid down. The Heston model parameter has the opposite sign convention; same idea.
smile depth (kurtosis)
0.50
flatdeep smile
The "wings" — far OTM strikes — sell for richer IV than the textbook flat-vol BS model predicts. The market is pricing in fat tails.
comparison · same distance from spot
$85 put (15% OTM)
41.1%
implied vol the market is asking
ATM
30.0%
reference
$115 call (15% OTM)
22.6%
implied vol the market is asking
read this →61.6% richer IV on the put side. A flat-IV Black-Scholes pricer mispriced both contracts. That's why our Heston SV bot exists — it captures this skew.
TAKEAWAY
"Black-Scholes is a beautiful model. Heston is a useful one."
"Will the price land in this band by expiry?" Black-Scholes, Monte Carlo, and an empirical fat-tailed model give you three different probabilities. Drag the band into the wings and watch them disagree. That gap is model risk — a real cost most retail platforms hide from you.
TERMINAL §08·INTERACTIVE
DRAG THE BAND
price distribution at expiry · spot $100iv 30% · 35d
Black-Scholesclosed-form analytical
54.5%
Assumes returns are perfectly lognormal. Fast, deterministic, ignores fat tails.
Monte Carlo4,000 simulated paths
52.8%
Same lognormal assumption as BS but via simulation. Will eventually converge to BS at infinite paths.
Empiricalfat-tailed mixture
57.3%
85% Gaussian + 15% Cauchy — simulates real-market kurtosis. Disagrees with BS at the wings.
read this → the three numbers should be close near the center, but as you push the band into the wings (deep OTM), empirical tends to give a higher probability than BS because real markets have fatter tails than lognormal admits. Stack probAll() in the Wedge tools to see all three at once on real positions.
Every bot has its own page with the math, when it shines, when it fails, and a live demo running on real data. Click any to dive in. The hand-written specialty essays are the most honest bot documentation in retail finance — we tell you when each one breaks.
The dataset card on every quant page lets you stress-test bots against any market. Drag each slider — the chart redraws deterministically. Same seed, same chart, every time. That's reproducibility nobody else in retail finance offers.
TERMINAL §10·INTERACTIVE
CLICK A KNOB → LEARN IT
dataset · synthetic mode
Symbol
Bars
Seed
Drift μ
Vol σ
click any knob to learn what it does →
AMZN · 120 bars · seed 11+13.42%
spot $256.32deterministic — same seed gives same chart, every time
§10→11
ELI12
§11
CHAPTER 11TEACHER MODE
Toggle on.The mathreads like a story.
↳Every bot ships hand-written explainers at age-12 level.
¶ 11.A
Every bot ships with a "Teacher" callout that explains its verdict at age-12 level. Hover any Greek on the trade page and you get a one-line explainer. The AI Teacher endpoint uses GPT-4o-mini when OPENAI_API_KEY is set; falls back to a hand-written mock if not.
◯ TEACHER · OFF
RSI 14 = 28.4. STATE: OVERSOLD.
3 reversion triggers in window.
Backtest return +4.2%.
BUY · 78% CONF
● TEACHER · ON
RSI is a 0-to-100 thermometer. Below 30 means everyone panicked and the price is probably going to bounce. The bot just spotted three of those bounces in your window — that's why it leans BUY with high confidence.
If you can writea JS function,you can write a bot.
↳Paste a function. Hot-load it. Backtest it.
¶ 12.A
Click + Import your bot in the bot library. Paste a function that takes candles + params and returns { verdict, summary, metrics }. It hot-loads into the workspace.
↳12 AI bots. Real Python NN. Honest fallback chip.
¶ 13.A
The 12 AI bots delegate to a Python service in ai quants/serve.py that runs trained neural networks. When the service is up, you see Source: Python NN on every card. When it's not, the bot falls back to a deterministic TS surrogate marked clearly Source: Mock.
0
trained ml models
0
bar lookback (transformer)
0.0
bs surrogate err %
0
ultra-tier accuracy %
01
Browser
React workbench
You click ▶ Run All in /quant. The bot's run() function builds a request body.
02
callApi()
POST /api/<endpoint>
aiBot() wrapper hits the FastAPI service (NEXT_PUBLIC_QUANTAI_URL) with an 8s timeout.
03
Python NN
13 trained models
serve.py loads the right surrogate, runs predict(), returns JSON. The card flips green.
Set in Fraunces & JetBrains Mono. Charts hand-rolled in SVG; every line you see drew itself in. The Hurst slider, vol smile, Greek surface, and probability comparison are all real Black-Scholes math — not a screenshot, not a mock. Built on a laptop in 2026.