2. The Trader's Toolkit: Candles, Timeframes & Indicators

Every trading platform speaks the same visual language: candlesticks, moving averages, volume, timeframes. A market maker must be bilingual — fluent in the chart, and fluent in the order flow hiding underneath it.

Part I · Chapter 2

Japanese candlesticks: four numbers, one story

Charts don't plot every trade — they compress them. A candlestick summarizes all trading inside a time window with four numbers, OHLC:

Open — the first trade price of the window;
High / Low — the extremes reached (drawn as the thin wicks or shadows);
Close — the last trade price. The thick body spans open→close: green/bullish when close > open, red/bearish when close < open.

diagram — anatomy of a candle & the three famous shapes

Left: the same window of trading drawn twice — buyers won (green: close above open) and sellers won (red: close below open). The wicks mark the extremes the price visited and abandoned. Right: the three shapes every platform trader names on sight — read them as flow summaries, not prophecies: a doji is a fought draw, a hammer is a rejected sell-off, an engulfing candle is one side taking control.

Parable · The rice merchant of Sakata In 1750s Osaka, rice trader Munehisa Homma grew rich at the Dōjima Exchange — the world's first organized futures market — and recorded prices as opens, highs, lows and closes, reading the "battle between buyers and sellers" in each day's shape. His notation became the candlestick chart, popularized in the West only in the 1990s by Steve Nison. Two and a half centuries later, every platform from Binance to Kalshi still renders markets in the script of an Edo-period rice merchant.

Traders name recurring candle shapes — a doji (open ≈ close: a tug-of-war that ended in a draw), a hammer (long lower wick: sellers pushed down, buyers slammed it back), an engulfing candle (a body that swallows the previous one: one side just took control). Read these as descriptions of order flow, not prophecies — academic evidence for candlestick patterns as standalone predictors is weak. Their real value is vocabulary: each shape summarizes who was hitting bids and lifting asks inside the window.

The market maker's reading of a candleA candle is the shadow that thousands of individual fills cast on the chart — and the market maker was the counterparty inside it. The wicks are where takers paid the most for immediacy (and where stale quotes got picked off); the body is the net drift the MM's inventory had to survive; the volume bar is how many times the spread was paid. Chartists read candles to predict; market makers read them to remember.

Timeframes: one market, many movies

The same tick stream renders into 1-minute, 1-hour, or daily candles — and the three movies can disagree: a brutal downtrend on the 5-minute chart can be an invisible blip on the daily. Platform traders practice multi-timeframe analysis: establish regime and trend on a higher timeframe, time entries on a lower one. Two facts matter mathematically:

Volatility scales with the square root of time. Under diffusion, the standard deviation over a window of length τ grows like σ·√τ — a 1% hourly vol is roughly a 4.9% daily vol (√24 ≈ 4.9). This is the same σ that will price your spread in Chapter 6; choosing the estimation window is choosing a timeframe.
Aggregation hides microstructure. A candle cannot show queue position, book imbalance, or who was maker and who was taker. Everything in Chapter 7 lives below the lowest timeframe your platform will draw.

The market maker's "timeframe" is the tick — but regime detection (is this a calm range or a trending, toxic hour?) is genuinely a higher-timeframe question, and that is exactly where the platform toolkit re-enters the MM's engine (layer L3 of Proposal F, Chapter 18).

Moving averages: the EMA you'll never stop using

A simple moving average (SMA) is the mean of the last N closes. An exponential moving average (EMA) weights recent prices more, via one cheap recursion:

EMA recursion (α = 2/(N+1)) EMA_t = α · P_t + (1 − α) · EMA_t−1 What the symbols mean EMA_t — the average's new value, right now · P_t — the latest price (the newest close) · EMA_t−1 — the average's previous value, one candle ago · α (alpha) — the weight given to the newest price, a number between 0 and 1; it comes from the window length N (e.g. N = 19 → α = 0.1, so each new price contributes 10% and the old average keeps 90%). In words: new average = a little bit of the newest price + mostly the old average.

Directional traders use crossovers — fast EMA crossing above a slow one (the 50/200-day version is the famous golden cross) — as trend signals. The market maker uses the same recursion for something deeper: it is an online filter that needs no history buffer, just one stored value. Inside a production quoting engine, EWMA (the same formula) estimates realized volatility from squared returns (the RiskMetrics tradition), smooths the fair-value anchor, tracks average fill rates and funding. When Chapter 6's formulas ask for σ, an EWMA of squared mid-price returns is usually what supplies it.

The difference between SMA and EMA is invisible in a formula and obvious in a picture. Hit the price with a sudden step and watch who reacts first:

interactive — SMA vs EMA: the lag laboratory

Window N: 14

SMA catches up after—

EMA catches up after—

Reading—

White: price. Blue: SMA(N) — every one of the last N bars votes equally, so it drags a rectangle of history behind it. Amber: EMA(N) — recent bars vote more (the inset shows each method's weights), so it leans toward the step immediately while the SMA waits for old prices to fall out of its window. The price of that speed: on the random walk, the EMA also chases more noise. Lag versus noise — the same contract every indicator signs, drawn for the two most-used averages on Earth.

Volume, VWAP, and the flow you want

Volume confirms what price claims: a move on heavy volume reflects broad repricing; the same move on thin volume is often just a hole in the book. The VWAP (volume-weighted average price), Σ(p·v)/Σv over the session, is the institutional benchmark: execution desks slice large parent orders into child orders across the day (VWAP/TWAP algos) precisely to avoid moving the market. Hear that as a market maker: schedule-driven institutional flow is uninformed at the tick scale — it trades because the clock says so, not because of what happens in the next minute. It is among the most benign flow you will ever be filled by, the polar opposite of Chapter 5's informed traders.

Support, resistance, and bands — microstructure in disguise

Support/resistance levels are real, but not mystical: they are prices where resting limit orders cluster (visible in the book as depth walls) and where stop orders accumulate just beyond. Breaching a level detonates those stops — stops are market orders waiting to be born — which is why "breakouts" accelerate and why an MM's jump detector should treat round numbers and prior highs with suspicion.
Bollinger Bands draw mean ± k·σ around an SMA. Look familiar? It is structurally the same object as the Avellaneda–Stoikov quote pair around a reservation price (Chapter 6) — a volatility-scaled envelope. The chartist trades the band; the maker is the band.
RSI and other oscillators measure short-horizon momentum vs. mean-reversion. The regime they sniff is existential for an MM: mean-reverting chop is the MM's paradise (both quotes fill repeatedly, inventory self-corrects), while trend is the MM's tax (one side fills relentlessly and inventory piles up against the move — the exact failure of naive quoting and grid bots, Chapter 15).

animation — why a "resistance level" breaks violently

The horizontal red band is a wall of resting sell orders at a round number — that is the "resistance" the chart trader draws. Each bounce eats part of the wall (watch it thin). Just above it, gray dots: dormant stop-loss buy orders from shorts. When the third push finally chews through the wall, the stops detonate in sequence — each one a fresh market buy — and the price gaps. Nothing mystical happened: liquidity ran out, then forced orders piled on. This is why an MM's jump detector treats prior highs and round numbers with suspicion.

Honest framingEvery indicator is a transformation of past price and volume — it contains no information the tape didn't already have, and most have near-zero standalone predictive power after costs. Use them as state descriptors (what regime am I in?) rather than signals (what happens next?). Anything you do trade on, backtest first — Chapter 18's proposals all ship with a testing plan for exactly this reason.

Chapter 1 introduced limit and market orders. Real platforms offer a longer menu, and three of these are load-bearing for everything in Part V:

Order type	What it does	Why an MM cares
Stop (stop-market)	Dormant until price touches the trigger, then fires a market order	Stop clusters are jump fuel: cascades of triggered stops cause the gaps your quotes must survive
Stop-limit	Trigger fires a limit order instead	Safer for users, but can fail to fill in a crash — liquidity still vanishes
Post-only	Rejected (or repriced) if it would cross the spread and execute as taker	The maker-only guarantee. On Kalshi or Hyperliquid, this flag is how "never pay taker fees" is enforced in code, not hope
IOC / FOK	Immediate-or-cancel (partial fills OK) / fill-or-kill (all or nothing)	The hedge leg's tools: take what's there now, never rest exposed (Proposal E)
GTC / GTD	Good-till-cancelled / till-date resting orders	The default lifetime of your quotes — paired with a dead-man's switch (Chapter 16)
Iceberg / reserve	Shows only a slice of the true size	Hides your depth from competitors; also hides theirs from your imbalance signals (Chapter 7)
OCO	One-cancels-the-other bracket (e.g., take-profit + stop-loss)	Retail risk plumbing; for an MM, brackets live in the engine, not the exchange
Amend / modify	Reprice or resize a resting order in place	Where supported with queue retention (Kalshi, Chapter 9), amend-don't-cancel preserves queue priority — an asset worth real money (Chapter 7)

Now watch all of it assemble. The figure below runs a single simulated tick stream — the raw trades a market maker actually experiences — and builds the chart a platform would show you: candles at your chosen timeframe, fast and slow EMAs, volume. Switch timeframes and watch the same history tell different stories; tighten the EMAs and watch them hug price and whipsaw.

interactive — from ticks to candles, EMAs & volume

Timeframe (ticks per candle)

Fast EMA: 9

Slow EMA: 21

Last candle O/H/L/C—

Fast EMA—

Slow EMA—

Regime read—

One tick stream, three timeframes. The amber line is the fast EMA, the blue line the slow; their crossovers are the classic trend signal — note how often they whipsaw on the fast timeframe and how late they arrive on the slow one. That lag-versus-noise trade-off is every indicator's contract. The volume bars at the bottom are spread crossings: each one was somebody paying a market maker for immediacy.

The bridge to market makingPlatforms sell you the map: candles, EMAs, RSI, levels. The market maker lives in the territory: ticks, queues, fills, inventory. The map still matters — vol estimates set your spread, regime reads set your aggression, VWAP schedules describe your best customers, stop clusters predict your jumps. Part II now goes underground into the territory itself: why the spread exists, and what it must pay for.