Introduction to Order Flow: Seeing Inside the Candles

A candlestick gives you four prices and a color. It doesn't say how many contracts traded at each price, or whether buyers or sellers were doing the hitting. Two candles with the same shape can have very different volume profiles, and order flow is the set of tools for looking at that difference.

It is also easy to over-read. These tools describe what already traded. Whether a pattern predicts anything in your market is a testable question, not a given, so this article covers how the pieces work and then how to check them.

Passive and aggressive orders

A market is an auction. Limit orders rest in the book at a price and wait, which supplies liquidity. Market orders take what's there: a buy market order lifts an offer, a sell market order hits a bid. Trades printing at the offer are usually classified as buyer-initiated, and trades printing at the bid as seller-initiated. Price moves when aggressive orders use up the resting size at a level, or when resting orders get cancelled.

Delta is aggressive buy volume minus aggressive sell volume over a chosen period, such as one bar. Positive delta means more volume traded at the offer. It's a classification of executed trades, so it depends on the quality of your tick data and bid/ask tagging.

A worked footprint example

A footprint chart splits each bar's volume by price, with sells at the bid on the left and buys at the offer on the right. This is a hypothetical ES bar to show the arithmetic.

Total selling is 835 and total buying is 782, so bar delta is negative 53. That single number looks mildly bearish, yet the top row shows heavy buying at the high and the bottom row shows heavy selling at the low. The distribution inside the bar carries information that the total hides, which is the argument for looking at a footprint at all.

Patterns traders look for, and their limits

Delta divergence. Price prints a new low while delta for the bar is positive. The reading is that sellers are pushing but buyers are absorbing. The limit: one bar of divergence is a single observation, and sellers can resume on the next bar.

Absorption. A lot of volume trades at a level, yet price doesn't move through it. This suggests large resting orders on the other side. The limit: you can't see who is behind them, and iceberg orders that show only part of their size look similar to genuine absorption.

Stacked imbalances. Several adjacent price levels where one side dominates by a chosen ratio. The ratio and the number of rows are settings you choose, so two traders can see different signals on the same data.

Trapped traders. Price breaks a level on strong delta and then reverses, leaving breakout buyers underwater. The story is plausible, but whether their stops add real selling is an assumption unless you've measured it.

None of these has a known win rate that I can cite, and I'd treat anyone who quotes one without methodology with suspicion.

The depth of market

The DOM shows displayed limit orders above and below the current price. It is useful for seeing where size is resting and how fast it refreshes. It is also incomplete. Displayed size can be cancelled before price arrives, hidden size doesn't show at all, and spoofing, the placement of orders with no intention to trade them, is prohibited yet still something you should guard against by not trading on a large order alone. A level that persists while it gets traded through tells you more than one that appears for a moment.

Using order flow with fixed levels

Key Price Levels are fixed reference prices anchored to the regular-session opening print. The standard ES and NQ map has four levels, built from full-session and after-hours travel, and they don't move intraday. They are not predictions, and not levels derived from institutional order flow. Because the anchor is the regular-session open (9:30 a.m. Eastern for US equity cash), a completed anchored map can't exist before that print. The method is on the KPL trading page.

Order flow can be a second lens at such a level: you watch how volume and delta behave as price reaches it, and you decide in advance what you'd treat as acceptance and what as rejection. That is a discretionary reading, and the result is not assured. If you want the logic of the automated version, read the KPL Bot specs instead of assuming what it checks.

What it takes to use these tools

You need a market data feed that provides depth and tick-level bid/ask data, and a platform that builds footprint or volumetric bars from it. NinjaTrader offers order flow tools, and licensing and data requirements change, so check its official help guide and your data provider's current terms rather than relying on a fixed price. Historical footprint also needs stored tick data, which affects how far back you can study.

How to practice without guessing

Watch first. For a few weeks, record what you see at your levels without trading: the level, the time, the delta and volume behavior as price touched it, and what price did over the next several minutes. Write the hypotheses down before the data arrives, such as 'absorption with delta turning positive precedes a bounce of at least N ticks,' so you can't redefine success afterward. Then compare against a baseline of touches that lacked the signal. A few dozen observations only tell you whether the idea is worth testing further. They are not proof, and SIM fills don't validate exchange queue position or slippage.