Is the Market Really Algorithmic? The ICT Delivery Thesis Examined

Is the Market Really Algorithmic?

Yes — in execution, modern markets are overwhelmingly algorithmic: most FX and equity order flow is generated or routed by machines. Whether one master algorithm "delivers" price to pre-selected liquidity, as ICT's IPDA thesis claims, is a separate and unproven question.

Those are two different claims, and most arguments about ICT collapse because they get treated as one. The first — algorithms dominate order flow — is an empirical fact documented by central banks and regulators. The second — a coordinated Interbank Price Delivery Algorithm with intent — has never been documented by anyone.

This page separates them: what ICT actually asserts, what the data confirms, where the thesis overreaches, and why it can still produce a working trading model even if the literal claim is wrong. If you want the citable answer to "is the market algorithmic," it is here, with the caveats attached.

What ICT Actually Claims: The IPDA Thesis, Stated Precisely

The Inner Circle Trader (ICT) thesis, taught by Michael J. Huddleston, holds that price is not primarily moved by open supply and demand. Instead, an Interbank Price Delivery Algorithm — IPDA — "delivers" price to liquidity that has already been booked in advance. The claim has four load-bearing parts:

• Price seeks liquidity. The algorithm targets pre-existing pools — clustered stops above Equal Highs and Equal Lows (EQH/EQL), old daily and weekly extremes — the current Draw on Liquidity (DOL).
• Delivery is time-based. The algorithm operates on schedules, concentrating in session windows ICT calls Kill Zones and in 20-minute Macro Times — so when matters as much as where.
• It references fixed lookbacks. The IPDA data ranges — the past 20, 40, and 60 trading days — define which highs, lows, and inefficiencies the algorithm will trade to next.
• It reprices to inefficiency. After a displacement, price returns to rebalance gaps such as the Fair Value Gap (FVG) before continuing toward the draw.

Stated strongly, this describes a real, singular entity: one algorithm at the interbank level delivering price with intent. Stated weakly, it says markets behave as if such an algorithm existed. Which version you evaluate determines the verdict, so this examination keeps them separate throughout.

What the Evidence Confirms: Market Execution Is Algorithmic

On the narrow question — is the market algorithmic in how it actually trades — the evidence is not close. Machines have dominated execution across major asset classes for over a decade, and the documentation comes from central banks, not trading gurus.

How Algorithmic Is the Market, by the Numbers?

• Foreign exchange. Bank for International Settlements research tracked algorithmic activity on EBS, a primary interbank venue, growing from roughly 2% of volume in 2004 to a clear majority of order submission within a decade. A 2020 BIS Markets Committee report estimated that execution algorithms alone handle roughly 10–20% of global spot FX turnover — on the order of $200–400 billion per day — before counting market-making and high-frequency strategies.
• Equities. Estimates vary by methodology, but widely cited figures place algorithmic participation in US equities well above half of total volume, with high-frequency strategies alone accounting for a substantial share.
• Futures and crypto. CME order flow is heavily automated, and crypto perpetuals trade 24/7 on venues where market-making bots provide the bulk of resting depth.

What are these algorithms? Execution algos slicing institutional orders against benchmarks like VWAP and TWAP; market-making algos quoting two-sided prices; arbitrage and high-frequency strategies; systematic funds executing signals. Different owners, different objectives, one shared trait: they are rule-based and they are the market's plumbing.

None of this is contested. When a trader says "the market is algorithmic" in the descriptive sense, they are correct — and can cite the BIS rather than a YouTube video. The dispute begins with the next step: going from many algorithms trade the market to one algorithm delivers the market.

The Critical Distinction: Many Competing Algorithms Are Not One Master Algorithm

The documented algorithmic market is a decentralized, adversarial ecosystem. Thousands of independent algorithms — owned by banks, market makers, funds, and brokers with directly opposing objectives — compete across EBS, LSEG Matching, CME, exchanges, and dozens of internalization pools. There is no shared controller; one bank's execution algo cannot know what a rival's will do next.

ICT's strong-form IPDA is categorically different: a single coordinating mechanism that books liquidity in advance and delivers price to it. No regulatory filing, market-structure document, exchange specification, or peer-reviewed study describes such an entity. And as usually stated, the claim is unfalsifiable: if price runs the stops, IPDA delivered; if it doesn't, IPDA is "seeking liquidity elsewhere." A thesis that fits every outcome forecasts none.

Note what the table does not say: it does not say ICT's observations are wrong. It says the proposed cause is undemonstrated. That gap matters, because the observations have a better-documented explanation.

Why the Delivery Thesis Still Works: Emergent Order From Competing Algorithms

Here is the productive part of the debate. You do not need a master algorithm to get ICT-style price behavior. Aggregate the documented ecosystem — competing algos plus predictable human stop placement — and it produces emergent regularities that look centrally delivered:

• Liquidity-seeking is engineered in. Execution algorithms and smart order routers are literally programmed to locate resting liquidity, because clustered stop orders above equal highs are the cheapest place to fill institutional size. A Liquidity Sweep happens because that is where the orders are — no intent required.
• Flows concentrate in time. Session opens, the 4pm London WM/R fix, 8:30am New York data releases, and expiry windows compress volume into known clock positions. Recurring time-price behavior — the thing kill zones map onto — falls out of the calendar, not a controller.
• Value reversion is benchmarked in. Trillions in institutional flow is executed against VWAP and arrival-price benchmarks, creating real, measurable pressure back toward value after a displacement. Opening ranges and prior-day extremes are hardcoded reference inputs in production algos.

Competing algorithms plus crowd behavior therefore yield a market that behaves as if price were delivered to liquidity — with no delivery algorithm anywhere in the stack. In that sense IPDA is a useful fiction: like describing traffic as a "flow" even though no one steers each car, it compresses real dynamics into one tractable story.

Watch it on a chart. EURUSD builds equal highs at 1.0848–1.0850 through the Asian session; breakout orders and short stops stack just above. At the London open, price runs to 1.0855, fills that resting liquidity, then displaces to 1.0818, leaving a fair value gap at 1.0832–1.0824. It retraces into the gap and sells off toward the pool of stops under 1.0790.

An ICT trader narrates that as IPDA taking buy-side then delivering to sell-side. A microstructure desk narrates it as stop-cluster liquidity absorbed by execution algos, momentum systems chasing, then benchmark-driven flow fading the extension back to value. Same tape, same trade, two vocabularies. That equivalence is why the model "works" without its metaphysics being true.

The Case For and Against the ICT Delivery Thesis

An honest examination has to make both sides as strong as possible. Here is each steelman.

The Steelman for ICT

• Obvious stops at obvious times get run with striking regularity. The Judas Swing — a false move at the session open before the true direction — recurs across decades of London and New York opens, in FX, indices, and crypto alike.
• Session structure is objectively real. Volatility and volume follow the clock; that is not an ICT invention, it is the reason any time-based framework works at all.
• Real algorithms genuinely reference old levels. VWAP, opening ranges, prior-day and multi-week highs and lows are standard inputs in institutional execution logic. "The algorithm looks back at prior ranges" is directionally true of actual production code — even if 20/40/60 days specifically is arbitrary.
• Engineered price movement has documented precedent. In 2014 the CFTC ordered five banks to pay over $1.4 billion for attempted manipulation of FX benchmark rates around the London fix. Coordinated liquidity-targeting is not a conspiracy theory; it has case numbers.

The Steelman Against IPDA

• Survivorship in pattern-reading. Sweeps that reverse become screenshots; sweeps that keep trending are forgotten. Unless every swept level is logged — hits and misses — the "uncanny" regularity cannot be measured, only felt.
• Ex-post narrative fitting. IPDA explains raids, continuations, and reversals equally well after the close. A model that cannot be surprised by any outcome carries no predictive information.
• No unique predictions. Everything IPDA predicts — sweeps, session moves, reversion to value — is equally predicted by standard microstructure, which additionally predicts things IPDA is silent on, like spread and depth dynamics.
• The lookbacks lack derivation. No published work shows the 20/40/60-day IPDA data ranges outperform 15, 30, or 50 days. Fixed sacred numbers that were never benchmarked are a classic overfitting signature.

The honest score: the phenomena ICT points at are real and tradable; the singular cause it proposes is undemonstrated and, as framed, untestable. The FX-fix cases prove episodes of engineered pricing occurred — they do not prove a permanent, market-wide delivery engine.

Trading an Algorithmic Market Without the Metaphysics

The tradable content of ICT survives either verdict. Whether stops get run because one algorithm delivers price or because a thousand algorithms hunt the same resting orders, the chart prints the same event — and the same entry. Practically:

• Trade observable events, not intent. A sweep of equal highs followed by displacement through structure is a fact on the chart; "IPDA wanted that liquidity" is a story about the fact. Execute on the fact.
• Keep the clock. Session-concentrated behavior is real regardless of cause. Restricting setups to windows where flows demonstrably cluster is defensible on the data alone.
• Test the sacred numbers. Run 20/40/60-day lookbacks against 15, 30, and 50 on your own instruments. If the edge exists only at ICT's exact windows, distrust it; if it is robust across nearby windows, you have found a genuine liquidity effect rather than a magic constant.
• Journal the misses. Log every swept level, including the ones that never reversed. Survivorship will otherwise manufacture conviction your data does not support.

This is also the sane way to use tooling: LiquidityScan's scanners flag the observable layer — sweeps, structure breaks, session ranges, gap fills — precisely because those events are testable no matter what you believe causes them.

So, is the market algorithmic? In execution, demonstrably yes — the BIS will back you up. As a single interbank machine that booked your stop loss in advance: unproven, unfalsifiable as stated, and unnecessary for the trade. Take the regularities seriously, hold the metaphysics loosely, and let your journal — not the narrative — decide what stays in your playbook.

Frequently Asked Questions

What percentage of trading is done by algorithms?

It varies by asset class and methodology. Widely cited estimates put algorithmic participation in US equities well above half of volume, and BIS research found the majority of order submission on primary interbank FX venues became algorithmic during the 2010s. Execution algorithms alone were estimated at roughly 10–20% of global spot FX turnover in 2020.

Who is supposed to operate the IPDA, according to ICT?

ICT attributes it to the interbank or institutional level but never names an operator, a vendor, or any documentation. That is the core evidentiary gap: real algorithmic infrastructure leaves regulatory and academic paper trails, while the IPDA exists only within ICT's own teaching materials.

Is the market rigged against retail traders?

Not personally. Algorithms target liquidity, and retail stops happen to cluster at predictable levels — above equal highs, below obvious swing lows — making them cheap fuel. Enforcement cases like the 2014 FX benchmark fines prove episodes of genuine manipulation occurred, but day-to-day stop runs are better explained by rule-based liquidity-seeking than by anyone hunting you specifically.

Can you trade ICT concepts without believing in IPDA?

Yes, and arguably you should. Liquidity sweeps, kill-zone timing, displacement, and reversion to fair value gaps are observable, backtestable regularities fully explainable by execution algorithms and crowd stop placement. Treating them as testable patterns rather than doctrine keeps your process falsifiable — which is exactly what the strong IPDA claim is not.

Related query paths

If this raised the right follow-up questions, these are the natural next reads, ordered from the core definition to application and verification.

• IPDA Explained: ICT's Price Delivery Algorithm — the full definition-level breakdown of the concept examined here.
• Why ICT Setups Need Both Time and Price — the time-based delivery logic in practice, whatever its cause.
• Liquidity Sweep Explained: The ICT Stop Hunt — the single most observable "delivery" event, mechanics and entries.
• Is the Forex Market Manipulated? The Institutional View — the manipulation question treated with the same evidence standard.
• How to Backtest an ICT Strategy the Right Way — how to test these regularities on your own data instead of trusting narratives.
• Is ICT Trading Real? A Skeptic's Guide to Smart Money — the broader methodology put under the same skeptical lens.
• Liquidity in ICT: Why Price Moves to Buy-Side and Sell-Side Pools — how it connects to liquidity in ict.