What it would cost to move the price

Manipulation risk is not a guess about intentions but a calculation: the cost of the move set against what depends on it.

This lesson has had no expert review. It was written for this platform and against the evidence it cites; nobody has gone through it independently.

Learning objectives

You can estimate what it costs to move a price source briefly.
You can say which design of a price source raises that cost and which does not.
You can check a fallback mechanism for whether it triggers when it matters.

Check your prior knowledge

Answer these for yourself before reading on. Wherever you hesitate is where this lesson pays off.

Where does a protocol take the price it values collateral with?
What changes when a price is averaged over a period?
What should happen when the source stops delivering data?

Core concept

Fact

The question is answerable, not speculative

Whether anyone wants to move a price source is unknown. What it would cost is computable: the price impact of a trade against the source's depth gives the cost, and that cost stands against the sum decided on the basis of the price. Both numbers are available today, and their ratio is the metric — not an assessment of anyone's intentions.

Interpretation

Averaging over time raises the cost, it does not remove it

If the price is averaged over several blocks, a move must be sustained across the whole period — which multiplies the cost and is the point of the design. But it also creates a second property: the averaged price follows a genuine move with a lag. In a fast market the protocol then values collateral at a price that no longer exists.

Risk

More sources help only if they are independent

A median of five sources sounds robust and is, as long as the five do not measure the same thing. If three of them take their data from the same venue, moving that one suffices to shift the median. So the usable figure is not the number of sources but the number of mutually independent markets beneath them.

Uncertainty

A fallback that never triggers is not one

Many designs provide for switching to a second source, or halting, on failure or implausible values. Whether that bites when it matters depends on the threshold: if it only triggers at a divergence larger than any realistic manipulation, it evidences care rather than protection. That threshold is rarely stated in the documentation and needs to be asked about.

Definitions

Price source: The market or service from which a protocol takes the price for its decisions.
Time-weighted price: A price averaged over a defined period instead of the last traded one.
Fallback: The designated response when the primary source fails or delivers implausible values.

Model

Which markets actually sit beneath the source — not: how many sources
Depth of those markets in the tight band
Over how many blocks is it averaged?
Which sum depends on that price?
At what divergence does the fallback bite?

Five figures that make up a manipulation estimate — The fifth figure is the one most often missing — and without it the fallback is not a checkable statement.

Formulas

Ratio of the cost of moving to the dependent sum

ratio = cost_of_moving / dependent_sum

cost_of_moving: Estimated price impact plus fees to shift the source by the required amount
dependent_sum: Sum decided on the basis of that price

Limit: The cost of moving is estimated from today's depth and changes with it; the ratio is therefore a snapshot. It also says nothing about whether an attempt is technically feasible — only whether it would pay.

Worked example

Two designs, the same sum behind them

Dependent sum: USD 40m of borrowing
Source A: last price from one pool, depth USD 1.4m
Source B: median of four independent markets, each averaged over 30 blocks
Fallback for B: halt on divergence above 25 %

For A a move in one shallow pool for a single block suffices. For B several independent markets would have to be moved at once and across 30 blocks. The 25 % threshold sits above what a worthwhile distortion typically needs.

B is considerably more expensive to move; its fallback, though, contributes nothing to that statement.

Reading: Two separate findings: B's design raises the cost considerably — that is the protection. The fallback at 25 % is documented and would probably not trigger on a worthwhile distortion; it should be recorded as such rather than counted as a second safeguard.

Retrieval

Five sources — redundancy is high.Three independent markets, since three providers measure the same thing.The number carries no meaning.

As a second safeguard that defuses the manipulation arithmetic.As a documented precaution that does not trigger on a worthwhile distortion below that threshold.As evidence that the source was examined.

Exercise on real data

Take the guiding questions and enumerate the genuinely independent markets beneath one price source.

Dimension 10: dependencies →

Application

Write the manipulation estimate for a collateralized position in five lines.

Related case studies

Institutional reading

Bank: Which of your positions hang on a price source whose depth you do not know?
Insurance: Under your wording, would a distorted price source be a loss event?

Metrics in this lesson

Liquidations

Key takeaways

Manipulation risk is a calculation from cost of moving and dependent sum, not a guess.
Averaging raises the cost and simultaneously creates a lag behind genuine moves.
What counts is independent markets, not data providers — and a threshold decides whether a fallback bites.

Evidence

EVD-2026-0004
Annual Review of Financial Economics — Smart Contracts and Decentralized Finance