The Rising MM 3: Statistical Advantages and Signal Design

Author: Dave

“Attack of MM 1: Market Maker Inventory Quoting System”

“Attack of MM 2: Market Maker Order Book and Order Flow”

The first two episodes mentioned order flow and inventory quoting. It sounds like market makers can only passively adjust, but do they have proactive strategies? The answer is yes. Today, we introduce statistical advantage and signal design, which are also the “micro alpha” pursued by market makers.

1. What is the alpha for market makers?

Micro alpha refers to the “conditional probability shift” of the next price movement direction / mid-price drift / trade asymmetry within an extremely short time scale (~100ms to ~10s). It’s important to note that, from the MM perspective, alpha is not trend prediction or guessing the magnitude of price moves; it only involves probability shifts. This is different from the alpha we often talk about in traditional finance. Let me explain in plain language:

A market maker’s statistical advantage can be understood as whether, within a very short time window, the order book state “tends to” cause the price to move in a certain direction first. If MM can successfully calculate the probability of the next millisecond’s price direction using some indicators, then they can: 1. Prefer to buy before a likely rise. 2. Cancel buy orders faster before a likely fall. 3. Reduce exposure during risky moments.

The financial basis for predicting the next price direction is: due to order flow, order book volume, order cancellations, and other factors (which will be discussed), the market in a very short moment is not a “random walk” Brownian motion, but has a directional component. The above sentence is the financial translation of the mathematical concept of “conditional probability.”

With these alphas, market makers can operate on the price directionally, finally earning the price difference rather than just the spread as a service fee.

2. Introduction to classic signals

2.1 Order Book Imbalance: OBI

OBI measures which side—buy or sell—is “more crowded” near the current price level. It is a standardized volume difference statistic.

The formula is straightforward: it’s a ratio of sums, indicating whether buy orders or sell orders dominate. OBI approaching 1 means almost all bid orders are on the buy side, with a thick order book below. Approaching -1 means the opposite. Approaching 0 indicates a balanced order book.

Note that OBI is a “static snapshot.” It’s a classic indicator but not very effective alone. It should be used together with order cancellations, order book slope, and other metrics.

2.2 Order Flow Imbalance (OFI)

OFI looks at who is actively attacking within a recent short period. OFI is the first derivative driver of price change because prices are driven by taker orders, not resting orders.

It resembles net buying or selling volume. In Kyle (1985) framework, ΔP ≈ λ⋅OFI, where λ is the tick depth, so OBI is a factor influencing price movement.

2.3 Queue Dynamics

Most exchanges now operate under continuous auction rules, following the best price and FCFS principles, so submitted orders are queued. The queue reflects the order book state, and abnormal states (including order cancellations and re-submissions) hint at directional price changes, i.e., micro alpha.

Pay attention to two types of queue phenomena:

1. Iceberg: Hidden orders

For example, only 10 lots are visible, but each time they are filled, another 10 lots are immediately replenished. The real intent might be 1000 lots. The malicious market manipulation method I introduced in the first episode—artificially lowering the cost basis—is essentially an iceberg order. In practice, some traders use iceberg orders to hide true order sizes.

2. Spoofing (Fake Orders)

Placing large orders on one side to create a “pressure illusion,” then quickly canceling them as the price approaches. Spoofing can distort OBI and slope indicators, artificially thickening the queue and increasing the risk of price movement. Some large spoof orders can scare the market and manipulate prices. For example, the London Stock Exchange reportedly caught a manipulator in 2015 who used spoofing in forex. In crypto, we can also manually spoof to manipulate market makers, but if the order gets filled, your exposure is significant.

2.4 Order Cancellation Ratio (Cancel Ratio)

The cancel ratio estimates liquidity “disappearance rate”:

Cancel↑ ⇒ Slope↓ ⇒ λ↑ ⇒ ΔP more sensitive. It is a leading indicator of OFI’s instability signals. CR→1: almost pure cancellations. CR→0: almost pure order additions. The mathematical formulas in this episode are simple; just look at the diagrams for interpretation.

CR↑ ⟹ passive side perceives increased future risk, but CR is not used alone; it’s combined with OFI and other metrics.

These may be old tricks in order book games, but the evolution of market making is rapid. Moreover, once stocks are on-chain, MM might also involve on-chain market making with JS and other tools. Still, these indicators are very useful and inspiring.

3. The absolute domain of market makers: speed

In movies, we often hear that a certain fund’s network speed is faster, making them more powerful. Many market makers move their servers closer to exchange servers—why? This article concludes with a discussion of the advantages of physical equipment and the “trading advantage” unique to crypto exchanges.

Latency arbitrage is not about predicting future prices but about executing trades at better prices before others “react.” In theoretical models: prices are continuous, and information is synchronized. But in reality: markets are event-driven, and information arrives asynchronously. Why is information asynchronous? Because receiving price signals from exchanges and sending order instructions both take time—limitations of the physical world. Even in fully compliant markets: different exchanges, data sources, matching engines, and geographic locations cause delays. Therefore, having more advanced equipment gives MM the initiative.

This tests the market maker’s own strength and is less about other players, so I believe it’s their absolute domain.

For example, if you want to sell a position, you quote the best market sell price, which theoretically can execute immediately. But if I also want to sell, and I see the price and quote faster than you, I might eat your order first, leaving your inventory unliquidated, causing your position to remain neutral and unable to recover. The actual situation is much more complex.

Interestingly, because there are no regulations yet, almost all crypto exchanges can give certain accounts priority execution rights—meaning some accounts can “cut in line.” This is especially common in smaller exchanges. It seems that in crypto, being “one of us” is as important as in scientific research. Safe execution is a critical step for alpha theory to transition into practical trading.

This episode attempts to write from the MM perspective. Actual operations are definitely more complex, such as dynamic queues, which involve many details in practice. Welcome teachers to comment.

Postscript: There is a regret about this article. The title “Field Expansion in Market Making” was originally intended to discuss dynamic hedging and options, as I believe these are the most conceptually challenging areas in market making, worthy of the “field expansion” title. But I spent a day working on it, half the article was written, and I couldn’t systematically explain this topic, so I switched to micro alpha. @agintender has an article covering many professional hedging concepts, I encourage everyone to check it out.