From Signal to Trade, Part 2: Reading Book Churn and Sweeps on the Screen
A book can be rewritten dozens of times in 100 milliseconds while nothing trades.
In Part 1, we turned book imbalance, VPIN and Market Resilience into rules for entry, sizing and re-entry. Part 1 also said this part would show two ratio tiles helping an execution engine pull resting orders in time. Our research did not support an order-pulling trigger. Method and limits, at the end, says why.
What the two tiles give a desk is context. The Order-to-Trade Ratio (OTR) flags a book that churns while nothing trades. The Trade-to-Order Ratio (TTO) makes many sweeps visible at a glance in VisualHFT Terminal. Read together, they show whether the book in front of you is being quoted or being traded.
Every number comes from one NASDAQ TotalView-ITCH day, 22 August 2023, 14 symbols. The full study is Six Study Tiles, One NASDAQ Day, sections 5 and 6.
1. Two Ratios Regulators Already Watch
OTR and TTO are institutional order-flow measures, not retail chart indicators.
Hasbrouck and Saar (2009) found that a large share of limit orders on an electronic book are fleeting. They are submitted and cancelled within seconds, without trading. OTR measures that churn: book changes per trade, in each 100 ms window by default.
Regulators watch it too. Friederich and Payne (2015) studied the Italian Stock Exchange’s penalty on high order-to-trade ratios. It was followed by a collapse in quoted depth and higher price impact, with spreads unchanged. In the EU, MiFID II requires trading venues to calculate the ratio of unexecuted orders to transactions (Commission Delegated Regulation (EU) 2017/566).
TTO is the mirror image: traded volume over book volume, in each 1-second window. The SEC’s MIDAS program publishes a trade-to-order volume ratio for U.S. stocks and exchanges. At high frequency, the ratio jumps when trading suddenly dominates book activity, as in a sweep that consumes displayed depth.
2. OTR: A Book Rewritten While Nothing Trades
We took OTR readings in their symbol’s top 1% for the day that held for at least half a second. In 40 of 61 (65.6% vs 33.1% at matched moments), the tape showed two prints or fewer within a second either side while the book changed at least four times. A spike marks a book rewritten with almost no trading.
IDXX, 13:14:55 to 13:15:07 ET. OTR (count) at 100 ms. Replayed near real time; one passage repeated at half speed.
On IDXX at 13:15:03.4 ET, the tile has read 62 since 13:15:00.6: 63 book changes in one 100 ms window, and no trade since 13:14:30. In that burst, the best offer stepped down from 486.71 to 486.66.
About five seconds after the jump, the quotes moved fast. By 13:15:07.3 ET, trades marked mostly as sells have printed, the best bid is 485.98, and the spread has widened from 0.24 to 0.47. Here the price fell, but a spike is context, not a direction.
What follows a spike (exploratory). A separate scan compared the 10 seconds after each top-1% reading with 50 random moments from the same symbol and half hour. On the held-out symbols, 69.8% of spikes vs 49.9% of random moments were followed by mid-price volatility above the symbol’s normal. Discovery agreed.
How a desk uses it. Read a spike as a message-intensity gauge. Hasbrouck and Saar’s fleeting orders are the reason to treat displayed size with care while the book churns. Use it as context for your own quoting rules, for example wider tolerances for a few seconds.
3. TTO: Sweeps at a Glance
AMZN, 13:09:30 to 13:09:40 ET. TTO at 1 second. Replayed near real time; one passage repeated at half speed.
On AMZN at 13:09:38 ET, the offer side shows 99 shares at 134.42 and 7,800 at 134.43. The TTO tile reads 0.0000.
By 13:09:39.5 ET, Time and Sales shows 19 prints, 18 marked as buys, one of them 5,600 shares at 134.43. Both offer levels are gone; the best offer is 134.44. The tile reads 0.2098, and the TTO line shoots far above every bump of the previous minute. Buyers swept the offer.
Across the day, we counted sweeps: at least 5 prints on the same side within 250 ms that move the mid. During 22.4% of sweeps vs 2.7% of quiet moments, TTO reached its own top 10% for the day and held there for at least half a second.
How a desk uses it.
- See sweeps at a glance. On a busy tape, the TTO line makes the moments when trading overwhelmed the book easy to spot while they are in the chart’s window.
- Keep it off fixed scales. The ratio is unbounded and can jump far above 1 when book activity is thin.
4. Two Context Rows for the Part 1 Desk Matrix
Part 1 built a matrix for entry, skew, sizing and re-entry. These two rows add context:
| Reading on screen | What it shows | Desk response |
|---|---|---|
| OTR at its symbol’s top 1% | The book is being rewritten with almost no trading. | Treat displayed size with care. Widen tolerances for a few seconds. |
| TTO spike | Trading overwhelmed book activity, as in a sweep that consumes displayed depth. | Note the sweep: its side from Time and Sales, the cleared levels from the ladder. Quoting decisions stay with the Part 1 tiles. |
5. Wiring It Into the Trigger Engine
The Trigger Engine walkthrough in Part 1 applies. Two rules fit. Both are our suggestions, not rules we tested.
- OTR context flag. A condition on OTR, operator
CrossesAbove, value the symbol’s top-1% level from a prior session. Its webhook goes to your system that widens tolerances for a few seconds. - TTO spike log. A condition on TTO,
CrossesAbove, value the symbol’s top-10% level from a prior session, posting to a desk channel or review log. It logs TTO spikes, not every sweep.
CrossesAbove fires on the update where the value crosses. Set a rate limit of seconds or longer so a flipping value cannot flood your endpoint. Test any rule the way we did: on your venue, at your aggregation level, against the cheapest rule you can think of, and with your own costs.
Method and limits
- Data. One day, 14 symbols, 10 price levels per side. The hold-out is other symbols on the same day, re-scored across several research rounds; capture rules were written after those results. Every test reads a value only from the first book update after it was computed.
- The OTR volatility scan is exploratory: written after the earlier results, no rule locked in advance, no free rival tested. Matched also on prior-10-second book activity, the hold-out finding was weaker: 62.8% vs 49.9% (p = 0.049).
- Spread widening after an OTR spike was not reliable across the day; the capture’s widening is one instance.
- OTR on fills and price impact. A plain count of book updates did as well or better in every case we could test; at 100 ms most windows hold no trade. Hence OTR as context, not a fill forecast.
- The 65.6% and 22.4% figures sit close to each ratio’s definition, so they describe what the tile shows, not a forecast. The 22.4% is the share of sweeps that drove TTO to its top 10%; it does not say how many TTO spikes were sweeps.
- TTO and fill cost. Read naively, passive fills during extreme TTO seemed to cost more, but the penalty belonged to the print’s place inside its sweep. Read causally, the sign reversed on the hold-out (p = 0.0005), and discovery did not confirm it. Hence no TTO trigger to pull resting orders.
- Not the regulatory ratios. The MiFID II ratio is per member, which a public feed cannot see. The TTO tile’s denominator counts cancels and updates too, so it differs from the SEC’s ratio.
- Captures and costs. Captures were chosen from setups that met each rule, favouring clear moves. No transaction costs were applied. The captures show the tiles as “OTR (count)” and “TTO (legacy)”; released builds label them “OTR” and “TTO”.
With both tiles on screen, your desk knows whether it is quoting into a book that is being rewritten or one that is being traded through.
Related reading
From Signal to Trade, Part 1: Turning Book Imbalance and Toxicity into Desk Rules
Why a 73.9% directional hit rate still loses to the spread when traded alone, how Touch-Weighted LOB Imbalance lifts reversal entries to 78.2%, and how to automate multi-tile alerts in the VisualHFT Trigger Engine.
Getting Started with the Open-Source VisualHFT Application
Build the open-source VisualHFT application, connect public market data, inspect its studies, and extend it with the public SDK templates.
VisualHFT Research, Issue 01: One NASDAQ day, six tiles
Issue 01 of VisualHFT Research, as a PDF: two settings of the LOB Imbalance tile compared on ten liquid NASDAQ symbols, a guide to reading the captures, and a map of the fourteen spreads.