PyPI - lobkit - Versions diffs - 0.2.1__tar.gz → 0.2.2__tar.gz - Mend

lobkit 0.2.1tar.gz → 0.2.2tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Files changed (22) hide show

{lobkit-0.2.1/src/lobkit.egg-info → lobkit-0.2.2}/PKG-INFO RENAMED Viewed

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lobkit
-Version: 0.2.1
+Version: 0.2.2
 Summary: Limit order book reconstruction, order-flow analytics and market-making simulation
 Author: Yousuf Hossain Shafeen
 License-Expression: MIT
@@ -170,35 +170,40 @@ simulated and live tables shows which effects are mechanical consequences of the
 and which require informed traders. Record at least 15 minutes; the script warns when there is
 too little data to say anything.
-Results for one session: BTCUSDT, 40 minutes on 2026-11-01 from 03:18 local time (76,296 quote
-updates, 5,813 trades). It was a quiet market: the price stayed within a 181 USDT range and the
-mid changed 688 times, about once every 3.5 seconds. That session is one sample, not a study, and
-the recording is not in the repository (market data is never committed), so treat these numbers
-as an illustration of the pipeline.
-| Result | Live, 1 s buckets | Simulated |
-|---|---|---|
-| Δmid on OFI | β = 0.221 USDT per unit, **R² = 0.56**, t = 55 | R² = 0.42 at 1 s (Study 5), 0.66 at 10 s |
-| Impact coefficient vs 1/depth | R² = 0.85 across depth quintiles | R² = 0.93 (10 s) |
-| P(next move up), imbalance < 0.1 → > 0.9 | 0.38 → 0.97, not monotone in between | monotone |
-| Next 1 s move from queue imbalance | out-of-sample R² 3.3% | 0.7% (next 10 s) |
-| Multi-level OFI, best 1 → 10 levels (5 s) | out-of-sample R² 0.50 → 0.40–0.49: no gain | 0.66 → 0.88, 1 → 5 levels (10 s) |
-| Micro-price vs mid, next price change | Stoikov −0.34% MSE, naive −0.32%: no real edge | Stoikov −6.6%, naive −1.3% |
+Results for one session, two instruments recorded side by side: BTCUSDT and ETHUSDT, 2 hours each
+on 2026-11-01 from 04:29 local time. BTCUSDT: 187,746 quote updates, 18,463 trades; the price
+stayed within a 162 USDT range and the mid changed 2,176 times, about once every 3.3 seconds.
+ETHUSDT: 127,805 quote updates, 14,136 trades; an 8 USDT range and 2,931 mid changes. Both books
+tick in cents, so a tick is 0.13 basis points of the BTC price but 0.44 of the ETH price, and the
+ETH best queues are deep (31.5 ETH on average against 2.5 BTC). That session is one sample, not a
+study, and the recordings are not in the repository (market data is never committed), so treat
+these numbers as an illustration of the pipeline.
+| Result | BTCUSDT, 1 s buckets | ETHUSDT, 1 s buckets | Simulated |
+|---|---|---|---|
+| Δmid on OFI | β = 0.203 USDT per BTC, **R² = 0.59**, t = 102 | β = 0.0008 USDT per ETH, **R² = 0.70**, t = 129 | R² = 0.42 at 1 s (Study 5), 0.66 at 10 s |
+| Impact coefficient vs 1/depth | R² = 0.91 across depth quintiles | R² = 0.87 | R² = 0.93 (10 s) |
+| P(next move up), imbalance < 0.1 → > 0.9 | 0.09 → 0.93, monotone | 0.09 → 0.84, dips at 0.5–0.6 | monotone |
+| Next 1 s move from queue imbalance | out-of-sample R² 2.0% | 4.6% | 0.7% (next 10 s) |
+| Multi-level OFI, best 1 → 10 levels (5 s) | out-of-sample R² 0.33 → 0.38 | 0.45 → 0.47 | 0.66 → 0.88, 1 → 5 levels (10 s) |
+| Micro-price vs mid, next price change | Stoikov −0.6% MSE, naive −0.03% | **Stoikov −10.8%**, naive −4.2% | Stoikov −6.6%, naive −1.3% |
 ![Binance OFI impact](https://raw.githubusercontent.com/hossainyousuf751/microstructure-lab/main/docs/binance_btcusdt_ofi_impact.png)
-**What survives contact with a real book.** The contemporaneous OFI relation is at least as
-strong live as in the simulator at 1-second buckets, and impact is still larger when the book is
-thin. Queue imbalance predicts the next move at the extremes, and on the live book it even has a
-little out-of-sample power over the next second.
-**What doesn't.** Deeper levels add nothing out of sample, while they explained a lot in the
-simulator. On BTCUSDT the levels behind the best quote are a cent apart and mostly small resting
-orders, so their flow carries little information. Neither micro-price beats the mid by more than
-a third of a percent. Most rows sit at a one-tick spread with the price unchanged. The 688 price
-changes include 323 updates where one side of the book thinned out for an instant and the spread
-jumped past 1 USDT (up to 10.8), and a squared-error comparison over so few moves is decided by
-those points. A busier session, or a robust loss, is the next step.
+**What survives contact with a real book.** The contemporaneous OFI relation is stronger live
+than in the simulator at 1-second buckets on both instruments, and impact is larger when the book
+is thin. Queue imbalance predicts the direction of the next move at the extremes, and has a little
+out-of-sample power over the next second.
+**What depends on the book.** On ETHUSDT the Stoikov micro-price forecasts the next price change
+with 10.8% less squared error than the mid, more than in the simulator, and the edge fades with
+the horizon (4.3% after 5 moves, 0.9% after 20). On BTCUSDT it does not: −0.6%. The BTC book
+often thins out for an instant: the spread was above 1 USDT (100 ticks) in 1,131 updates over 305
+episodes, up to 17.4 USDT, while the ETH spread never exceeded 0.45 USDT. Those blowouts dominate
+a squared-error comparison and wreck the naive micro-price (its gap to the mid has an
+out-of-sample R² of −0.77 for the next second on BTC). Deeper levels add only a few points of R²
+on either book, much less than in the simulator: the levels behind the best quote are a cent apart
+and mostly small resting orders.
 ## Study 4: the market maker inside the order book
@@ -252,24 +257,30 @@ other venue's order flow explain, and whether either venue's flow or return pred
 known: a simulated venue B that copies venue A 300 ms late. They find the delay exactly
 (correlation 1.00 at +300 ms) and the direction: only B's next move is predictable from A.
-Same session as Study 3 (40 minutes; 76,296 Binance and 24,969 Bybit quote updates):
+Same session as Study 3 (2 hours each; 187,746 Binance and 76,632 Bybit quote updates for
+BTCUSDT, 127,805 and 21,344 for ETHUSDT):
 | Target venue | R² own OFI | R² other venue's OFI | R² both | Next 1 s from other's OFI (OOS) | Next 1 s from other's return (OOS) |
 |---|---|---|---|---|---|
-| Binance | 0.56 | 0.16 | 0.56 | 0.014 | 0.009 |
-| Bybit | 0.58 | 0.28 | 0.60 | 0.004 | 0.004 |
+| Binance BTCUSDT | 0.59 | 0.23 | 0.59 | 0.059 | 0.064 |
+| Bybit BTCUSDT | 0.54 | 0.33 | 0.58 | 0.002 | 0.001 |
+| Binance ETHUSDT | 0.70 | 0.15 | 0.70 | 0.036 | 0.094 |
+| Bybit ETHUSDT | 0.31 | 0.43 | 0.51 | 0.001 | 0.003 |
 ![Lead-lag](https://raw.githubusercontent.com/hossainyousuf751/microstructure-lab/main/docs/cross_venue_lead_lag.png)
-**No exploitable lead at this resolution.** Returns are most correlated at 0 ms (0.09): the
-venues move within the same 50 ms. Each venue's own order flow explains its price far better than
-the other's. Neither venue's flow or return predicts the other's next second by more than about
-1% out of sample. There is more correlation at negative lags (Bybit first, summed 0.62 against
-0.20 for Binance first), and it persisted after the recorder's timestamp bug was fixed (see
-CHANGELOG), but Bybit's price changed only 218 times, so this is a hypothesis for a longer
-recording, not a finding. One limitation is built in: timestamps are local receive times, so each
-venue's network latency to the recording machine is part of the measured lag. A lead shorter than
-the difference in latencies can't be seen from one machine.
+**Bybit tends to move first, by tens of milliseconds.** Returns are most correlated at 0 ms (0.25
+for BTCUSDT, 0.18 for ETHUSDT): most moves land on both venues within the same 50 ms. What is left
+is lopsided in the same way on both instruments. The correlation at negative lags (Bybit first)
+sums to 1.00 for BTCUSDT and 1.13 for ETHUSDT, against 0.26 at positive lags; the strongest
+off-zero lag is Bybit 150 ms ahead on both. The regressions agree: Bybit's flow and return predict
+Binance's next second (out-of-sample R² 3.6–9.4%), Binance's predict Bybit's by 0.3% or less. On
+ETHUSDT, Binance's flow explains more of Bybit's price than Bybit's own flow does (0.43 against
+0.31). The earlier 40-minute session pointed the same way; two instruments over two hours make it
+a consistent pattern, but still one session. One limitation is built in: timestamps are local
+receive times, so each venue's network latency to the recording machine is part of the measured
+lag. A lead shorter than the difference in latencies can't be told apart from a faster connection
+to Bybit, and a recording from a second location is the way to check.
 ## Roadmap

{lobkit-0.2.1 → lobkit-0.2.2}/README.md RENAMED Viewed

@@ -130,35 +130,40 @@ simulated and live tables shows which effects are mechanical consequences of the
 and which require informed traders. Record at least 15 minutes; the script warns when there is
 too little data to say anything.
-Results for one session: BTCUSDT, 40 minutes on 2026-11-01 from 03:18 local time (76,296 quote
-updates, 5,813 trades). It was a quiet market: the price stayed within a 181 USDT range and the
-mid changed 688 times, about once every 3.5 seconds. That session is one sample, not a study, and
-the recording is not in the repository (market data is never committed), so treat these numbers
-as an illustration of the pipeline.
-| Result | Live, 1 s buckets | Simulated |
-|---|---|---|
-| Δmid on OFI | β = 0.221 USDT per unit, **R² = 0.56**, t = 55 | R² = 0.42 at 1 s (Study 5), 0.66 at 10 s |
-| Impact coefficient vs 1/depth | R² = 0.85 across depth quintiles | R² = 0.93 (10 s) |
-| P(next move up), imbalance < 0.1 → > 0.9 | 0.38 → 0.97, not monotone in between | monotone |
-| Next 1 s move from queue imbalance | out-of-sample R² 3.3% | 0.7% (next 10 s) |
-| Multi-level OFI, best 1 → 10 levels (5 s) | out-of-sample R² 0.50 → 0.40–0.49: no gain | 0.66 → 0.88, 1 → 5 levels (10 s) |
-| Micro-price vs mid, next price change | Stoikov −0.34% MSE, naive −0.32%: no real edge | Stoikov −6.6%, naive −1.3% |
+Results for one session, two instruments recorded side by side: BTCUSDT and ETHUSDT, 2 hours each
+on 2026-11-01 from 04:29 local time. BTCUSDT: 187,746 quote updates, 18,463 trades; the price
+stayed within a 162 USDT range and the mid changed 2,176 times, about once every 3.3 seconds.
+ETHUSDT: 127,805 quote updates, 14,136 trades; an 8 USDT range and 2,931 mid changes. Both books
+tick in cents, so a tick is 0.13 basis points of the BTC price but 0.44 of the ETH price, and the
+ETH best queues are deep (31.5 ETH on average against 2.5 BTC). That session is one sample, not a
+study, and the recordings are not in the repository (market data is never committed), so treat
+these numbers as an illustration of the pipeline.
+| Result | BTCUSDT, 1 s buckets | ETHUSDT, 1 s buckets | Simulated |
+|---|---|---|---|
+| Δmid on OFI | β = 0.203 USDT per BTC, **R² = 0.59**, t = 102 | β = 0.0008 USDT per ETH, **R² = 0.70**, t = 129 | R² = 0.42 at 1 s (Study 5), 0.66 at 10 s |
+| Impact coefficient vs 1/depth | R² = 0.91 across depth quintiles | R² = 0.87 | R² = 0.93 (10 s) |
+| P(next move up), imbalance < 0.1 → > 0.9 | 0.09 → 0.93, monotone | 0.09 → 0.84, dips at 0.5–0.6 | monotone |
+| Next 1 s move from queue imbalance | out-of-sample R² 2.0% | 4.6% | 0.7% (next 10 s) |
+| Multi-level OFI, best 1 → 10 levels (5 s) | out-of-sample R² 0.33 → 0.38 | 0.45 → 0.47 | 0.66 → 0.88, 1 → 5 levels (10 s) |
+| Micro-price vs mid, next price change | Stoikov −0.6% MSE, naive −0.03% | **Stoikov −10.8%**, naive −4.2% | Stoikov −6.6%, naive −1.3% |
 ![Binance OFI impact](https://raw.githubusercontent.com/hossainyousuf751/microstructure-lab/main/docs/binance_btcusdt_ofi_impact.png)
-**What survives contact with a real book.** The contemporaneous OFI relation is at least as
-strong live as in the simulator at 1-second buckets, and impact is still larger when the book is
-thin. Queue imbalance predicts the next move at the extremes, and on the live book it even has a
-little out-of-sample power over the next second.
-**What doesn't.** Deeper levels add nothing out of sample, while they explained a lot in the
-simulator. On BTCUSDT the levels behind the best quote are a cent apart and mostly small resting
-orders, so their flow carries little information. Neither micro-price beats the mid by more than
-a third of a percent. Most rows sit at a one-tick spread with the price unchanged. The 688 price
-changes include 323 updates where one side of the book thinned out for an instant and the spread
-jumped past 1 USDT (up to 10.8), and a squared-error comparison over so few moves is decided by
-those points. A busier session, or a robust loss, is the next step.
+**What survives contact with a real book.** The contemporaneous OFI relation is stronger live
+than in the simulator at 1-second buckets on both instruments, and impact is larger when the book
+is thin. Queue imbalance predicts the direction of the next move at the extremes, and has a little
+out-of-sample power over the next second.
+**What depends on the book.** On ETHUSDT the Stoikov micro-price forecasts the next price change
+with 10.8% less squared error than the mid, more than in the simulator, and the edge fades with
+the horizon (4.3% after 5 moves, 0.9% after 20). On BTCUSDT it does not: −0.6%. The BTC book
+often thins out for an instant: the spread was above 1 USDT (100 ticks) in 1,131 updates over 305
+episodes, up to 17.4 USDT, while the ETH spread never exceeded 0.45 USDT. Those blowouts dominate
+a squared-error comparison and wreck the naive micro-price (its gap to the mid has an
+out-of-sample R² of −0.77 for the next second on BTC). Deeper levels add only a few points of R²
+on either book, much less than in the simulator: the levels behind the best quote are a cent apart
+and mostly small resting orders.
 ## Study 4: the market maker inside the order book
@@ -212,24 +217,30 @@ other venue's order flow explain, and whether either venue's flow or return pred
 known: a simulated venue B that copies venue A 300 ms late. They find the delay exactly
 (correlation 1.00 at +300 ms) and the direction: only B's next move is predictable from A.
-Same session as Study 3 (40 minutes; 76,296 Binance and 24,969 Bybit quote updates):
+Same session as Study 3 (2 hours each; 187,746 Binance and 76,632 Bybit quote updates for
+BTCUSDT, 127,805 and 21,344 for ETHUSDT):
 | Target venue | R² own OFI | R² other venue's OFI | R² both | Next 1 s from other's OFI (OOS) | Next 1 s from other's return (OOS) |
 |---|---|---|---|---|---|
-| Binance | 0.56 | 0.16 | 0.56 | 0.014 | 0.009 |
-| Bybit | 0.58 | 0.28 | 0.60 | 0.004 | 0.004 |
+| Binance BTCUSDT | 0.59 | 0.23 | 0.59 | 0.059 | 0.064 |
+| Bybit BTCUSDT | 0.54 | 0.33 | 0.58 | 0.002 | 0.001 |
+| Binance ETHUSDT | 0.70 | 0.15 | 0.70 | 0.036 | 0.094 |
+| Bybit ETHUSDT | 0.31 | 0.43 | 0.51 | 0.001 | 0.003 |
 ![Lead-lag](https://raw.githubusercontent.com/hossainyousuf751/microstructure-lab/main/docs/cross_venue_lead_lag.png)
-**No exploitable lead at this resolution.** Returns are most correlated at 0 ms (0.09): the
-venues move within the same 50 ms. Each venue's own order flow explains its price far better than
-the other's. Neither venue's flow or return predicts the other's next second by more than about
-1% out of sample. There is more correlation at negative lags (Bybit first, summed 0.62 against
-0.20 for Binance first), and it persisted after the recorder's timestamp bug was fixed (see
-CHANGELOG), but Bybit's price changed only 218 times, so this is a hypothesis for a longer
-recording, not a finding. One limitation is built in: timestamps are local receive times, so each
-venue's network latency to the recording machine is part of the measured lag. A lead shorter than
-the difference in latencies can't be seen from one machine.
+**Bybit tends to move first, by tens of milliseconds.** Returns are most correlated at 0 ms (0.25
+for BTCUSDT, 0.18 for ETHUSDT): most moves land on both venues within the same 50 ms. What is left
+is lopsided in the same way on both instruments. The correlation at negative lags (Bybit first)
+sums to 1.00 for BTCUSDT and 1.13 for ETHUSDT, against 0.26 at positive lags; the strongest
+off-zero lag is Bybit 150 ms ahead on both. The regressions agree: Bybit's flow and return predict
+Binance's next second (out-of-sample R² 3.6–9.4%), Binance's predict Bybit's by 0.3% or less. On
+ETHUSDT, Binance's flow explains more of Bybit's price than Bybit's own flow does (0.43 against
+0.31). The earlier 40-minute session pointed the same way; two instruments over two hours make it
+a consistent pattern, but still one session. One limitation is built in: timestamps are local
+receive times, so each venue's network latency to the recording machine is part of the measured
+lag. A lead shorter than the difference in latencies can't be told apart from a faster connection
+to Bybit, and a recording from a second location is the way to check.
 ## Roadmap

{lobkit-0.2.1 → lobkit-0.2.2}/pyproject.toml RENAMED Viewed

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 [project]
 name = "lobkit"
-version = "0.2.1"
+version = "0.2.2"
 description = "Limit order book reconstruction, order-flow analytics and market-making simulation"
 readme = "README.md"
 license = "MIT"

{lobkit-0.2.1 → lobkit-0.2.2/src/lobkit.egg-info}/PKG-INFO RENAMED Viewed

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lobkit
-Version: 0.2.1
+Version: 0.2.2
 Summary: Limit order book reconstruction, order-flow analytics and market-making simulation
 Author: Yousuf Hossain Shafeen
 License-Expression: MIT
@@ -170,35 +170,40 @@ simulated and live tables shows which effects are mechanical consequences of the
 and which require informed traders. Record at least 15 minutes; the script warns when there is
 too little data to say anything.
-Results for one session: BTCUSDT, 40 minutes on 2026-11-01 from 03:18 local time (76,296 quote
-updates, 5,813 trades). It was a quiet market: the price stayed within a 181 USDT range and the
-mid changed 688 times, about once every 3.5 seconds. That session is one sample, not a study, and
-the recording is not in the repository (market data is never committed), so treat these numbers
-as an illustration of the pipeline.
-| Result | Live, 1 s buckets | Simulated |
-|---|---|---|
-| Δmid on OFI | β = 0.221 USDT per unit, **R² = 0.56**, t = 55 | R² = 0.42 at 1 s (Study 5), 0.66 at 10 s |
-| Impact coefficient vs 1/depth | R² = 0.85 across depth quintiles | R² = 0.93 (10 s) |
-| P(next move up), imbalance < 0.1 → > 0.9 | 0.38 → 0.97, not monotone in between | monotone |
-| Next 1 s move from queue imbalance | out-of-sample R² 3.3% | 0.7% (next 10 s) |
-| Multi-level OFI, best 1 → 10 levels (5 s) | out-of-sample R² 0.50 → 0.40–0.49: no gain | 0.66 → 0.88, 1 → 5 levels (10 s) |
-| Micro-price vs mid, next price change | Stoikov −0.34% MSE, naive −0.32%: no real edge | Stoikov −6.6%, naive −1.3% |
+Results for one session, two instruments recorded side by side: BTCUSDT and ETHUSDT, 2 hours each
+on 2026-11-01 from 04:29 local time. BTCUSDT: 187,746 quote updates, 18,463 trades; the price
+stayed within a 162 USDT range and the mid changed 2,176 times, about once every 3.3 seconds.
+ETHUSDT: 127,805 quote updates, 14,136 trades; an 8 USDT range and 2,931 mid changes. Both books
+tick in cents, so a tick is 0.13 basis points of the BTC price but 0.44 of the ETH price, and the
+ETH best queues are deep (31.5 ETH on average against 2.5 BTC). That session is one sample, not a
+study, and the recordings are not in the repository (market data is never committed), so treat
+these numbers as an illustration of the pipeline.
+| Result | BTCUSDT, 1 s buckets | ETHUSDT, 1 s buckets | Simulated |
+|---|---|---|---|
+| Δmid on OFI | β = 0.203 USDT per BTC, **R² = 0.59**, t = 102 | β = 0.0008 USDT per ETH, **R² = 0.70**, t = 129 | R² = 0.42 at 1 s (Study 5), 0.66 at 10 s |
+| Impact coefficient vs 1/depth | R² = 0.91 across depth quintiles | R² = 0.87 | R² = 0.93 (10 s) |
+| P(next move up), imbalance < 0.1 → > 0.9 | 0.09 → 0.93, monotone | 0.09 → 0.84, dips at 0.5–0.6 | monotone |
+| Next 1 s move from queue imbalance | out-of-sample R² 2.0% | 4.6% | 0.7% (next 10 s) |
+| Multi-level OFI, best 1 → 10 levels (5 s) | out-of-sample R² 0.33 → 0.38 | 0.45 → 0.47 | 0.66 → 0.88, 1 → 5 levels (10 s) |
+| Micro-price vs mid, next price change | Stoikov −0.6% MSE, naive −0.03% | **Stoikov −10.8%**, naive −4.2% | Stoikov −6.6%, naive −1.3% |
 ![Binance OFI impact](https://raw.githubusercontent.com/hossainyousuf751/microstructure-lab/main/docs/binance_btcusdt_ofi_impact.png)
-**What survives contact with a real book.** The contemporaneous OFI relation is at least as
-strong live as in the simulator at 1-second buckets, and impact is still larger when the book is
-thin. Queue imbalance predicts the next move at the extremes, and on the live book it even has a
-little out-of-sample power over the next second.
-**What doesn't.** Deeper levels add nothing out of sample, while they explained a lot in the
-simulator. On BTCUSDT the levels behind the best quote are a cent apart and mostly small resting
-orders, so their flow carries little information. Neither micro-price beats the mid by more than
-a third of a percent. Most rows sit at a one-tick spread with the price unchanged. The 688 price
-changes include 323 updates where one side of the book thinned out for an instant and the spread
-jumped past 1 USDT (up to 10.8), and a squared-error comparison over so few moves is decided by
-those points. A busier session, or a robust loss, is the next step.
+**What survives contact with a real book.** The contemporaneous OFI relation is stronger live
+than in the simulator at 1-second buckets on both instruments, and impact is larger when the book
+is thin. Queue imbalance predicts the direction of the next move at the extremes, and has a little
+out-of-sample power over the next second.
+**What depends on the book.** On ETHUSDT the Stoikov micro-price forecasts the next price change
+with 10.8% less squared error than the mid, more than in the simulator, and the edge fades with
+the horizon (4.3% after 5 moves, 0.9% after 20). On BTCUSDT it does not: −0.6%. The BTC book
+often thins out for an instant: the spread was above 1 USDT (100 ticks) in 1,131 updates over 305
+episodes, up to 17.4 USDT, while the ETH spread never exceeded 0.45 USDT. Those blowouts dominate
+a squared-error comparison and wreck the naive micro-price (its gap to the mid has an
+out-of-sample R² of −0.77 for the next second on BTC). Deeper levels add only a few points of R²
+on either book, much less than in the simulator: the levels behind the best quote are a cent apart
+and mostly small resting orders.
 ## Study 4: the market maker inside the order book
@@ -252,24 +257,30 @@ other venue's order flow explain, and whether either venue's flow or return pred
 known: a simulated venue B that copies venue A 300 ms late. They find the delay exactly
 (correlation 1.00 at +300 ms) and the direction: only B's next move is predictable from A.
-Same session as Study 3 (40 minutes; 76,296 Binance and 24,969 Bybit quote updates):
+Same session as Study 3 (2 hours each; 187,746 Binance and 76,632 Bybit quote updates for
+BTCUSDT, 127,805 and 21,344 for ETHUSDT):
 | Target venue | R² own OFI | R² other venue's OFI | R² both | Next 1 s from other's OFI (OOS) | Next 1 s from other's return (OOS) |
 |---|---|---|---|---|---|
-| Binance | 0.56 | 0.16 | 0.56 | 0.014 | 0.009 |
-| Bybit | 0.58 | 0.28 | 0.60 | 0.004 | 0.004 |
+| Binance BTCUSDT | 0.59 | 0.23 | 0.59 | 0.059 | 0.064 |
+| Bybit BTCUSDT | 0.54 | 0.33 | 0.58 | 0.002 | 0.001 |
+| Binance ETHUSDT | 0.70 | 0.15 | 0.70 | 0.036 | 0.094 |
+| Bybit ETHUSDT | 0.31 | 0.43 | 0.51 | 0.001 | 0.003 |
 ![Lead-lag](https://raw.githubusercontent.com/hossainyousuf751/microstructure-lab/main/docs/cross_venue_lead_lag.png)
-**No exploitable lead at this resolution.** Returns are most correlated at 0 ms (0.09): the
-venues move within the same 50 ms. Each venue's own order flow explains its price far better than
-the other's. Neither venue's flow or return predicts the other's next second by more than about
-1% out of sample. There is more correlation at negative lags (Bybit first, summed 0.62 against
-0.20 for Binance first), and it persisted after the recorder's timestamp bug was fixed (see
-CHANGELOG), but Bybit's price changed only 218 times, so this is a hypothesis for a longer
-recording, not a finding. One limitation is built in: timestamps are local receive times, so each
-venue's network latency to the recording machine is part of the measured lag. A lead shorter than
-the difference in latencies can't be seen from one machine.
+**Bybit tends to move first, by tens of milliseconds.** Returns are most correlated at 0 ms (0.25
+for BTCUSDT, 0.18 for ETHUSDT): most moves land on both venues within the same 50 ms. What is left
+is lopsided in the same way on both instruments. The correlation at negative lags (Bybit first)
+sums to 1.00 for BTCUSDT and 1.13 for ETHUSDT, against 0.26 at positive lags; the strongest
+off-zero lag is Bybit 150 ms ahead on both. The regressions agree: Bybit's flow and return predict
+Binance's next second (out-of-sample R² 3.6–9.4%), Binance's predict Bybit's by 0.3% or less. On
+ETHUSDT, Binance's flow explains more of Bybit's price than Bybit's own flow does (0.43 against
+0.31). The earlier 40-minute session pointed the same way; two instruments over two hours make it
+a consistent pattern, but still one session. One limitation is built in: timestamps are local
+receive times, so each venue's network latency to the recording machine is part of the measured
+lag. A lead shorter than the difference in latencies can't be told apart from a faster connection
+to Bybit, and a recording from a second location is the way to check.
 ## Roadmap