The fields nobody calculates for you.

Net pressure on the book across the top 10 levels: size added on the bid and removed from the ask counts as buying pressure, the reverse as selling pressure. Top-of-book OFI misses liquidity that moves one or two ticks behind the touch; summing across levels catches it.

# for level m, between book updates n-1 and n
e(n,m) =  qb(n,m)·1[Pb(n,m) ≥ Pb(n-1,m)] − qb(n-1,m)·1[Pb(n,m) ≤ Pb(n-1,m)]
        − qa(n,m)·1[Pa(n,m) ≤ Pa(n-1,m)] + qa(n-1,m)·1[Pa(n,m) ≥ Pa(n-1,m)]

OFI_m(t)    = Σ e(n,m) over updates in the 100 ms interval
ofi_ml_l10  = Σ OFI_m for m = 1..10
ofi_ml_norm = ofi_ml_l10 / mean depth of levels 1..10 over the last 60 s

Typical uses
Short-horizon mid-price prediction, execution timing, and lead-lag between venues listing the same contract.

Volume-synchronized probability of informed trading: the share of recent volume that was one-sided, measured in equal-volume buckets instead of clock time. We know the aggressor side of every trade, so buckets use the actual buy and sell volume and skip the bulk-classification approximation used when sides are unknown.

V        = trailing 30-day average daily volume / 50     # bucket size
VB(τ), VS(τ) = aggressor buy and sell volume in bucket τ
vpin     = Σ |VB(τ) − VS(τ)| / (50 · V)   over the last 50 buckets
vpin_cdf = percentile of vpin over the trailing 30 days

Typical uses
Throttling market making when flow turns one-sided, spread and risk-limit adjustments, and conditioning other signals on toxicity regimes.

The state of the queue at the best bid and ask, rebuilt order by order: how many orders are resting, how long they have been there, how long orders tend to live, and the empirical chance that an order joining the back of the queue now is filled within one second.

queue_bid_size, queue_bid_count = resting size and order count at best bid
queue_age_p50_ms   = median age of orders resting at the touch
queue_life_p50_ms  = median lifetime (add → cancel or full fill), trailing 1 h
queue_fill_1s_bid  = joins at best bid filled within 1 s
                     / joins at best bid,  trailing 1 h

Typical uses
Queue-position models for passive execution, fill-probability estimates in backtests, and spotting when resting liquidity turns over faster than usual.

How much displayed liquidity is pulled before anyone can trade against it. Computed separately for the touch and for the top five levels, over a trailing 10-second window, so a book that looks deep but cancels on contact shows up as a high ratio.

cxl_fill_l1 = Σ cancelled size at level 1 / Σ filled size at level 1
cxl_fill_l5 = Σ cancelled size in levels 1..5 / Σ filled size in levels 1..5
# trailing 10 s window; null when no fills occurred (never infinite)

Typical uses
Discounting displayed depth in slippage models, monitoring quote quality, and flagging intervals with unusual cancel activity for review.

How fast depth comes back after it is taken. A sweep is any 100 ms interval in which aggressive volume removes at least half of the top-five depth on one side. For each sweep we measure the time until that depth is 90% restored and the share restored after one second.

D(t)            = resting size in levels 1..5 on the swept side
sweep at ts     : D(ts) ≤ 0.5 · D(ts−)
resil_refill_ms = first t − ts with D(t) ≥ 0.9 · D(ts−)
resil_refill_1s = D(ts + 1 s) / D(ts−)
resil_med_5m    = median resil_refill_ms over sweeps in the last 5 min

Typical uses
Sizing child orders and spacing them in time, estimating how much market-maker capacity is present, and detecting thin regimes before they show up in spreads.

A rolling estimate of how far the mid price moves per unit of signed volume, in the spirit of Kyle's lambda. Regressing one-second mid returns on one-second signed notional over the last five minutes gives a live impact coefficient, with its fit quality alongside so you know when to trust it.

# k = 1-second buckets, trailing 300 buckets
r(k) = (mid(k) − mid(k−1)) / mid(k−1) × 10,000      # bps
Q(k) = Σ side · price · size  in bucket k, USD millions
r(k) = λ · Q(k) + ε(k)                              # OLS, no intercept
lambda_bps_per_musd = λ,  lambda_r2 = R²,  lambda_n = buckets with trades

Typical uses
Pre-trade impact estimates, strategy capacity, and comparing liquidity for the same contract across venues in one unit.

Post-trade price drift at fixed horizons, split by aggressor side: where the mid went 100 ms, 1 s, 5 s, 30 s and 60 s after buys, and after sells. Positive values mean the aggressor was right. Grid values are volume-weighted over the last minute of trades whose horizon has already elapsed.

s  = +1 for buy, −1 for sell
markout(h) = s · (mid(t + h) − mid(t)) / mid(t) × 10,000     # bps
h ∈ {100ms, 1s, 5s, 30s, 60s}

# grid row at T: trades with t + h ≤ T and t > T − h − 60 s
# per-trade event row: ts_available = t + 60 s

Typical uses
Adverse-selection measurement for makers, transaction cost analysis, and separating informed from uninformed flow by side and size.

Aggregated behaviour of attributed participants, on venues where our coverage includes L4. We do not publish per-participant rows in this field. It reports how concentrated trading is, how much of it comes from the largest participants, and which way they are leaning.

# trailing 1-minute window; top 10 ranked by volume over the trailing 1 h
part_active          = distinct participant_id with a fill
part_top10_share     = top-10 volume / total volume
part_top10_net_flow  = Σ side · size for top-10 taker fills
part_maker_hhi       = Σ (maker share of each participant)²
part_new_share       = volume from ids first seen in the last 24 h / total

Typical uses
Crowding and concentration monitors, tracking large-participant positioning, and measuring how many market makers are active at a given moment.