Super Bowl LX

Overview

Microstructure analysis compares the trading efficiency and liquidity characteristics of Kalshi and Polymarket using raw orderbook snapshots aligned at 1-second resolution. Each metric is computed over rolling windows to show how market quality evolves during the event.

Spread Efficiency (5-min SMA)

The bid-ask spread (best ask − best bid) is smoothed with a 5-minute simple moving average (300 data points at 1s resolution). Tighter spreads indicate more competitive quoting and lower transaction costs. Comparing smoothed spreads across exchanges reveals which platform consistently offers better execution prices.

Orderbook Depth

Total depth is computed by summing all resting volume across every price level on both sides of the orderbook:

Depth = Σ(yes bid volumes) + Σ(yes ask volumes) + Σ(no bid volumes) + Σ(no ask volumes)

Higher depth means larger orders can be absorbed without significant price impact. Depth is matched from each exchange's nearest raw snapshot using binary search on timestamp.

Realized Volatility

Rolling realized volatility is the standard deviation of log returns over a 5-minute window:

r[t] = ln(mid[t] / mid[t−1]),   RVol = stdev(r[t−299 .. t])

Higher volatility indicates faster or larger price movements. Differences between exchanges may reflect varying participant activity or information flow.

Volume-Weighted Effective Spread (VWES)

VWES measures the spread that traders actually face, weighted by available volume:

VWES = Σ(spread × total_volume) / Σ(total_volume)

where total_volume = bid volume + ask volume at each snapshot. Unlike the raw spread, VWES accounts for whether tight spreads are actually backed by meaningful liquidity. Computed over a rolling 5-minute window.

Liquidity Resilience (Depth Recovery Ratio)

Measures how quickly the orderbook recovers after volume is consumed:

Ratio = (bid_vol[t] + ask_vol[t]) / (bid_vol[t−60] + ask_vol[t−60])

Values near 1.0 indicate resilient liquidity (volume replenishes quickly). Sustained dips below 1.0 suggest liquidity is being consumed faster than it is replaced. The 60-second lookback captures recovery dynamics over a 1-minute horizon.

Assumptions & Limitations

• Kalshi volumes are in contracts (shares); Polymarket volumes are in USDC. Direct volume comparisons across exchanges reflect different units.
• Orderbook depth is parsed from snapshot JSON. If a snapshot is missing orderbook data, depth is recorded as zero.
• Rolling windows use a fixed 300-point (5-min) or 60-point (1-min) lookback. Rows before the window is full produce null values.
• Realized volatility uses log returns, which can be undefined when mid-prices are zero or missing.