Chapter 53

Kalshi Prediction Markets: Regulated Event Contracts

NotebookPython 3 (ipykernel)30 cells

Kalshi Prediction Markets: Regulated Event Contracts

Chapter 4: Fundamental and Alternative Data Docker image: ml4t

Purpose

Kalshi is the first CFTC-regulated prediction market in the US, offering binary contracts on economic, market, and policy events. This notebook loads real Kalshi OHLCV data and demonstrates how to build event probability indicators for ML feature engineering and regime detection.

Learning Objectives

After completing this notebook, you will be able to:

Understand Kalshi contract structure and implied probability pricing
Load and explore real OHLCV data from the Kalshi API
Build event probability indicators for ML pipelines
Assess prediction market data quality for systematic use

Cross-References

Upstream: data/prediction_markets/download.py (fetches data)
Downstream: Chapter 8 event features, macro regime indicators
Related: 13_polymarket_prediction_markets (crypto-based alternative)

In [1]python · cell 2

python

"""Kalshi Prediction Markets - build event probability indicators from regulated binary contracts."""

import warnings

warnings.filterwarnings("ignore")

import plotly.graph_objects as go
import polars as pl
from plotly.subplots import make_subplots

from data.prediction_markets.loader import load_kalshi
from utils.paths import get_output_dir
from utils.style import COLORS

In [2]python · cell 3

python

# Production defaults - Papermill injects overrides for CI

1. Kalshi Contract Structure

Kalshi contracts are binary event contracts that settle at $0 or $1. The contract price represents the market's implied probability of the event.

Feature	Description
Regulation	CFTC-regulated (legal in US)
Settlement	USD (real dollars)
Position Limit	$25,000 per contract
Trading Hours	24/7
Min Tick	$0.01

Ticker Format

KXFED-27APR-T4.25 decodes as:

KXFED: Federal Funds Rate series
27APR: April 2027 FOMC meeting
T4.25: threshold - contract pays $1 if rate is above 4.25%

The close price is the implied probability (0–1) that the rate will exceed the threshold at that meeting.

2. Load Kalshi Data

We load pre-downloaded OHLCV data from the Kalshi API. The download script (data/prediction_markets/download.py) fetches all configured economic series and stores them in canonical OHLCV format.

In [3]python · cell 6

python

df = load_kalshi()

print(f"Loaded {len(df):,} observations across {df['symbol'].n_unique()} contracts")
print(f"Date range: {df['timestamp'].min()} to {df['timestamp'].max()}")

df.group_by("symbol").len().rename({"len": "days"}).sort("symbol")

Output

Loaded 687 observations across 9 contracts
Date range: 2025-10-07 to 2025-12-31

shape: (9, 2)
┌───────────────────┬──────┐
│ symbol            ┆ days │
│ ---               ┆ ---  │
│ str               ┆ u32  │
╞═══════════════════╪══════╡
│ KXFED-27APR-T0.25 ┆ 73   │
│ KXFED-27APR-T2.25 ┆ 73   │
│ KXFED-27APR-T3.75 ┆ 73   │
│ KXFED-27APR-T4.00 ┆ 73   │
│ KXFED-27APR-T4.25 ┆ 67   │
│ KXFED-27JAN-T1.00 ┆ 86   │
│ KXFED-27JAN-T3.00 ┆ 86   │
│ KXFED-27MAR-T0.50 ┆ 78   │
│ KXFED-27MAR-T2.75 ┆ 78   │
└───────────────────┴──────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (9, 2)

symbol	days
str	u32
"KXFED-27APR-T0.25"	73
"KXFED-27APR-T2.25"	73
"KXFED-27APR-T3.75"	73
"KXFED-27APR-T4.00"	73
"KXFED-27APR-T4.25"	67
"KXFED-27JAN-T1.00"	86
"KXFED-27JAN-T3.00"	86
"KXFED-27MAR-T0.50"	78
"KXFED-27MAR-T2.75"	78

In [4]python · cell 7

python

df.head(10)

Output

shape: (10, 8)
┌────────────┬───────────────────┬──────┬──────┬──────┬───────┬────────┬─────────────────┐
│ timestamp  ┆ symbol            ┆ open ┆ high ┆ low  ┆ close ┆ volume ┆ category        │
│ ---        ┆ ---               ┆ ---  ┆ ---  ┆ ---  ┆ ---   ┆ ---    ┆ ---             │
│ date       ┆ str               ┆ f64  ┆ f64  ┆ f64  ┆ f64   ┆ f64    ┆ str             │
╞════════════╪═══════════════════╪══════╪══════╪══════╪═══════╪════════╪═════════════════╡
│ 2025-10-20 ┆ KXFED-27APR-T0.25 ┆ 0.43 ┆ 0.93 ┆ 0.0  ┆ 0.93  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-21 ┆ KXFED-27APR-T0.25 ┆ 0.93 ┆ 0.93 ┆ 0.93 ┆ 0.93  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-22 ┆ KXFED-27APR-T0.25 ┆ 0.93 ┆ 0.93 ┆ 0.93 ┆ 0.93  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-23 ┆ KXFED-27APR-T0.25 ┆ 0.93 ┆ 0.93 ┆ 0.01 ┆ 0.92  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-24 ┆ KXFED-27APR-T0.25 ┆ 0.92 ┆ 0.92 ┆ 0.92 ┆ 0.92  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-25 ┆ KXFED-27APR-T0.25 ┆ 0.92 ┆ 0.92 ┆ 0.0  ┆ 0.92  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-26 ┆ KXFED-27APR-T0.25 ┆ 0.92 ┆ 0.92 ┆ 0.92 ┆ 0.92  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-27 ┆ KXFED-27APR-T0.25 ┆ 0.92 ┆ 0.92 ┆ 0.92 ┆ 0.92  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-28 ┆ KXFED-27APR-T0.25 ┆ 0.92 ┆ 0.92 ┆ 0.92 ┆ 0.92  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-29 ┆ KXFED-27APR-T0.25 ┆ 0.92 ┆ 0.92 ┆ 0.92 ┆ 0.92  ┆ 0.0    ┆ monetary_policy │
└────────────┴───────────────────┴──────┴──────┴──────┴───────┴────────┴─────────────────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (10, 8)

timestamp	symbol	open	high	low	close	volume	category
date	str	f64	f64	f64	f64	f64	str
2025-10-20	"KXFED-27APR-T0.25"	0.43	0.93	0.0	0.93	0.0	"monetary_policy"
2025-10-21	"KXFED-27APR-T0.25"	0.93	0.93	0.93	0.93	0.0	"monetary_policy"
2025-10-22	"KXFED-27APR-T0.25"	0.93	0.93	0.93	0.93	0.0	"monetary_policy"
2025-10-23	"KXFED-27APR-T0.25"	0.93	0.93	0.01	0.92	0.0	"monetary_policy"
2025-10-24	"KXFED-27APR-T0.25"	0.92	0.92	0.92	0.92	0.0	"monetary_policy"
2025-10-25	"KXFED-27APR-T0.25"	0.92	0.92	0.0	0.92	0.0	"monetary_policy"
2025-10-26	"KXFED-27APR-T0.25"	0.92	0.92	0.92	0.92	0.0	"monetary_policy"
2025-10-27	"KXFED-27APR-T0.25"	0.92	0.92	0.92	0.92	0.0	"monetary_policy"
2025-10-28	"KXFED-27APR-T0.25"	0.92	0.92	0.92	0.92	0.0	"monetary_policy"
2025-10-29	"KXFED-27APR-T0.25"	0.92	0.92	0.92	0.92	0.0	"monetary_policy"

Data Integrity Check

These KXFED contracts are thinly traded: only a handful of bars carry any volume, and the rest are daily carry-forward snapshots. Two problems follow. First, some fields are corrupt: a price of exactly 0 (open, low, or close) while the high still holds the prior level is an ingestion artifact, since a live contract trades inside (0, 1) and never prints a true zero the same day it closes near 0.9. Second, on a zero-volume day there is no genuine intraday range at all, so a low far below the close is stale rather than traded. The one reliable field is close (the implied probability). We therefore carry the last valid close forward within each contract and rebuild every non-traded bar as a flat snapshot at that close, keeping raw OHLC only where the bar actually traded. Left unhandled, these artifacts inflate both the close-price range (misranking a near-certain contract as "most active") and the intraday range shown in Section 6.

In [5]python · cell 9

python

traded = pl.col("volume") > 0

# Report the clearest ingestion artifacts: a price field of exactly 0 with a positive high.
zero_price_mask = (pl.col("high") > 0.0) & (
    (pl.col("open") == 0.0) | (pl.col("low") == 0.0) | (pl.col("close") == 0.0)
)
artifact_bars = (
    df.filter(zero_price_mask)
    .select("timestamp", "symbol", "open", "high", "low", "close", "volume")
    .sort("timestamp", "symbol")
)
print(f"Zero-price artifact bars (a price field of 0 with a positive high): {artifact_bars.height}")
print(f"Genuinely traded bars (volume > 0): {df.filter(traded).height} of {df.height}")

# Repair: null any close that collapsed to 0, carry the last valid close forward (and
# backward for a leading gap) within each contract, then rebuild every non-traded bar as
# a flat snapshot at that close so no spurious intraday range survives.
df = (
    df.sort("symbol", "timestamp")
    .with_columns(
        pl.when((pl.col("close") == 0.0) & (pl.col("high") > 0.0))
        .then(None)
        .otherwise(pl.col("close"))
        .alias("close")
    )
    .with_columns(pl.col("close").forward_fill().over("symbol").alias("close"))
    .with_columns(pl.col("close").backward_fill().over("symbol").alias("close"))
    .with_columns(
        pl.when(traded).then(pl.col("open")).otherwise(pl.col("close")).alias("open"),
        pl.when(traded).then(pl.col("high")).otherwise(pl.col("close")).alias("high"),
        pl.when(traded).then(pl.col("low")).otherwise(pl.col("close")).alias("low"),
    )
)

artifact_bars

Output

Zero-price artifact bars (a price field of 0 with a positive high): 35
Genuinely traded bars (volume > 0): 5 of 687

shape: (35, 7)
┌────────────┬───────────────────┬──────┬──────┬─────┬───────┬────────┐
│ timestamp  ┆ symbol            ┆ open ┆ high ┆ low ┆ close ┆ volume │
│ ---        ┆ ---               ┆ ---  ┆ ---  ┆ --- ┆ ---   ┆ ---    │
│ date       ┆ str               ┆ f64  ┆ f64  ┆ f64 ┆ f64   ┆ f64    │
╞════════════╪═══════════════════╪══════╪══════╪═════╪═══════╪════════╡
│ 2025-10-08 ┆ KXFED-27JAN-T1.00 ┆ 0.89 ┆ 0.96 ┆ 0.0 ┆ 0.96  ┆ 0.0    │
│ 2025-10-08 ┆ KXFED-27JAN-T3.00 ┆ 0.61 ┆ 0.61 ┆ 0.0 ┆ 0.51  ┆ 0.0    │
│ 2025-10-15 ┆ KXFED-27JAN-T1.00 ┆ 0.96 ┆ 0.96 ┆ 0.0 ┆ 0.0   ┆ 0.0    │
│ 2025-10-15 ┆ KXFED-27JAN-T3.00 ┆ 0.51 ┆ 0.51 ┆ 0.0 ┆ 0.0   ┆ 0.0    │
│ 2025-10-15 ┆ KXFED-27MAR-T0.50 ┆ 0.96 ┆ 0.96 ┆ 0.0 ┆ 0.0   ┆ 0.0    │
│ …          ┆ …                 ┆ …    ┆ …    ┆ …   ┆ …     ┆ …      │
│ 2025-12-02 ┆ KXFED-27APR-T4.25 ┆ 0.01 ┆ 0.01 ┆ 0.0 ┆ 0.01  ┆ 0.0    │
│ 2025-12-11 ┆ KXFED-27APR-T4.00 ┆ 0.06 ┆ 0.06 ┆ 0.0 ┆ 0.06  ┆ 0.0    │
│ 2025-12-11 ┆ KXFED-27APR-T4.25 ┆ 0.01 ┆ 0.01 ┆ 0.0 ┆ 0.01  ┆ 0.0    │
│ 2025-12-17 ┆ KXFED-27APR-T4.00 ┆ 0.06 ┆ 0.06 ┆ 0.0 ┆ 0.06  ┆ 0.0    │
│ 2025-12-17 ┆ KXFED-27APR-T4.25 ┆ 0.01 ┆ 0.01 ┆ 0.0 ┆ 0.01  ┆ 0.0    │
└────────────┴───────────────────┴──────┴──────┴─────┴───────┴────────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (35, 7)

timestamp	symbol	open	high	low	close	volume
date	str	f64	f64	f64	f64	f64
2025-10-08	"KXFED-27JAN-T1.00"	0.89	0.96	0.0	0.96	0.0
2025-10-08	"KXFED-27JAN-T3.00"	0.61	0.61	0.0	0.51	0.0
2025-10-15	"KXFED-27JAN-T1.00"	0.96	0.96	0.0	0.0	0.0
2025-10-15	"KXFED-27JAN-T3.00"	0.51	0.51	0.0	0.0	0.0
2025-10-15	"KXFED-27MAR-T0.50"	0.96	0.96	0.0	0.0	0.0
…	…	…	…	…	…	…
2025-12-02	"KXFED-27APR-T4.25"	0.01	0.01	0.0	0.01	0.0
2025-12-11	"KXFED-27APR-T4.00"	0.06	0.06	0.0	0.06	0.0
2025-12-11	"KXFED-27APR-T4.25"	0.01	0.01	0.0	0.01	0.0
2025-12-17	"KXFED-27APR-T4.00"	0.06	0.06	0.0	0.06	0.0
2025-12-17	"KXFED-27APR-T4.25"	0.01	0.01	0.0	0.01	0.0

3. Contract Universe

All contracts are from the KXFED (Federal Reserve) series, covering different rate thresholds for upcoming FOMC meetings. Each threshold represents a different market expectation about the terminal rate.

In [6]python · cell 11

python

contracts = (
    df.sort("timestamp")
    .group_by("symbol")
    .agg(
        pl.col("close").last().alias("latest_prob"),
        pl.col("close").first().alias("initial_prob"),
        pl.col("volume").sum().alias("total_volume"),
        pl.col("timestamp").min().alias("first_date"),
        pl.col("timestamp").max().alias("last_date"),
        pl.len().alias("observations"),
    )
    .sort("symbol")
)
contracts

Output

shape: (9, 7)
┌──────────────┬─────────────┬──────────────┬──────────────┬────────────┬────────────┬─────────────┐
│ symbol       ┆ latest_prob ┆ initial_prob ┆ total_volume ┆ first_date ┆ last_date  ┆ observation │
│ ---          ┆ ---         ┆ ---          ┆ ---          ┆ ---        ┆ ---        ┆ s           │
│ str          ┆ f64         ┆ f64          ┆ f64          ┆ date       ┆ date       ┆ ---         │
│              ┆             ┆              ┆              ┆            ┆            ┆ u32         │
╞══════════════╪═════════════╪══════════════╪══════════════╪════════════╪════════════╪═════════════╡
│ KXFED-27APR- ┆ 0.92        ┆ 0.93         ┆ 0.0          ┆ 2025-10-20 ┆ 2025-12-31 ┆ 73          │
│ T0.25        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27APR- ┆ 0.74        ┆ 0.75         ┆ 0.0          ┆ 2025-10-20 ┆ 2025-12-31 ┆ 73          │
│ T2.25        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27APR- ┆ 0.14        ┆ 0.14         ┆ 2.0          ┆ 2025-10-20 ┆ 2025-12-31 ┆ 73          │
│ T3.75        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27APR- ┆ 0.06        ┆ 0.06         ┆ 0.0          ┆ 2025-10-20 ┆ 2025-12-31 ┆ 73          │
│ T4.00        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27APR- ┆ 0.01        ┆ 0.01         ┆ 0.0          ┆ 2025-10-20 ┆ 2025-12-31 ┆ 67          │
│ T4.25        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27JAN- ┆ 0.95        ┆ 0.89         ┆ 0.0          ┆ 2025-10-07 ┆ 2025-12-31 ┆ 86          │
│ T1.00        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27JAN- ┆ 0.39        ┆ 0.61         ┆ 312.0        ┆ 2025-10-07 ┆ 2025-12-31 ┆ 86          │
│ T3.00        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27MAR- ┆ 0.95        ┆ 0.96         ┆ 1.0          ┆ 2025-10-14 ┆ 2025-12-31 ┆ 78          │
│ T0.50        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27MAR- ┆ 0.51        ┆ 0.59         ┆ 0.0          ┆ 2025-10-14 ┆ 2025-12-31 ┆ 78          │
│ T2.75        ┆             ┆              ┆              ┆            ┆            ┆             │
└──────────────┴─────────────┴──────────────┴──────────────┴────────────┴────────────┴─────────────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (9, 7)

symbol	latest_prob	initial_prob	total_volume	first_date	last_date	observations
str	f64	f64	f64	date	date	u32
"KXFED-27APR-T0.25"	0.92	0.93	0.0	2025-10-20	2025-12-31	73
"KXFED-27APR-T2.25"	0.74	0.75	0.0	2025-10-20	2025-12-31	73
"KXFED-27APR-T3.75"	0.14	0.14	2.0	2025-10-20	2025-12-31	73
"KXFED-27APR-T4.00"	0.06	0.06	0.0	2025-10-20	2025-12-31	73
"KXFED-27APR-T4.25"	0.01	0.01	0.0	2025-10-20	2025-12-31	67
"KXFED-27JAN-T1.00"	0.95	0.89	0.0	2025-10-07	2025-12-31	86
"KXFED-27JAN-T3.00"	0.39	0.61	312.0	2025-10-07	2025-12-31	86
"KXFED-27MAR-T0.50"	0.95	0.96	1.0	2025-10-14	2025-12-31	78
"KXFED-27MAR-T2.75"	0.51	0.59	0.0	2025-10-14	2025-12-31	78

4. Probability Evolution

The implied probability for each contract evolves over time as the market incorporates new information about Fed policy. Higher thresholds have lower probabilities (less likely the rate exceeds a high level).

In [7]python · cell 13

python

# Volume is near zero across these contracts, so "most active" means the widest range
# in implied probability (on the cleaned data), which flags the genuine battleground
# thresholds rather than a data glitch.
price_range = (
    df.group_by("symbol")
    .agg((pl.col("close").max() - pl.col("close").min()).alias("range"))
    .sort("range", descending=True)
)

top_contracts = price_range.head(3)["symbol"].to_list()

fig = go.Figure()
palette = [COLORS["blue"], COLORS["amber"], COLORS["copper"]]

for sym, color in zip(top_contracts, palette, strict=False):
    data = df.filter(pl.col("symbol") == sym).sort("timestamp").to_pandas()
    fig.add_trace(
        go.Scatter(
            x=data["timestamp"],
            y=data["close"],
            mode="lines",
            name=sym,
            line=dict(color=color, width=2),
        )
    )

fig.update_layout(
    title="Battleground Fed-rate thresholds hover near even odds while far thresholds stay pinned",
    xaxis_title="Date",
    yaxis_title="Implied Probability",
    yaxis=dict(tickformat=".0%", range=[0, 1.05]),
    height=400,
    legend=dict(orientation="h", yanchor="bottom", y=1.02, xanchor="center", x=0.5),
)

fig.show()

Output

The contracts with rate thresholds near the current rate show the most price movement - these are the "battleground" levels where the market is genuinely uncertain. Contracts far from the current rate trade near 0 or 1 with little movement.

5. Multi-Threshold View

Looking at all thresholds for a single meeting gives a snapshot of the market's full probability distribution over rate outcomes.

In [8]python · cell 16

python

# Group by meeting date prefix
meetings = {}
for sym in df["symbol"].unique().to_list():
    # KXFED-27APR-T4.25 → 27APR
    parts = sym.split("-")
    if len(parts) >= 3:
        meeting = parts[1]
        meetings.setdefault(meeting, []).append(sym)

fig = make_subplots(
    rows=len(meetings),
    cols=1,
    shared_xaxes=True,
    subplot_titles=[f"Meeting: {m}" for m in sorted(meetings.keys())],
    vertical_spacing=0.08,
)

for i, (meeting, symbols) in enumerate(sorted(meetings.items()), 1):
    for sym in sorted(symbols):
        data = df.filter(pl.col("symbol") == sym).sort("timestamp").to_pandas()
        threshold = sym.split("-T")[-1] if "-T" in sym else sym
        fig.add_trace(
            go.Scatter(
                x=data["timestamp"],
                y=data["close"],
                mode="lines",
                name=f"T{threshold}",
                showlegend=(i == 1),
            ),
            row=i,
            col=1,
        )
    fig.update_yaxes(tickformat=".0%", range=[0, 1.05], row=i, col=1)

fig.update_layout(
    height=250 * len(meetings),
    title_text="Each FOMC meeting's threshold ladder maps the market's full rate distribution",
)

fig.show()

Output

6. Day-over-Day Probability Dynamics

These markets trade on only a handful of days in the sample; the rest are zero-volume carry-forward snapshots with no genuine intraday range. Intraday high/low is therefore not a meaningful signal here. What moves is the day-over-day implied probability (the close path), so we summarize each contract by how many days it actually traded and by the volatility of its daily probability changes. Contracts whose threshold sits near the expected rate show the largest daily moves.

In [9]python · cell 18

python

df_enriched = df.sort("symbol", "timestamp").with_columns(
    pl.col("close").diff().over("symbol").alias("prob_change"),
)

activity = (
    df_enriched.group_by("symbol")
    .agg(
        (pl.col("volume") > 0).sum().alias("traded_days"),
        pl.col("volume").sum().round(1).alias("total_volume"),
        pl.col("prob_change").std().round(4).alias("daily_prob_vol"),
        pl.col("prob_change").abs().max().round(4).alias("max_daily_move"),
    )
    .sort("daily_prob_vol", descending=True, nulls_last=True)
)

activity

Output

shape: (9, 5)
┌───────────────────┬─────────────┬──────────────┬────────────────┬────────────────┐
│ symbol            ┆ traded_days ┆ total_volume ┆ daily_prob_vol ┆ max_daily_move │
│ ---               ┆ ---         ┆ ---          ┆ ---            ┆ ---            │
│ str               ┆ u32         ┆ f64          ┆ f64            ┆ f64            │
╞═══════════════════╪═════════════╪══════════════╪════════════════╪════════════════╡
│ KXFED-27JAN-T3.00 ┆ 3           ┆ 312.0        ┆ 0.0157         ┆ 0.1            │
│ KXFED-27JAN-T1.00 ┆ 0           ┆ 0.0          ┆ 0.0083         ┆ 0.07           │
│ KXFED-27MAR-T2.75 ┆ 0           ┆ 0.0          ┆ 0.0074         ┆ 0.04           │
│ KXFED-27APR-T3.75 ┆ 1           ┆ 2.0          ┆ 0.0044         ┆ 0.02           │
│ KXFED-27APR-T2.25 ┆ 0           ┆ 0.0          ┆ 0.0039         ┆ 0.02           │
│ KXFED-27APR-T4.00 ┆ 0           ┆ 0.0          ┆ 0.0038         ┆ 0.02           │
│ KXFED-27MAR-T0.50 ┆ 1           ┆ 1.0          ┆ 0.0034         ┆ 0.02           │
│ KXFED-27APR-T0.25 ┆ 0           ┆ 0.0          ┆ 0.0031         ┆ 0.02           │
│ KXFED-27APR-T4.25 ┆ 0           ┆ 0.0          ┆ 0.0018         ┆ 0.01           │
└───────────────────┴─────────────┴──────────────┴────────────────┴────────────────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (9, 5)

symbol	traded_days	total_volume	daily_prob_vol	max_daily_move
str	u32	f64	f64	f64
"KXFED-27JAN-T3.00"	3	312.0	0.0157	0.1
"KXFED-27JAN-T1.00"	0	0.0	0.0083	0.07
"KXFED-27MAR-T2.75"	0	0.0	0.0074	0.04
"KXFED-27APR-T3.75"	1	2.0	0.0044	0.02
"KXFED-27APR-T2.25"	0	0.0	0.0039	0.02
"KXFED-27APR-T4.00"	0	0.0	0.0038	0.02
"KXFED-27MAR-T0.50"	1	1.0	0.0034	0.02
"KXFED-27APR-T0.25"	0	0.0	0.0031	0.02
"KXFED-27APR-T4.25"	0	0.0	0.0018	0.01

In [10]python · cell 19

python

ad = activity.to_pandas()
fig = go.Figure(go.Bar(x=ad["symbol"], y=ad["daily_prob_vol"], marker_color=COLORS["blue"]))
fig.update_layout(
    title="Near-the-money Fed thresholds carry the most daily probability movement",
    xaxis_title="Contract",
    yaxis_title="Std. of daily probability change",
    height=400,
)
fig.update_xaxes(tickangle=-45)
fig.show()

Output

7. Event Indicators for ML

Transform Kalshi probabilities into ML-ready features. Since close is already the implied probability, we derive momentum, volatility, and regime indicators directly.

In [11]python · cell 21

python

LOOKBACK = 5
VOL_WINDOW = 10

kalshi_features = df.sort("symbol", "timestamp").with_columns(
    (pl.col("close") - pl.col("close").shift(LOOKBACK).over("symbol")).alias("prob_momentum"),
    pl.col("close").diff().rolling_std(VOL_WINDOW).over("symbol").alias("prob_volatility"),
    pl.when(pl.col("close").rolling_std(VOL_WINDOW).over("symbol") > 0)
    .then(
        (pl.col("close") - pl.col("close").rolling_mean(VOL_WINDOW).over("symbol"))
        / pl.col("close").rolling_std(VOL_WINDOW).over("symbol")
    )
    .otherwise(0.0)
    .alias("prob_zscore"),
    ((pl.col("close") > 0.8) | (pl.col("close") < 0.2)).cast(pl.Int8).alias("high_confidence"),
    (pl.col("high") - pl.col("low")).alias("uncertainty"),
)

print(f"Feature matrix: {kalshi_features.shape}")
kalshi_features.select(
    "timestamp", "symbol", "close", "prob_momentum", "prob_volatility", "high_confidence"
).head(10)

Output

Feature matrix: (687, 13)

shape: (10, 6)
┌────────────┬───────────────────┬───────┬───────────────┬─────────────────┬─────────────────┐
│ timestamp  ┆ symbol            ┆ close ┆ prob_momentum ┆ prob_volatility ┆ high_confidence │
│ ---        ┆ ---               ┆ ---   ┆ ---           ┆ ---             ┆ ---             │
│ date       ┆ str               ┆ f64   ┆ f64           ┆ f64             ┆ i8              │
╞════════════╪═══════════════════╪═══════╪═══════════════╪═════════════════╪═════════════════╡
│ 2025-10-20 ┆ KXFED-27APR-T0.25 ┆ 0.93  ┆ null          ┆ null            ┆ 1               │
│ 2025-10-21 ┆ KXFED-27APR-T0.25 ┆ 0.93  ┆ null          ┆ null            ┆ 1               │
│ 2025-10-22 ┆ KXFED-27APR-T0.25 ┆ 0.93  ┆ null          ┆ null            ┆ 1               │
│ 2025-10-23 ┆ KXFED-27APR-T0.25 ┆ 0.92  ┆ null          ┆ null            ┆ 1               │
│ 2025-10-24 ┆ KXFED-27APR-T0.25 ┆ 0.92  ┆ null          ┆ null            ┆ 1               │
│ 2025-10-25 ┆ KXFED-27APR-T0.25 ┆ 0.92  ┆ -0.01         ┆ null            ┆ 1               │
│ 2025-10-26 ┆ KXFED-27APR-T0.25 ┆ 0.92  ┆ -0.01         ┆ null            ┆ 1               │
│ 2025-10-27 ┆ KXFED-27APR-T0.25 ┆ 0.92  ┆ -0.01         ┆ null            ┆ 1               │
│ 2025-10-28 ┆ KXFED-27APR-T0.25 ┆ 0.92  ┆ 0.0           ┆ null            ┆ 1               │
│ 2025-10-29 ┆ KXFED-27APR-T0.25 ┆ 0.92  ┆ 0.0           ┆ null            ┆ 1               │
└────────────┴───────────────────┴───────┴───────────────┴─────────────────┴─────────────────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (10, 6)

timestamp	symbol	close	prob_momentum	prob_volatility	high_confidence
date	str	f64	f64	f64	i8
2025-10-20	"KXFED-27APR-T0.25"	0.93	null	null	1
2025-10-21	"KXFED-27APR-T0.25"	0.93	null	null	1
2025-10-22	"KXFED-27APR-T0.25"	0.93	null	null	1
2025-10-23	"KXFED-27APR-T0.25"	0.92	null	null	1
2025-10-24	"KXFED-27APR-T0.25"	0.92	null	null	1
2025-10-25	"KXFED-27APR-T0.25"	0.92	-0.01	null	1
2025-10-26	"KXFED-27APR-T0.25"	0.92	-0.01	null	1
2025-10-27	"KXFED-27APR-T0.25"	0.92	-0.01	null	1
2025-10-28	"KXFED-27APR-T0.25"	0.92	0.0	null	1
2025-10-29	"KXFED-27APR-T0.25"	0.92	0.0	null	1

In [12]python · cell 22

python

# Feature distributions for the most active contract
active_sym = top_contracts[0]
active_features = kalshi_features.filter(
    (pl.col("symbol") == active_sym) & pl.col("prob_momentum").is_not_null()
)

fig = make_subplots(
    rows=1,
    cols=2,
    subplot_titles=("Probability Momentum", "Probability Z-Score"),
)

fig.add_trace(
    go.Histogram(
        x=active_features["prob_momentum"].to_list(),
        nbinsx=30,
        name="Momentum",
        marker_color=COLORS["blue"],
    ),
    row=1,
    col=1,
)

fig.add_trace(
    go.Histogram(
        x=active_features["prob_zscore"].to_list(),
        nbinsx=30,
        name="Z-Score",
        marker_color=COLORS["slate"],
    ),
    row=1,
    col=2,
)

fig.update_layout(
    title=f"Probability momentum and z-score center near zero for {active_sym}",
    height=350,
    showlegend=False,
)

fig.show()

Output

8. Data Quality Assessment

Two checks matter for prediction-market data: whether any bars were corrupt (caught and repaired at load), and how much genuine price variation each contract carries.

In [13]python · cell 24

python

print(f"Zero-price artifact bars caught and repaired at load: {artifact_bars.height}")
artifact_bars

Output

Zero-price artifact bars caught and repaired at load: 35

shape: (35, 7)
┌────────────┬───────────────────┬──────┬──────┬─────┬───────┬────────┐
│ timestamp  ┆ symbol            ┆ open ┆ high ┆ low ┆ close ┆ volume │
│ ---        ┆ ---               ┆ ---  ┆ ---  ┆ --- ┆ ---   ┆ ---    │
│ date       ┆ str               ┆ f64  ┆ f64  ┆ f64 ┆ f64   ┆ f64    │
╞════════════╪═══════════════════╪══════╪══════╪═════╪═══════╪════════╡
│ 2025-10-08 ┆ KXFED-27JAN-T1.00 ┆ 0.89 ┆ 0.96 ┆ 0.0 ┆ 0.96  ┆ 0.0    │
│ 2025-10-08 ┆ KXFED-27JAN-T3.00 ┆ 0.61 ┆ 0.61 ┆ 0.0 ┆ 0.51  ┆ 0.0    │
│ 2025-10-15 ┆ KXFED-27JAN-T1.00 ┆ 0.96 ┆ 0.96 ┆ 0.0 ┆ 0.0   ┆ 0.0    │
│ 2025-10-15 ┆ KXFED-27JAN-T3.00 ┆ 0.51 ┆ 0.51 ┆ 0.0 ┆ 0.0   ┆ 0.0    │
│ 2025-10-15 ┆ KXFED-27MAR-T0.50 ┆ 0.96 ┆ 0.96 ┆ 0.0 ┆ 0.0   ┆ 0.0    │
│ …          ┆ …                 ┆ …    ┆ …    ┆ …   ┆ …     ┆ …      │
│ 2025-12-02 ┆ KXFED-27APR-T4.25 ┆ 0.01 ┆ 0.01 ┆ 0.0 ┆ 0.01  ┆ 0.0    │
│ 2025-12-11 ┆ KXFED-27APR-T4.00 ┆ 0.06 ┆ 0.06 ┆ 0.0 ┆ 0.06  ┆ 0.0    │
│ 2025-12-11 ┆ KXFED-27APR-T4.25 ┆ 0.01 ┆ 0.01 ┆ 0.0 ┆ 0.01  ┆ 0.0    │
│ 2025-12-17 ┆ KXFED-27APR-T4.00 ┆ 0.06 ┆ 0.06 ┆ 0.0 ┆ 0.06  ┆ 0.0    │
│ 2025-12-17 ┆ KXFED-27APR-T4.25 ┆ 0.01 ┆ 0.01 ┆ 0.0 ┆ 0.01  ┆ 0.0    │
└────────────┴───────────────────┴──────┴──────┴─────┴───────┴────────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (35, 7)

timestamp	symbol	open	high	low	close	volume
date	str	f64	f64	f64	f64	f64
2025-10-08	"KXFED-27JAN-T1.00"	0.89	0.96	0.0	0.96	0.0
2025-10-08	"KXFED-27JAN-T3.00"	0.61	0.61	0.0	0.51	0.0
2025-10-15	"KXFED-27JAN-T1.00"	0.96	0.96	0.0	0.0	0.0
2025-10-15	"KXFED-27JAN-T3.00"	0.51	0.51	0.0	0.0	0.0
2025-10-15	"KXFED-27MAR-T0.50"	0.96	0.96	0.0	0.0	0.0
…	…	…	…	…	…	…
2025-12-02	"KXFED-27APR-T4.25"	0.01	0.01	0.0	0.01	0.0
2025-12-11	"KXFED-27APR-T4.00"	0.06	0.06	0.0	0.06	0.0
2025-12-11	"KXFED-27APR-T4.25"	0.01	0.01	0.0	0.01	0.0
2025-12-17	"KXFED-27APR-T4.00"	0.06	0.06	0.0	0.06	0.0
2025-12-17	"KXFED-27APR-T4.25"	0.01	0.01	0.0	0.01	0.0

In [14]python · cell 25

python

quality_df = (
    df.group_by("symbol")
    .agg(
        pl.len().alias("observations"),
        pl.col("volume").mean().round(3).alias("avg_volume"),
        (pl.col("close").max() - pl.col("close").min()).round(3).alias("price_range"),
    )
    .sort("price_range", descending=True)
)
quality_df

Output

shape: (9, 4)
┌───────────────────┬──────────────┬────────────┬─────────────┐
│ symbol            ┆ observations ┆ avg_volume ┆ price_range │
│ ---               ┆ ---          ┆ ---        ┆ ---         │
│ str               ┆ u32          ┆ f64        ┆ f64         │
╞═══════════════════╪══════════════╪════════════╪═════════════╡
│ KXFED-27JAN-T3.00 ┆ 86           ┆ 3.628      ┆ 0.23        │
│ KXFED-27JAN-T1.00 ┆ 86           ┆ 0.0        ┆ 0.08        │
│ KXFED-27MAR-T2.75 ┆ 78           ┆ 0.0        ┆ 0.08        │
│ KXFED-27APR-T3.75 ┆ 73           ┆ 0.027      ┆ 0.04        │
│ KXFED-27APR-T2.25 ┆ 73           ┆ 0.0        ┆ 0.03        │
│ KXFED-27APR-T4.00 ┆ 73           ┆ 0.0        ┆ 0.03        │
│ KXFED-27APR-T0.25 ┆ 73           ┆ 0.0        ┆ 0.02        │
│ KXFED-27MAR-T0.50 ┆ 78           ┆ 0.013      ┆ 0.02        │
│ KXFED-27APR-T4.25 ┆ 67           ┆ 0.0        ┆ 0.01        │
└───────────────────┴──────────────┴────────────┴─────────────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (9, 4)

symbol	observations	avg_volume	price_range
str	u32	f64	f64
"KXFED-27JAN-T3.00"	86	3.628	0.23
"KXFED-27JAN-T1.00"	86	0.0	0.08
"KXFED-27MAR-T2.75"	78	0.0	0.08
"KXFED-27APR-T3.75"	73	0.027	0.04
"KXFED-27APR-T2.25"	73	0.0	0.03
"KXFED-27APR-T4.00"	73	0.0	0.03
"KXFED-27APR-T0.25"	73	0.0	0.02
"KXFED-27MAR-T0.50"	78	0.013	0.02
"KXFED-27APR-T4.25"	67	0.0	0.01

In [15]python · cell 26

python

qd = quality_df.to_pandas()
fig = go.Figure(go.Bar(x=qd["symbol"], y=qd["price_range"], marker_color=COLORS["blue"]))
fig.update_layout(
    title="After repair, close-price range concentrates in near-the-money thresholds",
    xaxis_title="Contract",
    yaxis_title="Close price range",
    height=400,
)
fig.update_xaxes(tickangle=-45)
fig.show()

Output

After repairing the corrupt bars, price variation reflects real market movement. Volume is near zero across the universe, so the range in implied probability, not turnover, is the useful activity signal: contracts whose thresholds sit near the expected rate move the most and carry the richest information for ML features.

9. Save Enriched Data

In [16]python · cell 29

python

output_dir = get_output_dir(4, "kalshi")
output_dir.mkdir(parents=True, exist_ok=True)

output_file = output_dir / "kalshi_features.parquet"
kalshi_features.write_parquet(output_file)

print(f"Saved {len(kalshi_features)} observations to {output_file}")

Output

Saved 687 observations to /tmp/ml4t-test-output/ch04_kalshi/kalshi_features.parquet

Key Takeaways

Implied probability = close price: Kalshi contract prices directly encode the market's probability estimate for the event, no transformation needed
Threshold structure: Multiple contracts per meeting create a full probability distribution over rate outcomes - richer than a single forecast
Feature engineering: Momentum, volatility, and z-score of probability paths provide regime-detection signals for rate-sensitive strategies
Liquidity caveat: Economic event contracts are still early-stage; volume is thin compared to traditional derivatives markets
Screen for corrupt bars first: thin, carry-forward markets are prone to ingestion artifacts (a positive high with a zero close). Detect and repair them before ranking or feature engineering, or a single bad bar distorts both.

Next: See 13_polymarket_prediction_markets for the higher-liquidity crypto-based alternative and cross-platform comparison.

Chapter 53

Kalshi Prediction Markets: Regulated Event Contracts

NotebookPython 3 (ipykernel)30 cells

Kalshi Prediction Markets: Regulated Event Contracts

Chapter 4: Fundamental and Alternative Data Docker image: ml4t

Purpose

Learning Objectives

After completing this notebook, you will be able to:

Understand Kalshi contract structure and implied probability pricing
Load and explore real OHLCV data from the Kalshi API
Build event probability indicators for ML pipelines
Assess prediction market data quality for systematic use

Cross-References

Upstream: data/prediction_markets/download.py (fetches data)
Downstream: Chapter 8 event features, macro regime indicators
Related: 13_polymarket_prediction_markets (crypto-based alternative)

In [1]python · cell 2

python

"""Kalshi Prediction Markets - build event probability indicators from regulated binary contracts."""

import warnings

warnings.filterwarnings("ignore")

import plotly.graph_objects as go
import polars as pl
from plotly.subplots import make_subplots

from data.prediction_markets.loader import load_kalshi
from utils.paths import get_output_dir
from utils.style import COLORS

In [2]python · cell 3

python

# Production defaults - Papermill injects overrides for CI

1. Kalshi Contract Structure

Kalshi contracts are binary event contracts that settle at $0 or $1. The contract price represents the market's implied probability of the event.

Feature	Description
Regulation	CFTC-regulated (legal in US)
Settlement	USD (real dollars)
Position Limit	$25,000 per contract
Trading Hours	24/7
Min Tick	$0.01

Ticker Format

KXFED-27APR-T4.25 decodes as:

KXFED: Federal Funds Rate series
27APR: April 2027 FOMC meeting
T4.25: threshold - contract pays $1 if rate is above 4.25%

The close price is the implied probability (0–1) that the rate will exceed the threshold at that meeting.

2. Load Kalshi Data

We load pre-downloaded OHLCV data from the Kalshi API. The download script (data/prediction_markets/download.py) fetches all configured economic series and stores them in canonical OHLCV format.

In [3]python · cell 6

python

df = load_kalshi()

print(f"Loaded {len(df):,} observations across {df['symbol'].n_unique()} contracts")
print(f"Date range: {df['timestamp'].min()} to {df['timestamp'].max()}")

df.group_by("symbol").len().rename({"len": "days"}).sort("symbol")

Output

Loaded 687 observations across 9 contracts
Date range: 2025-10-07 to 2025-12-31

shape: (9, 2)
┌───────────────────┬──────┐
│ symbol            ┆ days │
│ ---               ┆ ---  │
│ str               ┆ u32  │
╞═══════════════════╪══════╡
│ KXFED-27APR-T0.25 ┆ 73   │
│ KXFED-27APR-T2.25 ┆ 73   │
│ KXFED-27APR-T3.75 ┆ 73   │
│ KXFED-27APR-T4.00 ┆ 73   │
│ KXFED-27APR-T4.25 ┆ 67   │
│ KXFED-27JAN-T1.00 ┆ 86   │
│ KXFED-27JAN-T3.00 ┆ 86   │
│ KXFED-27MAR-T0.50 ┆ 78   │
│ KXFED-27MAR-T2.75 ┆ 78   │
└───────────────────┴──────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (9, 2)

symbol	days
str	u32
"KXFED-27APR-T0.25"	73
"KXFED-27APR-T2.25"	73
"KXFED-27APR-T3.75"	73
"KXFED-27APR-T4.00"	73
"KXFED-27APR-T4.25"	67
"KXFED-27JAN-T1.00"	86
"KXFED-27JAN-T3.00"	86
"KXFED-27MAR-T0.50"	78
"KXFED-27MAR-T2.75"	78

In [4]python · cell 7

python

df.head(10)

Output

shape: (10, 8)
┌────────────┬───────────────────┬──────┬──────┬──────┬───────┬────────┬─────────────────┐
│ timestamp  ┆ symbol            ┆ open ┆ high ┆ low  ┆ close ┆ volume ┆ category        │
│ ---        ┆ ---               ┆ ---  ┆ ---  ┆ ---  ┆ ---   ┆ ---    ┆ ---             │
│ date       ┆ str               ┆ f64  ┆ f64  ┆ f64  ┆ f64   ┆ f64    ┆ str             │
╞════════════╪═══════════════════╪══════╪══════╪══════╪═══════╪════════╪═════════════════╡
│ 2025-10-20 ┆ KXFED-27APR-T0.25 ┆ 0.43 ┆ 0.93 ┆ 0.0  ┆ 0.93  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-21 ┆ KXFED-27APR-T0.25 ┆ 0.93 ┆ 0.93 ┆ 0.93 ┆ 0.93  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-22 ┆ KXFED-27APR-T0.25 ┆ 0.93 ┆ 0.93 ┆ 0.93 ┆ 0.93  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-23 ┆ KXFED-27APR-T0.25 ┆ 0.93 ┆ 0.93 ┆ 0.01 ┆ 0.92  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-24 ┆ KXFED-27APR-T0.25 ┆ 0.92 ┆ 0.92 ┆ 0.92 ┆ 0.92  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-25 ┆ KXFED-27APR-T0.25 ┆ 0.92 ┆ 0.92 ┆ 0.0  ┆ 0.92  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-26 ┆ KXFED-27APR-T0.25 ┆ 0.92 ┆ 0.92 ┆ 0.92 ┆ 0.92  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-27 ┆ KXFED-27APR-T0.25 ┆ 0.92 ┆ 0.92 ┆ 0.92 ┆ 0.92  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-28 ┆ KXFED-27APR-T0.25 ┆ 0.92 ┆ 0.92 ┆ 0.92 ┆ 0.92  ┆ 0.0    ┆ monetary_policy │
│ 2025-10-29 ┆ KXFED-27APR-T0.25 ┆ 0.92 ┆ 0.92 ┆ 0.92 ┆ 0.92  ┆ 0.0    ┆ monetary_policy │
└────────────┴───────────────────┴──────┴──────┴──────┴───────┴────────┴─────────────────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (10, 8)

timestamp	symbol	open	high	low	close	volume	category
date	str	f64	f64	f64	f64	f64	str
2025-10-20	"KXFED-27APR-T0.25"	0.43	0.93	0.0	0.93	0.0	"monetary_policy"
2025-10-21	"KXFED-27APR-T0.25"	0.93	0.93	0.93	0.93	0.0	"monetary_policy"
2025-10-22	"KXFED-27APR-T0.25"	0.93	0.93	0.93	0.93	0.0	"monetary_policy"
2025-10-23	"KXFED-27APR-T0.25"	0.93	0.93	0.01	0.92	0.0	"monetary_policy"
2025-10-24	"KXFED-27APR-T0.25"	0.92	0.92	0.92	0.92	0.0	"monetary_policy"
2025-10-25	"KXFED-27APR-T0.25"	0.92	0.92	0.0	0.92	0.0	"monetary_policy"
2025-10-26	"KXFED-27APR-T0.25"	0.92	0.92	0.92	0.92	0.0	"monetary_policy"
2025-10-27	"KXFED-27APR-T0.25"	0.92	0.92	0.92	0.92	0.0	"monetary_policy"
2025-10-28	"KXFED-27APR-T0.25"	0.92	0.92	0.92	0.92	0.0	"monetary_policy"
2025-10-29	"KXFED-27APR-T0.25"	0.92	0.92	0.92	0.92	0.0	"monetary_policy"

Data Integrity Check

In [5]python · cell 9

python

traded = pl.col("volume") > 0

# Report the clearest ingestion artifacts: a price field of exactly 0 with a positive high.
zero_price_mask = (pl.col("high") > 0.0) & (
    (pl.col("open") == 0.0) | (pl.col("low") == 0.0) | (pl.col("close") == 0.0)
)
artifact_bars = (
    df.filter(zero_price_mask)
    .select("timestamp", "symbol", "open", "high", "low", "close", "volume")
    .sort("timestamp", "symbol")
)
print(f"Zero-price artifact bars (a price field of 0 with a positive high): {artifact_bars.height}")
print(f"Genuinely traded bars (volume > 0): {df.filter(traded).height} of {df.height}")

# Repair: null any close that collapsed to 0, carry the last valid close forward (and
# backward for a leading gap) within each contract, then rebuild every non-traded bar as
# a flat snapshot at that close so no spurious intraday range survives.
df = (
    df.sort("symbol", "timestamp")
    .with_columns(
        pl.when((pl.col("close") == 0.0) & (pl.col("high") > 0.0))
        .then(None)
        .otherwise(pl.col("close"))
        .alias("close")
    )
    .with_columns(pl.col("close").forward_fill().over("symbol").alias("close"))
    .with_columns(pl.col("close").backward_fill().over("symbol").alias("close"))
    .with_columns(
        pl.when(traded).then(pl.col("open")).otherwise(pl.col("close")).alias("open"),
        pl.when(traded).then(pl.col("high")).otherwise(pl.col("close")).alias("high"),
        pl.when(traded).then(pl.col("low")).otherwise(pl.col("close")).alias("low"),
    )
)

artifact_bars

Output

Zero-price artifact bars (a price field of 0 with a positive high): 35
Genuinely traded bars (volume > 0): 5 of 687

shape: (35, 7)
┌────────────┬───────────────────┬──────┬──────┬─────┬───────┬────────┐
│ timestamp  ┆ symbol            ┆ open ┆ high ┆ low ┆ close ┆ volume │
│ ---        ┆ ---               ┆ ---  ┆ ---  ┆ --- ┆ ---   ┆ ---    │
│ date       ┆ str               ┆ f64  ┆ f64  ┆ f64 ┆ f64   ┆ f64    │
╞════════════╪═══════════════════╪══════╪══════╪═════╪═══════╪════════╡
│ 2025-10-08 ┆ KXFED-27JAN-T1.00 ┆ 0.89 ┆ 0.96 ┆ 0.0 ┆ 0.96  ┆ 0.0    │
│ 2025-10-08 ┆ KXFED-27JAN-T3.00 ┆ 0.61 ┆ 0.61 ┆ 0.0 ┆ 0.51  ┆ 0.0    │
│ 2025-10-15 ┆ KXFED-27JAN-T1.00 ┆ 0.96 ┆ 0.96 ┆ 0.0 ┆ 0.0   ┆ 0.0    │
│ 2025-10-15 ┆ KXFED-27JAN-T3.00 ┆ 0.51 ┆ 0.51 ┆ 0.0 ┆ 0.0   ┆ 0.0    │
│ 2025-10-15 ┆ KXFED-27MAR-T0.50 ┆ 0.96 ┆ 0.96 ┆ 0.0 ┆ 0.0   ┆ 0.0    │
│ …          ┆ …                 ┆ …    ┆ …    ┆ …   ┆ …     ┆ …      │
│ 2025-12-02 ┆ KXFED-27APR-T4.25 ┆ 0.01 ┆ 0.01 ┆ 0.0 ┆ 0.01  ┆ 0.0    │
│ 2025-12-11 ┆ KXFED-27APR-T4.00 ┆ 0.06 ┆ 0.06 ┆ 0.0 ┆ 0.06  ┆ 0.0    │
│ 2025-12-11 ┆ KXFED-27APR-T4.25 ┆ 0.01 ┆ 0.01 ┆ 0.0 ┆ 0.01  ┆ 0.0    │
│ 2025-12-17 ┆ KXFED-27APR-T4.00 ┆ 0.06 ┆ 0.06 ┆ 0.0 ┆ 0.06  ┆ 0.0    │
│ 2025-12-17 ┆ KXFED-27APR-T4.25 ┆ 0.01 ┆ 0.01 ┆ 0.0 ┆ 0.01  ┆ 0.0    │
└────────────┴───────────────────┴──────┴──────┴─────┴───────┴────────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (35, 7)

timestamp	symbol	open	high	low	close	volume
date	str	f64	f64	f64	f64	f64
2025-10-08	"KXFED-27JAN-T1.00"	0.89	0.96	0.0	0.96	0.0
2025-10-08	"KXFED-27JAN-T3.00"	0.61	0.61	0.0	0.51	0.0
2025-10-15	"KXFED-27JAN-T1.00"	0.96	0.96	0.0	0.0	0.0
2025-10-15	"KXFED-27JAN-T3.00"	0.51	0.51	0.0	0.0	0.0
2025-10-15	"KXFED-27MAR-T0.50"	0.96	0.96	0.0	0.0	0.0
…	…	…	…	…	…	…
2025-12-02	"KXFED-27APR-T4.25"	0.01	0.01	0.0	0.01	0.0
2025-12-11	"KXFED-27APR-T4.00"	0.06	0.06	0.0	0.06	0.0
2025-12-11	"KXFED-27APR-T4.25"	0.01	0.01	0.0	0.01	0.0
2025-12-17	"KXFED-27APR-T4.00"	0.06	0.06	0.0	0.06	0.0
2025-12-17	"KXFED-27APR-T4.25"	0.01	0.01	0.0	0.01	0.0

3. Contract Universe

In [6]python · cell 11

python

contracts = (
    df.sort("timestamp")
    .group_by("symbol")
    .agg(
        pl.col("close").last().alias("latest_prob"),
        pl.col("close").first().alias("initial_prob"),
        pl.col("volume").sum().alias("total_volume"),
        pl.col("timestamp").min().alias("first_date"),
        pl.col("timestamp").max().alias("last_date"),
        pl.len().alias("observations"),
    )
    .sort("symbol")
)
contracts

Output

shape: (9, 7)
┌──────────────┬─────────────┬──────────────┬──────────────┬────────────┬────────────┬─────────────┐
│ symbol       ┆ latest_prob ┆ initial_prob ┆ total_volume ┆ first_date ┆ last_date  ┆ observation │
│ ---          ┆ ---         ┆ ---          ┆ ---          ┆ ---        ┆ ---        ┆ s           │
│ str          ┆ f64         ┆ f64          ┆ f64          ┆ date       ┆ date       ┆ ---         │
│              ┆             ┆              ┆              ┆            ┆            ┆ u32         │
╞══════════════╪═════════════╪══════════════╪══════════════╪════════════╪════════════╪═════════════╡
│ KXFED-27APR- ┆ 0.92        ┆ 0.93         ┆ 0.0          ┆ 2025-10-20 ┆ 2025-12-31 ┆ 73          │
│ T0.25        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27APR- ┆ 0.74        ┆ 0.75         ┆ 0.0          ┆ 2025-10-20 ┆ 2025-12-31 ┆ 73          │
│ T2.25        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27APR- ┆ 0.14        ┆ 0.14         ┆ 2.0          ┆ 2025-10-20 ┆ 2025-12-31 ┆ 73          │
│ T3.75        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27APR- ┆ 0.06        ┆ 0.06         ┆ 0.0          ┆ 2025-10-20 ┆ 2025-12-31 ┆ 73          │
│ T4.00        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27APR- ┆ 0.01        ┆ 0.01         ┆ 0.0          ┆ 2025-10-20 ┆ 2025-12-31 ┆ 67          │
│ T4.25        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27JAN- ┆ 0.95        ┆ 0.89         ┆ 0.0          ┆ 2025-10-07 ┆ 2025-12-31 ┆ 86          │
│ T1.00        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27JAN- ┆ 0.39        ┆ 0.61         ┆ 312.0        ┆ 2025-10-07 ┆ 2025-12-31 ┆ 86          │
│ T3.00        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27MAR- ┆ 0.95        ┆ 0.96         ┆ 1.0          ┆ 2025-10-14 ┆ 2025-12-31 ┆ 78          │
│ T0.50        ┆             ┆              ┆              ┆            ┆            ┆             │
│ KXFED-27MAR- ┆ 0.51        ┆ 0.59         ┆ 0.0          ┆ 2025-10-14 ┆ 2025-12-31 ┆ 78          │
│ T2.75        ┆             ┆              ┆              ┆            ┆            ┆             │
└──────────────┴─────────────┴──────────────┴──────────────┴────────────┴────────────┴─────────────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (9, 7)

symbol	latest_prob	initial_prob	total_volume	first_date	last_date	observations
str	f64	f64	f64	date	date	u32
"KXFED-27APR-T0.25"	0.92	0.93	0.0	2025-10-20	2025-12-31	73
"KXFED-27APR-T2.25"	0.74	0.75	0.0	2025-10-20	2025-12-31	73
"KXFED-27APR-T3.75"	0.14	0.14	2.0	2025-10-20	2025-12-31	73
"KXFED-27APR-T4.00"	0.06	0.06	0.0	2025-10-20	2025-12-31	73
"KXFED-27APR-T4.25"	0.01	0.01	0.0	2025-10-20	2025-12-31	67
"KXFED-27JAN-T1.00"	0.95	0.89	0.0	2025-10-07	2025-12-31	86
"KXFED-27JAN-T3.00"	0.39	0.61	312.0	2025-10-07	2025-12-31	86
"KXFED-27MAR-T0.50"	0.95	0.96	1.0	2025-10-14	2025-12-31	78
"KXFED-27MAR-T2.75"	0.51	0.59	0.0	2025-10-14	2025-12-31	78

4. Probability Evolution

In [7]python · cell 13

python

# Volume is near zero across these contracts, so "most active" means the widest range
# in implied probability (on the cleaned data), which flags the genuine battleground
# thresholds rather than a data glitch.
price_range = (
    df.group_by("symbol")
    .agg((pl.col("close").max() - pl.col("close").min()).alias("range"))
    .sort("range", descending=True)
)

top_contracts = price_range.head(3)["symbol"].to_list()

fig = go.Figure()
palette = [COLORS["blue"], COLORS["amber"], COLORS["copper"]]

for sym, color in zip(top_contracts, palette, strict=False):
    data = df.filter(pl.col("symbol") == sym).sort("timestamp").to_pandas()
    fig.add_trace(
        go.Scatter(
            x=data["timestamp"],
            y=data["close"],
            mode="lines",
            name=sym,
            line=dict(color=color, width=2),
        )
    )

fig.update_layout(
    title="Battleground Fed-rate thresholds hover near even odds while far thresholds stay pinned",
    xaxis_title="Date",
    yaxis_title="Implied Probability",
    yaxis=dict(tickformat=".0%", range=[0, 1.05]),
    height=400,
    legend=dict(orientation="h", yanchor="bottom", y=1.02, xanchor="center", x=0.5),
)

fig.show()

Output

5. Multi-Threshold View

Looking at all thresholds for a single meeting gives a snapshot of the market's full probability distribution over rate outcomes.

In [8]python · cell 16

python

# Group by meeting date prefix
meetings = {}
for sym in df["symbol"].unique().to_list():
    # KXFED-27APR-T4.25 → 27APR
    parts = sym.split("-")
    if len(parts) >= 3:
        meeting = parts[1]
        meetings.setdefault(meeting, []).append(sym)

fig = make_subplots(
    rows=len(meetings),
    cols=1,
    shared_xaxes=True,
    subplot_titles=[f"Meeting: {m}" for m in sorted(meetings.keys())],
    vertical_spacing=0.08,
)

for i, (meeting, symbols) in enumerate(sorted(meetings.items()), 1):
    for sym in sorted(symbols):
        data = df.filter(pl.col("symbol") == sym).sort("timestamp").to_pandas()
        threshold = sym.split("-T")[-1] if "-T" in sym else sym
        fig.add_trace(
            go.Scatter(
                x=data["timestamp"],
                y=data["close"],
                mode="lines",
                name=f"T{threshold}",
                showlegend=(i == 1),
            ),
            row=i,
            col=1,
        )
    fig.update_yaxes(tickformat=".0%", range=[0, 1.05], row=i, col=1)

fig.update_layout(
    height=250 * len(meetings),
    title_text="Each FOMC meeting's threshold ladder maps the market's full rate distribution",
)

fig.show()

Output

6. Day-over-Day Probability Dynamics

In [9]python · cell 18

python

df_enriched = df.sort("symbol", "timestamp").with_columns(
    pl.col("close").diff().over("symbol").alias("prob_change"),
)

activity = (
    df_enriched.group_by("symbol")
    .agg(
        (pl.col("volume") > 0).sum().alias("traded_days"),
        pl.col("volume").sum().round(1).alias("total_volume"),
        pl.col("prob_change").std().round(4).alias("daily_prob_vol"),
        pl.col("prob_change").abs().max().round(4).alias("max_daily_move"),
    )
    .sort("daily_prob_vol", descending=True, nulls_last=True)
)

activity

Output

shape: (9, 5)
┌───────────────────┬─────────────┬──────────────┬────────────────┬────────────────┐
│ symbol            ┆ traded_days ┆ total_volume ┆ daily_prob_vol ┆ max_daily_move │
│ ---               ┆ ---         ┆ ---          ┆ ---            ┆ ---            │
│ str               ┆ u32         ┆ f64          ┆ f64            ┆ f64            │
╞═══════════════════╪═════════════╪══════════════╪════════════════╪════════════════╡
│ KXFED-27JAN-T3.00 ┆ 3           ┆ 312.0        ┆ 0.0157         ┆ 0.1            │
│ KXFED-27JAN-T1.00 ┆ 0           ┆ 0.0          ┆ 0.0083         ┆ 0.07           │
│ KXFED-27MAR-T2.75 ┆ 0           ┆ 0.0          ┆ 0.0074         ┆ 0.04           │
│ KXFED-27APR-T3.75 ┆ 1           ┆ 2.0          ┆ 0.0044         ┆ 0.02           │
│ KXFED-27APR-T2.25 ┆ 0           ┆ 0.0          ┆ 0.0039         ┆ 0.02           │
│ KXFED-27APR-T4.00 ┆ 0           ┆ 0.0          ┆ 0.0038         ┆ 0.02           │
│ KXFED-27MAR-T0.50 ┆ 1           ┆ 1.0          ┆ 0.0034         ┆ 0.02           │
│ KXFED-27APR-T0.25 ┆ 0           ┆ 0.0          ┆ 0.0031         ┆ 0.02           │
│ KXFED-27APR-T4.25 ┆ 0           ┆ 0.0          ┆ 0.0018         ┆ 0.01           │
└───────────────────┴─────────────┴──────────────┴────────────────┴────────────────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (9, 5)

symbol	traded_days	total_volume	daily_prob_vol	max_daily_move
str	u32	f64	f64	f64
"KXFED-27JAN-T3.00"	3	312.0	0.0157	0.1
"KXFED-27JAN-T1.00"	0	0.0	0.0083	0.07
"KXFED-27MAR-T2.75"	0	0.0	0.0074	0.04
"KXFED-27APR-T3.75"	1	2.0	0.0044	0.02
"KXFED-27APR-T2.25"	0	0.0	0.0039	0.02
"KXFED-27APR-T4.00"	0	0.0	0.0038	0.02
"KXFED-27MAR-T0.50"	1	1.0	0.0034	0.02
"KXFED-27APR-T0.25"	0	0.0	0.0031	0.02
"KXFED-27APR-T4.25"	0	0.0	0.0018	0.01

In [10]python · cell 19

python

ad = activity.to_pandas()
fig = go.Figure(go.Bar(x=ad["symbol"], y=ad["daily_prob_vol"], marker_color=COLORS["blue"]))
fig.update_layout(
    title="Near-the-money Fed thresholds carry the most daily probability movement",
    xaxis_title="Contract",
    yaxis_title="Std. of daily probability change",
    height=400,
)
fig.update_xaxes(tickangle=-45)
fig.show()

Output

7. Event Indicators for ML

Transform Kalshi probabilities into ML-ready features. Since close is already the implied probability, we derive momentum, volatility, and regime indicators directly.

In [11]python · cell 21

python

LOOKBACK = 5
VOL_WINDOW = 10

kalshi_features = df.sort("symbol", "timestamp").with_columns(
    (pl.col("close") - pl.col("close").shift(LOOKBACK).over("symbol")).alias("prob_momentum"),
    pl.col("close").diff().rolling_std(VOL_WINDOW).over("symbol").alias("prob_volatility"),
    pl.when(pl.col("close").rolling_std(VOL_WINDOW).over("symbol") > 0)
    .then(
        (pl.col("close") - pl.col("close").rolling_mean(VOL_WINDOW).over("symbol"))
        / pl.col("close").rolling_std(VOL_WINDOW).over("symbol")
    )
    .otherwise(0.0)
    .alias("prob_zscore"),
    ((pl.col("close") > 0.8) | (pl.col("close") < 0.2)).cast(pl.Int8).alias("high_confidence"),
    (pl.col("high") - pl.col("low")).alias("uncertainty"),
)

print(f"Feature matrix: {kalshi_features.shape}")
kalshi_features.select(
    "timestamp", "symbol", "close", "prob_momentum", "prob_volatility", "high_confidence"
).head(10)

Output

Feature matrix: (687, 13)

shape: (10, 6)
┌────────────┬───────────────────┬───────┬───────────────┬─────────────────┬─────────────────┐
│ timestamp  ┆ symbol            ┆ close ┆ prob_momentum ┆ prob_volatility ┆ high_confidence │
│ ---        ┆ ---               ┆ ---   ┆ ---           ┆ ---             ┆ ---             │
│ date       ┆ str               ┆ f64   ┆ f64           ┆ f64             ┆ i8              │
╞════════════╪═══════════════════╪═══════╪═══════════════╪═════════════════╪═════════════════╡
│ 2025-10-20 ┆ KXFED-27APR-T0.25 ┆ 0.93  ┆ null          ┆ null            ┆ 1               │
│ 2025-10-21 ┆ KXFED-27APR-T0.25 ┆ 0.93  ┆ null          ┆ null            ┆ 1               │
│ 2025-10-22 ┆ KXFED-27APR-T0.25 ┆ 0.93  ┆ null          ┆ null            ┆ 1               │
│ 2025-10-23 ┆ KXFED-27APR-T0.25 ┆ 0.92  ┆ null          ┆ null            ┆ 1               │
│ 2025-10-24 ┆ KXFED-27APR-T0.25 ┆ 0.92  ┆ null          ┆ null            ┆ 1               │
│ 2025-10-25 ┆ KXFED-27APR-T0.25 ┆ 0.92  ┆ -0.01         ┆ null            ┆ 1               │
│ 2025-10-26 ┆ KXFED-27APR-T0.25 ┆ 0.92  ┆ -0.01         ┆ null            ┆ 1               │
│ 2025-10-27 ┆ KXFED-27APR-T0.25 ┆ 0.92  ┆ -0.01         ┆ null            ┆ 1               │
│ 2025-10-28 ┆ KXFED-27APR-T0.25 ┆ 0.92  ┆ 0.0           ┆ null            ┆ 1               │
│ 2025-10-29 ┆ KXFED-27APR-T0.25 ┆ 0.92  ┆ 0.0           ┆ null            ┆ 1               │
└────────────┴───────────────────┴───────┴───────────────┴─────────────────┴─────────────────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (10, 6)

timestamp	symbol	close	prob_momentum	prob_volatility	high_confidence
date	str	f64	f64	f64	i8
2025-10-20	"KXFED-27APR-T0.25"	0.93	null	null	1
2025-10-21	"KXFED-27APR-T0.25"	0.93	null	null	1
2025-10-22	"KXFED-27APR-T0.25"	0.93	null	null	1
2025-10-23	"KXFED-27APR-T0.25"	0.92	null	null	1
2025-10-24	"KXFED-27APR-T0.25"	0.92	null	null	1
2025-10-25	"KXFED-27APR-T0.25"	0.92	-0.01	null	1
2025-10-26	"KXFED-27APR-T0.25"	0.92	-0.01	null	1
2025-10-27	"KXFED-27APR-T0.25"	0.92	-0.01	null	1
2025-10-28	"KXFED-27APR-T0.25"	0.92	0.0	null	1
2025-10-29	"KXFED-27APR-T0.25"	0.92	0.0	null	1

In [12]python · cell 22

python

# Feature distributions for the most active contract
active_sym = top_contracts[0]
active_features = kalshi_features.filter(
    (pl.col("symbol") == active_sym) & pl.col("prob_momentum").is_not_null()
)

fig = make_subplots(
    rows=1,
    cols=2,
    subplot_titles=("Probability Momentum", "Probability Z-Score"),
)

fig.add_trace(
    go.Histogram(
        x=active_features["prob_momentum"].to_list(),
        nbinsx=30,
        name="Momentum",
        marker_color=COLORS["blue"],
    ),
    row=1,
    col=1,
)

fig.add_trace(
    go.Histogram(
        x=active_features["prob_zscore"].to_list(),
        nbinsx=30,
        name="Z-Score",
        marker_color=COLORS["slate"],
    ),
    row=1,
    col=2,
)

fig.update_layout(
    title=f"Probability momentum and z-score center near zero for {active_sym}",
    height=350,
    showlegend=False,
)

fig.show()

Output

8. Data Quality Assessment

Two checks matter for prediction-market data: whether any bars were corrupt (caught and repaired at load), and how much genuine price variation each contract carries.

In [13]python · cell 24

python

print(f"Zero-price artifact bars caught and repaired at load: {artifact_bars.height}")
artifact_bars

Output

Zero-price artifact bars caught and repaired at load: 35

shape: (35, 7)
┌────────────┬───────────────────┬──────┬──────┬─────┬───────┬────────┐
│ timestamp  ┆ symbol            ┆ open ┆ high ┆ low ┆ close ┆ volume │
│ ---        ┆ ---               ┆ ---  ┆ ---  ┆ --- ┆ ---   ┆ ---    │
│ date       ┆ str               ┆ f64  ┆ f64  ┆ f64 ┆ f64   ┆ f64    │
╞════════════╪═══════════════════╪══════╪══════╪═════╪═══════╪════════╡
│ 2025-10-08 ┆ KXFED-27JAN-T1.00 ┆ 0.89 ┆ 0.96 ┆ 0.0 ┆ 0.96  ┆ 0.0    │
│ 2025-10-08 ┆ KXFED-27JAN-T3.00 ┆ 0.61 ┆ 0.61 ┆ 0.0 ┆ 0.51  ┆ 0.0    │
│ 2025-10-15 ┆ KXFED-27JAN-T1.00 ┆ 0.96 ┆ 0.96 ┆ 0.0 ┆ 0.0   ┆ 0.0    │
│ 2025-10-15 ┆ KXFED-27JAN-T3.00 ┆ 0.51 ┆ 0.51 ┆ 0.0 ┆ 0.0   ┆ 0.0    │
│ 2025-10-15 ┆ KXFED-27MAR-T0.50 ┆ 0.96 ┆ 0.96 ┆ 0.0 ┆ 0.0   ┆ 0.0    │
│ …          ┆ …                 ┆ …    ┆ …    ┆ …   ┆ …     ┆ …      │
│ 2025-12-02 ┆ KXFED-27APR-T4.25 ┆ 0.01 ┆ 0.01 ┆ 0.0 ┆ 0.01  ┆ 0.0    │
│ 2025-12-11 ┆ KXFED-27APR-T4.00 ┆ 0.06 ┆ 0.06 ┆ 0.0 ┆ 0.06  ┆ 0.0    │
│ 2025-12-11 ┆ KXFED-27APR-T4.25 ┆ 0.01 ┆ 0.01 ┆ 0.0 ┆ 0.01  ┆ 0.0    │
│ 2025-12-17 ┆ KXFED-27APR-T4.00 ┆ 0.06 ┆ 0.06 ┆ 0.0 ┆ 0.06  ┆ 0.0    │
│ 2025-12-17 ┆ KXFED-27APR-T4.25 ┆ 0.01 ┆ 0.01 ┆ 0.0 ┆ 0.01  ┆ 0.0    │
└────────────┴───────────────────┴──────┴──────┴─────┴───────┴────────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (35, 7)

timestamp	symbol	open	high	low	close	volume
date	str	f64	f64	f64	f64	f64
2025-10-08	"KXFED-27JAN-T1.00"	0.89	0.96	0.0	0.96	0.0
2025-10-08	"KXFED-27JAN-T3.00"	0.61	0.61	0.0	0.51	0.0
2025-10-15	"KXFED-27JAN-T1.00"	0.96	0.96	0.0	0.0	0.0
2025-10-15	"KXFED-27JAN-T3.00"	0.51	0.51	0.0	0.0	0.0
2025-10-15	"KXFED-27MAR-T0.50"	0.96	0.96	0.0	0.0	0.0
…	…	…	…	…	…	…
2025-12-02	"KXFED-27APR-T4.25"	0.01	0.01	0.0	0.01	0.0
2025-12-11	"KXFED-27APR-T4.00"	0.06	0.06	0.0	0.06	0.0
2025-12-11	"KXFED-27APR-T4.25"	0.01	0.01	0.0	0.01	0.0
2025-12-17	"KXFED-27APR-T4.00"	0.06	0.06	0.0	0.06	0.0
2025-12-17	"KXFED-27APR-T4.25"	0.01	0.01	0.0	0.01	0.0

In [14]python · cell 25

python

quality_df = (
    df.group_by("symbol")
    .agg(
        pl.len().alias("observations"),
        pl.col("volume").mean().round(3).alias("avg_volume"),
        (pl.col("close").max() - pl.col("close").min()).round(3).alias("price_range"),
    )
    .sort("price_range", descending=True)
)
quality_df

Output

shape: (9, 4)
┌───────────────────┬──────────────┬────────────┬─────────────┐
│ symbol            ┆ observations ┆ avg_volume ┆ price_range │
│ ---               ┆ ---          ┆ ---        ┆ ---         │
│ str               ┆ u32          ┆ f64        ┆ f64         │
╞═══════════════════╪══════════════╪════════════╪═════════════╡
│ KXFED-27JAN-T3.00 ┆ 86           ┆ 3.628      ┆ 0.23        │
│ KXFED-27JAN-T1.00 ┆ 86           ┆ 0.0        ┆ 0.08        │
│ KXFED-27MAR-T2.75 ┆ 78           ┆ 0.0        ┆ 0.08        │
│ KXFED-27APR-T3.75 ┆ 73           ┆ 0.027      ┆ 0.04        │
│ KXFED-27APR-T2.25 ┆ 73           ┆ 0.0        ┆ 0.03        │
│ KXFED-27APR-T4.00 ┆ 73           ┆ 0.0        ┆ 0.03        │
│ KXFED-27APR-T0.25 ┆ 73           ┆ 0.0        ┆ 0.02        │
│ KXFED-27MAR-T0.50 ┆ 78           ┆ 0.013      ┆ 0.02        │
│ KXFED-27APR-T4.25 ┆ 67           ┆ 0.0        ┆ 0.01        │
└───────────────────┴──────────────┴────────────┴─────────────┘

.dataframe > thead > tr, .dataframe > tbody > tr { text-align: right; white-space: pre-wrap; } shape: (9, 4)

symbol	observations	avg_volume	price_range
str	u32	f64	f64
"KXFED-27JAN-T3.00"	86	3.628	0.23
"KXFED-27JAN-T1.00"	86	0.0	0.08
"KXFED-27MAR-T2.75"	78	0.0	0.08
"KXFED-27APR-T3.75"	73	0.027	0.04
"KXFED-27APR-T2.25"	73	0.0	0.03
"KXFED-27APR-T4.00"	73	0.0	0.03
"KXFED-27APR-T0.25"	73	0.0	0.02
"KXFED-27MAR-T0.50"	78	0.013	0.02
"KXFED-27APR-T4.25"	67	0.0	0.01

In [15]python · cell 26

python

qd = quality_df.to_pandas()
fig = go.Figure(go.Bar(x=qd["symbol"], y=qd["price_range"], marker_color=COLORS["blue"]))
fig.update_layout(
    title="After repair, close-price range concentrates in near-the-money thresholds",
    xaxis_title="Contract",
    yaxis_title="Close price range",
    height=400,
)
fig.update_xaxes(tickangle=-45)
fig.show()

Output

9. Save Enriched Data

In [16]python · cell 29

python

output_dir = get_output_dir(4, "kalshi")
output_dir.mkdir(parents=True, exist_ok=True)

output_file = output_dir / "kalshi_features.parquet"
kalshi_features.write_parquet(output_file)

print(f"Saved {len(kalshi_features)} observations to {output_file}")

Output

Saved 687 observations to /tmp/ml4t-test-output/ch04_kalshi/kalshi_features.parquet

Key Takeaways

Implied probability = close price: Kalshi contract prices directly encode the market's probability estimate for the event, no transformation needed
Threshold structure: Multiple contracts per meeting create a full probability distribution over rate outcomes - richer than a single forecast
Feature engineering: Momentum, volatility, and z-score of probability paths provide regime-detection signals for rate-sensitive strategies
Liquidity caveat: Economic event contracts are still early-stage; volume is thin compared to traditional derivatives markets
Screen for corrupt bars first: thin, carry-forward markets are prone to ingestion artifacts (a positive high with a zero close). Detect and repair them before ranking or feature engineering, or a single bad bar distorts both.

Next: See 13_polymarket_prediction_markets for the higher-liquidity crypto-based alternative and cross-platform comparison.