Details
At a glance
We're in early access. Our column is planned; only the free sample downloads are live today.
| Tardis.dev | marketdata.co (planned, early access) | |
|---|---|---|
| Asset classes | Crypto only: spot, perpetuals, futures, options, plus prediction markets [1] | Crypto plus US stocks, options, forex, futures, indices |
| Delivery | Daily gzip CSV files per exchange, data type, day and symbol; raw replay API; self-hosted Tardis Machine [1] | REST, WebSocket, CSV and Parquet dataset files |
| Pricing model | Subscriptions from $350 a month (Academic) to $6,000 a month (Business, all exchanges); minimum order $300 [2] | Plans from Free to $249 a month; datasets per file, e.g. crypto trades $3 per pair-month |
| History | From 2019-03-30 at the earliest exchanges [1] | Varies by product and venue; published per product page |
| File downloads | Yes, by subscription; their FAQ says they don't sell single instruments or date ranges [3] | Per pair, per month or per GB, bought once and kept |
| Free data | First day of every month free for every exchange and data type [1] | Free sample files now; free first day of each month planned for trades |
Sources: [1] https://docs.tardis.dev/downloadable-csv-files [2] https://tardis.dev/#pricing [3] https://docs.tardis.dev/faq/billing-and-subscriptions.md
Details
Where they are strong
Tardis records raw exchange feeds tick by tick, including every L2 order book update, and keeps two timestamps on every row: the exchange's and their own receive time. Coverage runs to 50+ exchanges. Their docs are among the best in crypto data, with incident logs, contract spec history and machine-readable pages. If you're building an execution simulator that replays full depth across many venues, Tardis is hard to replace, and we don't plan to match their full incremental L2 history on self-serve plans.
Side by side
Where we differ
Buy a slice, not a subscription. Crypto trades are planned at $3 per pair per month, so one month of BTC-USDT trades costs $3, not the $300 Tardis minimum. Order book files are planned at $1 per GB, with a cost estimate before checkout.
Bars as files. Tardis computes OHLCV only inside Tardis Machine. We plan bar datasets from 1 second to 1 day at $5 per pair-year, plus daily bars on the Free plan.
More than crypto. The same schema covers stocks, options, forex and futures, so a cross-asset study doesn't need a second vendor and a second loader.
Live data hosted for you. Tardis leaves real-time streaming to open-source clients that connect to exchanges directly. Our live trade stream is planned on Pro.
Samples today. Our Binance BTC and ETH trade, bar, funding and depth samples are on the sample data page.
Watch out
Migration notes
Field mapping from Tardis CSV data types to our crypto schemas. We kept the names close on purpose so your loaders need little change.
| Tardis field | marketdata.co field | Note |
|---|---|---|
exchange |
exchange |
|
symbol (e.g. BTCUSDT) |
symbol (e.g. BTC-USDT) |
Dash between base and quote |
timestamp (µs since epoch) |
timestamp |
ISO 8601 UTC string |
local_timestamp |
received_timestamp |
|
id (trades) |
trade_id |
|
side (trades) |
aggressor_side |
buy or sell |
amount |
size |
Base-asset amount |
| none | notional |
price times size, added for you |
funding_rate, predicted_funding_rate, open_interest, mark_price, index_price, last_price |
same names | Our funding and OI schema |
funding_timestamp |
next_funding_time |
|
side (liquidations): buy / sell |
side: short / long |
Ours names the position that was closed |
is_snapshot (L2) |
record_type |
snapshot or update |
Example: reading a Tardis trades file and one of ours.
import pandas as pd
def load_tardis(path):
df = pd.read_csv(path)
df["timestamp"] = pd.to_datetime(df["timestamp"], unit="us", utc=True)
return df.rename(columns={"id": "trade_id", "side": "aggressor_side", "amount": "size"})
def load_marketdata(path): # our trades file, CSV or Parquet
df = pd.read_parquet(path) if path.endswith(".parquet") else pd.read_csv(path)
df["timestamp"] = pd.to_datetime(df["timestamp"])
return df
FAQ
Frequently asked questions
Is marketdata.co cheaper than Tardis.dev?
For small, targeted pulls, yes: one pair-month of trades is planned at $3, against a $300 minimum order at Tardis. For full L2 depth across dozens of exchanges, Tardis is built for that job and we don't plan to compete on it.
Do you have order book history like Tardis?
We plan top of book and depth bands on Pro, 25-level L2 snapshots on Business, and order book files at $1 per GB. Full incremental L2 is Enterprise only.
Which exchanges will you cover?
Our current samples come from Binance. The full exchange list will be published on the crypto tick data page before launch.
Is marketdata.co live?
Not yet. Free samples are downloadable now, and the API and dataset store open in stages during early access.
Details
Try it
Download the BTC-USDT trades sample and run it through your Tardis loader with the renames above.