Skip to content
All library documents

Choosing Market Data for Bitcoin Strategy Backtests

Article Quant Q&A · Author: PeeS
Read the original

Summary

The discussion considers what market data to collect when building a Bitcoin price feeder for backtesting. Its practical recommendation is to begin with level-one data, such as trades and best bid and offer, and add order-book data only when a specific strategy or later use case requires it. If a strategy uses trade prices alone, that simpler feed may suffice; OHLC bars can also be derived from trade-level observations.

The answer offers a scope-management principle rather than a tested trading method: collecting more data than needed can consume time and encourage infrastructure work before the research question is clear. It also cautions against treating Bitcoin markets as equivalent to foreign exchange. No empirical comparison of data feeds, backtest results, or detailed definitions of tick data are provided, so the appropriate data resolution still depends on the strategy and intended execution assumptions.

Key ideas

  • Begin with level-one market data and expand collection when a concrete use case calls for it.
  • Trade-price strategies may not need order-book data.
  • OHLC bars can be constructed from trade observations.
  • Collecting unnecessary data can divert effort from research.
  • Bitcoin markets should not automatically be modeled as foreign exchange.

Tags

CryptoBacktestingMarket microstructure
Full text
# Newbie Quant: Bulding price feeder to securities master db


# Newbie Quant: Bulding price feeder to securities master db












First of all, warm hello to all. I am newbie and i admit it, but with at least 15+ years of exp in C++. Working in IB - derivatives mainly, but unfortunately on the business side not trading at all. Got interested in Quant trading very much.

Intro: I am learning very hard to understand the basics of quant trading currently building my first out of 2 applications. The Feeder - which generally speaking retrieves tick data (OHLC, price and volume) from Bitcoin Exchanges with a resolution of 1 minute (queries the markets every minute and stores the tick values in the database). No place to mention theory i read about in books and blogs.

Question: Is it enough to get the tick values from exchanges to be able to backtest strategies ? Should i consider downloading orderbooks too?

Reason: I am interested in backtesting and eventually quant trading Bitcoins, no options, futures or any derivatives. I understand BTC markets as FX.

## Answer by madilyn (score 2, accepted)

https://quant.stackexchange.com/a/19009

Whether to store L1 (trades/BBO only), OHLC and order book depends on your downstream application. I encourage you to start with L1 (easy to store) and then think about what to do as your use cases evolve. If your trading strategy only uses trade prices, then you are fine with L1. And OHLC can be backed out from L1.

It is very tempting to store more data than you actually need and overoptimize your data infrastructure and get lost in shifting objectives, and in my experience it gets overly time-consuming.

Lastly, FX and Bitcoin are very different markets. I would encourage you think carefully about this assumption.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.