Advertisement IC Markets

Tick data is not free, even when the download is

Share, develop, and backtest custom MQL4/MQL5 Expert Advisors, Python data-scraping scripts, trading bots, and automated market alert systems.
Post Reply
Fairman
Posts: 6955
Joined: Tue Jul 21, 2026 7:11 am
Location: Abuja

Tick data is not free, even when the download is

Post by Fairman »

Tick data is not free, even when the download is

Downloading years of tick data for a major pair is easy. Trusting it is another matter.

Gaps are common. A feed that dropped for a few minutes leaves holes, and a backtest simply jumps across them. Some sources fill the gap with synthetic ticks, which look like data but are invented.

Timestamps can be off. A source that records time in local time instead of UTC, or shifts at daylight saving changes, misaligns events with the news calendar.

Spreads may be missing or fixed. Some free datasets hold only bid prices, leaving the ask to be guessed with a constant spread that never existed.

Volume is usually the number of quotes, which may differ widely between providers.

The fix is checking. Compare the dataset with another source for a few known events, such as a major data release or a known spike, and look at whether the sequences match. Then count the gaps and check the timestamps around daylight saving.

A clean approach is building a small checklist: timestamps in UTC, no gaps over a set length, spreads present and varying, no negative spreads, no jumps beyond a threshold. Any dataset passing all five earns a measure of trust.

What checks have been run on tick data before trusting a test built on it?
It’s Fairman :geek:
Recommended broker for automated trading & scalping IC Markets
Amanhasnoname
Posts: 2060
Joined: Fri Oct 30, 2026 9:45 am

Re: Tick data is not free, even when the download is

Post by Amanhasnoname »

Free tick files often look more complete than they are, so start with provenance and coverage before running a strategy. Check the timezone against known market events, watch for duplicate or missing timestamps, flat stretches, impossible bid/ask values, and whether the file is bid-only. A fixed spread assumption can make a scalper's fills look cleaner than they could be live.

Then test the same interval against an independent feed around quiet periods and sharp moves, but don't treat agreement as proof: providers may share upstream sources or smoothing. Count gaps by session and volatility, and document how missing segments are handled. Dropping bad data is safer than quietly filling it with synthetic ticks unless the method is explicit.

For validation, replay realistic variable spreads, commissions, latency and slippage, and verify timestamp conversion through daylight-saving changes. Compare results with a bar-based check and inspect a sample of raw quotes manually. The key pitfall is letting a backtest silently skip the worst gaps or use an ask price that was never available; make those assumptions visible in the report.
Post Reply