Backtesting & data integrity · Python
Tickbloom scores your market data before your strategy ever sees it — gaps, duplicates, look-ahead leakage, survivorship bias — and writes the audit trail your prop firm will ask for.
Typical time a solo quant spends rebuilding ETL and cleanup for each new data provider.
A single shifted index is enough to turn a losing strategy into a beautiful, fictional equity curve.
What most traders can produce when a prop firm asks why a specific fill looked like an outlier.
What's in the box
Tickbloom doesn't try to replace QuantConnect or Lean. It does one thing: make sure the data feeding your strategy is clean, documented, and auditable end to end. The layers run in order — each only makes sense once the one before it has passed.
Databento and CSV today; Polygon.io and IQFeed next. Pull by symbol, date range, and venue without re-learning an API. Symbol metadata, exchange calendars, splits, and contract rolls attach to every batch automatically.
tickbloom.load("ES.c.0", "2024-07-01:2024-09-30")Gaps, duplicates, zero-volume prints, out-of-order timestamps, and session-border drift are validated against configurable thresholds. The Integrity Score turns "is this data clean?" into a number you can put in a document and defend.
tickbloom.audit(data).score # 98.4A static AST scanner flags places where a future-looking feature can contaminate results — a negative shift, a full-series fit(), a resample that leaks the close. Runtime guards catch the rest, and delisted names are restored to the universe so your equity curve stops lying.
tickbloom.scan("strategies/") → 2 findingsAn AutoAudit bundle: order log, fill breakdown, slippage attribution, and equity curve, exported in the format prop firms and allocators ask for. When someone asks why a trade was an outlier, you send the file instead of writing an apology.
tickbloom.export("trades/2024", fmt="review")Reads your findings and proposes targeted changes: tune the fill model, adjust slippage assumptions, handle a tail event. Then it runs the regression suite against your documented history and shows the delta. You approve before anything is written. Optional, and off by default.
tickbloom.agent.propose(report) # dry-runThe workflow
No Docker, no separate server, no DevOps. Every step returns a structured object you can hand to an auditor, an allocator, or your future self six months from now.
A standard pip install inside your existing virtualenv. Pure Python, pandas and pyarrow only.
pip install tickbloom
Load from one or more providers in a single call. Metadata and calendars attach automatically.
df = tb.load("ES.c.0",
"2024-Q3")Integrity checks plus bias scans in one command. Output is a structured report, not a printout.
rep = tb.audit(df) rep.score # 98.4
Produce the AutoAudit bundle. Every order, fill, and equity tick logged with replay support.
tb.export("trades/24",
fmt="review")Who it's for
If you write strategies in Python, want data confidence before you risk capital, and would rather pay than lose another month to cleanup — this is for you.
Writes strategies from a notebook. Needs the data clean before any capital moves.
Budget that can't stretch to an enterprise platform, but still needs an institutional audit trail.
Trading someone else's capital under review, with rules that require evidence for every outlier.
I used to rebuild ETL from scratch for every new data provider. Now the tool documents itself, and I spend the week on the strategy instead.
Pricing
Monthly, cancel anytime. Every tier runs the same loader and the same integrity suite — higher tiers add surface area, reporting formats, and support, never a better version of the core checks.
For the individual quant validating strategies before risking their own capital.
For funded prop traders and small desks who owe someone else an explanation.
For funds who need the audit trail to survive an allocator's due diligence.
14-day trial, no card. Data-provider subscriptions (e.g. Databento) are billed by the provider, not by us.
FAQ
Didn't find your answer? Email us — a human replies, usually within a day.
Loader protocol, and a community connector is a pull request — no fork required. The integrity suite works on any source that returns the normalized frame.t depends on an index greater than t. Static findings are advisory; runtime violations fail the audit.score = 100 - sum(penalties), where each check has a documented maximum weight and a tolerance. Rate-based checks use exponential saturation, so the first defects cost the most and no single defect maxes out a check. Every weight and tolerance is recorded in the report manifest and can be overridden in tickbloom.toml — so you can say "we used these thresholds" rather than "the vendor decided".Contact
Tickbloom is new and we're onboarding a first cohort of funds and prop traders by hand. Email us and you'll get subscription details plus a live walkthrough on your own data.
We do not store your strategy code or your trade orders. Runs execute client-side; the system records only audit metadata — Tickbloom version, data source identifier, thresholds, and check results — to make the audit trail reproducible.