Do I need to know how to code?

No. Describe the strategy in plain English and the AI writes the full file, compiles it, and fixes its own errors until it runs. Every #[param] in the file becomes a control in the UI — range minutes, stop distance, risk-reward — so you tune and re-run without ever opening the code. When you do want to read or edit it, it’s a real file that’s yours.

What data do backtests actually run on?

Licensed CME market data: 16 years of tick history across 48 futures markets, plus L2 order-book depth. Fills are simulated at tick level — and with book-walk fills when you want to know whether your size actually gets filled at the touch — not approximated from minute bars.

How does the optimizer avoid overfitting?

Five modes — exhaustive grid, bayesian search, genetic (CMA-ES), walk-forward and rolling walk-forward — fan out across up to 64 cloud workers. Out-of-sample windows, overfit detection and parameter-stability views make sure the winning trial is an edge, not a coincidence.

What if my strategy doesn’t hold up?

Then you found out in a few minutes instead of over a blown prop challenge or months of live trading — which is most of the point. Plenty of strategies look good in-sample and fall apart out-of-sample; the platform is built to show you that early, and to show you why: which parameter was fragile, which session was carrying the P&L, where the fills were unrealistic. Often that diagnosis is what turns an idea that didn’t work into one that does. We can’t promise your strategy will be profitable, and nobody honestly can — what we can do is make being wrong cheap and fast.

What happens when I deploy a bot?

It gets its own security-sandboxed container with kernel-level isolation, running 24/7 whether your laptop is open or not. Every decision, order and fill is logged per bot; you can edit a running bot’s parameters without wiping its history; and equity history survives redeploys, so each bot keeps one honest, unbroken curve.

Which AI models can I use?

Roboquant AI is included and runs on plan credits — tuned for strategy code, tool calls and long research loops. You can also switch to Claude, GPT or Gemini per conversation, set thinking effort from none to extra-high per message, or bring your own OpenAI, Anthropic or OpenRouter keys and pay your provider directly. Keys stay encrypted.

What can I do in a research pod?

Anything scientific Python can do: pandas, NumPy, scikit-learn, statsmodels against the whole data store — trades, quotes, book depth and FRED macro series — with zero setup. Findings render into the chat as charts, tables and interactive reports. Pods hold no data credentials; every read is mediated, capped and audited, and raw data never leaves the pod.

Can I automate my TradingView strategies?

Yes, on Tradovate, with Roboquant Connect — a separate product at connect.roboquant.dev. Point your TradingView alerts at your Connect webhook and it places the orders on your connected Tradovate accounts, with a kill switch, daily trade and loss limits, and an allowed-symbols list in the middle. That includes prop-firm accounts issued on Tradovate, but check your firm’s own rules on automation and third-party hosting before you connect one, because they differ and some prohibit it.

What does RoboCharts include?

A full futures charting terminal on real-time CME data: footprint bars, volume profile and L2 book visuals, custom indicators written in Python, bar-by-bar (and tick-by-tick) session replay, and a strategy tester that runs on the chart itself — trades, P&L and your bot’s own drawings appear inline. Every timeframe from 1-minute to daily, 16 years of history.

Algo Trading
on Steroids

Backtest millions of CME ticks in seconds, then deploy the exact strategy you tested - powered by Roboquant's proprietary Test-to-Trade Engine.

Start free Take the free course on Whop

Roboquant AI in action

Real code. Yours to read, edit and own.

Watch the AI write, compile, test, diagnose and optimize a strategy using the same tools as the platform.

Features Pricing Docs Sign in

A trading AI that writes strategies that compile.

Roboquant's AI writes your strategy, compiles it and fixes compile errors before you backtest — with the option to use your own model keys.

Choose any model. Roboquant AI is included — or turn any frontier model into a quant.

Roboquant AI

es_trades_2026.parquet

orb_regimes_ml.html

docs.roboquant.dev

FRED: CPI · payrolls

Add knowledge. The data store, docs and your research — one call away.

Roboquant AI [09:31]

Your ORB v2 tested +31.8% yesterday.
I kept the walk-forward results and the Friday finding — want to add the score filter we discussed and re-run it?

Built-in memory. It remembers every backtest and finding across sessions.

Add tools. The whole platform is callable — backtests, optimizer, deploys.

run_backtest

get_trades

optimize_strategy

write_file

deploy_bot

search_docs

Every fill simulated on real ticks.

Minute bars flatter a strategy — they fill you at prices that were never available. This one is filled tick by tick, over six years of ES, in the platform's own results card.

ORB Breakout — ES · ES 5m · ticks · 2020-06-01 → 2026-06-21

Return+3155.45%

Net P&L+$315,545.00

Win Rate47.8%

Max DD-19.0%

Cumulative PnL

Total Trades1264

Sharpe1.72

Sortino2.41

Profit Factor1.78

Calmar4.10

Expectancy+$249.60

Avg Win+$1,192.00

Avg Loss-$613.00

Avg R:R1.94

Win Streak9

Loss Streak7

Drawdown

Drawdown periods & recovery

Based on closed-trade equity, matching the drawdown chart above. Recovery is measured from trough to previous peak; time underwater is peak to recovery. Durations are elapsed calendar time. Ongoing means not recovered by the end of the saved run.

Periods: 18
Longest underwater: 255d 12h
Avg. recovery: 36d 12h
Ongoing at run end: None

Deepest periods first. Average recovery includes recovered periods only.

Drawdown periods ranked by depth
Peak	Trough	Recovered	Depth	Decline	Recovery	Underwater
11 Feb 202207:00	25 Jul 202214:00	24 Oct 202220:00	−18.95%	164d 6h	91d 6h	255d 12h
12 Nov 202001:00	11 Feb 202107:00	25 Apr 202108:00	−13.55%	91d 6h	73d	164d 6h
11 Nov 202401:00	17 Dec 202413:00	28 Feb 202513:00	−9.87%	36d 12h	73d	109d 12h
31 May 202320:00	07 Jul 202308:00	06 Oct 202314:00	−7.72%	36d 12h	91d 6h	127d 18h
06 Oct 202114:00	24 Oct 202120:00	30 Nov 202107:00	−4.81%	18d 6h	36d 12h	54d 18h

−18.95% drawdownRecovered

Peak: 11 Feb 202207:00
Trough: 25 Jul 202214:00
Recovered: 24 Oct 202220:00
Time underwater: 255d 12h
Decline: 164d 6h
Recovery time: 91d 6h

−13.55% drawdownRecovered

Peak: 12 Nov 202001:00
Trough: 11 Feb 202107:00
Recovered: 25 Apr 202108:00
Time underwater: 164d 6h
Decline: 91d 6h
Recovery time: 73d

−9.87% drawdownRecovered

Peak: 11 Nov 202401:00
Trough: 17 Dec 202413:00
Recovered: 28 Feb 202513:00
Time underwater: 109d 12h
Decline: 36d 12h
Recovery time: 73d

−7.72% drawdownRecovered

Peak: 31 May 202320:00
Trough: 07 Jul 202308:00
Recovered: 06 Oct 202314:00
Time underwater: 127d 18h
Decline: 36d 12h
Recovery time: 91d 6h

−4.81% drawdownRecovered

Peak: 06 Oct 202114:00
Trough: 24 Oct 202120:00
Recovered: 30 Nov 202107:00
Time underwater: 54d 18h
Decline: 18d 6h
Recovery time: 36d 12h

1–5 of 18 periods

Win / Loss

48%

52%

604 wins660 losses

Monthly Returns (%)

	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec	Year total
2020	-	-	-	-	+0.0	+0.0	+4.1	+6.2	+0.0	+8.2	-3.4	+0.0	+40.3
2021	-6.3	+0.0	+1.1	+9.5	+10.1	+0.0	+11.4	+2.1	+0.0	-4.8	+5.7	+0.0	+99.1
2022	+4.0	+0.0	-3.1	+1.2	-2.2	+0.0	-1.1	+6.7	+0.0	+12.6	+0.8	+0.0	+33.2
2023	+1.7	+0.0	+9.7	-0.5	+6.4	+0.0	+7.1	-3.2	+0.0	-4.2	+9.5	+0.0	+64.4
2024	+8.4	+7.6	+0.0	+10.4	-1.5	+0.0	+0.8	+8.0	+0.0	+3.1	-4.4	+0.0	+85.8
2025	+3.9	+6.5	+0.0	+2.1	+14.1	+0.0	+7.3	-2.5	+0.0	+4.8	+3.4	+0.0	+134.1
2026	+2.2	+5.6	+0.0	+7.3	+0.0	-	-	-	-	-	-	-	+22.3

Net P&L by period

Closed-trade P&L after commission, grouped by exit time in ET. Weeks start Monday. Only periods with closed trades are shown.

Total net P&L +$315,545.00

Yearly closed-trade net P&L
Period	Net P&L	Trades	Win rate	Avg / trade
2026 8 trades · 75.0% wins Avg / trade +$7,421.13	+$59,369.00	8	75.0%	+$7,421.13
2025 20 trades · 85.0% wins Avg / trade +$7,624.45	+$152,489.00	20	85.0%	+$7,624.45
2024 20 trades · 75.0% wins Avg / trade +$2,625.30	+$52,506.00	20	75.0%	+$2,625.30
2023 20 trades · 60.0% wins Avg / trade +$1,198.20	+$23,964.00	20	60.0%	+$1,198.20
2022 20 trades · 65.0% wins Avg / trade +$463.70	+$9,274.00	20	65.0%	+$463.70
2021 20 trades · 80.0% wins Avg / trade +$695.50	+$13,910.00	20	80.0%	+$695.50
2020 11 trades · 81.8% wins Avg / trade +$366.64	+$4,033.00	11	81.8%	+$366.64

Newest periods first. First and last periods may cover only part of the calendar period.

6 years

of ES, 2020 → 2026

1,264

trades taken

$250

expectancy per trade

tick fills

entries filled at tick level

One engine from backtest to live trading.

Backtesting, optimization and live trading run on the same compiled Rust engine — so brackets and trailing stops behave consistently from test to execution.

Same backtest, three engines — racing in real time6 years of ES ticks · 2.14B events · single core

The grid advances at each engine's real throughput — Roboquant's 520M ticks/s is measured on real CME ES trade data.[ ~470× faster ]

Compiled, not interpreted. Six years of tick-by-tick ES resolves in seconds; an optimizer trial averages a quarter second.

Tick, depth and order-book fills. Bar-level for speed, tick-level for truth, L2 book-walk when size matters.

One engine, tester to live. Same engine, different feed — brackets and trailing stops behave identically in both.

500M+

ticks per second, measured

Rust

compiled engine core

L1 · L2

trade and depth simulation

1 engine

backtest, optimize and live

200 tick-level trials. 54.7 seconds.

Sweeping parameters is how strategies get overfitted — so every trial is scored on data it never saw. Six parameters over 3.5 years of ticks, in the platform's own optimizer view.

New optimizationORB Breakout — ES · orb_final.rqc

Jan 2023 – Jun 2026 · tick data

Parameters to sweep

Range duration

5–60 min · increments of 5

Stop loss

1–10 pts · increments of 0.5

Risk / reward

1.0–3.0 · increments of 0.25

Take-profit method

Risk / reward · range multiple · fixed

Trailing trigger

4–16 pts · increments of 2

grid200 trials · 64 workersWalk-forward off Run

About 1 combinations

parallel cloud workers per sweep

54.7s

for 200 tick-level trials

5 modes

grid, bayesian, genetic, walk-forward, rolling

887

trades simulated per trial

Charts that show what your bot is thinking.

A full futures charting terminal on real-time CME data — and the only one where your strategy draws its own analysis on the chart.

ES · 1m · strategy replay — opening-range breakout, historical session

1m5m15m1h

CME · real-time

Market data © Chicago Mercantile Exchange Inc.

Session closed — waiting for the next open…

Strategy drawings on the chart. Your bot’s zones, levels, entries and exits render live — you see the setup it saw.

Order-flow: footprint and depth. Footprint bars, volume profile and L2 book visuals on real CME feeds.

Custom indicators in Python. Write any indicator with tunable parameters; it renders like a native one.

It finds what the equity curve hides.

On one of our research runs the curve looked fine — until the analytics split it by weekday. The edge died on Fridays, and a one-line session filter fixed it.

P&L by day of week — opening-range research run, computed by the platform

Mon

+$1,525

Tue

+$1,875

Wed

+$3,987.5

Thu

+$8,750

Fri

-$8,850

Thursdays made +$8,750 across 28 trades. Fridays lost -$8,850 across 25. One line — skip Fridays — and the strategy that looked mediocre was worth keeping. This is the kind of thing you find here, or you find it with real money.

Risk, quantified. VaR $300 per trade at 95%, Kelly-optimal size 1.8% — computed on every run.

Time-of-day and seasonality. P&L by hour, weekday and month — schedule effects show up before you trade them.

MAE / MFE trade efficiency. How far trades ran against you before working — the raw material for better stops.

Every bot keeps one unbroken track record.

One click deploys the file you just validated into an isolated container on your Tradovate account. Edit its parameters, redeploy it, come back in six months — the equity history never resets. The longer it runs, the more you actually know.

deployment · orb_final.rqc · running on Tradovate

orb_final.rqcES · 5mTradovate · Apex 50krunning · 6d 4h

Session P&L

+$0.00

Equity

$51,846.20

Trades

Win rate

—

Position

Flat

Max DD

−4.1%

Session equity — simulated

bot logs — every decision, order and filllive

09:30:00session open — ORB window 09:30–10:00 ET

Risk caps enforced server-side. Loss limits and kill switches hold even if your strategy, browser or connection doesn’t.

A private, sandboxed container per bot. Kernel-level isolation, no shared runtime, running 24/7.

Live P&L from the broker. What you see is what the account says — not what the bot believes.

TradingView alerts, executed on Tradovate.

Roboquant Connect is a separate product: point your TradingView alerts at a Connect webhook and it places the orders on your Tradovate accounts — including prop-firm accounts issued on Tradovate — with risk controls in the middle. Check your firm's rules on automation and third-party hosting before you connect it.

Roboquant Connect dashboard — executions, positions and account snapshot

Multiple accounts. One alert can place orders on every Tradovate account you connect.

Daily limits. Per-account daily trade and daily-loss limits block new entries once hit.

Kill switch and allowed symbols. Stop every order at once, and only trade the symbols you allow.

Visit Roboquant Connect

Run Roboquant from ChatGPT, Claude or Cursor.

Connect the Roboquant MCP server once, and the AI you already talk to can build, backtest and monitor your strategies.

Backtest my opening-range strategy on ES over the last 2 years and tell me if it's worth deploying.

ChatGPT can make mistakes. Check important info.

Ask ChatGPT5.6 Sol

strategies_createbacktests_runbacktests_chart_imageoptimizations_rundeployments_listdocs_get+28 more · mcp.roboquant.dev · OAuth

Works where you work. Claude, ChatGPT, Claude Code, Cursor — one OAuth connect, no keys to paste.

The full platform, as tools. Strategies, backtests, optimizer, deployments — callable by your AI.

Chart vision. The AI can look at your backtest charts, not just the numbers.

More honest backtests start with better data.

A backtest is only as honest as the data under it. Licensed CME market data—not delayed quotes or minute bars from a charting API. Access up to sixteen years of OHLCV and tick history, with order-book depth on higher tiers, stored next to the engine so backtests read it in seconds.

1mBars

OHLCV candles, one row per minute — the classic view.

L1Trades

Every trade event, tick by tick — time, price, size, side.

09:31:04.1826941.2512

09:31:04.0086941.003

09:31:03.8776941.2541

09:31:03.6106941.007

L2Depth — MBP-10

Ten price levels of resting liquidity, on every change.

L3Order flow — MBO

The full book: every individual order at every price level.

One trading day of ES, four resolutions.

Most platforms stop at the first row. Your edge usually lives in the other three.

1-minute bars

open, high, low, close, volume

events

Trades — L1 ticks

every fill, with price, size and side

≈

events · 300× bars

Book updates — L2 depth

ten price levels, every change (MBP-10)

≈

events · 10,000× bars

Order events — L3 flow

every add, modify and cancel (MBO)

≈

events · 8,000× bars

Counts from the store: 9.4M ES trades in July 2026 (~430k per session); 3.0 GB of raw MBP-10 book updates on Aug 3 alone. Multipliers are relative to one day of 1-minute bars.

How far back it goes.

Tick and bar history from July 2010. Full order-book history lands daily and keeps accumulating.

Bars + ticks

July 2010 → today

L2 depth

daily drops, growing

L3 order flow

daily drops, growing

20102014201820222026

48 CME futures markets.

Indices to crypto, full-size and micros — every one with tick-level history.

IndexESNQRTYYMMESMNQM2KMYM

RatesZNZBZFZTSR3SR1

EnergyCLNGRBHOMCL

MetalsGCSIHGPLMGCSIL

FX6E6J6B6A6C6S6N6M6LE7M6EM6AM6BCNH

CryptoBTCETHMBTMET

AgsZCZSZWHELEGF

CME futures markets

16 yrs

of tick history

512 GB

compressed market data

resolutions, bars to MBO

Simple pricing. Start free.

Roboquant 2.0 is live. Everyone can start free. Pick monthly or annual billing inside the app.

2 months free

AI credits cover strategy generation, compilation and tool calls. BYOK is available on Pro and above; requests billed to your provider do not consume Roboquant credits.

Start here

Free

Prove the first idea

$0/forever

Free plan

Create a strategy and run your first evidence-backed test.

Start free

Roboquant 2.0 is live. Everyone can start free.

2,000 AI credits once

5 OHLCV backtests per day

1 year of OHLCV history

Guided AI strategy builder
All Roboquant strategy templates
One year of OHLCV history
5 standard backtests per day
1 small optimization per day (one parameter, grid)
Basic performance metrics
Historical/T+1 RoboCharts
Interactive Replay preview
Community support

See all plan features

Guided AI strategy builder
All Roboquant strategy templates
One year of OHLCV history
5 standard backtests per day
1 small optimization per day (one parameter, grid)
Basic performance metrics
Historical/T+1 RoboCharts
Interactive Replay preview
Community support

Starter

Build and validate

$49/mo

Billed monthly

For turning an idea into a strategy you can validate seriously.

Get Starter