AI writes the algorithm. We test it on real futures data and score it out of 10. Score 7 or higher and it is ready to run.
3futures markets
15 yearsof market data
7+score means ready to run
Choose how your AI works.
$99/month
Bring your own AI
Connect Claude Code, Codex, Cursor or another coding agent through MCP and OAuth. Includes $99 of compute every month. Your coding subscription covers the AI.
I built 3 versions and tested them on 11 years of NQ. The opening range version scored 7.4.
Make the stop tighter.
Ask for a change
Your idea
7.4Ready to run
Opening range pullback
1,412 trades, PF 1.62
6.2Lead
Gap and go
868 trades, PF 1.21
4.1Rejected
First hour fade
2,040 trades, PF 0.94
Example session, not real results.
Fifteen years of history, down to every order.
No data feed to buy or set up. It all stays on our servers.
Need more tick or MBO history? Buy a bigger package. For pricing, contact support.
OHLC barsOpen, high, low, close
15 years
Tick dataEvery trade
1 year
MBO dataEvery order in the book
3+ months
15 years agoToday
Most ideas only look good.
A quick backtest flatters almost any idea. Ours adds real costs, blocks lookahead and runs our extensive scoring framework to tell if the strategy really holds up, before you trade it.
A quick backtest+$84,200
After our scoring framework−$6,900
Lookahead found
Costs ignored
Overfitted
2.1out of 10 Rejected
Example, not a real strategy.
Score 7 or higher? It is ready to run.
Every strategy the AI builds goes through our extensive scoring framework: costs, lookahead, overfitting, walk-forward and more. A 7 or 8 means it held up across all of it.
Opening range pullback
NQ futures, 2015 to 2026, 1,412 trades
7.4Ready to run
Lookahead10.0
Cost stress9.1
Overfitting8.8
Walk-forward7.2
Monte Carlo6.4
Parameter stability4.0
Example report, not a real strategy.
<5Unproven0 to 5.0
5.1Common5.1 to 5.9
6.0Uncommon6.0 to 6.9
Ready to run
7.0Rare7.0 to 7.9
8.0Epic8.0 to 8.9
9.0Legendary9.0 to 10.0
7 or higher means ready to run. Rarity follows the final, capped score, and incomplete evidence stays provisional. See how scoring works
Results from our own strategies.
Score
9.1
Net result
+$269,965
Profit factor
1.40
Win rate
41%
Trades
1,450
Max drawdown
$16,630
20162026
2016+$5,750
2017+$4,460
2018+$18,455
2019+$6,015
2020+$18,625
2021+$21,850
2022+$55,855
2023+$28,185
2024+$33,340
2025+$30,850
2026+$46,580
Score
9.0
Net result
+$216,370
Profit factor
1.40
Win rate
41%
Trades
1,125
Max drawdown
$19,785
20162026
2016+$7,075
2017+$1,155
2018+$10,930
2019+$6,265
2020+$25,705
2021+$18,940
2022+$41,010
2023+$23,725
2024+$36,880
2025+$26,110
2026+$18,575
Score
8.2
Net result
+$44,155
Profit factor
1.75
Win rate
64%
Trades
200
Max drawdown
$6,440
20202026
2020+$5,566
2021+$2,635
2022+$3,669
2023+$1,188
2024+$13,357
2025+$8,484
2026+$9,255
Score
8.0
Net result
+$47,045
Profit factor
1.54
Win rate
51%
Trades
247
Max drawdown
$11,420
20202026
2020+$1,390
2021+$2,850
2022+$9,015
2023+$2,705
2024+$13,040
2025+$13,075
2026+$4,970
Backtested results per 1 contract after trading costs, through mid to late 2026. Backtests are hypothetical and have limits: no real money was at risk in them, and they are chosen with hindsight. Most ideas fail, and these are survivors. Past results do not predict future results, and trading futures can lose you money.
Scored 7 or higher? Pick how to run it.
Both roads open soon.
7.4Ready to run
Recommended
We run it.
Deploy on our platform
Hosted and connected to Topstep and Tradovate. Nothing trades without your approval.
Latency to the brokerHandled by us
BandwidthHandled by us
Stays onlineHandled by us
For experts only
You run it.
Export and run it yourself
Runs on your own server. For experts only.
Latency to the brokerHome: too slowVPS: fast
BandwidthHome: sharedVPS: high
Stays onlineHome: until it sleepsVPS: all day
Bars are an illustration, not a measurement.
Your latency to the brokers, right now
Topstep
Tradovate
Measured from this browser, only when you press the button.
What is BenchAlgo?
BenchAlgo is where you turn trading ideas into tested algorithms with AI. Describe an idea in plain words, your AI writes the algorithm, and we test it on real CME market data for ES, NQ and GC with trading costs, check it for lookahead and overfitting, and score it out of 10 with our extensive scoring framework. A score of 7 or higher means the strategy is ready to run. Read the product guide
Serious about finding your edge?
Stop clicking buy and sell and hoping for progress. Turn your ideas into clear rules, test them on real market data, and find out what actually holds up.
Open BenchAlgoPrivate preview. Registration is currently closed.
def on_bar(ctx): if ctx.signal: ctx.submit("buy")Illustration, not a real trade.