Options Strategy Hub — SPY options backtesting with AI strategy analysis

Four labels, never blurred

Every AI statement says what kind of claim it is

Options Strategy Hub computes the statistics itself. The AI only interprets them, and the app checks every citation against the run before showing it to you. It can still be wrong: treat every statement as research, not advice.

Fact from the run

Recorded by the backtest, or exact arithmetic on it: counts, P&L, dates, trade IDs.

Entries in the 1 PM hour won 1 of 14 trades and lost $762.

Derived observation

A statistic the app calculated. It describes this sample; it doesn't prove a cause.

Winners entered with stronger relative volume (1.8× vs 1.3×), d = 1.73.

Hypothesis to test

An idea worth trying. It is never called "better" until it has been backtested.

Stop taking new entries after 1:30 PM.

Actual test result

Only allowed when a backtest of the changed strategy actually ran. Otherwise it is relabelled a hypothesis.

v2 vs v1 on the same period: 55 trades vs 96.

Example statements come from the sample run shown in the screenshots below.

The research loop

From a sentence to a tested strategy version

Describe an idea, run it, ask why it worked or failed, test one change, and keep every version. The original strategy is never overwritten.

STEP 1 · DESCRIBE

Say what you want to trade, in plain English

The AI gets the live indicator catalog and writes the strategy in the app's own format. It then goes through the same strict validation as anything you type, so a typo'd indicator or a stop loss written as 50 instead of 0.50 is rejected with a fix you can send back.

Create a strategy with AI panel with a plain-English idea typed into the Your idea box.

STEP 2 · BUILD

Review it in a builder that reads like sentences

Entry rules on any timeframe from 1 to 60 minutes, trend-regime filters, Greek filters for contract selection, stops, targets, trailing and breakeven exits, and time and risk limits. A live summary checks the configuration as you go.

104 rule operands, from VWAP distance to opening-range levels
Delta, gamma, vega, theta and IV filters
Optional market-structure gate: observe or enforce

Strategy builder showing direction, trend regime and an entry rule, with a strategy summary panel on the right.

STEP 3 · ASK

Ask what happened, and get an answer with receipts

Answers are organised as what happened, what appears different, what we could test next and what we still don't know. Each statement is labelled and linked to the numbers it came from. Invented figures and unknown trade IDs are flagged.

"Why did March perform poorly?"
"Are afternoon entries hurting the strategy?"
"Was profit concentrated in a few trades?"

AI Strategy Analysis answer: a summary, then labelled Fact from the run statements about time-of-day results, each checked against the run.

STEP 4 · REFINE

Get research cards, not a silently rewritten strategy

Each proposal states the observation, the evidence, the exact field it would change, what it might help, what could be lost, and how much the sample can support it. Choose Test this change and the app saves a new version beside the original, validated and ready to run. Nothing changes until you choose it.

Refinement card: Stop entering after 1:30 PM, marked hypothesis to test with medium confidence and 33 trades, showing the change from 15:30 to 13:30 and a Test this change button.

STEP 5 · COMPARE

Run the new version and see exactly what changed

Compare versions side by side. The app lists every rule and setting that differs, from each run's recorded configuration. A better number on the same dates is still one sample, so the app says so and points you to an out-of-sample re-test.

Compare page showing v1 and v2 of VWAP Pullback Calls side by side: net P&L, trades, win rate, profit factor, drawdown and averages.

Built against curve-fitting

An assistant that argues for caution

"Improve my strategy" is easy to turn into an optimizer that memorises the past. Options Strategy Hub applies these checks itself, whatever the AI claims. They reduce the risk of fooling yourself; they cannot remove it.

Confidence follows sample size

Under 10 trades, confidence is capped at low. Under 30, at medium. The cap is shown next to what the AI claimed.

sample_size < 10 → low

One change at a time

Proposals that change several things at once, or jump a threshold far from its current value, are flagged as hard to attribute.

15:30 → 13:30

Concentration warnings

If the whole profit came from one trade or one day, every proposal carries that warning.

net without best trade ≤ 0

The strategy must match

A change is rejected if its "from" value doesn't match your strategy. An AI that misread the strategy can't slip a change through.

from must equal current

Nothing is overwritten

Refinements become new versions: v2, v3… The original strategy and its backtests are untouched.

VWAP Pullback Calls v2

Lineage you can audit

Each version records its parent, the run that suggested it, and the exact change, which the app computes itself.

parent_strategy_hash

Export bundle

One zip for people, programs and AI

Every run exports as a self-describing bundle. Spreadsheet users, Python notebooks and AI assistants each get a format made for them, with the same numbers underneath.

For you

A plain-English README, headline results already calculated, and every report as a spreadsheet.

For your code

Full datasets as Parquet with stable column names such as net_pnl and trade_id.

For an AI assistant

A compact analysis context and one structured record per trade, with credentials and file paths removed.

options_strategy_hub_run.zip ├─ README.md start here ├─ run_summary.json headline results ├─ strategy.json edit and re-import ├─ export_manifest.json what's inside, keys, units ├─ trades.csv daily_summary.csv … ├─ reports/ exit reason, month, DTE, │ winners vs losers, skips ├─ data/ Parquet, canonical names │ ├─ trades.parquet │ ├─ skipped_setups.parquet │ └─ option_selection_debug.parquet └─ ai/ ├─ analysis_context.json └─ trades.jsonl

Pricing

A real backtester on Basic. The whole research loop on Pro.

Both plans backtest SPY options on SPY historical data from your own Databento account: the app shows Databento's quoted cost of each run before you approve a maximum, and Databento bills that data to you separately from your Options Strategy Hub subscription. Subscriptions renew automatically; cancel anytime from the billing portal and keep access to the end of the paid period. Payments are handled by Stripe. Prices in USD, taxes may apply.

Basic

$59/month · or $588/year

For building and testing strategies by hand, inspecting every trade, and using AI to understand what happened.

500 standard backtests a month
25 one-second replay runs a month
25 saved strategies · 1 job at a time
Strategy builder, calls and puts, 0DTE and 1DTE
Indicators, regime filters, stops, targets, time exits
Trade-level results, equity and P&L analytics, CSV exports
AI strategy analysis
AI refinement (limited)
Winner vs loser and regime analysis
Optimizer and walk-forward

Start with Basic

Pro

Founder

FIRST 100 CUSTOMERS

$699for your first year

Every Pro feature for $699 for the first year. After that it renews at the normal Pro annual price of $948/year.

Everything in Pro
Founder price locked for year one
Renews on Pro annual, cancel anytime

Claim a founder seat

Compare plans	Basic	Pro
Standard backtests	500/month	2,500/month
1-second execution replay	25 runs/month	250 runs/month
Saved strategies	25	150
Concurrent jobs	1	3
AI strategy analysis	Included	Full
AI strategy refinement	Limited	Full
Winner vs loser, attribution, regime, entry/exit	Not included	Included
Strategy comparisons	Basic	Full
Optimizer	Not included	25 jobs/month
Walk-forward analysis	Not included	8 jobs/month
Execution diagnostics and advanced exports	Not included	Included

A standard backtest covers up to about six months of sessions with a normal strike window; larger runs count as more than one. Same-day and live data are not included in either plan. Founder seats are limited to 100 and allocated by Stripe in the order checkouts complete. See the Refund & Cancellation Policy.

Questions

Frequently asked

Does the AI calculate my P&L?

No. The backtest engine calculates every number. The AI reads the app's pre-computed evidence and explains it. If it cites a figure that isn't in your run, the statement is flagged. The AI can still misread evidence or propose a worse strategy; nothing it says is investment advice.

Which AI assistant does it use?

When live AI is enabled on the server, requests go to our configured provider (Anthropic) with only your run's computed evidence and your question, as the Privacy Policy describes. Otherwise the app shows you the prompt to paste into an assistant you already use, and checks the reply the same way.

Can the AI change my strategy?

Not by itself. Proposals stay proposals until you choose one. Then the app validates the change and saves it as a new version beside the original.

What does Options Strategy Hub simulate?

Buying single SPY options, calls or puts, from same-day expiry out to weekly, on historical data. SPY is the only underlying in the Founding Beta — no QQQ, IWM, single stocks or futures. It doesn't sell options or trade spreads in the main backtester. Contract selection uses your DTE, strike window, premium range, spread limit and Greek filters. It never places real orders and does not connect to a broker.

Where does the market data come from, and what does it cost?

From your own Databento account. You connect your Databento API key once; each backtest then fetches exactly the SPY 1-minute bars and SPY option-chain sessions it needs into an isolated workspace and deletes them afterwards. Before anything is fetched, the app asks Databento for the price of that data, shows it to you, and only runs after you approve a maximum — it never spends past that maximum. 1-second execution adds a separate, capped amount. These Databento charges are billed by Databento under your own agreement with them and are separate from your Options Strategy Hub subscription; we cannot refund them.

How are fills modelled?

Entries are assumed to fill at the ask (or between mid and ask) and exits at the bid, plus the slippage and commission you configure. The optional 1-second replay prices entries and exits off the second-by-second historical quote tape with a configurable delay. These are modelling assumptions, not prices you could have traded at: spreads, liquidity, market impact and latency make live fills differ, sometimes materially. See the Risk & Backtesting Disclosure.

What counts as one backtest?

A standard backtest covers up to about six months of trading sessions with a normal strike window. Longer periods, the full option chain, several expiries or many rule conditions count as more than one, so the monthly allowance reflects the work each run takes. The Account page shows your usage.

Can I change or cancel my plan?

Yes. Upgrade, switch between monthly and yearly, update your card or cancel from the billing portal. A cancelled plan stays active until the end of the period you paid for and does not renew. Refund terms are in the Refund & Cancellation Policy.

Is this investment advice?

No. Options Strategy Hub is research and simulation software made by SpyWave LLC. It is not a broker, investment adviser or trade-execution platform, and nothing on this site or in the app is investment advice. Backtests are hypothetical: they describe what a set of rules did on historical data and don't predict future results. Options, and 0DTE options in particular, can lose their entire value.

Can I delete my account and take my data with me?

Yes. Account → Privacy & Data lets you download everything the app holds about you (strategies, backtests, cost history, consent record) as a zip and delete the account yourself. The Privacy Policy explains exactly what is removed and what must be retained.

Test a SPY options idea on historical quotes. Then ask what happened.