Historical Options-Chain Data Vendors — Raxx Use-Case Comparison

Status: research-only. This document does NOT constitute legal or tax advice. Before executing any data licensing agreement, consult a licensing attorney and a CPA licensed in Pennsylvania on COGS treatment. Last updated: 2026-04-23. Vendor prices and terms sourced as of that date — SaaS pricing and redistribution terms change frequently; verify freshness before any financial commitment.

TL;DR

Alpaca Algo Trader Plus (your current $90.75/mo subscription) does NOT solve the blocker. Historical options data on Alpaca goes back only to February 2024 — roughly 26 months of history as of today. The markov-fit backtest requires Jan 2007 – Dec 2025 (~18 years, covering 2008–09 stress, 2020 COVID, 2022 bear). The blocker is real.

ORATS at the $199/mo tier (Data API + historical) is the best-fit option for Raxx's strategy scope, with genuine 2007-to-present coverage, purpose-built iron-condor backtest tooling, and a confirmed use by spintwig (the practitioner ORATS cites as the closest analog to Application A in markov-fit-analysis.md). However, ORATS's standard terms prohibit redistribution of raw data to third parties under a single-user license. A commercial/SaaS arrangement requires a separate enterprise agreement — that conversation must happen with ORATS's sales team before any licensing commitment. Do not assume the $199/mo retail API covers SaaS use.

The second-best path, if ORATS's enterprise terms are prohibitive, is Databento OPRA for raw chain history (historical pay-as-you-go) combined with ORATS for the backtester-as-a-service model (ORATS backtester runs on their servers; Raxx presents results — different license surface than redistributing raw data). A licensing attorney should review both paths before contracting.

1. Executive Summary

Top recommendation: ORATS Data API (historical tier)

• Price to investigate: $199/mo (historical) + potentially a one-time $2,000 data export for the full 2007-to-present archive — amounts are from third-party sources and must be verified directly with ORATS; see §2 for sourcing.
• Why it fits: 2007-to-present options chain history; purpose-built iron condor backtest engine; same vendor used by spintwig (the exact practitioner source in markov-fit-analysis.md §6.3). Coverage verified for SPX/SPY/mega-cap universe.
• Hard gate: Standard retail API terms appear to restrict use to a single user for internal business purposes only, prohibiting redistribution to third parties (including Raxx's end users). Enterprise/commercial licensing terms are not publicly disclosed. Must be negotiated before adoption.

Runner-up: Databento OPRA historical (pay-as-you-go)

• Price to investigate: Historical data pay-as-you-go (pricing per dataset/query); live data (not needed for backtest-only use case) starts at $199/mo Standard plan. Historical archive starts March 28, 2023 — insufficient on its own for the 2007-to-present backtest requirement.
• Why it's a runner-up: Raw tick-level OPRA data, clean API, transparent pricing. History only goes to March 2023, meaning a multi-source strategy (Databento for recent + historicaloptiondata.com or CBOE DataShop for 2007–2022) would be needed. More assembly work but potentially cleaner commercial terms for internal backtesting use vs. redistribution.
• Hard gate: Databento distinguishes "internal use" from "redistribution." Using their data to run Raxx's MBT backtester internally (not displaying raw chain data to users) may qualify as internal use. A licensing attorney should assess.

Alpaca re-investigation verdict

Your current $90.75/mo Algo Trader Plus subscription does not solve the blocker. Historical options data on Alpaca covers only from February 2024 forward. The 2007–2019 training window specified in markov-fit-analysis.md §3, Application A, requires 13 years of data unavailable through Alpaca. Additionally, Alpaca's support documentation explicitly states that API data cannot be redistributed for business purposes — so even if history were adequate, Raxx's SaaS redistribution use case would require a separate commercial arrangement.

Total cost of ownership — worked estimate

The table below uses ORATS at the $199/mo historical API tier as the scenario. These are estimates pending enterprise pricing confirmation.

Key takeaway: At scale (100+ users), a $199/mo data license costs under $2/user and is easily absorbed by the existing margin. It shifts break-even from ~14–22 users to ~26–34 users — a meaningful but not fatal change. The real constraint is the license negotiation, not the dollar amount.

These figures use the cost structure from docs/business/founders-pro-pricing-rationale.md §2. Verify current ORATS enterprise pricing before incorporating into the model. Bring to the CPA for COGS classification.

2. Vendor Matrix

Prices marked with † were sourced from third-party review articles, community sources, or indirect search-result summaries as of 2026-04-23. Verify each directly with the vendor before any commitment. Prices marked with ‡ were sourced from the vendor's own documentation pages as indexed in search results as of this date.

3. Use-Case Fit Analysis

Raxx's specific requirements

From agent_prompt.txt and markov-fit-analysis.md:

• Strategy: Iron condor, credit spread, cash-secured put
• Underlyings: SPX/SPY (index), QQQ, IWM, mega-cap equities (AAPL, MSFT, etc.)
• Parameters: IVR > 50; 30–45 DTE; 15–20 delta; credit ≥ 30% width; 50% profit take; 2× stop; 21 DTE exit
• Backtest window: Jan 2007 – Dec 2025 (training: 2007–2019; test: 2020–2025)
• Data needed per trade: strike, expiry, bid/ask, greeks (delta, IV), underlying price — consistent with NBBO mid fill modeling per mbt-paper-trading-engine.md §4
• Frequency: EOD snapshots at a minimum; intraday beneficial for fill simulation

Vendor fit by requirement

Conclusion: For the markov-fit backtest (Application A, markov-fit-analysis.md), ORATS is the highest-fit single vendor. It has the history depth, the iron condor tooling, and the exact track record (spintwig backtests). The gap is exclusively on the commercial SaaS redistribution license. DeltaNeutral/historicaloptiondata is a viable raw-data supplement for pre-2023 history if Databento is the primary API layer. OptionsDX is useful only for SPX/SPY-specific backtests; it does not cover the broader equity universe needed for a production MQ-A signal.

4. License-Risk Summary

Understanding OPRA redistribution fees (applies to Databento, Polygon.io, and any OPRA-sourced vendor)

The Options Price Reporting Authority (OPRA) charges fees for redistribution of their consolidated options data. Key thresholds as of 2026-04-23:

• Redistributor flat fee: $1,500/mo to anyone who displays OPRA data externally to customers, in an app, tool, or website. Applies regardless of user count. Source: https://www.marketdata.app/education/options/opra-fees/
• Per-user display fee (real-time, non-professional): $1.25/user/mo. Source: https://www.marketdata.app/education/options/opra-fees/
• Per-user display fee (professional): $31.50/user/mo. Source: https://www.marketdata.app/education/options/opra-fees/
• Historical / delayed data (>15 min): No per-user display fee from OPRA. Source: https://www.marketdata.app/education/options/opra-fees/
• Non-display (backend) use: Still requires licensing even if data is not shown to users. Source: https://www.marketdata.app/education/options/opra-fees/

Implication for Raxx's MBT backtester: If Raxx uses historical options chain data exclusively for internal backtesting — running simulations on the server, never displaying raw chain data to users — this may qualify as "non-display" use rather than "redistribution." Non-display use still requires licensing from OPRA (or the OPRA- accredited vendor), but avoids the $1,500/mo redistributor fee and per-user display fees. A licensing attorney must confirm this interpretation before any contract.

Vendor-specific license flags

Red flags to raise with a licensing attorney before any contract

1. Does Raxx's use of historical options chain data to run MBT's backtest engine qualify as "internal non-display use" under OPRA licensing rules, avoiding the $1,500/mo redistributor fee?
2. Do ORATS's enterprise/commercial terms permit Raxx to run backtests on behalf of paying users (i.e., user triggers mbt.run_backtest on Raxx's server, ORATS data is consumed server-side, and the backtest result — not raw data — is returned to the user)?
3. Does Polygon.io's business/enterprise tier permit the same server-side-only use case, and what is the enterprise price?

5. Alpaca Re-Investigation

What we currently pay for

• Subscription: Algo Trader Plus at $90.75/mo (confirmed from dashboard 2026-04-23; see docs/business/founders-pro-pricing-rationale.md §9.1)
• Included per documentation:
• Real-time SIP data (stocks + options)
• Real-time options chains with greeks
• Historical bars and trades (stocks: multi-year; options: Feb 2024 – present only)
• 10,000 API calls/min
• Unlimited WebSocket symbols

The hard gap

Alpaca's historical options data begins February 2024 only. Source: https://docs.alpaca.markets/docs/historical-option-data

The markov-fit backtest requires: - Training window: Jan 2007 – Dec 2019 (from markov-fit-analysis.md §3, Application A) - Test window: Jan 2020 – Dec 2025 - Minimum: ~18 years of EOD options chain snapshots for SPX/SPY and liquid equities

Alpaca provides ~26 months of historical options data (Feb 2024 – Apr 2026) against a requirement of ~216 months. This is an 88% data gap by time period.

The redistribution constraint

Alpaca's support documentation explicitly states: "You cannot redistribute Alpaca API data for business purpose." Source: https://alpaca.markets/support/redistribute-alpaca-api

Even setting aside the history gap, the standard Alpaca terms prohibit the commercial redistribution scenario implied by Raxx's SaaS model. Note: "redistribution" in the context of MBT's internal backtester (where options data is consumed server-side and a backtest result is returned to the user) may or may not constitute "redistribution" under Alpaca's specific definition — a question for a licensing attorney. However, given the history gap is fatal on its own, this is a secondary concern.

Verdict

The blocker is real. Alpaca Algo Trader Plus does not solve it. A supplemental historical options data source is required for any backtest window predating February 2024. The existing $90.75/mo subscription remains valuable for:

• Real-time options quotes and chains (for MBT's live paper-trading fill simulation)
• Real-time market data for MQ-A regime signal inputs (VIX, SPY)
• Post-Feb-2024 historical options data as supplemental confirmation data

But it cannot replace a dedicated historical options data vendor for the markov-fit backtest or any backtesting covering market stress periods prior to 2024.

6. Cost Delta to Raxx Pricing

Base case from docs/business/founders-pro-pricing-rationale.md §2:

• Break-even (lean fixed costs): ~14 users
• Break-even (with attorneys/CPA active): ~22 users
• Margin at 100 users: ~$13.53/user/mo
• Margin at 250 users (Founders cap): ~$16.30/user/mo

Scenario: Add ORATS at $199/mo (retail historical tier, pending enterprise confirmation)

This is an illustrative estimate only. Enterprise pricing may differ materially.

At 250 users (the Founders cap), a $199/mo data license costs $0.80/user/mo — well within the existing margin. It shifts Founders break-even from ~14–22 users to ~25–33 users. That is material at the pre-launch stage but manageable once the cohort reaches 30+ paying users.

This analysis is sensitive to enterprise pricing. If ORATS's enterprise/commercial tier (required for SaaS use) costs $500/mo or more, the numbers change:

At $1,000/mo+ (Databento Plus tier), the economics become strained at the Founders $19/mo price point. At $500/mo or below, the model remains viable.

Flag for the CPA: A data licensing fee for a SaaS product is typically classified as COGS (cost of goods sold) on a gross-margin view of the business, similar to Heroku and Alpaca. Confirm classification and treatment for the PA LLC before committing.

7. Questions for Attorney and CPA

Questions for a licensing attorney (IP / contract focus)

1. OPRA non-display vs. redistribution classification. If Raxx's MBT engine consumes historical options chain data server-side — no raw data is displayed to users; only computed backtest results (P&L, win rate, drawdown) are returned — does this qualify as "non-display use" under OPRA's fee schedule, avoiding the $1,500/mo redistributor fee? Cite: https://www.marketdata.app/education/options/opra-fees/

2. ORATS enterprise terms review. If Raxx executes ORATS's enterprise/commercial agreement, what specific provisions would govern: (a) running backtests on behalf of paying users (server-side computation, results returned); (b) caching ORATS data in Raxx's own database for repeated backtest access; (c) displaying derived metrics (e.g., "regime state" probabilities derived from ORATS IV data) in Antlers? Which of these uses requires a redistribution license vs. an internal-use license? Source: https://orats.com/terms-conditions

3. Polygon.io business/enterprise tier review. Same questions as #2 for Polygon.io's enterprise tier, if ORATS terms are not acceptable. Polygon.io's standard terms explicitly prohibit commercial use and redistribution; is the enterprise tier's scope sufficient for Raxx's use case, and at what price? Source: https://polygon.io/legal/market-data-terms-of-service

4. Alpaca redistribution clarification. Does Alpaca's prohibition on "redistributing for business purposes" apply to: (a) raw options chain data passed through to a user's browser, or (b) server-side use of data to simulate fills in MBT? The distinction matters for the MBT real-time fill simulation path (not the backtest path). Source: https://alpaca.markets/support/redistribute-alpaca-api

Questions for a CPA (PA-licensed, SaaS/digital-goods focus)

1. COGS classification for data licensing fees. Is a monthly options data licensing fee (e.g., ORATS at $199–$500/mo) appropriately classified as COGS or as an operating expense for a Pennsylvania LLC treated as a disregarded entity or S-Corp? How does this classification affect the gross margin view in founders-pro-pricing-rationale.md?

2. Variable vs. fixed cost treatment. A data licensing fee that does not scale with user count (fixed monthly fee) is a semi-fixed cost, similar to Alpaca. Confirm whether the IRS and PA FTB treat a flat-rate data subscription differently from a per-unit COGS for purposes of Schedule C or Form 1120-S.

3. Prepaid data archive classification. If Raxx purchases a one-time historical data archive (e.g., ORATS one-time export cited at ~$2,000† for full 2007-to-present history) in addition to a monthly subscription, is that a capital expenditure to be amortized or an expensed operating cost? What amortization schedule applies?

8. Recommendation

This section presents research findings only. It does NOT constitute legal or business advice. Take this packet to a licensing attorney and CPA before any contract commitment.

Research finding: A two-step process is required

Step 1 — Commercial license confirmation (licensing attorney, 1–2 weeks). Before any vendor selection can be finalized, a licensing attorney must confirm whether Raxx's specific use case (server-side backtest computation, results returned to users, no raw data displayed) qualifies as "internal use" or triggers "redistribution" fees under each candidate vendor's terms. This is the true blocker — not the dollar amount.

Step 2 — Vendor selection (after license path is confirmed). If ORATS enterprise terms permit server-side SaaS use at under ~$500/mo, ORATS is the recommendation: - Coverage: 2007 – present, matching the backtest requirement exactly - Quality: same data source used by spintwig (the practitioner reference in markov-fit-analysis.md §6.3) - Built-in tooling: iron condor backtester with 300M+ pre-computed runs available - Integration path: REST API aligns with Raxx's Flask/Celery stack

If ORATS enterprise terms are not viable (prohibitive cost or licensing posture), the fallback path is: 1. Databento OPRA for historical data Mar 2023 – present (PAYG, pending attorney confirmation of "internal use" classification) 2. DeltaNeutral / historicaloptiondata.com for 2002–2022 archive (flat file, per-dataset purchase, pending confirmation of commercial use terms) 3. These two sources would provide coverage from 2002 to present at potentially lower total cost, at the expense of more data assembly work.

IEX Cloud: Disqualified — service has shut down. Polygon.io (standard): Disqualified under standard terms for commercial use. Tradier: Not a historical data provider; powered by ORATS underneath; not a path. Bloomberg / Refinitiv / FactSet: Out of scope at current scale. OptionMetrics IvyDB: Academic posture; commercial license bespoke; investigate only if ORATS and Databento both fail. OptionsDX: SPX/SPY only; no API; flat-file only; acceptable for narrow SPX-only backtest validation but not for the full Raxx equity universe.

Summary of required actions

1. Engage Matthew Crosby (IP, already engaged — per reference_attorneys.md) or a contract/licensing attorney to review ORATS's enterprise terms and confirm the OPRA non-display use classification. Timeline: before any data vendor contract.
2. Request ORATS enterprise pricing. Contact: https://orats.com/faq (contact form); mention SaaS backtesting use case, ~250 end users, server-side only computation. Do not sign the standard retail API agreement for SaaS use.
3. Bring the cost delta analysis (§6 above) to the CPA for COGS classification before incorporating the data line item into Raxx's pricing model or Founders Pro economics.
4. Once license path is confirmed, file an implementation card with product-manager to obtain ORATS test credentials, validate data coverage against the 2007–2025 backtest spec, and integrate the historical chain feed into mbt.run_backtest.

9. Jurisdiction Flags

• Federal (OPRA): OPRA fee schedule is set by the Options Price Reporting Authority (a national market system plan under SEC oversight). The $1,500/mo redistributor fee and per-user display fees are federal in nature — they apply regardless of Raxx's state of incorporation. Source: https://www.opraplan.com/
• Delaware / Pennsylvania: Entity formation state does not change OPRA fee applicability. It is a federal market data licensing regime.
• EU / International: Most US options data vendors (ORATS, Databento, CBOE DataShop) do not explicitly flag EU data-transfer restrictions in publicly available terms. However, if Raxx accepts EU users, GDPR Article 46 (data transfers to third countries) may interact with vendor data processing agreements. This is a separate flag from OPRA licensing. Confirm with the licensing attorney before onboarding EU users.

10. Timing / Deadlines

No hard regulatory deadlines apply directly to vendor selection. However:

• Dependency on markov-fit backtest (from markov-fit-analysis.md): Application A (Regime-Aware Entry Gate) is Priority 1. The data licensing review is the stated hard blocker before any backtest work begins. Resolving this within 2–4 weeks unblocks ~3–4 weeks of data-scientist backtest work, then 3–4 weeks of feature development. Total estimated delay to regime-gate feature: ~2–3 months from today if licensing confirmation is prompt.
• ORATS pricing freshness: Vendor SaaS pricing changes without notice. The $199/mo figure for the historical API tier comes from third-party review sources (FlashAlpha, Bullish Bears) and must be verified directly with ORATS before any contract.
• Databento pricing freshness: The June 3, 2025 pricing restructure (Standard at $199/mo, live data only; historical PAYG preserved) is sourced from Databento's own blog but should be re-verified since it was published before today's date.

Sources

Before acting on any vendor recommendation in this document, consult: - A licensing attorney (IP / contract focus) for OPRA classification and vendor terms review (Matthew Crosby is currently engaged for IP — confirm whether data licensing is in scope, or identify a contract specialist). - A CPA licensed in Pennsylvania for COGS classification of a recurring data licensing fee and one-time archive purchase treatment. Jurisdiction: federal (OPRA), Pennsylvania (entity), potential EU flag if international users onboarded.