Methodology — How Polymarket Bots Ranks (In Full)

The five criteria (100 points)

Every product ranked on polymarketbots.io is scored against the same five-axis rubric. Weights are fixed across products. A product's final rating is the weighted average, normalised to a 5-star scale for display.

30%

On-chain verifiability

Polymarket settles on-chain. We use that. Builder-tagged wallets, processed volume, wallet-level P&L attribution — if it can be traced, we trace it.

Full credit: Public, builder-tagged wallet that anyone can verify against Polymarket's Builder API. Volume and P&L reproducible on-chain or via the public Polymarket Builder leaderboard. Vendor stats consistent with on-chain data.
Zero credit: Vendor stats with no traceable wallet. Hyper-precise self-reported numbers ("1.2% verified daily ROI", "60.8% win rate over 102 picks") published without methodology or wallet attribution. Vendor self-ranking inside their own content.
What would change a rank: A vendor publishing a builder-tagged wallet moves up materially. A vendor whose claimed wallet stops matching their published stats moves down or gets a verification-flag.

20%

Data infrastructure

Polymarket has no commercial historical-data archive — you record it continuously or you don't have it. Depth and freshness of order-book history is a moat.

Full credit: Multi-billion-row order-book archive captured continuously, with provenance and capture dates. Freshness measured in seconds, not days. Replayable for backtesting at quote-tick resolution.
Zero credit: Vendor depends on sporadic Polymarket-API pulls or third-party aggregators with unclear provenance. Backtesting limited to OHLCV-style summaries without order-book microstructure.
What would change a rank: Continuous capture starts moving the needle as the archive ages. Documented data gaps or capture downtime move it down.

20%

Execution speed

Signal-to-fill latency, gasless support, wallet isolation. Bots running off-shelf RPC providers without colocation are capped.

Full credit: Colocated execution near Polymarket's order-flow infrastructure. Documented signal-to-fill latency in milliseconds. Gasless trading support. Per-strategy wallet isolation for blast radius.
Zero credit: Off-shelf RPC, generic cloud-region execution, no documented latency. Trade execution on retail-style RPC providers shared with thousands of other consumers.
What would change a rank: Moving execution from generic cloud to a colocated venue is an immediate upgrade. A documented loss event traced to slow execution is a downgrade.

15%

Strategy flexibility

Pre-built vs custom strategies, real-time authoring, backtesting depth, scripting support, walk-forward validation.

Full credit: Pre-built strategies plus a real-time authoring layer (e.g., live strategy authoring as a third mode beyond pre-game and in-play). Walk-forward validation. Drag-and-drop or scripting interface, user's choice.
Zero credit: Fixed strategy presets with no customisation. No backtesting interface. No way to validate a strategy before deploying capital.
What would change a rank: Shipping a real-time authoring mode or a quality backtesting interface is an upgrade. Removing previously available customisation is a downgrade.

15%

Transparency

Published methodology, openness about losses, identifiable founders, public engineering write-ups. Vendor-published stats without methodology are treated as unsourced.

Full credit: Public methodology page (separate from marketing copy). Identifiable team with verifiable bylines. Published losses alongside wins. Engineering blog or whitepaper that lets a third party reproduce claims.
Zero credit: Anonymous team. Stats without methodology. Only wins published, never losses. "Verified by our internal team" as the entire transparency story.
What would change a rank: Naming founders, publishing methodology, or shipping a postmortem of a losing strategy moves a vendor up. Vendor moving from public to anonymous moves them down.

Exclusions

Several "Polymarket bots" appear in marketing surfaces but fail our verification gate. They are not ranked. The disqualifying patterns are specific and applied consistently:

No traceable wallet. If a vendor publishes performance stats but cannot point at a Polymarket builder-tagged wallet whose on-chain activity reproduces those stats, we cannot rank them. We do not bury — we omit.
Stats inconsistent with on-chain data. If the vendor wallet exists but its on-chain volume or P&L does not match published claims, the product is excluded until the discrepancy is reconciled.
Pure vendor-self-ranking inside vendor blogs. Products whose only "ranking" evidence is their own listicle on their own marketing surface ("we tested #1!") do not clear the bar. We treat that as marketing, not evidence.
Anonymous teams + unverifiable claims. Anonymous teams alone are not disqualifying (open-source projects with pseudonymous maintainers are fine). Anonymous teams plus unverifiable performance claims are.

Conflicts of interest

Minmax.one is our our editorial pick. The same organisation that operates this editorial site also operates minmax.one. This is the biggest single conflict on the site and we treat it accordingly: disclosed prominently on every page that mentions minmax, including the homepage card, the full review, all comparison pages, and this methodology page. Minmax's position is scored against the same five criteria as every other product — the ownership relationship does not raise or lower the rank, but you deserve to know it exists.
Affiliate programs are rare in this category. Polymarket bots mostly do not run affiliate programs the way centralised-crypto bots do. Where any product on this site has an affiliate relationship, the affiliate link is tagged rel="sponsored" and visible by inspection.
On-chain verification via the Polymarket Data API. Vendor performance claims are cross-checked against the public Polymarket Builder leaderboard at data-api.polymarket.com/v1/builders/leaderboard. Where a vendor does not surface a Builder-tagged wallet we can trace, we flag the unverifiability rather than rely on vendor-published numbers.

Update cadence

Every product is re-scored quarterly. The "Updated" date on each review reflects the most recent full re-score. We also re-score off-cycle when one of these catalysts fires:

A vendor wallet stops producing on-chain activity matching their marketing claims.
A documented architecture change — a vendor switching from one execution model to another, a new data-capture stack, a different AI/ML pipeline.
A change in Polymarket Builder API status that materially affects the bot's ability to operate.
A regulatory change affecting Polymarket access in a major jurisdiction.
A public incident — security, custody, or strategy failure — relevant to a ranked product.

What would change our minds about a rank

Per-criterion, the "what would change a rank" line above tells you the move. At the page-as-a-whole level, the single thing that would most rapidly move a product up our ranking is publishing a builder-tagged wallet and inviting independent verification of its P&L. The single thing that would most rapidly move a product down is unexplained divergence between published stats and on-chain activity.