How we score Polymarket tools
Every ranking on this site is computed from per-dimension scores using the weights below — no hand-picked orderings without a stated rationale, no scores without a written justification.
The rubric
| Dimension | Weight | What it measures |
|---|---|---|
| Security & Custody | 20% | Who holds the keys, whether you can export them, how permissions are scoped, 2FA, operational history, and audits. It carries the heaviest weight because every automated tool needs authority to sign orders. |
| Execution Speed | 15% | Documented execution architecture and speed claims, vendor and third-party. No tool in this niche publishes audited latency, so no tool scores a 10. |
| Configurability & Risk Controls | 15% | Depth of documented controls: sizing modes, odds filters, per-market and daily caps, take-profit/stop-loss/trailing exits. |
| Feature Richness | 15% | Coverage of the 11-capability matrix used across this site — copying, automation, exits, auto-claim, discovery, group features, referrals. |
| Reliability & Uptime | 10% | Verifiable order flow, incident record, candor about failure modes, status/uptime signals. |
| Platform Surface & Mini App | 10% | Where you can actually use it: chat bot, Mini App, web, desktop, mobile — and how complete each surface is. |
| Track Record & Reputation | 10% | Time in market, incident history, entity transparency, independent coverage. |
| Fees & Value | 5% | Effective cost at realistic odds — flat rates, curves, subscriptions — and fee transparency. |
Weights in code: Security & Custody 20% · Execution Speed 15% · Configurability & Risk Controls 15% · Feature Richness 15% · Reliability & Uptime 10% · Platform Surface & Mini App 10% · Track Record & Reputation 10% · Fees & Value 5% (sums to 100). The weighted overall is computed at build time from these scores — it is never typed in by hand.
Evidence rules
- Every factual claim about custody, fees, capabilities, or incidents carries a source URL with the date we checked it.
- Negative claims about a competitor without a dated source don't ship. Where sources conflict (it happens), we publish the conflict, attributed to each side.
- Language discipline: "documented" (in vendor docs), "advertised" (marketing claims), "reported" (third-party). We don't run hands-on latency benchmarks yet, and we don't pretend to.
- Every dimension score has a written rationale, visible on the tool's review.
- Facts are re-verified at least quarterly; every change lands in the changelog.
What we refuse to publish
- Fabricated user testimonials, star ratings, or user counts.
- Profit claims or "risk-free" framing — no tool on this site guarantees profits, and prediction-market trading can lose money.
- Review/AggregateRating structured data — as a site with a disclosed connection to one listed product, self-serving review markup would violate Google's guidelines, so our build fails if it's ever added.
- Unsourced negative claims about any listed product.
The conflict of interest, stated plainly
Prediction Bot Review is operated by the team behind PolyBot, one of the tools reviewed on this site. We may earn when readers choose PolyBot. Every verdict follows our published methodology, and factual claims about competitors carry dated sources.
Prediction Bot Review exists because the team behind PolyBot believes it wins an honest comparison. The scoring rubric, the evidence rules, and the fact that competitors sometimes win categories on this site are how we make that belief checkable. If you find an error, the about page explains how to reach us; corrections are logged publicly.