How is TradeWell accuracy measured?

Live rankers are evaluated on NDCG@30. When the evaluated list is shorter than 30 names, the product reports NDCG@20. Research architectures use walk-forward out-of-sample testing before any promotion to the public board. TradeWell does not publish fabricated win-rate or IC figures as product claims.

NDCG@30 is a ranking metric that scores how well the top thirty names in a list align with realized outcomes for that horizon. It rewards putting better-performing names higher in the list. If fewer than 30 names were evaluated, TradeWell reports NDCG@20 instead.

Does TradeWell publish a win rate?

No. TradeWell does not invent or publish fabricated win-rate, hit-rate, or IC percentages as product claims on marketing pages.

Is TradeWell investment advice?

No. TradeWell publishes research scores and forecasts, not investment advice. Nothing on the site or in the app is a recommendation to buy or sell any security.

How TradeWell accuracy is measured

Searching for how TradeWell accuracy is measured, or how the TW Quant Score is evaluated? This page is the plain-language answer: live rankings are scored with NDCG@30 (or NDCG@20 when the evaluated list is shorter than 30), research models must clear walk-forward out-of-sample tests before promotion, and marketing pages do not invent win rates.

TW Quant Score is research, not investment advice. It is not a guarantee of future performance and does not tell you what to buy or sell. Past model performance does not guarantee future results. TradeWell does not publish fabricated accuracy percentages as product claims.

What “accuracy” means here

TradeWell ranks stocks for a stated horizon. The useful question is not “did one price target nail the exact close?” — that is a noisy ask — but “did the top of the list put stronger-realized names higher?” That is why the production evaluation metric for live rankers is NDCG@30 (normalized discounted cumulative gain at thirty). When fewer than 30 names were evaluated, the product reports NDCG@20 instead of collapsing the headline to NDCG@10.

NDCG@30 rewards putting better-performing names higher in the top thirty for that horizon. It is a ranking metric, not a single-ticker win/loss badge and not a fabricated “beats the market X% of the time” claim.

Live board vs research program

Live rankers — the scores you see in the app — are evaluated on NDCG@30 (NDCG@20 if n_predicted < 30). Realized returns feed back into training.
Research architectures (foundation models, transformers, state-space models, and related challengers) use walk-forward out-of-sample testing before any model can reach live rankings — see the dedicated walk-forward validation explainer.
Promotion is operator-only. A research name is not “live” because it appears in a paper, a Slack card, or a challenger job. Never auto-promote.

What we refuse to invent

Marketing and App Store copy do not invent win rates, hit rates, or IC percentages as product claims. If a number is missing, we say missing. If a research fold is warming, we say warming. See the fuller model list and signal framing on the methodology page.

How this connects to ten horizons

Each horizon (1D through 5Y) has its own TW Quant Score, expected return, and model-implied price target. A strong 1W rank does not imply a strong 1Y rank — and accuracy is measured per horizon, not mashed into one fake all-time number. See the ten prediction horizons explainer, or screen by horizon in the stock screener by time horizon.

How TradeWell accuracy is measured — FAQ

How is TradeWell accuracy measured?: Live rankers are evaluated on NDCG@30. When the evaluated list is shorter than 30 names, the product reports NDCG@20. Research architectures use walk-forward out-of-sample testing before any promotion to the public board. TradeWell does not publish fabricated win-rate or IC figures as product claims.
What is NDCG@30?: A ranking metric that scores how well the top thirty names in a list align with realized outcomes for that horizon. Higher placement of better-performing names scores higher. Lists shorter than 30 report NDCG@20.
Does TradeWell publish a win rate?: No. We do not invent or publish fabricated win-rate, hit-rate, or IC percentages as product claims on marketing pages.
Is TradeWell investment advice?: No. TradeWell publishes research scores and forecasts, not investment advice, and does not tell users what to buy or sell.

Open the web app Get the iOS app