The edge is only half the desk. This page documents the other half: how much of your risk capital each idea is worth, where a loss is cut, what break-even actually is at our payoff ratio, and the losing streaks you should expect if the model is working correctly.
Full Kelly maximises long-run growth on the assumption your probabilities are exactly right. They never are. Every published weight is a quarter-Kelly fraction of risk capital, floored at 0.5% and capped at 6%. That cap exists so a single mis-calibrated idea can bruise an account but cannot end one.
Each idea publishes an invalidation price at the expected-move distance divided by 1.6. That asymmetry is the whole risk model: a winner realises roughly 1.6 times what a loser costs, which is why a hit rate in the mid-50s compounds instead of bleeding. Trading the entry without the invalidation is a different strategy with a different, worse expectancy.
At that payoff ratio the desk breaks even at a 38.46% hit rate. Every performance number we publish is quoted against that bar, not against a coin flip, because 50% is the wrong benchmark for an asymmetric payoff and quoting it would flatter the model.
A mid-50s hit rate produces losing runs. Across the validation sample, strings of six consecutive losing conviction ideas occur regularly and strings of nine are not rare. If a two-week cold streak would change how you size, the honest answer is that this desk is not sized for you yet.
Ideas sharing a sector, a macro sensitivity or the same volatility exposure are not independent bets. Weights are trimmed when the card concentrates, so thirteen published ideas never represent thirteen units of uncorrelated risk.
Every accuracy figure on the site carries the sample it came from and a 95% confidence interval. A hit rate without a sample size is a number chosen for how it looks, and you should treat it that way everywhere you see one — including here.
Ascend publishes quantitative research, not investment advice. Nothing here is a recommendation to buy or sell any security, we are not a registered investment adviser, and past model performance does not predict future returns. Every idea can lose. Size positions accordingly and never risk capital you cannot afford to lose.