Contrarian
An approach that bets against the prevailing view or the statistical prior; LDBD quantifies that prior as base_rate_up and adds scoring weight when you go against it.
Formula
direction_sign = +1 if you predict 'down', -1 if you predict 'up'
contrarian_multiplier = 1 + 2 × (base_rate_up − 0.5) × direction_sign
(final value clamped to 0.5 ~ 2.0)Contrarian investing means standing on the opposite side of the crowd — selling when everyone is piling in to buy, buying when everyone is dumping in fear. The idea is simple; the hard parts are two. First, figuring out where 'the crowd' actually sits right now, and second, checking whether the bet against it ever actually paid off.
LDBD pins down that 'crowd direction' as base_rate_up — the historical probability that a given asset rose over a given timeframe. Predict the opposite of that prior and you've made a contrarian bet. LDBD rewards it in two ways. The legacy Total Score multiplies your points by a contrarian_multiplier that grows the harder you push against a strong prior, and the Elo skill rating structurally pays out more whenever you're right on a call the odds said was unlikely.
To be honest about the limits, though: the headline leaderboard metric, rate, doesn't apply that multiplier at all — a correct call counts for exactly the return it realized. What still makes LDBD useful to a contrarian is that every prediction resolves against real adjusted-close prices, dividends and splits included. So you can see whether betting against the tide actually earned its keep from the track record itself, not from conviction.
Example
For an asset that usually rises, base_rate_up = 0.7, predicting 'down' gives 1 + 2 × (0.7 − 0.5) × 1 = 1.4x, while going with the trend and predicting 'up' shrinks it to 0.6x. In Elo terms that down call has an expected hit rate of 1 − 0.7 = 0.3, so being right pays K × (1 − 0.3) = 0.7K — far more than the 0.3K a with-the-crowd call would earn.
How LDBD uses it
The contrarian_multiplier lives only in the legacy Total Score (a secondary, engagement-oriented metric). It's computed at submission time from base_rate_up and your direction, clamped to [0.5, 2.0], and frozen onto the prediction. The primary metric, rate (annualized directional log return), ignores it entirely. The Elo skill rating is where beating a low-probability contrarian call is rewarded structurally, through expected = base_rate_up when you predict up, or 1 − base_rate_up when you predict down. When an asset has fewer than 100 historical price samples, base_rate_up falls back to its sector average.
FAQ
Does contrarian investing always win?
No. Going against the crowd only pays when the crowd turns out to be wrong. LDBD's multiplier, up to 2x, scales a contrarian bet in both directions: a correct one earns more, but a wrong one loses more, because the sign of the score comes from whether you were right and the multiplier only sizes it.
How does LDBD decide what 'the consensus' is?
It stands in for the crowd with base_rate_up — how often the asset has historically risen over that timeframe — rather than reading live sentiment or order flow. With fewer than 100 historical periods it falls back to the sector average.