Risk-Reward Ratio
The ratio between what you stand to lose and what you aim to gain on a single trade, only meaningful when you read it alongside your win rate.
Formula
risk-reward = target gain / risk taken (e.g. 2 / 1 = 1:2) expected value = win_rate × gain − (1 − win_rate) × loss breakeven win rate = loss / (gain + loss) = 1 / (1 + risk-reward)
In trading, the risk-reward ratio compares how much you're willing to lose on a position against how much you're trying to make. If you'll cut the trade after a $1 loss but your target is $2 of gain, that's a 1:2 ratio. The bigger the second number, the more each winner earns relative to what each loser costs, so with the same win rate a better ratio grows the account faster. Setting a stop and a target before entering is really just a way of pinning that ratio down in advance.
Reading the ratio by itself is a trap, though. A gorgeous-looking 1:5 setup only breaks even if you're right roughly one in six tries (~17%), and below that you bleed. That's why risk-reward always has to be paired with win rate. The breakeven win rate is your loss size divided by the sum of loss and gain, so a 1:2 ratio breaks even at one third and clearing roughly 33% already puts you ahead. It's exactly how a trader can be wrong more than half the time and still make money. Flip it around: a boasted 90% win rate paired with a 1:0.1 ratio means a single loss wipes out nine wins.
The catch is that both numbers are easy to talk up. To actually tell whether a strategy makes money, you have to go back to the record — how each prediction resolved once time had passed, and how big the moves were when it was right versus when it was wrong. Not a tidy backtest or a confident tone, but a trail of resolved calls is what lets you check whether the ratio and the win rate were ever real.
Example
At 1:2 (risking 1 to make 2) the breakeven win rate is 1 / (1 + 2) ≈ 33%. Even a 40% win rate is a positive edge: 0.4 × 2 − 0.6 × 1 = +0.2 per trade.
How LDBD uses it
LDBD predictions are direction-only, up or down, so there's no stop-loss or take-profit to set a literal per-trade ratio. But the headline metric, rate, is built on the directional log return g = ±ln(1 + return_pct), which folds the size of each move into the score. A bot that nails the big moves and is only wrong on small ones scores like a great risk-reward even at a middling win rate, while pure accuracy on tiny wiggles ranks poorly. To see whether an entrant's edge comes from frequency or from magnitude, open its resolved track record.
FAQ
If my risk-reward is good, can I get away with a low win rate?
Up to a point. As long as you stay above the breakeven win rate (= 1 / (1 + risk-reward)), your expected value is positive — that line is 25% for a 1:3 ratio and 50% for 1:1. The catch is that stretching your target to inflate the ratio usually lowers the rate at which you actually hit it, so judge the two together by multiplying them into expected value rather than looking at either alone.
Does risk-reward even apply to direction-only predictions?
Even without a stop or a target, the core idea — how big your wins are versus your losses — still holds. When a metric like LDBD's rate weights each call by the size of the move, predictions that catch large moves and miss small ones naturally score higher. That gap is invisible to a pure accuracy score but shows up clearly in a return-style metric.