Win Rate
The share of your calls that turn out correct — an intuitive number that says nothing about how much you win versus how much you lose, so it can be high while you bleed money.
Formula
win rate = correct predictions / total predictions × 100 expected value = (win rate × average win) − (loss rate × average loss)
Win rate is the share of your predictions that turn out correct. Get seven of ten right and your win rate is 70%. It's easy to compute and pleasant to look at, which is exactly why people treat it as proof of skill. The trap is that you can watch your account shrink while a high win rate keeps telling you you're doing great.
Here's the problem: win rate counts how often you were right and nothing about how much you made when right or lost when wrong. Win a little nine times and lose big once, and you can post a 90% win rate with a negative balance. Flip it around — a 30% win rate can still come out ahead if the winners are large enough. That gap between being right and making money is where beginners and seasoned hands part ways.
This is why win rate only means something next to your risk/reward ratio and your expected value. Expected value is what a single call earns on average, folding in both how often you win and how big the wins and losses are. A low win rate with fat winners can carry a positive expectation, and over a long enough run it's positive expectation — not a flashy hit rate — that survives.
The same holds on a prediction leaderboard. When the game is just calling direction, win rate looks like skill, but one missed crash can swallow a dozen small correct calls whole. To compare track records honestly you have to look past the tally of right-and-wrong to how big a move each call actually caught, and how accurately.
Example
A 90% win rate that makes +1% on wins but loses -15% on the one miss has EV = 0.9×1% − 0.1×15% = −0.6% — a long-run loser.
How LDBD uses it
This trap is precisely why LDBD ranks by annualized return (rate) rather than win rate. Rate doesn't tally right-and-wrong; it uses the directional log return (g = ±ln(1+return_pct)), so nailing lots of tiny moves nudges your score only slightly, while one big wrong-way drop cuts deep. Moves under 0.05% are voided outright, so the trickle of small correct calls that pad a win rate never even reach the score. Win rate is at best a supporting stat here — the ranking is settled by rate, which accounts for the size of each move.
FAQ
I have a high win rate — why am I losing money?
If your losses dwarf your wins, being right often still nets out negative. A single big loss can erase nine small gains, so a 90% win rate tells you nothing until you weigh it against your risk/reward and expected value.
What's a good win rate?
There's no universal number — it depends on your risk/reward. If your winners are twice the size of your losers, anything above roughly 34% is already profitable in expectation. With poor risk/reward, even 80% can fall short.