#1
Betting size instead of edge
Position size should follow the gap between your estimate and the price, not how confident you feel. Run it through Kelly and cut the result.
Short answer
Most prediction-market losses are not forecasting failures — they are sizing, spread and resolution failures. Fix those three and an average forecaster stops bleeding.
#1
Position size should follow the gap between your estimate and the price, not how confident you feel. Run it through Kelly and cut the result.
#2
A 10¢ contract needs a 10% hit rate. The favourite-longshot bias exists because that rate is almost always lower than it looks.
#3
You can be right about the world and lose on a technicality. Read the source, the deadline and the ambiguity clause before entering.
#4
Your EV must be computed at the price you actually get. On thin markets the spread eats small edges whole.
#5
Round-trip costs apply per trade. Twenty trades a day at 1% friction is a 20% daily drag on a stake's worth of capital.
#6
Leaderboards rank by profit, which one lucky bet can produce. Copy hit rate within a category, not headline P&L.
#7
Adding is only correct when the price moved and your estimate did not. If the news changed your estimate, adding is denial with extra steps.
#8
Without recorded probability estimates you can never tell whether you are calibrated. Fifty logged trades is the minimum honest sample.
#9
Better tooling removes friction and shows you flow. It does not supply a view. The view is still your job.
Position sizing. Traders with a genuine edge still blow up by betting far past Kelly, because variance on binary contracts is brutal over short samples.
Before buying anything under 20¢, write down the hit rate it requires and ask whether your history supports it. Most people's does not.
Yes, but for discipline rather than speed: seeing exposure, spread and flow in one place prevents most of the errors on this list.
Fix the biggest one first: size your next position properly.
Gravia access is invite-gated
1,500+ traders · $2.0M volume