How to Trade the Sports Analysis

Sports betting on Convex works like any other market — the AI looks for situations where the bookmaker's odds don't reflect the true probability of an outcome. When the market is wrong, that's an edge. When the market is right, you stay out.

How the edge is calculated

For every match, the AI estimates the true probability of each outcome (home win, draw, away win) using league standings, home/away records, head-to-head history, and form data.

It then compares its fair probability to the bookmaker's implied probability (derived from the odds). The difference is the edge. For example: if the AI thinks the home team has a 55% chance of winning, but the odds imply only 47.6%, that's a +7.4% edge.

Only edges above 5% with HIGH confidence are turned into recommendations. This is deliberately conservative — the system would rather miss a marginal bet than take a bad one.

An edge is not a guarantee. A 7% edge means that over hundreds of similar bets, you should profit. On any single bet, you can still lose. The edge only works with volume and discipline — just like in mean-reversion trading strategies.

Understanding confidence levels

HIGH — The AI has strong data: good sample size, clear form patterns, no major unknowns. These are the only bets that generate recommendations.

MEDIUM — Reasonable analysis but some uncertainty (small sample, recent managerial change, key player injury unknown). Not recommended as bets but shown for reference.

LOW — Too much uncertainty. The AI acknowledges it doesn't have enough data to form a view. Never bet on these.

Position sizing: the Kelly criterion

Convex uses a fractional Kelly criterion (quarter-Kelly) to size bets. This mathematical formula calculates the optimal bet size based on your edge and the odds. Using quarter-Kelly is more conservative than full Kelly, which helps survive the inevitable losing streaks.

The bet is further capped at 2% of your bankroll per bet. This means even if the model sees a huge edge, it won't risk more than 2% on a single match. This is essential discipline — the same logic behinddrawdown management in financial markets.

Treat your sports bankroll separately from your trading capital. Decide in advance how much you're allocating to sports betting and don't dip into other funds when you hit a losing streak.

What makes a good bet

Edge above 5%
HIGH confidence from the AI
Clear reasoning that you can verify (e.g. "Home team has won 8 of last 10 at home; away team has worst away record in the league")
No major warnings (injuries, weather, managerial changes)
Match in a league with reliable data (top European leagues have the best data coverage)

Key terms you'll see

Edge — The difference between the AI's estimated probability and the bookmaker's implied probability. Same concept as arbitrage in financial markets.
Kelly Criterion — A formula for sizing bets optimally based on your edge. Full Kelly maximises growth but is volatile; Convex uses quarter-Kelly for safety.
Drawdown — The decline from your peak bankroll to a trough. In sports, a 15-20% drawdown from a losing streak is normal even with a genuine edge.
Sharpe Ratio — A measure of risk-adjusted return. Useful for comparing your sports returns to your financial trading returns on a like-for-like basis.
Implied Probability — The probability embedded in the odds. Decimal odds of 2.10 imply a 47.6% chance. Related to implied volatility in options — both measure what the market "expects."

What NOT to do

Don't bet on every recommendation. The system produces bets based on statistical edge, but you should still apply common sense. If a key factor has changed since the data was last updated (breaking team news, last-minute injury), skip the bet.

Don't increase stakes after losses. "Chasing" — betting bigger to recover losses — is the single fastest way to blow up a sports bankroll. Stick to the Kelly sizing. This is the sports equivalent of doubling down after a drawdown.

Don't bet on draws unless the edge is very large. Draws are the hardest outcome to predict and the model has the lowest confidence on draw selections.

Don't ignore the warnings. If the AI flags "key player fitness uncertain" or "recent managerial change," that uncertainty is real and can invalidate the statistical model — just like a black swan event invalidates financial models.

Don't bet with money you can't afford to lose. Even with a genuine edge, variance is real. You can have 10 losing bets in a row and the edge is still working — but only if you're still solvent at bet 11.

Tracking your results

The system tracks every recommendation outcome automatically. Over time, check your calibration on the Calibration page — this shows whether the AI's confidence levels are accurate. If HIGH confidence bets are hitting at 60%+ and the average edge is positive, the system is working.

If hit rates drop below 50% for an extended period, something may have changed (league dynamics, data quality). Don't blindly follow — review the results periodically.

← Crypto Predictions →