V8 Predicts

Reading Performance

How to read a published record, and which number to distrust first.

Lesson 1

Sample size comes first

After this you can: Ask for the denominator before reacting to any performance figure.

Betting results are noisy enough that short records routinely look impressive by luck alone. A 20 percent return over fifteen bets tells you almost nothing; the same figure over eight hundred would be remarkable.

So the first question about any published number is how many settled results it rests on. A record without its denominator is a claim, not evidence, and the absence of that number is usually deliberate.

This cuts both ways. An early losing stretch is equally uninformative, and reading it as proof that a method is broken is the same error in the other direction.

Key takeaways

  • Always read a performance figure together with its sample size.
  • Small samples produce impressive and disastrous records for the same reason.
  • A missing denominator is itself information about the publisher.

Terms used: Sample size, ROI, Variance, Confidence interval

Lesson 2

Win rate is the wrong headline

After this you can: Explain why win rate alone cannot tell you whether a record is any good.

Win rate is the most quoted betting statistic and the least informative, because the price decides what win rate is required. Seventy percent on heavy favourites can lose money. Forty percent on longer prices can be strongly profitable.

Return on the amount staked is the more honest headline, since it accounts for what was risked to achieve the record. Pushes and voids should be reported separately rather than folded in, because hiding them flatters the win rate.

The most useful habit is to convert a claimed win rate into the break-even rate for the prices involved. If a service advertises 58 percent at prices that need 60, the advertised number is a losing record.

Key takeaways

  • Win rate is meaningless without the prices it was achieved at.
  • Return per unit staked is the fairer summary figure.
  • Compare any claimed win rate against the break-even rate for those prices.

Terms used: Win rate, ROI, Break-even probability, Push