Sample Size and Variance

This page explains why a few hundred graded picks, ours included, is a smaller piece of evidence than it looks.

The break-even number

At -110 you need to win 52.38 percent of your bets to break even. That number comes straight from the price: 110 divided by 210. Everything about handicapping is the attempt to clear it by a few points. A bettor winning 55 percent long-term is doing genuinely well. A bettor claiming 65 percent over a meaningful sample is making a claim that almost nothing in the public record supports.

Why a few hundred bets proves little

A coin-flip bettor with no skill at all will, over two hundred bets, land above 55 percent a meaningful fraction of the time purely by chance. Run enough people through two hundred bets and some of them will finish with records that look like expertise. Those are the records that get screenshotted. Nothing about the arithmetic distinguishes them from skill at that sample size, which is why the only honest thing to say about a two-hundred-bet record is that it is consistent with skill and also consistent with luck.

What this means for our numbers

As of 4 October 2026 the published record stood at 109-87 across 196 graded picks: a 55.6 percent win rate, 3.2 points above the break-even number, for roughly 12 units at exact -110 pricing. We think that is a decent start. We do not think it is proof of anything, and anyone presenting a figure like that as proof is either not thinking about variance or is counting on you not to. Check the record page for the current figures.

Streaks are not information

A seven-game winning run and a seven-game losing run are both completely ordinary at any win rate near the break-even point. They feel like information because human pattern recognition is aggressive and because the people selling you something benefit from your treating them as information. Neither streak should change your staking, and if a run of results makes you want to bet bigger, that impulse is the thing to watch rather than the streak.

How to read any published record

Ask four questions. How many graded bets, not how many wins. Over what period, and does it include the bad months. At what price, and could you have taken it. Were the grading rules fixed in advance or decided afterwards. A record that answers all four and still looks modest is worth more than a spectacular one that answers none.

Common Questions

How many bets before a record is meaningful?

Statisticians generally want thousands before separating a small edge from noise with confidence. A few hundred narrows the range but leaves it wide. Treat any record below about a thousand graded bets as suggestive rather than conclusive.

Is a 60 percent win rate realistic?

Over a short run, yes, through variance alone. Sustained over thousands of bets against standard pricing, it would be exceptional to the point that you should want a great deal of evidence before believing it.

Does betting more sports improve the sample?

It increases volume but it also mixes different markets with different efficiencies into one number. A combined record across sports can hide that one sport is carrying everything, which is why we publish the per-sport breakdown.

Should variance stop me betting?

It should stop you betting money you need, and it should set your stake sizing. Bet small enough that a fifteen-bet losing stretch, which will happen, is an inconvenience rather than a problem.

For entertainment purposes only. Must be 21+. Please gamble responsibly. If gambling stops being fun, call or text 1-800-MY-RESET (1-800-697-3738).