How to Track Sports Predictions Honestly

Most people who make sports predictions believe they are better than they are, and the reason is not dishonesty. Memory keeps the hits and quietly drops the misses, the near-misses become “basically right”, and the stake on the confident calls grows in the retelling. A tracking system exists to take memory out of the loop. This guide describes one that works, the arithmetic behind it, and how the Sportslyx pick tracker automates most of it.

Rule one: record before the start

A prediction recorded after the first quarter is not a prediction. The first rule is mechanical: the read, the side, the price and the stake are written down before the event begins, and nothing is edited afterward. Sportslyx enforces this with a cutoff — a tracked pick cannot be placed once a game is more than fifteen minutes past its scheduled start. The breakdown stays readable all game; only the act of logging a read closes.

The price matters as much as the side. A favorite at -120 and the same favorite at -180 are different decisions, and a record that stores only “Team A” cannot compute anything useful later.

Rule two: fix the stake

Variable stakes make a record unreadable. If you risk twice your usual amount on the calls you feel good about and half of it on the rest, your results measure your feelings more than your method. The standard fix is a flat stake: every read risks the same amount. Sportslyx grades the company board at a flat stake for exactly this reason — one large result cannot dominate a month, and a hit rate and an ROI computed at a flat stake describe the method rather than the mood. Your own tracked picks carry the stake you enter and your net result is shown in money, which is why keeping that stake steady is rule two.

Rule three: separate hit rate from ROI

Hit rate and return on investment answer different questions. Hit rate is wins divided by wins plus losses — pushes are excluded, because a push is not a decision that resolved. ROI is total profit or loss divided by total stake, expressed as a percentage. A method that takes short favorites can post a 68% hit rate and a negative ROI; a method that takes long prices can post 40% and a positive ROI.

Neither number is enough alone. A high hit rate with a negative ROI means the prices were too short for the accuracy; a positive ROI on a small sample may be one large result. Publish both, with the settled count beside them, and read the count first.

Rule four: one period, chosen in advance

A record that starts “since my method changed” or “from the week things clicked” is a story, not a record. Choose the period before you see the results and never move it. Sportslyx uses one calendar month in UTC across the whole product — the public record, the per-sport pages and the personal tracker all read the same months — so a figure on one page is the same figure on every other page. Rolling windows and all-time totals are not shown next to it, because two periods under one label are how numbers get chosen after the fact.

Rule five: write the grading rules down

Edge cases decide more records than skill does. What happens to a soccer draw? A postponed game? A player prop when the player is scratched? An overtime loss in hockey? Every one of these must have an answer before the result is known. The Sportslyx rules, published on the methodology page:

• Game reads settle on the final score, including overtime and shootouts where the market settles that way
• A soccer draw is a push — stated openly, because it flatters the record against a draw-no-bet price
• Player props — covered for MLB only — settle from the official box score; a player who did not play is void, not a loss
• Cancelled fights, no-contests and postponed games are void and removed
• Voids and pushes are excluded from hit rate; settled volume shows them separately

The point is not that these are the only defensible rules. It is that they are fixed, visible and applied by a settlement job that reads official results rather than by a person deciding what counts.

How much volume is enough?

Less than most people expect, and more than most records contain. The standard error of a hit rate near 55% is roughly the square root of 0.55 × 0.45 ÷ n. The table shows the width of a 95% interval — the range within which the true rate probably sits — for different sample sizes.

Fifty results cannot tell a 45% method from a 65% one. Even five hundred leave several points of doubt. A monthly figure is a window into a method, not a verdict on it, and past results do not predict future results.

Closing line value

There is one measure that says something on a much smaller sample: whether the price moved toward your side between the moment you logged the read and the start of the event. Consistently getting a better price than the closing line is evidence that your information was ahead of the market, and it shows up long before results do. It is not a substitute for a graded record — the market can be wrong too — but it is the quickest honest signal available.

What the Sportslyx tracker does for you

• Closes the log fifteen minutes after the scheduled start, so every tracked pick is a real prediction
• Grades automatically from official final scores and box scores via scheduled settlement jobs
• Applies the published rules for pushes, voids and props without a human deciding after the fact
• Measures hit rate, ROI and settled volume over one calendar month (UTC), per sport — the company board at a flat stake, your own picks as a net result at the stake you entered
• Shows your record and the company board’s record side by side, computed with the same definitions

You can log your own reads from any game page, follow the board’s reads, or run a Studio model and let its runs be graded — the same rules apply to all three. The predictions boards show every read before it is settled.