A clean betting log is supposed to ruin your best excuses. It should stop a bettor from forgetting losses, moving the reference price, or calling a lucky result proof of skill. Closing-line value helps, but only when the execution and closing prices describe the same market and both timestamps are honest.
Why CLV matters more than W/L
Closing-line value is a price audit. It compares the market position accepted at execution with a fixed pregame closing reference for the same event, market, side, line type, and settlement rules. If any of those fields differ, the comparison is noise.
This makes CLV useful before a large settled sample exists, but it does not make CLV a profit guarantee. A bettor can beat one reference and still pay a poor price elsewhere. A market can move for reasons unrelated to the original thesis. Keep the reference source and cutoff visible.
The CLV guide covers the mechanics. The operational rule is stricter: no matching market, no comparison; no approved timestamp, no reference.
The database problem: why serious tracking breaks down
Most people who try to track CLV fail for a boring reason: the data model is harder than the betting model. A bet looks simple — team, line, stake, result. A complete CLV record needs the event, market, side, execution line, execution price, execution timestamp, sportsbook, closing line, closing price, and closing timestamp.
What to track, exactly
Every bet row needs at minimum:
- Market identity: event, date, market type, side, and line.
- Execution: price taken, sportsbook, stake, and timestamp.
- Close: reference sportsbook or consensus, closing line, closing price, and capture timestamp.
- Model context: probability at execution, edge estimate, model version, and reason code.
- Settlement: result, settlement status, and payout.
- Process tags: sport, market family, confidence bucket, and whether the bet followed the model or was discretionary.
The one rule
Record the wager when it is placed. Do not rebuild the entry from memory after settlement. A bettor who remembers only the winning price is not tracking; they are editing. The bet-tracking guide explains why complete capture matters more than a clever dashboard.
How Shark Snip should automate it
The useful workflow is capture, match, grade, and review. The builder or another tool may reduce manual work, but automation cannot repair an unmatched market or invent a missing timestamp. Honest empty states are part of the record.
Visualising CLV distributions
The live-pregame outcome view currently supplies no non-null CLV rows for this article, so there is no distribution to render. That empty state is honest: no matching real rows means no chart, no average, and no performance claim.
When the view does contain qualified rows, publish the full distribution with its market, prediction window, sample, and provenance. A mean without the shape can hide a few outliers, and a chart without the eligibility rule can hide omitted decisions.
Fractional Kelly: optimal sizing for humans
The Kelly Criterion converts an estimated probability and offered price into a bankroll fraction. It does not know whether the probability estimate is calibrated, whether several wagers are correlated, or whether the bettor can follow the plan through a drawdown.
full Kelly fraction = (b × p − q) / b
fractional stake = declared fraction × full Kelly fractionThe Kelly guide derives the equation. Operationally, keep the sizing rule, probability estimate, price, and correlation assumptions in the receipt. When any input is weak, the cure is less exposure.
Why not full Kelly?
Because the formula is only as good as the edge estimate. The calculator can be exact while the probability is wrong. A declared fractional rule can reduce exposure to that error, but this article does not prescribe one universal fraction or publish a made-up growth path.
Pyramid leagues as a forcing function for discipline
Knowing the right process is not the same as following it. A competition or public record can make risk limits visible, but the current product contract must define what is measured and how standings work.
The discipline feedback loop
External accountability can make exceptions visible. It does not prove the model has an edge. The Gridiron surface is the authority for any current promotion, relegation, or reward rule.
How promotion and relegation can change behavior
A transparent leaderboard can reward steady process and expose risk-taking, but only when the scoring rule, reset window, and complete decision set remain visible. Without those fields, placement has no reliable meaning.
Bankroll heatmaps: visualizing variance + drawdown
A current balance hides the path that produced it. Two strategies can finish at the same balance while taking very different risks. A drawdown view should show peaks, troughs, stake changes, and the model version active at each decision.
How to read the heatmap
Read across weeks to see the streak structure. Read across rows to separate weekly returns from drawdown against the running peak. Then check whether the chart is built from real receipts rather than a hand-drawn “expected” path.
The article has no qualifying real bankroll series to publish, so no heatmap is rendered. When data is available, the benchmark must use the same strategy, sizing rule, and prediction window. An unavailable chart is better than a smooth fake one.
The full Snip → Leaderboard workflow
The workflow has separate stations so a later result cannot rewrite an earlier decision.
Station 1: Capture
Record the event, market, side, execution line, price, source, stake, timestamp, and eligibility rule when the wager is placed. Coverage matters because omitted wagers bias every later summary.
Station 2: Score
Match the approved pregame closing reference for the same market. Market references must pass the provenance_tier='pregame' allow-list. Postgame corrections and merely similar markets do not qualify.
Station 3: Size
Apply the declared sizing rule to the execution-time probability and price. Log the fraction and final stake. Keep model confidence separate from stake size so a risk decision cannot masquerade as a stronger prediction.
Station 4: Grade
Settle the wager under the published rules and keep pushes, voids, and missing references explicit. graded_at is the batch processing stamp; it is never the wager date or a performance axis.
Station 5: Rank
Read price quality beside the complete settled record and bankroll exposure. The leaderboards are useful only when they preserve the same window, sample, market, and provenance.
The loop
Use each settled window to review the data contract, model drift, execution, and sizing. Do not react to one result. Change a rule only when the evidence is defined before the next decision.
Common mistakes
Mistake 1: Selectively logging wins
A complete log begins at execution, before the result can influence memory. If a wager is missing, mark the record incomplete rather than publishing a flattering summary.
Mistake 2: Ignoring the offered price
A raw win-loss line cannot describe profit or break-even requirements without the prices accepted. Keep the price attached to every decision and do not convert it into a generic stake convention.
Mistake 3: Ignoring CLV entirely
Results alone make it harder to distinguish a good process from a lucky one. CLV adds a price-process diagnostic, but only when the closing reference is matched and pregame.
Mistake 4: Blending model quality with stake size
A model can select reasonable prices while an aggressive sizing rule creates unacceptable risk. Report selection and exposure separately so each failure has a visible cure.
Mistake 5: No timestamp at execution
Without an execution time, the bettor cannot prove which information or price was available. Do not reconstruct it later. Mark the receipt incomplete.
Mistake 6: Conflating model bets with "feel" bets
Model and discretionary decisions need separate eligibility labels. Mixing them makes both records harder to diagnose.
Mistake 7: Stopping the log mid-streak
Pausing capture when results are ugly permanently biases the history. Keep the rule running or mark the window incomplete. A partial sample should never be presented as the full strategy record.
Manual tracking vs automated tracking
A spreadsheet is sufficient when it captures every field consistently. Software becomes useful when it reduces missing rows and locks the reference process, not because automation makes an unmatched or postgame price valid.
The manual template
At minimum: date, sport, market, side, execution line, execution price, sportsbook, stake, timestamp, closing line, closing price, closing timestamp, model probability, reason code, result, and CLV status.
The edge decays when capture is delayed
If you try to enter everything on Sunday night, the execution price is gone and the remembered timestamp is unreliable. Capture at bet time or mark the receipt incomplete.
How long until you know if you have an edge?
No fixed sample settles that question for every market. More observations help only when the market definition, eligibility rule, and data timing remain stable. Publish uncertainty and an honest empty state rather than inventing a threshold after seeing the curve.
Frequently asked questions
What does CLV stand for?
Closing-line value compares your execution with a fixed, matching pregame closing reference.
Is CLV more important than win rate?
They answer different questions. CLV audits price process; a settled record audits outcomes. Read them together.
What fraction of Kelly should I use?
Use a declared rule sized for model uncertainty, correlated exposure, and your ability to follow it. There is no universal fraction.
How much drawdown is normal?
Compare the live path with a sourced historical distribution for the same strategy and sizing rule. Without that reference, no universal band is honest.
Can I track bets in a spreadsheet?
Yes. The tool matters less than complete receipts, matched markets, and fixed timestamps.
The handbook in one page
- Capture every wager at execution.
- Freeze the closing benchmark before grading.
- Accept only approved live-pregame provenance.
- Keep execution, reference, and processing clocks separate.
- Record the sizing rule and uncertainty.
- Read CLV beside a properly windowed settled record.
Bottom line
The point of the log is not to make the bettor look sharp. It is to make self-deception expensive. Complete receipts, honest empty states, fixed market references, and declared sizing rules do that better than a smooth chart.
Bet responsibly. Set deposit limits, take regular breaks, and never chase losses. If betting stops being fun, take a step back.
