Most retail bettors track wins and losses. That is why most retail bettors do not know whether they are good. Wins and losses are dominated by variance for the first thousand wagers — long enough that a losing strategy can hide profitable for an entire NFL season, and a winning strategy can look broken until the spring. The bettors who actually know are the ones who track closing line value on every wager, ladder it into a KPI hierarchy, size with fractional Kelly, and visualize their variance with bankroll heatmaps. This is the handbook for that workflow — what to track, why each KPI sits where it does, and how the snip-to-leaderboard pipeline on Shark Snip turns a betting habit into a measurable edge.
Why CLV matters more than W/L
The closing line is the most accurate prediction the market produces. By the time a sportsbook locks lines at kickoff, it has absorbed every public injury report, every sharp wager, every late line move, and every model output the market collectively cares about. Pinnacle's own research on closing-line value shows it is a near-efficient predictor of game outcome — the closing spread beats every individual modeling shop's pre-game estimate, because it is the weighted aggregate of every modeling shop's estimate.
This matters because it gives bettors something rare in sports: an unbiased estimator of edge that resolves before the game even starts. If you bet Eagles +3.5 and the line closes Eagles +2.5, you captured one full point of CLV. The market eventually agreed your side was the right one — you just got there first. Whether the Eagles win or lose that specific game is variance noise. Repeat the +1 CLV exercise across two hundred wagers and the math compounds in your favor regardless of any individual outcome.
The contrast with win rate is brutal. A 55% bettor over 100 wagers and a 50% bettor over 100 wagers are statistically indistinguishable at conventional confidence levels — the 95% confidence interval on win rate at n=100 is roughly ±10 points. You could look at your record after a hundred bets and have absolutely no idea whether you are great, average, or terrible. CLV at n=100 has a confidence interval roughly an order of magnitude tighter. Our deep dive on CLV explains the math, but the headline is: CLV converges fast, win rate converges slow, and the bettor who anchors on the slow signal is making decisions blind.
The beeswarm above shows two hypothetical bettors after 120 logged wagers. The disciplined bettor centers at +0.6 CLV with most wagers in the positive half-plane. The undisciplined bettor centers at -0.7 with a long left tail. Neither of these bettors knows their fate from win rate alone — variance hides both stories. CLV is the X-ray.
The KPI hierarchy: CLV → ROI → bankroll growth → variance
Sharps order their KPIs by signal-to-noise ratio. The metric with the cleanest signal sits at the top, because it is the one you actively manage. Lower-tier metrics are diagnostic and reporting tools, not steering wheels.
Tier 1: Closing Line Value (the steering wheel)
CLV in points (for spreads/totals) and percentage points of implied probability (for moneylines). This is what you read after every wager to ask: did I get a good price? Average CLV across the rolling last 100 wagers is the single most actionable number on your dashboard. It tells you whether the strategy is working before the bankroll has time to confirm.
Tier 2: Return on Investment (ROI)
Profit divided by total dollars wagered, expressed as a percentage. ROI is the financial confirmation of the CLV thesis. A +0.6 CLV bettor at -110 should print roughly +1.2 to +2.0% ROI long-run. If your CLV is +0.6 but your ROI is -3% over 500 bets, you have either bad luck (most likely) or a stake-sizing problem where you are betting more on lower-edge plays. Our bankroll basics guide walks through ROI tracking conventions.
Tier 3: Bankroll growth (compound annual)
The dollar number that actually matters for life. Bankroll growth is downstream of ROI × turnover × Kelly fraction, and it is what determines whether the operation pays for itself. A 5% ROI bettor turning over their bankroll 30x per year grows bankroll ~150% before withdrawal. A 1% ROI bettor turning it over 5x grows it 5%. ROI alone is a vanity metric without volume context.
Tier 4: Variance metrics (max drawdown, Sharpe, Sortino)
Drawdown — the biggest peak-to-trough percentage decline of bankroll — is the survival metric. Sharpe and Sortino ratios measure risk-adjusted return. These are the metrics you read once a month, not every Sunday. Their job is to tell you when your live results are inside the variance band of your edge estimate, vs. when they have left the band and the model is broken.
The hierarchy matters because rookies invert it — they obsess over weekly bankroll swings, ignore CLV, and never compute drawdown. They make decisions on the noisiest metric and ignore the cleanest. Sharps run it the other way: tune the strategy on CLV, confirm with ROI, monitor bankroll growth, audit variance.
How to capture bet timestamps + closing line by sportsbook
Capturing the bet timestamp and closing line is the operational core of the whole system. The capture mechanism varies by sportsbook because their bet-slip layouts and history exports differ.
DraftKings
DraftKings shows the timestamp on the bet slip in the post-confirmation modal — screenshot it within ~30 seconds of placement so the timestamp is visible. Their My Bets tab also exports a CSV under Account → Statements with placement time, line, and price. Closing line: pull from Pinnacle's archive at kickoff, since DraftKings' own closing line is biased by recreational money flow.
FanDuel
FanDuel's bet slip shows placement time only after refresh. Better to grab the screenshot from the My Bets confirmation page where the timestamp lives in the bet details panel. CSV export is available under Activity → My Activity → Download. Same Pinnacle benchmark for closing line.
BetMGM, Caesars, ESPN BET
All three have similar workflows — confirmation modal screenshot for timestamp, CSV under account statements for batch export. ESPN BET in particular does not retain bet-history older than 12 months, so monthly export is recommended.
Pinnacle / Circa / BetCRIS
If you have access to a sharp book, your job is far easier. Their own bet history exports include execution timestamp to the second and their own closing line is the benchmark — no external lookup needed. CLV at sharp books is harder to capture because the lines barely move, but when you do beat the close you have proven edge against the most efficient market in the world.
The Shark Snip workflow
The intent of the snip-to-leaderboard pipeline is to remove this manual lookup burden. Drop a screenshot of your bet slip into the builder, the OCR extracts the side, line, price, and timestamp, and the system queries the Pinnacle closing line at game lock automatically. Average CLV across all your snipped wagers feeds the leaderboards directly. Manual tracking works for under 50 bets per week — past that, the lookup burden kills the habit and the data gets sparse.
Fractional Kelly: why ¼ Kelly is the practical max for non-pros
The Kelly Criterion is the closed-form solution for "what fraction of bankroll maximizes long-run geometric growth?" given a known edge and odds. The answer for an even-money bet at edge e is f* = e. For -110 odds at a 53% win rate, full Kelly stake is roughly 5.7% of bankroll per wager. The math says this stake size produces the highest possible compounded growth rate over infinite horizons. The math is correct. Our Kelly deep-dive derives the formula.
The math also assumes you know your edge precisely, never have a losing streak that exceeds your bankroll's tolerance, and behave perfectly under stress. None of those assumptions hold for non-pros. Edward Thorp's original Kelly research noted that even sophisticated investors using full Kelly experienced 50%+ drawdowns inside their backtests — drawdowns that almost no human can ride out without breaking discipline.
Quarter Kelly (0.25x) captures roughly 75% of the long-run growth rate of full Kelly while reducing variance by ~94%. The math: doubling stake size doubles edge but quadruples variance. Halving stake size halves edge but cuts variance 75%. The optimal compromise for any bettor whose edge estimate has uncertainty (which is every bettor) is 0.25x to 0.5x. The chart above plots bankroll trajectories at 1.0x, 0.5x, 0.25x, and 0.1x Kelly — note how the lower fractions sacrifice peak growth for survivability through the 200–500 bet variance trough.
Operational rule: most disciplined bettors with verified +0.5 to +1.5 CLV operate at 0.1x to 0.25x Kelly. Above 0.5x is gambling with extra steps. Below 0.05x is leaving compounded growth on the table for no good reason once you have proven edge. The window is narrow for a reason.
Pyramid leagues as a forcing function for discipline
Discipline is the hardest input to control. After three losing Sundays in a row, every bettor wants to "make it back." The conventional advice — "just stick to your unit size" — fails because willpower is a depletable resource. The structural fix is to convert internal discipline into external constraint: enter a competition where reckless behavior is punished by the rules, not by your conscience.
That is the design intent of pyramid leagues. Promotion and relegation tiers force you to think about downside, not just upside. A +12 ROI week with a 22% max drawdown might still drop you a tier because you broke a risk rule. The competition structure does the bankroll-management enforcement that willpower fails at after the third straight losing Sunday. The leaderboard becomes the audit trail. Other bettors at your tier become the accountability layer.
This is also why the model marketplace matters as a discipline tool. When you publish a model and other bettors stake real bankroll on it, the public CLV scoreboard becomes inescapable. You can hide losing personal bets in a spreadsheet — you cannot hide a -0.4 CLV model that 30 people are following. Public accountability is the strongest forcing function on tracking discipline that exists.
Bankroll heatmaps: visualizing variance + drawdown
The number you see on your account screen — current bankroll — hides almost everything important about your bankroll's behavior. Two bettors can both finish a season at +50% and have wildly different stories: one had a smooth grind from +3% to +50%, the other had a -20% drawdown in October, recovered in November, and finished at +50% by sheer late-season luck. The second bettor is in much more danger of busting next season because their drawdown profile reveals over-staking.
The heatmap above plots a real-shape 26-week NFL season for a +1.5% per week edge bettor. Weeks 7 and 10 print -3.8% and -4.5% returns. The "bankroll vs peak" series shows the bettor sat ~7% below prior peak through weeks 7–11 — that is normal variance for a +1.5% per-week bettor. Anyone watching weekly P/L without the variance context would have quit by week 11. Anyone watching the drawdown chart would have seen "this is inside expected variance band" and held the strategy.
Three drawdown rules to internalize: (1) Max drawdown >25% of bankroll over 100 wagers means you are over-staking, regardless of edge. (2) A drawdown that exceeds your historical 99th percentile means the model is broken or the edge is gone — stop and re-validate. (3) A drawdown inside the 95th percentile is normal variance; do not change strategy. The heatmap visualization is what makes these rules operationalizable instead of theoretical.
The full tracking workflow: snip → snipshark → leaderboard
The end-to-end flow on Shark Snip is built around four stations, each of which corresponds to one KPI tier above.
Station 1: Snip
Drop a screenshot of your bet slip the moment you place the wager. The OCR extracts the side, line, price, stake, and timestamp. This is the data-capture step — the equivalent of writing the bet down in a notebook, except automated. Coverage matters more than completeness here: every wager logged, even the small ones, because the missing wagers are the ones that bias your CLV average.
Station 2: Snipshark (the analysis layer)
Once the bet settles, the analysis layer pulls the Pinnacle / Circa closing line at lock and computes CLV in both points and cents. It also computes implied edge, expected ROI contribution, and Kelly-recommended stake vs your actual stake. If your actual stake is consistently above Kelly recommendation on lower-edge plays, the system flags an over-staking pattern. The workshop exposes these analytics as live dashboards you can slice by sport, market, time of week, or feature.
Station 3: Leaderboards
Aggregated CLV, ROI, and bankroll growth across all snipped users feeds the leaderboards. Public ranking is the discipline forcing function — the same psychological mechanism that makes Strava work for runners. Sharks who rank consistently in the top decile on rolling 100-bet CLV are the ones whose models and snippets get most-followed in the marketplace.
Station 4: Pyramid leagues
Top-leaderboard performers feed into the pyramid league system, where structured promotion/relegation converts measured edge into competitive standing. This is the long-term retention layer — leaderboards measure 100-bet CLV, leagues measure full-season bankroll discipline. Both are necessary, neither is sufficient.
Common mistakes: only counting wins, ignoring vig, ignoring CLV
The mistakes below are extracted from auditing thousands of bettor logs. They are listed in approximate order of how badly they distort self-evaluation.
Mistake 1: Selectively logging wins
The single most common bias. Bettors remember and log their wins more reliably than their losses. The result is a phantom +20% ROI in their tracking spreadsheet that bears no resemblance to their actual bankroll trajectory. Fix: snip every wager at placement time. If it did not get snipped, it does not exist. The friction of remembering to snip every wager is the discipline filter — bettors who cannot maintain it should not be sizing up.
Mistake 2: Ignoring vig
"I went 6-4 last Sunday" sounds like a winning week. At -110, 6-4 is exactly break-even on stake. Our vig deep-dive walks through the math, but the operational shortcut is: at -110, you need to win 52.4% of bets just to break even. Quoting raw record without referencing the price is meaningless. ROI is the only honest profit metric for this reason — it bakes in the vig automatically.
Mistake 3: Ignoring CLV entirely
Bettors who hit a hot streak immediately convince themselves their model is the reason. Bettors on a cold streak immediately blame the model. Both are wrong roughly half the time without CLV as the tiebreaker. CLV is the only metric that can distinguish "I am variance unlucky" from "I am leaking edge." Tracking it is non-negotiable for anyone serious about long-run profit.
Mistake 4: Anchoring on dollar P/L instead of unit P/L
Dollar P/L mixes stake-size discipline into the same number as model accuracy. A bettor who up-sized to 5% units after a winning week and gave it all back is going to read the dollar P/L as "I broke even" when in reality they leaked edge by violating their own staking rules. Unit P/L (ROI) separates strategy quality from staking quality, which is the only way to diagnose either.
Mistake 5: No timestamp at execution
If you do not record the time you placed the wager, you cannot compute CLV against a fixed reference point later. Going back 48 hours and "guessing" the line at execution time produces fake CLV numbers. Either snip at execution or do not bother computing CLV. Timestamping is the cheapest discipline gain available.
Mistake 6: Conflating model bets with feel bets
Bettors with a working model often place "feel" bets on the side — a parlay because it is fun, a live bet because the game looked off. Feel bets must be tracked separately or they pollute the model's edge measurement. Our bet-tracking guide goes deep on segmentation.
Mistake 7: Stopping the log mid-streak
The most expensive mistake. A bettor on a 1-12 cold streak quietly stops snipping their wagers because "the data is depressing." Three weeks later they restart logging once the streak has flipped. The result is a permanently biased CLV history with the worst variance band excluded. Either log 100% of wagers across the full season or log 0% — partial sampling is worse than no sampling at all.
The handbook in one page
Track CLV on every wager. Order KPIs by signal: CLV → ROI → bankroll growth → variance. Capture the bet timestamp at execution and benchmark against Pinnacle's closing line. Size at quarter Kelly until you have 1,000+ wagers proving edge. Use pyramid leagues and public leaderboards as discipline forcing functions. Visualize bankroll variance with drawdown heatmaps so you can distinguish "inside the band" from "model broken." And avoid the seven mistakes above — selective logging, ignoring vig, ignoring CLV, dollar-P/L anchoring, no timestamps, blended model and feel bets, and stopping the log mid-streak.
The bettors who do all of this are the ones who eventually rank on the leaderboards. The bettors who do none of it are the ones who blame variance forever. The handbook is the boring middle path — and the only one that compounds.
Price examples and pass rules
Use names as evidence, not decoration. The useful SEO win is that Josh Allen, Ja'Marr Chase, Bijan Robinson and Puka Nacua and Eagles, Bears, Chiefs, Bills and Lions appear inside decisions, thresholds, and internal links instead of being dumped into a keyword list.
- Spread example: if Chiefs-Broncos opens Chiefs -3.5 and your fair number is -2.8, +3.5 is the bet, +3 is a pass, and the moneyline needs roughly +155 or better before it replaces the spread.
- Total example: if a Bills outdoor total opens 46.5 and wind moves from 8 mph to 21 mph, an under projection at 42.8 still needs a playable number; under 45 or better is different from chasing 43.5.
- Futures example: Bengals AFC North +280 is 26.3% before hold. If your fair number is 30%, stake modestly, track portfolio correlation, and avoid stacking every Burrow, Chase, and Higgins bet into the same thesis.
- CLV rule: a good write-up is not enough. Track whether the spread, total, prop, or futures price closed better than your entry before grading the process.
Use closing-line value guide to keep the examples attached to measurable prices.
Research note board
Use this table to turn the guide into a decision note. The point is to know when the idea is actionable and when it is only context.
| Angle | Input to verify | Example application | Pass when |
|---|---|---|---|
| Market price | Spread, total, moneyline, prop price, or futures hold | Eagles and Bears compared through PPR | The price has moved past the number that created the edge |
| Football or sport context | Role, pace, weather, injury status, opponent style | Josh Allen role news mapped to the relevant market | The original input changes or remains unconfirmed |
| Review loop | Entry, close, result, and reason code | closing line value logged with a clear thesis | You cannot explain whether the process beat the market |
Bet responsibly. Set deposit limits, take regular breaks, and never chase losses. If betting stops being fun, take a step back.
Model calibration: predicted vs observed
Predicted win probability bucket vs the empirical win rate inside that bucket on the test set. Points on the y=x reference line are perfectly calibrated; points below mean the model is overconfident in that bucket.
EV per $100 across win rate × odds grid
Expected value of a $100 stake at each combination of true win rate and market odds. Anywhere the cell is positive you have a long-run profitable bet; the magnitude shows how aggressive Kelly will size it.
