97 non-obvious advantages that separate elite practitioners from everyone else.
Standard statistics (frequentist) assumes large samples. Sports gives you 38 EPL games per team, ~10 CL games, and sometimes far fewer. In small-N environments, frequentist models produce unstable estimates and throw away real data as "outliers." Bayesian models treat every observation as fixed truth that updates the distribution — exactly the right behavior when you have almost no data.
Most bettors assume a full season of data overrides priors. In college football (12 games), even after a full season "we're still regressing a large amount to our priors." Without strong priors, a model would pick the Minnesota Vikings to win the Super Bowl based on a negative point differential.
The default bettor instinct is to hedge for "safety," but every hedge adds a transaction and every transaction adds vig. The only mathematically justified hedges are: (1) final leg of a large parlay where the math supports locking in profit, and (2) a line that moved badly against you, suggesting the market knows something you didn't, so you wash the trade.
Parlays are compounding leverage — "magnificently efficient" in both directions. With +EV bets they accelerate wealth. With neutral or negative EV bets they're "magnificently efficient wealth destroyers." Since most bettors can't prove they have positive EV, accumulators amplify ignorance, not edge.
When a live position is enormous relative to bankroll, a negative-EV hedge maximizes expected bankroll growth. Rufus calculated a $100K+ hedge on a -2.5% edge line because his live position ($1K to win $300K) was too large a fraction of a $2M bankroll. The correct framework is expected bankroll growth, not expected value.
Beyond full Kelly, BOTH the growth rate AND the probability of reaching a target before ruin decline simultaneously. There is zero rational justification for exceeding Kelly. Half Kelly gives roughly 2/3 of growth but only 1/2 the variance — a tradeoff most practitioners should take. Markowitz independently confirmed this finding.
Models rate promoted teams based on prior-league performance, which is categorically non-transferable. Championship quality does not equal EPL quality. The model error is not random — it's directionally biased (overrating promoted teams) for a predictable 5-10 game window while "half these guys barely know their teammates."
When a new data source or methodology becomes mainstream, the market overcorrects toward it — and simple older approaches temporarily outperform. When xG models became popular in hockey, Corsi (simpler shot-attempt counting) outperformed them for years. The overcorrection creates the opposite edge: fade the new thing while everyone else is adopting it.
Andrew Mack built a 25-variable rolling ridge regression that outperformed a 200-day moving average by one-third of 1%. Before declaring any model has edge, benchmark it against the simplest possible baseline. The added overfitting risk from complexity may exceed the marginal improvement.
Three structural forces systematically overprice longshots: bookmaker risk aversion (post-Leicester compression), punter risk aversion (lottery-ticket psychology), and odds myopia. The excitement premium is real — you're paying for the thrill. Backing favorites in outright markets is the actual edge.
In prior Super Bowls, halftime set lists were leaked on social media days before the game, accurate "to a tea." NFL/production teams, angry about leaks, now deliberately plant false information to identify leakers. "Insider information" on halftime show props circulating on social media could be deliberately false.
Unlike financial markets where equity risk premium and volatility risk premium persist because they compensate for real economic risk, sports betting has NO risk premia. Every sports betting edge requires a market mistake — someone has to be wrong for you to profit. This means every profitable model has a ticking clock; edges erode as others discover them.
The common belief that bookmakers set odds to get equal action on both sides is wrong. On a single-market basis, action is NOT balanced. If bookmakers moved draw odds in football to balance money, sharps would immediately exploit the distorted line. Instead, bookmakers profit via the law of large numbers across thousands of markets — accepting imbalanced positions on individual events.
The widely-taught "start in low-liquidity markets" strategy has a critical caveat: approximately 80% of identified edges in smaller/illiquid markets are eliminated by widened bid-ask spreads. Player prop vig is often 8%+ (vs. 2-4% on main lines). You need more edge in thin markets, not just any edge.
"People's read of the incentives is often overblown." The market already prices obvious motivation asymmetries. Backing the "hungry team" in a narrative-driven spot often means paying for an edge that's already in the line. Being contrarian about obvious motivation narratives is frequently more profitable.
Tuning model drift factors to match bookmaker odds is circular — you're forcing your model to agree with the market, which is the exact thing you're trying to beat. The whole point of a model is independent assessment. Calibrating to market prices produces a model that confirms the market rather than finding where it's wrong.
Opta — the official Premier League data provider with the best raw data in football — lost $4,735 on a $10k bankroll across one EPL season using Kelly staking. Credibility, data quality, and brand prestige have zero correlation with betting model accuracy. Most "credible" models are marketing exercises, not sharps-tested.
If your model says a bet has 10% edge, the real edge is probably 4% after accounting for market information you haven't incorporated. The market already knows most of what you know. Rufus's calibration: 70% market / 30% own model in college football. Most bettors never regress their model to the market, leading to oversized bets on phantom edges.
The standard affiliate model pays tipsters a share of their subscribers' LOSSES to the bookmaker. The worse your tips, the more affiliate revenue you earn. When evaluating betting advice, check whether the revenue model profits from your success or your failure.
120 tipsters submitted pre-existing records showing ~17% profit on turnover across 24,000 tips. Under independent verification on 90,000+ tips: ~0.4-0.5% profit — almost complete regression. They submitted because they were winning (survivorship bias), not because they were skilled. Any unverified track record should be discounted by ~95%.
A winning bettor (+7.7% neutral ROI, +14 on 182 bets) paid £1700 in vig over a season. No bonus scheme compensates for that. At double the vig, that same winning bettor goes negative. The math is unambiguous: margin structure matters more than any promotion.
Burnley's 18-goal defensive overperformance exceeds what Leicester achieved while winning the Premier League (~11 goals in StatsBomb data). When a non-elite team's overperformance exceeds what champions sustain, regression is near-certain. This turns "regression" from a vague concept into a quantified, actionable threshold.
A player's point total depends on two independent sources of variance: efficiency (points per minute) and playing time (minutes). Point-estimate models collapse this into one number and miss the variance from playing time fluctuation. A player averaging 1.2 pts/min who plays anywhere from 15-35 minutes has enormous outcome variance that a single "projected 28 points" can't capture.
When your model shows a huge edge (e.g., 29% on a single game), naive model-market averaging is dangerous because the situations where your model disagrees most with the market are precisely the situations where YOUR model is most likely wrong. A horse your model prices at 3/1 going off at 25/1 is a debugging signal, not confirmation of a huge opportunity.
High-parity leagues increase result variance independently of analytical edge. Losing streaks get longer even when your reads are correct. Most bettors interpret parity-driven cold streaks as "my model is broken" and change approach — when the real problem is insufficient sizing discipline for the environment.
Reducing costs (vig) is the highest-leverage activity available because the return is certain and compounds across every bet. No model improvement — which is uncertain in both magnitude and direction — can match the certain compounding of a 1% vig reduction applied to hundreds of bets per season.
Different risk contexts require fundamentally different sizing tools. In the probability regime (bounded losses, repeated bets), use Kelly to win the most. In the plausibility regime (model could be systematically wrong), use fractional Kelly to lose the least. In the possibility regime (tail risks that could wipe you out), use survival-first sizing. The market-model discrepancy tells you which regime you're in.
When a coaching change swaps in a manager whose system is fundamentally incompatible with the squad built for the predecessor, 4-8 weeks of predictable underperformance follows. This isn't generic "transition difficulty" — it's a specific mechanic: players trained in one system physically cannot execute an opposite one without retraining. The fade has a predictable duration based on tactical distance.
NIL compensation has dramatically expanded the talent gap between power programs and mid-major schools. Historical first-round upset rates (approximately 1-in-3 for 12-seeds, 1-in-5 for 13-seeds) were built on pre-NIL parity. If the structural change is durable, the market continues pricing historical rates while true rates have shifted — making favorite-side early-round bets systematically underpriced.
P&L feedback makes bettors more risk-averse when winning (cut winners short, cash out early) and less risk-averse when losing (let losers run, chase losses). This doesn't just hurt individual bet sizing — it systematically destroys the payoff profile of the entire strategy, turning a profitable edge into a losing one by asymmetrically clipping the tails.
Winning produces cruise control; losing triggers deep investigation where model improvements actually come from. The asymmetric learning cycle (losing → diagnosis → improvement vs. winning → coasting → stagnation) means the bettor who systematically investigates losses compounds improvement, while the bettor who celebrates wins stagnates.
In-play betting compresses all biases (recency, loss aversion, chasing) into minutes instead of days. A bad beat in the 60th minute triggers the same emotional cascade as a week-long losing streak — but with a live market still open to act on impulsively. The speed advantage in live markets belongs to automated systems, not humans operating under emotional duress.
The entire 50-100% probability range maps to just 1.0-2.0 in decimal odds, while a mere 5% band (20-25%) spans a full unit (4.0-5.0). A 70% chance priced at 1.43 "feels" uncomfortably short to both bettors AND bookmakers, creating real value on favorites purely from the format's perceptual distortion.
What looks like irrational longshot bias in prediction markets is actually efficient pricing of fee structure. Biden at 94 cents on PredictIt election morning wasn't a 94% probability — with 10% profit commission + 5% withdrawal fee, buying at 94 cents yields only 0.4% return. A 5-6 cent band at each tail is a "no man's land" where fees prevent price correction.
24-25 EPL home-away goal difference collapsed to 0.09 — essentially zero HFA, vs. a historical norm of 0.32-0.41. This was the second-lowest on record behind COVID empty-stadium games. This single structural shift was "the main culprit for underperformance vs. model." Meanwhile Championship maintained strong HFA in the same season. HFA is league-specific AND season-specific.
Markets are slightly biased toward quality and reputation over situational advantage. Home underdogs who are close-to-competitive get systematically underpriced, especially in Championship where HFA is strong and parity is high. Ted explicitly structures his approach around hunting for these spots.
Rufus Peabody's COVID decomposition found that crowd noise accounts for only ~15% of total home field advantage. Familiarity and routine (~50%) and reduced travel fatigue (~35%) persist without fans. This means models using "zero HFA" for empty stadiums or neutral sites are deeply wrong — 85% of HFA survives even without fans.
When you place a bet and the line doesn't move in your direction — or stays flat — that's information. The market disagrees with you. Rufus explicitly notes that his public picks are the ones with market resistance, "which probably means they're not quite as strong." Conversely, line moving toward you after betting = strong confirmation.
Markets underweight CB absences relative to forward absences. "CBs are a big deal" — defensive injuries cause systemic team deterioration because defensive organization is collective and fragile, while attacking quality is more individual and replaceable.
Some players' absence breaks the entire tactical system — it's a binary on/off switch, not a marginal adjustment. Brentford's Norgaard, Arsenal's Saka, Norwich's Sargent — their absence doesn't just remove one player, it collapses the way the team functions. Standard per-player injury discounts completely miss this non-linear cliff.
Lines are initially set by algorithms, then adjusted as human information filters in. The exploitable gap is between machine pricing and reality that only humans can see — injuries announced late, coaching changes not yet priced, financial crises, player chemistry shifts. The machines are good at what they do, but they're blind to anything not in the data.
Outright markets represent less than 1% of football betting volume. Capital gets locked for 9 months, punters can't recycle stakes. Because of this, bookmakers almost certainly don't build sophisticated models for these markets — they use simple approximations and risk-management shading. This is a rare case where the bookmaker is explicitly NOT trying to be efficient.
The main game spread is the at-the-money strike; alternate lines are OTM/ITM strikes. Vig increases as you move away from the main line — just like implied volatility increases for options strikes further from ATM. When alternate line pricing is inconsistent with the vol surface implied by the efficient main line, the bookmaker has mispriced the distribution tail.
The most robust, enduring inefficiencies in betting markets are driven by behavioral psychology — because "people as a whole have a tremendous amount of trouble seeing both sides of something." Consensus one-sidedness, narrative-driven pricing, and recency bias create durable mispricing categories that survive algorithmic improvements because they're structural to human cognition.
As more bettors acquire similar skills, information, and rating systems, they cancel each other out — what remains in results is noise. A bettor clearly profitable in 2010 may find the exact same approach breakeven in 2025 — not because they got worse, but because everyone else got better. The response is not "try harder at the same thing" but "find a new dimension of edge."
Bookmakers allocate analytical resources proportional to betting volume. EPL main lines get the most sophisticated modeling; corners, player props, and lower-league totals get far less. The mispricing in low-liquidity markets isn't just larger — it's structurally persistent because the books will never invest in fixing it.
Deep sport knowledge (watching games, understanding rotations, knowing player tendencies) has limited value in main lines where it's already priced in — but has high value in thin markets where the bookmaker's model doesn't capture it. A basketball watcher who understands minute rotations has an edge in player props that they don't have in the point spread.
Super Bowl prop betting evolved from 30%+ opener edges (early 2010s, where receiver lines opened 13+ yards off fair value) to edge concentrated in three specific niches: (1) props that only appear at the Super Bowl (books have no weekly pricing history), (2) game-day action (~95% of Rufus's volume is Saturday/Sunday as public distortions create late opportunities), and (3) the "no" side of yes/no props (public hammers "yes," inflating the line).
EPL team ratings stay flat (99% of initial) across a season. Championship ratings end at 149% of initial dispersion, League One at 168%, League Two at 182%. This means a static-rating model works fine for EPL but is systematically wrong for lower divisions where 17-25% of teams are new each season and quality shifts dramatically.
Books set prop lines using the mean (average). But skewed distributions (receiving yards, points scored, rushing yards) always have a median below the mean — big games pull the mean up while most games are below average. In-game injuries can only hurt, never help, a player's total. The structural result: overs are systematically overpriced and unders systematically underpriced.
A binary garbage-time cutoff (e.g., "filter all plays under 5% WP") creates discontinuous functions and throws away real signal. Teams performing well in garbage time exhibit real skill (moving the ball, completing passes) — it's just lower-leverage. A continuous leverage multiplier that drops off steeply near blowout territory captures this signal while the public's binary "doesn't count" filter discards it.
When a famous coach "unlocks" a QB (or any high-narrative story emerges), the market's line adjustment outruns the data signal. After 2 strong games with Cam Newton + Belichick, the market moved NE to -6.5; Rufus's model had -2.9. A well-calibrated model should only move ~2 points after 2 NFL games — a 3.5-point gap was entirely narrative-driven.
When vig is heavily skewed to one side of a line (-127/+109 vs. neutral -105/-105), the skew itself tells you where sharp money has already landed. A bet that's +EV at neutral vig may have its edge already captured via the bookmaker's vig adjustment. Reading the vig shape is reading the market's real opinion.
Financial markets have genuine risk premia — compensation for bearing systematic risk that cannot be diversified away (equity risk premium, VRP). These are structurally stable because they're tied to economic fundamentals. Sports betting has NO equivalent. Every sports betting edge is a distortion or inefficiency — which means it will be competed away as other bettors find it. Sports edges erode faster than financial edges because they have no structural reason to persist.
Sufferball teams (Forest, Burnley) suppress totals AND cap favorites' spread coverage. Chaos teams (West Ham, 31+ shots/match) inflate Overs. The game PROFILE created by the style matchup is more predictable than the winner. Style analysis shifts you from "who wins?" to "what kind of game will this be?" — a much more answerable question.
Some coaches systematically stop attacking once ahead — Corberan's WBA took zero shots from the 70th minute onward with a lead. This creates great aggregate xGD but systematically fails to cover large handicaps and accumulates draws. xG models see a "good team" while the behavioral pattern says "reliable to cover -0.5 but will never cover -1.5."
"Models will take ages to catch up." The bet isn't "my model is more accurate overall" — it's "I spotted this team's improvement or decline 3 weeks before the market did." The gap between reality and market perception IS the edge, and it's time-bounded. By the time models catch up, the value is gone.
"Find the unfancied team near the bottom who is slightly better than people think, and ride that horse again and again." Bookie models overweight historical/legacy ratings and adjust slowly. Teams like Oxford or Plymouth after a coaching change become systematically undervalued because models still price them as cellar dwellers. This is Ted's most repeated and profitable pattern.
European teams in American summer heat can't even train. 3pm kickoffs in Charlotte, Cincinnati, and Miami systematically degraded European team performance in the CWC. With the 2026 World Cup in North America, this pattern will recur — and it's unlikely to be priced correctly because it's unprecedented at this scale.
Everyone focuses on picking winners. The hidden lever is that the SAME winning bets produce losses at high-vig books. A bettor who is +7.7% before vig can be negative after vig if their bookmaker charges double the industry minimum. Margin is the first-order variable, not analytical skill.
On a 50/50 event at $100 stakes over 10,000 bets: odds of 1.95 produces -$20,000; odds of 2.05 produces +$15,000. A 5-cent difference in decimal odds — invisible on any single bet — compounds into a $35,000 gap. This makes the case for obsessive line shopping more visceral than any theoretical argument.
The main spread in major sports receives the most analytical attention from bookmakers and sharp bettors. Alternate lines (and their implied distribution shape) often receive much less. This creates a second-order inefficiency: the median is correct, but the variance structure is mispriced. A model that produces full distributions can find value in the alternate market even when the main spread has no edge.
When a key playmaker (QB, center, point guard) is injured, the distribution of outcomes widens — the game becomes more random because the remaining players have higher uncertainty. The market adjusts the median line for the expected average impact, but often fails to widen the distribution appropriately. The alternate lines and totals remain priced as if the game were "normal variance."
Teams trailing early rack up flattering xG while chasing — shots accumulated after falling behind look impressive on paper but are low-quality, game-state-inflated numbers. If you don't strip out these score effects, you systematically overrate losing teams and underrate dominant ones who concede garbage-time chances.
A team that "looks terrible on xG but keeps winning" could be elite goalkeeping (partially sustainable over 1-2 seasons) or random finishing variance (fully regresses within weeks). PSxG+/- (Post-Shot Expected Goals plus/minus) isolates the goalkeeper's contribution. Without this decomposition, all defensive overperformance looks identical — but the sustainability profiles are completely different.
Value exists precisely where most people don't want to bet. The willingness to bet on ugly teams at ugly prices in ugly situations is itself a structural advantage because the supply of willing bettors on those sides is lower, leaving more value unclaimed. Ted repeatedly takes bets he personally hates because the math demands it.
If you quote a probability on an outcome, you are logically bound to accept a bet at any favorable odds relative to that probability. If you say something is 60% and refuse even-money, your probability is either dishonest or doesn't incorporate real-world factors like model uncertainty. A forecast without skin in the game is just an opinion with math around it.
Outrights have LARGER per-bet edge than match bets (less efficient market), but match bets recycle capital across 40 game weeks while outrights lock it for 9 months. Cap outrights at 5-10% of bankroll — optimize total season profit, not edge per bet. The market with the biggest edge per bet is NOT always the one that produces the most total profit.
Kelly criterion requires knowing your precise edge per bet — which is inherently uncertain. Variable sizing amplifies errors in edge estimation: if you overestimate edge on one bet, Kelly overallocates and destroys returns. Flat staking is the "minimum regret" strategy that protects against the unknowable.
For distant events, leave ~12% probability mass unallocated as an uncertainty cushion. A basketball game tonight warrants 0.5-1% cushion; NFL this weekend 2-3%; an election months away 12%+. The cushion represents things that COULD happen but can't be priced — coaching changes, injuries, scandals, rule changes. Only bet when your edge exceeds the cushion.
When a strong-performing team fires its coach, bookmakers immediately slash ratings. But actual performance under a new coach often reverts to squad quality, not to zero. The exploitable window is betting the team at deflated odds during the immediate post-sacking panic, before metrics stabilize.
Two bettors with identical modeling skills will have very different results if one has found a real causal effect and the other hasn't. The effect is the durable competitive advantage — it represents genuine market mispricing. The model is just infrastructure to exploit it. Most bettors focus 90% of their energy on models and 10% on effects, when the ratio should be reversed.
There are two distinct phases of development: sponge (absorb everything, find edges, explore) and operator (reduce noise, narrow focus, execute with conviction). The transition signal is clear — you have a validated effect. Most bettors never make the transition, staying in perpetual learning mode that prevents committed execution.
A billionaire quit betting despite affording infinite losses: "It occupies too many brain cycles." The biggest risk for semi-professional bettors isn't bankroll ruin — it's attention ruin. The bettor who controls their mental investment makes better decisions both in betting and outside it. The meta-skill is managing your attention budget, not just your bankroll.
When a bettor's identity is fused with their professional label ("I am a sports bettor"), losing a bet feels like losing self. This directly prevents objective edge evaluation, makes it impossible to quit a dead market, and blocks necessary breaks. The diagnostic is simple: can you answer "who are you?" without referencing your profession?
Continuous medium-intensity grinding produces wasted motion and prevents strategic thinking. The most effective working mode is hyper-focus sprints (peak season, new model development) followed by genuine reset periods (no screens, different stimulation). Rufus came back from Burning Man with "a bunch of ideas" and renewed motivation — the reset produced more edge than grinding through would have.
Lines are widest early (high vig), then compress as match day approaches (low vig but smart money has moved the price). The 1-2 day pre-match window captures value AFTER vig compression but BEFORE hedge fund-scale liquidity arrives to move the lines. This "Goldilocks zone" is a deliberate timing strategy, not an accident.
Professional timing strategy is bifurcated by bet type. Sides/totals should be bet early when your model has conviction (racing analytical consensus that forms quickly). Props should be bet late on game day (waiting for recreational money to inflate popular player overs, creating under value). Most bettors use a single timing approach for all bets.
During halftime, prediction market second-half prices settle before sportsbooks finalize their lines. A model that predicts where the second-half line will open — even an "antiquated model that hasn't been updated in 10 years" — generates edge because the structural timing gap exists regardless of model quality.
After the Trump assassination attempt, 15-20 minutes passed before anything appeared on Twitter. The "French whale" who bet $50-80M on Polymarket commissioned private polls (~$1M cost) asking who neighbors would vote for, capturing the shy voter effect. Being first to proprietary information is worth more than being better at analyzing public information — the analytical edge is already competed away.
In automated prediction markets, bots are significant participants. They execute with predictable logic that can be reverse-engineered by studying their fill patterns, price triggers, and failure modes. "The best trader will always be better than the best bot" — because bots can't handle novel situations and their deterministic responses can be exploited once mapped.
"Don't Give a Fuck" isn't casual observation — it's a systematic filter that professionals apply formally. Teams that have clinched safety get categorized as DGF, and their opponents' lines are adjusted. The asymmetry between a DGF team and one still fighting creates persistent, predictable mispricing that models can't see.
Market-Implied ratings (what the market thinks) and xG models (what the data shows) approach team quality from completely different angles. When both agree, confidence is high. When they diverge, the divergence itself is the signal — investigate WHY they disagree, because one of them is seeing something the other can't.
Some defensive styles suppress opponent quality without generating big xGD advantages — they're invisible to models but visible to someone who watches closely. Ted explicitly overrides his own models when pattern of play tells a different story. This is the rarest and most valuable edge: knowing when to trust what you see over what the numbers say.
When losing $300K in the first 5 weeks of college basketball, Rufus checked CLV (approximately -0.5%) and concluded: "Worst case, we're going to lose a half percent going forward. We're not just throwing darts." He held course and finished the season at +5% ROI. Checking win rate during drawdowns leads to panic; checking CLV leads to correct decisions.
Rufus's default is to assume any potential override factor is noise — no adjustment. Injuries are the primary exception. "Improved culture," "new coach philosophy," narrative factors — all noise until measured. The bettor who tracks every override and compares override prediction vs. model prediction to actual outcome discovers that most overrides destroy value.
Outrights are tradeable positions in a portfolio, not lottery tickets. Ted bought Inter at +975, then hedged against PSG in the final to lock in leveraged profit while leaving upside. This is qualitatively different from "I bet my team to win" — it's using futures markets as financial instruments with entry, exit, and hedge strategies.
Tracking bets AGAINST specific teams revealed that "against Manchester City was the most profitable PL pattern." The asymmetry of who you're fading matters as much as who you're backing — some teams are systematically overrated by the market, making every bet against them +EV regardless of who they're playing.
Sharp bettors deliberately place recreational-looking bets at square books to maintain access. Deposit with credit card, bet random accumulators early, save sharp picks for discount books. This is a meta-strategy that only matters once you're actually winning — and it's invisible to beginners because they never face the problem it solves.
Every strategy has a set of conditions under which it performs. Testing a strategy unconditionally across all environments misses the signal that it only works some of the time — and deploying it in the wrong environment is what creates the most damage. The professional who conditions their strategy on validated environmental filters avoids deploying their edge when the environment is hostile.
"Teams That Actually Give a Fuck — this is weirdly the most important variable." In league play, motivation is roughly equal. In tournaments, it's the single biggest factor. Check transfers and pre-tournament preparations — teams collecting appearance fees play fundamentally differently than teams going for glory.
"The bets went 3-5 but the PROCESS outcomes went 6-2." Elite bettors rate every bet thumbs-up/thumbs-down on process quality independent of outcome. This is how you distinguish skill from luck — and it prevents the two most common errors: abandoning a winning strategy during a cold streak, and persisting with a broken strategy during a lucky hot streak.
Championship lines on Fridays carry 14-20 cents of vig vs. 10 cents on match day — nearly double. WHEN you bet matters almost as much as WHAT you bet. A profitable system can become unprofitable purely through bet timing.
The largest professional betting syndicates (Gelco: 80 people, $1.2B/year, $180M net) don't primarily profit from picking winners — they break even on bets and profit from rebate structures. Tracks charge ~15% commission; rebate companies provide a 5% discount (10% instead of 15%). Meanwhile, retail bettors face an effective 18% blended take because the pool includes professional volume at reduced rates.