Chapter 22 Quiz: Modeling Emerging Markets
Test your understanding of esports modeling, golf strokes gained, player props, futures markets, and niche sport approaches.
Question 1. What are the three largest esports betting markets by volume, and what game format does each use?
Answer
The three largest are: (1) Counter-Strike 2 (CS2) -- a 5v5 tactical first-person shooter played in best-of-one or best-of-three map series with rounds to 16 per map; (2) League of Legends (LoL) -- a 5v5 MOBA played best-of-one in regular season and best-of-five in playoffs, with individual games lasting 25-45 minutes; (3) Dota 2 -- a 5v5 MOBA with similar structure to LoL but more complex drafting and different gameplay mechanics.Question 2. What is a "patch" in esports, and why does it create both a modeling challenge and a betting opportunity?
Answer
A patch is a developer-released modification to game balance, maps, or mechanics. It creates a modeling challenge because historical data may become less relevant after a significant patch, as strategies and characters that were previously effective may be weakened or strengthened. It creates a betting opportunity because: (1) the market may be slow to incorporate the patch's impact on team strength; (2) teams that benefit from the patch may be temporarily undervalued; (3) modelers who quickly identify the meta shift can gain an edge before the market adjusts.Question 3. Explain the map veto system in CS2 and why it matters for match prediction.
Answer
In a best-of-three CS2 match, each team bans maps, then picks maps from the seven-map competitive pool. This creates a strategic element because teams have strong map preferences and weaknesses. Modeling which maps will be played and each team's win probability on each specific map is essential because a team might be a 60% favorite on their best map but only 40% on their worst. The overall series probability depends critically on which maps are played, making map-specific ratings and veto prediction key components of accurate match prediction.Question 4. What is the strokes gained framework, and how is a single shot evaluated?
Answer
Strokes gained measures performance relative to the PGA Tour average by comparing expected strokes remaining before and after each shot. For a single shot: SG = (Expected strokes from start position) - (Expected strokes from end position) - 1. If a golfer takes a shot from a position where the average player needs 3.8 strokes to hole out, to a position where the average needs 2.5, they used 1 stroke but "gained" 3.8 - 2.5 - 1.0 = 0.3 strokes relative to the field average.Question 5. Name the four components of the strokes gained decomposition in golf.
Answer
The four components are: (1) SG: Off the Tee (SG:OTT) -- driving performance on par 4s and par 5s; (2) SG: Approach (SG:APP) -- approach shots into the green; (3) SG: Around the Green (SG:ARG) -- chips, pitches, and bunker shots near the green; (4) SG: Putting (SG:PUTT) -- performance on the putting surface. Total SG equals the sum of all four components.Question 6. What is "course fit" in golf modeling, and how does it affect tournament predictions?
Answer
Course fit measures how well a golfer's specific skill profile matches a course's demands. Different courses place different demands on the four SG components. A long, open course rewards SG:OTT, while a short, tight course with small greens rewards SG:ARG and SG:APP. Course fit analysis uses regression to determine which SG components are most predictive at each venue, then weights golfer profiles accordingly. This can significantly alter predictions: a golfer strong off the tee but weak around the green will be overvalued at a short-game-demanding course if course fit is ignored.Question 7. Why is field strength adjustment necessary when comparing strokes gained data across tournaments?
Answer
Because strokes gained is measured relative to the field average, and field quality varies enormously between tournaments. A golfer gaining 2.0 strokes per round at The Masters (a 90-player elite field) is performing at a much higher absolute level than one gaining 2.0 strokes at a weaker-field event. Without adjustment, raw SG from weak-field events appears inflated. The adjustment adds a field strength offset based on the difference between the average world ranking of the actual field and a reference field, normalizing all SG data to a common baseline.Question 8. What five adjustments does a player prop projection system typically apply to a baseline projection?
Answer
The five typical adjustments are: (1) Pace adjustment -- scaling for the expected number of possessions or plays in the game; (2) Game total/script adjustment -- accounting for the implied total and spread, which affect playing time and strategy; (3) Blowout risk -- reducing projections when a large spread suggests starters may rest in the fourth quarter; (4) Matchup adjustment -- using Defense vs. Position (DvP) ratings to account for opponent defensive quality at the relevant position; (5) Usage/role adjustment -- modifying for teammate absences or role changes that affect the player's opportunity share.Question 9. What is the Defense versus Position (DvP) rating, and how is it used in prop modeling?
Answer
DvP measures how many stats (points, yards, etc.) an opposing defense allows to players at a specific position relative to the league average. A DvP of +0.05 means the opponent allows 5% more than average to that position. It is used as a multiplicative adjustment to the player's baseline projection: $\hat{Y}_{\text{matchup}} = \hat{Y}_{\text{base}} \times (1 + \text{DvP})$. This captures the opponent's defensive weakness or strength at the specific position being modeled.Question 10. Why is correlation between props critical for same-game parlay pricing?
Answer
SGPs combine multiple props from the same game, and the correct price depends on the correlation between legs. If the sportsbook overstates the correlation between positively correlated props (e.g., QB passing yards and WR receiving yards, which are trivially correlated), the parlay price is too low. If it understates correlation between negatively correlated props (e.g., two receivers' targets on the same team), the price is too high. Bettors with accurate correlation models can identify when the SGP price is better than the true joint probability implies.Question 11. What are the four common types of futures bets in sports betting?
Answer
The four common types are: (1) Championship futures -- betting on which team wins the championship (Super Bowl, NBA Finals, etc.); (2) Division/conference winners -- betting on divisional or conference outcomes; (3) Award futures -- MVP, Rookie of the Year, and other individual awards; (4) Season win totals -- over/under on a team's regular-season wins.Question 12. How do you extract implied probabilities from a futures market and remove the overround?
Answer
First, convert each outcome's American odds to raw implied probability: for positive odds, $p = 100 / (odds + 100)$; for negative odds, $p = |odds| / (|odds| + 100)$. The overround is the sum of all raw probabilities minus 1 (e.g., if they sum to 1.30, the overround is 30%). The simplest vig removal method is multiplicative normalization: divide each raw probability by the total sum. More sophisticated methods include the power method (finding an exponent k where $\sum p_i^k = 1$) and Shin's method (which accounts for the favorite-longshot bias).Question 13. What is the full hedge calculation for a futures position?
Answer
If you hold a futures bet with potential payout P (including stake) and the opponent in the final is at decimal odds D, the full hedge bet amount is $H = P / D$. This equalizes profit regardless of outcome. If the original bet wins, profit = P - original stake - H. If the opponent wins, profit = H * D - original stake - H. Both paths yield approximately equal profit. A partial (50%) hedge uses H/2, which guarantees less but preserves more upside if the original bet wins.Question 14. Why is "entry timing" important for futures betting?
Answer
The optimal time to bet futures is when the market has not yet incorporated information the model has. Pre-season futures have the widest margins but the most uncertainty. In-season futures are more efficient but have less uncertainty. The sweet spot is often early in the season, when a few weeks of data have significantly updated the model but the market is still partially anchored to pre-season assessments. Additionally, the market is slowest to adjust immediately after major roster changes (trades, injuries), creating temporary mispricings.Question 15. What is the "favorite-longshot bias" in futures markets?
Answer
The favorite-longshot bias is the empirical observation that sportsbooks tend to apply more margin to longshot outcomes than to favorites. In a futures market, a +5000 longshot may have more vig embedded in its price than a +200 favorite. This means that simple multiplicative normalization (which assumes proportional vig distribution) understates the true probability of favorites and overstates the probability of longshots. Methods like Shin's method account for this by modeling the market as containing a proportion of insider money that distorts prices.Question 16. What are three strategies for overcoming data scarcity in niche sport modeling?
Answer
Three strategies are: (1) Transfer learning -- adapting modeling frameworks from well-studied sports to niche sports with similar structure (e.g., tennis Elo to table tennis Elo); (2) Bayesian methods with informative priors -- incorporating domain knowledge about player quality even before observing results, using priors based on rankings, junior results, or tier of competition; (3) Proxy metrics -- using easily observable statistics (three-dart average in darts, break rate in snooker) as proxies for underlying ability when more sophisticated metrics are unavailable.Question 17. How do thin market sportsbooks typically set their lines, and what patterns does this create?
Answer
In thin markets, sportsbooks often set lines algorithmically using simple models or by copying from third-party odds feeds. This creates exploitable patterns: (1) slow adjustment to form changes because the book does not closely track player performance; (2) incorrect head-to-head pricing when there are strong historical patterns the algorithm misses; (3) stale prices from delayed odds feed updates; (4) mispriced correlations between related markets (moneyline vs. total) because derivative markets receive even less attention.Question 18. What is the "roster change discount" in esports modeling?
Answer
The roster change discount reduces confidence in a team's rating after a player change, proportional to the importance of the changed player. Since esports rosters are small (5 players), replacing one player can fundamentally alter team capabilities. The discount is implemented by increasing the Glicko-2 rating deviation equivalent. An in-game leader (IGL) change in CS2 warrants a larger discount than replacing a utility player; a mid-lane change in LoL warrants more than a support change. The team's predictions are pulled toward uncertainty until post-change results are observed.Question 19. What is the Monte Carlo approach to golf tournament simulation?
Answer
The approach: (1) For each golfer, compute a course-fit-adjusted expected strokes gained per round; (2) For each simulation, sample round scores from a distribution calibrated to the golfer's SG profile and historical variance; (3) After 36 simulated holes, apply the cut rule (top 65 and ties); (4) For surviving golfers, simulate rounds 3 and 4; (5) Record finish positions; (6) Repeat thousands of times to build probability distributions for all outcomes -- win probability, top-5/10/20 finishes, and make-cut probability.Question 20. How does the back-to-back penalty affect NBA player prop projections?
Answer
Back-to-back games (playing on consecutive days) reduce player minutes by approximately 2-3 minutes on average, with corresponding reductions in all counting stats. The prop model applies a multiplicative penalty (typically 5-7%) to the baseline projection: $\text{projection} \times (1 - b2b\_penalty)$. This adjustment is valuable because: (1) the market may not fully account for schedule effects; (2) the effect varies by player age and role (older stars lose more minutes); (3) some sportsbooks set lines before knowing the starting lineup, which may involve rest decisions.Question 21. What makes golf outright winner markets particularly inefficient?
Answer
Golf outright markets are among the most inefficient because: (1) With 144-player fields, the favorite typically has only 10-15% implied probability, creating a long tail of opportunities; (2) Course fit effects can make a 100/1 golfer meaningfully more likely to win than the market implies; (3) The field size means sportsbooks cannot devote deep analysis to each golfer; (4) Derivative markets (top-5, top-10, top-20, make/miss cut, matchup bets) are priced from the same underlying model but offer different risk-return profiles; (5) Week-to-week variance in golf is high, creating more noise for the market to parse.Question 22. What is the typical overround for NFL Super Bowl futures compared to a regular game spread?
Answer
NFL Super Bowl futures typically carry an overround of 20-40% (meaning the sum of all implied probabilities is 1.20 to 1.40), compared to approximately 4-5% overround on a standard game spread at a reduced-juice book (or about 10% at a standard -110/-110 book). The much higher futures overround reflects the sportsbook's higher cost of carrying long-term risk, the wider margins charged for markets with less sharp action, and the difficulty of managing inventory across 30+ outcomes.Question 23. What is Shin's method for vig removal, and when should it be preferred over multiplicative normalization?
Answer
Shin's method models the overround as arising from a proportion of "insider" bettors who know the true outcome, which distorts prices by applying more margin to longshots. The method estimates this insider proportion (z) and uses it to adjust probabilities in a way that removes more vig from longshots and less from favorites. It should be preferred over multiplicative normalization when: (1) the market has many outcomes (e.g., 30+ team futures); (2) there is reason to believe the favorite-longshot bias is present; (3) accurate probability estimation for longshot outcomes matters (e.g., for identifying value on +5000 teams).Question 24. Describe the concept of "model transferability" and give an example.
Answer
Model transferability is the practice of adapting a modeling framework developed for one sport to another sport with similar structure. The mathematical structure of the model transfers even when the parameters must be re-estimated. For example, a surface-specific tennis Elo system and a map-specific CS2 Elo system share the same mathematical structure (maintaining separate ratings for each context, blending with overall rating based on sample size). Only the K-factors, blending weights, and initialization differ. Similarly, an NBA player prop model can transfer to handball player props with sport-specific adjustments to pace calculation and stat distributions.Question 25. Why does the chapter recommend that futures bets should have "at least double the minimum edge threshold of game-by-game bets"?