Chapter 22 Exercises: Modeling Emerging Markets

Part A: Conceptual Questions (Exercises 1-8)

Exercise 1. Explain why esports betting markets are structurally less efficient than traditional sports markets. Identify at least four specific factors that contribute to this inefficiency and describe how a quantitative bettor can exploit each one.

Exercise 2. The concept of "patch effects" is unique to esports. Describe what a game patch is, how it can invalidate historical data, and propose a systematic framework for handling patch transitions in a predictive model. Compare this challenge to the closest analogue in traditional sports.

Exercise 3. In the strokes gained framework for golf, performance is decomposed into four components: Off the Tee, Approach, Around the Green, and Putting. Explain why this decomposition is analytically powerful for tournament prediction. How does course fit analysis leverage this decomposition, and why is it more informative than using overall strokes gained alone?

Exercise 4. Player prop markets have exploded in popularity. Discuss why sportsbooks offer such wide margins on props compared to sides and totals. From the bettor's perspective, explain why prop markets paradoxically represent some of the best opportunities despite higher margins.

Exercise 5. Futures markets tie up capital for extended periods. Describe the concept of "time value of money" as it applies to futures betting. Calculate the minimum edge needed for a futures bet that locks up capital for five months, assuming the bettor could alternatively find 2% edge bets with weekly turnover during that period.

Exercise 6. The chapter argues that niche sports specialization follows the same logic as niche market specialization in finance. Develop this analogy: what is the niche sports equivalent of a market maker, an informed trader, a liquidity provider, and an arbitrageur? How does the "cost of information" differ between betting on the NFL versus betting on Russian table tennis?

Exercise 7. Explain the concept of "model transferability" across sports. Give three specific examples of model frameworks that can transfer from a well-studied sport to a niche sport with minimal structural modification. For each, identify what transfers directly and what must be re-estimated.

Exercise 8. Same-game parlays (SGPs) combine multiple prop bets from the same game. Sportsbooks price SGPs by assuming some correlation structure between the components. Explain why the sportsbook's assumed correlation is often wrong, and describe two specific scenarios where this creates exploitable value for a bettor with a proper correlation model.

Part B: Calculations (Exercises 9-15)

Exercise 9. A CS2 team has the following map-specific Elo ratings: Mirage 1650, Inferno 1580, Nuke 1720, Overpass 1600. Their overall Elo is 1640, and they have played 30 maps on each map. Using a blending threshold of 20, calculate their blended ratings on each map. Then calculate the win probability against a team with overall Elo 1600 and blended Mirage rating 1620.

Exercise 10. A golfer has the following strokes gained profile: SG:OTT = +0.8, SG:APP = +0.6, SG:ARG = +0.3, SG:PUTT = +0.1. A course has the weight profile: OTT = 0.30, APP = 0.30, ARG = 0.18, PUTT = 0.22. (a) Calculate the course-fit-adjusted expected strokes gained. (b) Calculate the same golfer's course-fit score at a course with weights OTT = 0.15, APP = 0.35, ARG = 0.28, PUTT = 0.22. (c) At which course does this golfer have the larger advantage, and why?

Exercise 11. A player prop model projects LeBron James to score 28.5 points with a standard deviation of 8.2 points. The sportsbook line is 26.5 points with Over -115 and Under -105. (a) Calculate the model's probability of the Over. (b) Convert -115 to implied probability. (c) Calculate the expected value of betting the Over. (d) Determine the Kelly fraction for this bet.

Exercise 12. An NFL Super Bowl futures market has the following odds: Team A +350, Team B +600, Team C +900, Team D +1500, Field +200. (a) Calculate the raw implied probabilities and the total overround. (b) Apply multiplicative normalization to remove the vig. (c) If your model assigns Team C a 12% championship probability, what is the edge and expected value?

Exercise 13. You placed a $200 futures bet on a team at +1500 before the season. The team has now reached the conference championship, and the opponent is listed at +140 (decimal 2.40). (a) Calculate the full hedge bet amount to equalize profit. (b) Calculate the guaranteed profit regardless of outcome. (c) Calculate a 50% partial hedge and the resulting profit range.

Exercise 14. In a League of Legends match, Team A has a gold difference at 15 minutes averaging +1,200 across the season, while Team B averages +400. Historical analysis shows that gold difference at 15 minutes has a correlation of 0.55 with match outcome. If Team A's base Elo win probability is 58%, apply a simplified gold-advantage adjustment (1 percentage point per 500 gold differential advantage) to estimate the adjusted win probability.

Exercise 15. A thin-market rating system for professional darts uses an adaptive K-factor based on rating deviation. Player A has a rating of 1680 with RD of 80, and Player B has a rating of 1620 with RD of 150. The base K-factor is 40, and the adaptive formula is $K = K_{\text{base}} \times RD/100 \times \max(0.5, 2.0 - \text{matches}/20)$. Player A has 25 matches and Player B has 8 matches. Calculate the K-factor for each player, the expected outcome, and the rating updates if Player A wins.

Part C: Programming (Exercises 16-20)

Exercise 16. Build a complete CS2 match prediction system that maintains map-specific Elo ratings for 10 teams across 7 maps. Implement the map veto simulation (each team bans their weakest map, picks their strongest remaining map, with a decider from the remaining pool). Process a simulated season of 100 best-of-three matches and evaluate prediction accuracy using log-loss. Compare the map-aware system to a simple overall-Elo-only approach.

Exercise 17. Implement the full golf tournament Monte Carlo simulator. Create a field of 30 golfers with realistic strokes gained profiles, define two different course profiles, and simulate 50,000 tournaments at each course. Report each golfer's win probability, top-5 probability, and make-cut probability. Identify golfers whose relative ranking changes most between the two courses (the biggest "course fit" effects).

Exercise 18. Build a player prop projection system for NBA points. Create a PlayerPropModel class that takes baseline stats, game environment (pace, total, spread), and matchup factors (DvP rating) as inputs and outputs a projected mean and standard deviation. Then build a screening function that takes a slate of 10 games with 20 player props each and identifies the top 5 highest-EV bets. Include proper handling of blowout risk and back-to-back games.

Exercise 19. Implement a futures market analyzer that: (a) extracts implied probabilities from a market of 30+ outcomes using both multiplicative and power normalization; (b) compares to model probabilities and identifies value bets with edge > 2%; (c) calculates Kelly fractions for each value bet; (d) simulates the season outcome 10,000 times and reports the distribution of betting profits under full Kelly and quarter-Kelly staking.

Exercise 20. Build a versatile thin-market rating system for professional darts. The system should: (a) use adaptive K-factors based on rating deviation; (b) track head-to-head records and apply adjustments when sufficient history exists; (c) increase rating deviation over time for inactive players; (d) produce predictions with confidence levels based on combined rating deviations; (e) compare predictions to simulated market odds and flag value bets. Test with a simulated season of 200 matches among 40 players.

Part D: Analysis (Exercises 21-25)

Exercise 21. Analyze the impact of roster changes on CS2 team performance. Using the CS2MatchPredictor class, simulate a team replacing one player (reduce overall Elo by a configurable amount, increase the roster change penalty). Track how many maps it takes for the team's predictions to recover to pre-change accuracy. Test with replacement penalties of 5%, 10%, and 20%, and discuss the implications for betting on teams immediately after roster changes.

Exercise 22. Conduct a course fit sensitivity analysis for golf. Using the GolfTournamentSimulator, run tournaments at five different course profiles that progressively shift from "driving course" (high OTT weight) to "short game course" (high ARG weight). Track how the win probability for a long-hitting, weak-short-game golfer changes across courses. Quantify the maximum probability differential and discuss when course fit analysis provides the most betting value.

Exercise 23. Evaluate the correlation structure in NBA player props. Create a hypothetical correlation matrix between points, rebounds, assists, three-pointers made, and minutes for a star player. Using this matrix, calculate the joint probability of a two-leg SGP (Over points + Over rebounds) under both the independent assumption and the correlated assumption. What is the percentage error from ignoring correlation?

Exercise 24. Build a futures timing analysis. Simulate an NBA season where your model's championship probability for a specific team evolves from 8% (pre-season) to 15% (mid-season) to 5% (after a key injury). Track how the sportsbook's odds evolve (with a lag) and identify the optimal entry and exit (hedge) points. Calculate the total return from a timing-aware strategy versus a single pre-season bet.

Exercise 25. Compare the efficiency of three different betting markets for the same sport (e.g., darts): moneyline, handicap, and over/under totals (e.g., total 180s). Using the ThinMarketModel, generate predictions for 100 matches and simulate market prices for each bet type with different levels of noise. Determine which market type offers the most consistent edge and discuss why sportsbooks might price certain derivative markets less accurately.

Part E: Research (Exercises 26-30)

Exercise 26. Research the current state of esports data availability. For CS2, League of Legends, and Dota 2, identify: (a) the primary free data sources with their coverage and granularity; (b) available paid APIs and their pricing; (c) the lag between live events and data availability; (d) the specific metrics that are most predictive of match outcomes according to published research. Propose an optimal data pipeline for a bettor specializing in one title.

Exercise 27. Investigate the "favorite-longshot bias" in golf outright winner markets. Research whether sportsbooks systematically overvalue or undervalue longshot golfers (100/1 or longer). Propose a methodology for testing this using historical odds and tournament results. Discuss how the bias, if it exists, could be exploited using the strokes gained framework presented in this chapter.

Exercise 28. Research the growth and regulation of prop betting markets across major US sportsbooks. Identify which states restrict player props, what the typical hold percentages are compared to sides and totals, and whether any academic studies have evaluated the efficiency of prop markets. Discuss the implications of increasing regulatory scrutiny on prop betting for quantitative bettors.

Exercise 29. Study the phenomenon of "steam moves" in niche sports markets. Research how sharp money flows into thin markets (darts, table tennis, handball) and the typical speed at which lines adjust. Propose a strategy for detecting and following steam moves in these markets, and discuss the practical challenges (betting limits, account restrictions) that limit exploitation.

Exercise 30. Design a complete multi-market betting operation that spans three of the emerging markets covered in this chapter. Your design should specify: (a) which three markets you would target and why; (b) the data sources and model architecture for each; (c) the bankroll allocation across markets; (d) the risk management framework for handling correlation between markets; (e) the operational infrastructure needed (data feeds, betting accounts, automation). Estimate the expected annual ROI and the capital required to generate a specified income level.