Case Study: The Decay and Renewal of a Betting Edge --- Adapting Through Three Market Regimes

Executive Summary

This case study examines the five-year betting career of an experienced sports bettor ("Morgan") who maintained profitability through three distinct market regimes by systematically detecting edge decay and adapting their approach. Beginning in 2020 with a model-based edge in NBA totals, Morgan achieved a +4.2% ROI in the first two years before the edge eroded as sportsbooks improved their totals modeling. Rather than continuing to bet a declining edge, Morgan detected the decay through rolling CLV analysis, paused betting, and rebuilt around a new edge source: in-game rest and pace modeling that the market was slow to incorporate after the NBA's schedule compression following the COVID-shortened season. When that edge also decayed by 2024, Morgan pivoted again to cross-sport portfolio optimization. The case demonstrates that sustainable profitability in sports betting requires not just building an edge, but recognizing when it is dying and having the discipline and skill to find a replacement.

Background

The Bettor's Profile

Morgan is a former quantitative finance professional who transitioned to full-time sports betting in 2020. Morgan's background in options pricing --- specifically in modeling the decay of edge in rapidly adapting markets --- provided an unusual advantage: the ability to think systematically about when an edge is temporary and how to detect its expiration.

Morgan's betting philosophy: "Every edge has an expiration date. The question is not whether it will expire, but when, and whether you will notice in time."

The Starting Point: 2020

Morgan began with a $50,000 bankroll and a focus on NBA totals. The thesis was straightforward: the NBA's restart after the COVID shutdown created unusual schedule dynamics (back-to-backs, compressed rest, bubble environment) that affected scoring patterns in predictable ways that the market was slow to price.

The Analysis

Regime 1: The Pace-Rest Model (2020-2022)

The Edge Source

Morgan built a pace-rest model that predicted game totals by combining:

1. Team pace ratings --- adjusted possessions per 48 minutes, weighted by recency
2. Rest differentials --- the difference in days of rest between teams, with a nonlinear response (2+ days of rest advantage had diminishing returns)
3. Altitude/travel adjustment --- teams playing at altitude (Denver) or after cross-country travel scored differently
4. Back-to-back fatigue --- teams on the second night of a back-to-back scored approximately 2.3 fewer total points than their baseline

Morgan used these features in a gradient-boosted regression model that output predicted game totals. The model's predictions were then compared to sportsbook totals to identify value.

Value Identification Process

For each NBA game, Morgan:

1. Generated a model-predicted total with a confidence interval
2. Compared the model total to the sharpest available line (Pinnacle closing)
3. Compared the model total to soft book lines at six US-regulated sportsbooks
4. Flagged games where the soft book total deviated from the model total by more than 1.5 points

Results: 2020-2021 Season

The edge was concentrated in specific situations:

The back-to-back and altitude edges were strong because these situations create predictable scoring pattern deviations that some sportsbooks were slow to incorporate into their totals lines.

Results: 2021-2022 Season

The edge had already begun to shrink. The CLV declined from +3.1% to +1.8%, and the ROI from +5.2% to +2.1%. Morgan's rolling CLV analysis told the story:

Root Cause Analysis

Morgan investigated the decay and identified two causes:

1. Sportsbook model improvement. Two of the six sportsbooks Morgan used had clearly improved their totals lines. The average discrepancy between Morgan's model and these books' lines narrowed from 2.1 points in early 2021 to 0.8 points by spring 2022.

2. Back-to-back scheduling changes. The NBA reduced the frequency of back-to-back games starting in 2021-22, shrinking Morgan's highest-edge betting pool (98 bets in 2020-21 to 62 in 2021-22).

The Decision Point: March 2022

With rolling CLV below +1.0% and trending downward, Morgan faced a critical decision. The options were:

1. Continue betting the current model at reduced stakes
2. Pause betting and retrain the model
3. Pivot to a new edge source

Morgan chose option 3, reasoning that the fundamental edge (back-to-back fatigue mispricing) was closing not because of model error but because the market had adapted. Retraining the same model on better data would not restore an edge that the market itself had eliminated.

Morgan stopped live betting NBA totals in April 2022 and spent three months analyzing new potential edge sources.

Regime 2: The In-Game Pace Adjustment Model (2022-2024)

The New Edge Source

Morgan's analysis of the 2021-22 season revealed a new opportunity: in-game pace adjustments. When one team fell behind by 10+ points in the first half, their second-half pace increased significantly (playing faster to catch up), which inflated second-half scoring. The pre-game total reflected average pace, not the actual in-game pace adjustment.

This created value in two markets:

1. Live totals. When a game was lopsided at halftime, the live total often underestimated second-half scoring because it assumed pace would normalize. Morgan bet live overs in blowouts.
2. Second-half totals. Some books offered second-half totals that did not fully account for pace adjustments.

Morgan built a pace-adjustment model that predicted second-half scoring given first-half score differential, team pace tendencies, and home/away status.

Results: 2022-2023 Season

The live totals edge was particularly strong because:

• Live markets are inherently less efficient than pre-game markets (less time for sharp bettors to correct mispricings)
• The pace-adjustment phenomenon is well-documented in basketball analytics but was not yet fully incorporated into sportsbook live models
• Morgan had a speed advantage: the pace-adjustment calculation took seconds, while the market took minutes to adjust

Results: 2023-2024 Season

Again, the edge was decaying. Morgan's rolling analysis:

Root Cause Analysis: Second Decay

1. Sportsbook live model improvements. The major sportsbooks had invested heavily in live pricing engines. DraftKings and FanDuel both publicly announced improvements to their live-game algorithms in 2023.
2. Latency arbitrage crackdown. Several books implemented stricter latency requirements for live bets, reducing Morgan's speed advantage.
3. Competitor activity. Morgan suspected that other sophisticated bettors had independently identified the pace-adjustment edge, increasing the speed at which the market corrected mispricings.

The Second Pivot: March 2024

Morgan again faced the three options: continue, retrain, or pivot. This time, Morgan chose a more radical transformation.

Regime 3: Cross-Sport Portfolio Optimization (2024-Present)

The New Strategy

Rather than seeking a single large edge in one market, Morgan diversified across four sports with smaller but more sustainable edges:

The thesis: each individual edge was smaller than Morgan's previous single-market edges, but the portfolio was more diversified and therefore more resilient. If one edge decayed, the others would continue producing value while Morgan developed a replacement.

Portfolio Risk Analysis

Morgan used portfolio theory (Chapter 4 concepts applied to betting):

The diversified portfolio nearly doubled the Sharpe ratio and increased the probability of annual profitability from 68% to 89%, despite the average per-bet edge being only marginally higher.

Results: 2024-2025 (Through February 2025)

The NFL totals edge was the strongest, driven by weather modeling that the market continued to underweight. The NHL edge was the weakest and Morgan was monitoring it for potential removal from the portfolio.

Five-Year Cumulative Analysis

*Through February 2025.

Morgan's five-year profit of $58,070 on an initial $50,000 bankroll represents a 116% total return. Critically, this return was not achieved by finding one edge and riding it forever. It required three pivots, multiple model iterations, and the discipline to stop betting when the edge was dying.

Key Findings

1. Every edge decayed, without exception. Morgan's back-to-back fatigue edge lasted approximately 18 months. The live pace-adjustment edge lasted approximately 18 months. Neither was permanent. The question was always "when" not "if."

2. Rolling CLV detected decay 2-3 months before ROI did. In both Regime 1 and Regime 2, CLV declined steadily over several months before ROI turned negative. A bettor tracking only ROI would have continued betting unprofitably for 2-3 additional months.

3. The market adapts faster each cycle. Regime 1's edge lasted roughly 18 months at meaningful CLV (+2%+). Regime 2 also lasted about 18 months but peaked at lower CLV (+2.4% vs. +3.1%). This acceleration is consistent with sportsbooks investing more in quantitative pricing.

4. Diversification extends the portfolio's useful life. By spreading across four sports, Morgan reduced the damage from any single edge decaying. Even if the NHL edge disappears entirely, the portfolio remains viable.

5. The willingness to stop is the most important skill. Morgan stopped betting NBA totals twice --- once after detecting pace-rest edge decay, and once after detecting live pace-adjustment decay. Each pause lasted 2-3 months and involved zero revenue. Most bettors cannot tolerate this. They continue betting a decaying edge, compounding losses that erode the bankroll needed for the next opportunity.

The Python Analysis

The accompanying code (case-study-code.py) implements the edge decay detection framework used in this case study, including:

1. Rolling CLV tracker with configurable windows and linear trend detection.
2. Edge decay significance test using linear regression on window means with p-value threshold.
3. Cross-sport portfolio optimizer that computes weighted CLV, portfolio variance, and Sharpe ratio.
4. Market regime detection algorithm that identifies structural breaks in CLV time series.
5. Monte Carlo simulation comparing single-market versus diversified portfolio outcomes.

Discussion Questions

1. Morgan detected edge decay through rolling CLV analysis. What if CLV itself becomes a less reliable signal (e.g., if closing lines become more efficient)? Design an alternative early-warning system.

2. The cross-sport portfolio approach trades edge magnitude for diversification. Is there a principled way to determine the optimal number of sports/markets to include? What are the diminishing returns of adding a fifth or sixth sport?

3. Morgan paused betting for 2-3 months during each transition. If Morgan had a team of analysts, how would the transition strategy change? Could the pause be eliminated?

4. The NHL edge (+0.7% CLV) is the weakest in the portfolio. At what CLV level should Morgan remove it? Consider the opportunity cost of analyst time versus the diversification benefit.

5. Morgan's career shows declining peak CLV across regimes (+3.1%, +2.4%, +1.1%). If this trend continues, at what point does sports betting cease to be profitable after accounting for the time investment? How should Morgan plan for this scenario?