This project will build on the tournament data used in Project 1 to evaluate how each chess player performed relative to what their pre-tournament rating would predict. Rather than comparing players only by their final tournament points, I will calculate an expected score for each player based on the rating differences between that player and each opponent. I will then compare expected and actual scores to identify the five players who most overperformed and the five who most underperformed expectations.
I will begin with the cleaned tournament information from Project 1, including each player’s pair number, name, pre-rating, actual tournament score, and opponents. Because expected performance depends on each individual matchup, I will match every opponent pair number back to that opponent’s pre-tournament rating rather than relying only on the player’s average opponent rating.
For each game, I will calculate the rating difference between the player and the opponent and convert that difference into an expected scoring probability using a documented ELO method. I plan to use a consistent expected-score calculation for every matchup and cite the source used for the formula. The expected probabilities from all of a player’s games will then be added together to produce that player’s expected tournament score.
Next, I will compare each player’s actual tournament score with the calculated expected score. I will define the performance difference as actual score minus expected score. A positive difference will indicate that a player scored more points than expected, while a negative difference will indicate that the player scored fewer points than expected. I will sort these differences to identify the five largest positive values and the five largest negative values.
The final analysis will present the overperformers and underperformers in tables and explain what the difference between actual and expected score represents. I will also validate several intermediate calculations manually so that the final rankings are not based on an opponent-matching or rating-calculation error.
Before interpreting the final results, I will check that each player’s expected tournament score is within a reasonable range based on the number of games played. I will also manually verify selected player-opponent matchups by comparing the player’s rating, opponent rating, rating difference, and expected probability. Finally, I will confirm that the calculated performance difference is consistently defined as actual score minus expected score before selecting the top five overperformers and underperformers.
One challenge will be correctly connecting every opponent number in the tournament record to the appropriate player’s pre-rating. An incorrect match would affect the expected score for that game and therefore the player’s total expected score. I will check that opponent identifiers exist in the player table and that the number of matched games is consistent with the tournament records.
Another issue is that the tournament file contains round entries that may not represent a normal game against a rated opponent, such as byes or other special round codes. These records will need to be identified so that they are not accidentally treated as opponent pair numbers. I will determine how these cases should contribute to the actual and expected tournament scores and document that decision in the final report.
The rating data may also contain formatting such as provisional-rating indicators. I will need to make sure the numeric pre-rating is extracted consistently before calculating rating differences.
Finally, different ELO implementations can produce slightly different expected probabilities. I will therefore use one reasonably sourced method consistently for all players, clearly document the formula or probability conversion used, and cite the source. This will make the calculations reproducible and make it clear how the expected scores were produced.