Worked case study · 6 minute read
A 1–4 rivalry record can hide a points advantage
Lanterns lose four of five meetings with Ravens. Add the scores and Lanterns lead by 25 points. Both statements are correct—and neither is enough on its own to judge the matchup.
Four narrow losses, one large win
Consider five fictional meetings between Lanterns and Ravens. Lanterns lose the first four by one, two, three and four points. They win the fifth by 35. Treat these as five completed games across an illustrative history, with the same scoring rules and no ties.
| Meeting | Lanterns points | Ravens points | Lanterns result | Margin |
|---|---|---|---|---|
| 1 | 121 | 122 | Loss | −1 |
| 2 | 116 | 118 | Loss | −2 |
| 3 | 132 | 135 | Loss | −3 |
| 4 | 124 | 128 | Loss | −4 |
| 5 | 155 | 120 | Win | +35 |
A rivalry table sorted by low win rate may put Ravens near the top of Lanterns’ difficult-opponent list. That is a valid description of the results. The average margin adds a different perspective: one large victory outweighs four small defeats when the scores are summed.
Neither value replaces the other. The league awards a win for each game, not for the aggregate five-game score. Lanterns’ points advantage does not turn any of those four losses into wins.
Win rate describes frequency; margin describes size
Aggregate margin = −1 −2 −3 −4 +35 = +25
Average margin = +25 ÷ 5 = +5.0 points per meeting
The low win rate comes from losing more games. The positive average margin comes from the size of the one win. If we reduce the five margins to a single number, we lose the pattern that explains their disagreement.
A simple sensitivity check makes that visible. Across only the first four games, Lanterns have no wins and trail by ten total points, averaging −2.5 per game. Adding the fifth changes the average margin to +5.0 while leaving the win rate at just 20%. One observation has moved the scoring story substantially.
This is why a useful rivalry view should retain the meeting count and game-level scores. “Ravens are Lanterns’ nemesis” is memorable league banter. “Ravens won four of five recorded meetings, with four narrow wins and one large loss” describes the evidence more accurately.
Two new meetings can change a small sample quickly
Now extend the example with two invented results: Lanterns win 127–122 and then 140–130. Their record becomes 3–4. Their total scoring margin becomes +40, but the most striking change is in the win rate.
| Included meetings | Lanterns record | Win rate | Aggregate margin | Mean margin |
|---|---|---|---|---|
| First five | 1–4 | 20.0% | +25 | +5.0 |
| All seven | 3–4 | 42.9% | +40 | +5.7 |
Two games move the observed win rate by 22.9 percentage points. That is arithmetic, not proof that Lanterns suddenly became a better manager. In the five-game sample, each result represents one-fifth of the data. Adding games changes both the numerator and the denominator.
A longer history gives more observations, but it can introduce new complications: different managers, roster rebuilds, scoring changes and playoff formats. More games do not automatically create a clean comparison of the same underlying contest.
Make sure the history is the history you meant to include
Before interpreting a live Sleeper rivalry, check which seasons were found, whether the same people were identified across those seasons and whether the chosen report includes regular season, playoffs or both. A replacement manager inheriting a roster is not automatically the same person for an all-time rivalry.
Our league-history guide explains how linked seasons and manager identities are used. Sleeper’s league response includes a previous-league reference, which can help connect seasons; missing links or unavailable older records can leave the report incomplete.
Scoring context also matters. A 30-point margin under one scoring format is not directly equivalent to 30 points under another. If a league switches to superflex or changes passing-touchdown values, pooled raw margins can obscure that change. Review the scoring-formats guide before comparing distant seasons.
For an incomplete history, use wording such as “recorded meetings in the available seasons.” That keeps a technically correct subtotal from being mistaken for an exhaustive lifetime record.
A nemesis record is not evidence of a special predictive effect
These results do not tell us why Ravens won four games. Close results could reflect player outcomes, lineup choices, different team strength or ordinary variation. The score table alone cannot separate those explanations.
It also cannot establish that the identity of the opponent changes what Lanterns will score next week. A head-to-head history records past outcomes; a forecast would need current roster information, league settings and an explicit predictive method.
The synthetic numbers demonstrate that two summaries can tell different stories. They do not establish the frequency of this pattern in actual leagues or the reliability of a particular rivalry threshold. No real-league sample was analysed for this page.
Read rivalry results alongside the rest of the report
- Check the number of meetings and seasons included before reading the win percentage.
- Compare wins with average margin, then inspect the individual games that create the gap.
- Separate regular-season and playoff results when the distinction matters to your question.
- Confirm manager identity and scoring continuity across seasons.
- Use weekly all-play and schedule analysis to put each meeting in its league-wide context.
The public example league demonstrates the last step using seven weeks of scores and a schedule swap. It does not include this fictional multi-meeting history. For example, Lanterns’ week-one loss to Ravens in that separate demo is also a second-place weekly score—context a rivalry win-loss line would omit.
Data and sources
The five meetings and two later results on this page are separate synthetic teaching data, printed in full above. Totals, percentages and means are derived from those scores; no real manager’s history was retrieved.
The Sleeper API documentation for weekly matchups describes team points, roster IDs and shared matchup IDs. That is the structure used for live league analysis; this worked example uses no API data. Was It Luck? is independent of Sleeper.
Try the analysis before entering a username
The example report includes standings, weekly scores and a schedule swap. Open it directly, then use the same questions to investigate your own Sleeper league.