The following page contains my 2025-26 season predictions for each NBA team, and my new LAKER Estimated RAPTOR player ratings.1 For more info about how it works, see below. Also check out all the updating forecasts here.

Navigate to…

• 2025-26 NBA forecast

• Today’s game predictions

• 2025-26 NBA LAKER leaderboard

• 2025-26 NBA Player Percentiles

• Historical NBA data spreadsheet

  ---

🏀 2025-26 NBA forecast 📈

🏀 Upcoming Games 👀

🏀 2025-26 NBA LAKER leaderboard 📈

🏀 2025-26 NBA Elo ratings and form report 📈

🏀 2025-26 NBA player percentiles 📈

Check out what every player is good (or bad) at in my latest percentile grades:

🏀 2025-26 NBA Percentiles 📊

How the NBA Elo forecast model works

This year’s NBA forecast introduces a new twist on traditional Elo — one that better reflects the different dynamics of the regular season and the playoffs.

Like before, the model maintains ratings for every team based on how they’ve performed and who they’ve played, adjusting for the quality of the opposition and home-court advantage. These ratings are converted into win probabilities for each game, and the rest of the season is simulated thousands of times to project standings and championship odds.

But instead of a rating based on rolling efficiency ratings, I now track two versions of Elo for every team — one tuned for the regular season and another for the playoffs.

The regular-season Elo evolves primarily based on regular-season games. It’s designed to reflect teams’ day-to-day form — the things that fluctuate across an 82-game grind, with all the scheduling, injuries and random swings that come with it. These ratings react a bit more quickly to new information, giving extra weight to recent games while still accounting for margin of victory and opponent strength.

The playoff Elo, meanwhile, updates more heavily from postseason games and moves more slowly during the regular season. This track captures a longer-term view of team quality — the kind that tends to matter in May and June, when rotations tighten and effort levels spike. It places more emphasis on playoff performance and continuity, rewarding teams that have proven themselves under postseason conditions.

The two systems also interact. Regular-season results slightly influence playoff ratings (and vice versa), so that each track informs the other without hurting their different purposes. Over time, this helps the model distinguish between teams that are built to thrive in the regular season versus those that are better suited for playoff basketball.

Each offseason, ratings are partially regressed toward the league average before the new season begins — reflecting that rosters change and teams rarely pick up exactly where they left off. I also blend in converted Elos via preseason odds (over/unders for regular-season Elo; title odds for the playoffs) Once the season starts, the dual Elo system quickly recalibrates based on new results.

As in past versions, the game-by-game win probabilities are fed into a Monte Carlo simulation that plays out the rest of the schedule thousands of times to estimate playoff odds, Finals probabilities and title chances for each team.

Limitations of the model

No model can perfectly capture how a team’s true strength evolves over a long season. This system focuses only on the scoreboard — who won, by how much, and where the game was played — and doesn’t yet account for lineup changes, trades or injuries in real time.

To help bridge that gap, my default Composite Mode blends the dual-track Elo projections with implied betting-market probabilities from FanDuel, which tend to reflect real-world adjustments for player availability and other contextual factors.

If you prefer to see how the numbers behave in isolation, the Stats-Only Mode removes those betting inputs, leaving a purely results-based forecast grounded in the dual-track Elo system.

Historical NBA Data Spreadsheet

This season, paid subscribers will have access to a constantly updating spreadsheet of LAKER and a variety of other advanced player metrics, plus historical team metrics, going back to the 1976 NBA-ABA merger — and team Elo rating season summaries for all of NBA history.

Subscribe now

For an example of what is contained in this spreadsheet, see last year’s LAKER Estimated RAPTOR page. The link for the Elo archive folder and this year’s sheet is below the paywall:

Read more

The following table is a list of MLB teams with their Elo ratings and their predicted wins throughout the 2026 season. Find the full list of Elo games here. See last year’s version here. Also, check out the other forecast models here.

Navigate to…

• 2026 MLB Elo ratings and win projections

• Today’s game predictions

• 2025 MLB Wins Above Replacement leaderboard

• Historical WAR data spreadsheet

⚾ 2026 MLB Elo ratings and win projections📈

Note: These ratings start the 2026 season as a regressed version of last year’s final ratings.

⚾ Today's schedule 👀

📈 2026 MLB Wins Above Replacement leaderboard ⚾

📊 Historical WAR Data Spreadsheet

Similar to in the NBA and NHL, this season paid subscribers will have access to a constantly updating spreadsheet of JEFFBAGWELL Wins Above Replacement and a variety of other advanced player metrics, plus historical team metrics and seasonal team Elo rating summaries for all of MLB history.

Subscribe now

The link for 2026’s sheet is below the paywall:

Read more

This page contains a list of WNBA teams with their Elo ratings for the 2026 season. (Note: This is basic Elo, so no adjustments for trades, injuries, etc. for the moment.) I also list the win probabilities generated by those ratings, plus a WNBA player ranking that ports over my LAKER ratings from the NBA — here is a full explainer of how those work.

For last year’s landing page, click here.

Table of Contents

• 2026 WNBA Elo ratings and forecast

• Win probabilities and point spreads for today’s games

• 2026 WNBA LAKER player ratings ✨Coming Soon!✨

• Historical Elo and Advanced Metrics Spreadsheet

🏀 2026 WNBA Elo ratings and forecast 📈

🏀 Today's games 👀

🏀 2026 WNBA LAKER player ratings 📈

✨Coming Soon!✨

📊 Historical Elo and Advanced Metrics Spreadsheet

Similar to in the NBA, NHL, MLB and the NFL, this season paid subscribers will have access to a constantly updating spreadsheet of LAKER and a variety of other advanced player metrics, plus historical team metrics and seasonal team Elo rating summaries for all of WNBA history.

Subscribe now

The link for the folder is below the paywall:

Read more

Throughout the 2026 NASCAR Cup Series season, I’ll be updating the following table to keep tabs on the fastest drivers in the sport, week in and week out. For more information about how it works, scroll down:

Navigate to…

• 🏁 2026 NASCAR Cup Series driver ratings 🚗

• How the rolling Driver Ratings work

  • How the playoff odds work

• How Adjusted Points+ Index works

• How the Track Scouting system works

• Historical NASCAR Data Spreadsheet

  ---

If you just want the TL:DR? Every driver is assigned points based on how they finish, and I track their average points per race (relative to average) with weight also given to previous seasons — and also changing weights by track similarity — to predict performance at this week’s upcoming track.

Subscribe now

Just like in 2025, the overall rankings are based on a rolling average of NASCAR’s official Driver Rating metric, with modifications for each track type.

🏁 2026 NASCAR Cup Series driver ratings 🚗

How it works:

⛽ Rolling Driver Ratings

These ratings work in a fashion similar to the rolling NFL quarterback ratings I helped develop for FiveThirtyEight. Each driver is assigned a base rating and a track type-specific modified rating before each race, which are blended together to predict his or her Driver Rating in that race. After the race, the base and track type ratings are adjusted by weighting their actual driver rating from the race into their previous rating.

Specifically, each driver’s rating and modifier are updated via:

Rating_new = update_rate*race_Driver_Rating + (1 - update_rate)*Rating_old

The update rates are as follows:

To make predictions before each race, we blend base ratings with track type-modified ratings like so:

Predicted_Driver_Rating = blend_rate*track_modifier + (1 - blend_rate)*base_rating

Here are the specific blend rates:

Finally, there are special rules for dealing with new drivers and transitions between seasons.

When a driver makes his or her debut (whether overall or on a particular track type), they are assigned a pre-race rating of 45.0 for their base and/or track modifier. To begin a new season, a driver’s base rating and modifiers for ovals and short tracks are simply their last rating from the previous season in each category, but we regress his/her restrictor plate modifier toward the mean of 70.0 by 25 percent, and their road-course modifier toward the mean of 70.0 by 5 percent.

⛽ In-Season Challenge and playoff advancement odds

To estimate each driver’s odds of advancing through the In-Season Challenge and/or NASCAR playoffs — and ultimately winning the title — I used a Monte Carlo simulation model powered by the aforementioned rolling track-type-adjusted driver ratings. For each round of the bracket, the model looks at the race’s track type and converts every driver’s projected rating into an expected finish percentile using a logistic regression trained on historical NASCAR data from Racing-Reference.info (dating back to 2005). This regression captures how well pre-race ratings typically correlate with actual finishing position, and the resulting percentile reflects the likelihood of a strong performance in that specific race’s context.

From there, I added a randomizing element to each driver’s expected performance based on the standard deviation of the residuals in the logistic regression — effectively simulating the week-to-week chaos and uncertainty that makes NASCAR so unpredictable. This random “noise” was drawn from a normal distribution, calibrated to the model’s historical error, and applied on the logit scale before being transformed back into a percentile. In a given race, the driver with the higher randomized percentile “wins” the matchup and advances in the bracket of the In-Season Challenge — or, for the playoffs, the entire field is rank-ordered by their randomized percentiles to simulate finishing positions.

This process was repeated 2,000 times to simulate a full playoff structure, generating win probabilities and round-by-round advancement odds for each driver. Because each race uses a different track type rating — including ovals, road courses, short tracks and “restrictor plate”/drafting-style superspeedways — the model properly accounts for driver strengths and weaknesses over the upcoming schedule.

⛽ Adjusted Points+ Index

One mainstay of this system is my Adjusted Points index (Pts+) metric. Adjusted Points gives credit to drivers for finishing in each position, based on the following curve:

The average of those points per race is then scaled relative to the Cup Series norm, so that an average driver is always 100.

⛽ Track Scouting Ratings

The driver rankings then look at this across multiple seasons, in the manner of the classic “Marcel” projection system. (I used a version of this for my 2024 NASCAR season preview as well.) The following weights are applied to a driver’s total races and Pts+ from each season:

• 2026 season weight: 6.0

• 2025 season weight: 2.4

• 2024 season weight: 1.2

• 2023 season weight: 0.4

The result is a rating that uses multiple seasons’ worth of data — giving more weight to 2026 — while regressing small-sample drivers toward a low Pts+ index.

For this week’s track projections, I also use a 6-3-1 weighting scheme for each driver’s performance at each track, where races at that specific track receive a weight of 6, primary similar tracks get a weight of 3 and secondary similar tracks get a weight of 1. Those are combined with the recency weighting from the overall ratings to get ratings at each track. The “Diff.” category represents the gap between weighted projected performance at that track and the driver’s overall rating. (This will tell you who does better at this track than usual.)

📊 Historical NASCAR Data Spreadsheet

Similar to in the NBA, MLB and NHL, this season paid subscribers will have access to a constantly updating spreadsheet of NASCAR (primarily Cup Series) data, including yearly registers of Adjusted Points+ index, head-to-head W-L records between drivers and their teammates/similar grid cars/car manufacturers/etc. since 1972, and average Driver Rating (split by track type, etc.) and rolling driver ratings by race since 2005.

Subscribe now

The link for this year’s sheet is below the paywall:

Read more

Throughout the 2025 NASCAR Cup Series season, I’ll be updating the following table to keep tabs on the fastest drivers in the sport, week in and week out. For more information about how it works, scroll down:

Navigate to…

• 🏁 2025 NASCAR Cup Series driver ratings 🚗

• 🏁 2025 NASCAR Cup Series Playoff Odds (✨New!✨)

  • Next Race’s Leverage Index 😰 (✨New!✨)

• 🏁 2025 NASCAR In-Season Challenge odds

• How the rolling Driver Ratings work

  • How the playoff odds work

• How Adjusted Points+ Index works

• How the Track Scouting system works

• Historical NASCAR Data Spreadsheet

  ---

If you just want the TL:DR? Every driver is assigned points based on how they finish, and I track their average points per race (relative to average) with weight also given to previous seasons — and also changing weights by track similarity — to predict performance at this week’s upcoming track.

Subscribe now

As a new twist for 2025, the overall rankings are based on a rolling average of NASCAR’s official Driver Rating metric, with modifications for each track type.

🏁 2025 NASCAR Cup Series driver ratings 🚗

Update (Wednesday, May 7, 2025, 11:00 AM): The predicted track type ratings in the table have been adjusted to blend with the overall base rating in a way that would best predict the next race at that track type, rather than simply listing the track-type adjustments in isolation.

🏁 2025 NASCAR Cup Series playoff odds 🏁

🏁 2025 NASCAR In-Season Challenge odds

How it works:

⛽ Rolling Driver Ratings

These ratings work in a fashion similar to the rolling NFL quarterback ratings I helped develop for FiveThirtyEight. Each driver is assigned a base rating and a track type-specific modified rating before each race, which are blended together to predict his or her Driver Rating in that race. After the race, the base and track type ratings are adjusted by weighting their actual driver rating from the race into their previous rating.

Specifically, each driver’s rating and modifier are updated via:

Rating_new = update_rate*race_Driver_Rating + (1 - update_rate)*Rating_old

The update rates are as follows:

To make predictions before each race, we blend base ratings with track type-modified ratings like so:

Predicted_Driver_Rating = blend_rate*track_modifier + (1 - blend_rate)*base_rating

Here are the specific blend rates:

Finally, there are special rules for dealing with new drivers and transitions between seasons.

When a driver makes his or her debut (whether overall or on a particular track type), they are assigned a pre-race rating of 45.0 for their base and/or track modifier. To begin a new season, a driver’s base rating and modifiers for ovals and short tracks are simply their last rating from the previous season in each category, but we regress his/her restrictor plate modifier toward the mean of 70.0 by 25 percent, and their road-course modifier toward the mean of 70.0 by 5 percent.

⛽ In-Season Challenge and playoff advancement odds

To estimate each driver’s odds of advancing through the In-Season Challenge and/or NASCAR playoffs — and ultimately winning the title — I used a Monte Carlo simulation model powered by the aforementioned rolling track-type-adjusted driver ratings. For each round of the bracket, the model looks at the race’s track type and converts every driver’s projected rating into an expected finish percentile using a logistic regression trained on historical NASCAR data from Racing-Reference.info (dating back to 2005). This regression captures how well pre-race ratings typically correlate with actual finishing position, and the resulting percentile reflects the likelihood of a strong performance in that specific race’s context.

From there, I added a randomizing element to each driver’s expected performance based on the standard deviation of the residuals in the logistic regression — effectively simulating the week-to-week chaos and uncertainty that makes NASCAR so unpredictable. This random “noise” was drawn from a normal distribution, calibrated to the model’s historical error, and applied on the logit scale before being transformed back into a percentile. In a given race, the driver with the higher randomized percentile “wins” the matchup and advances in the bracket of the In-Season Challenge — or, for the playoffs, the entire field is rank-ordered by their randomized percentiles to simulate finishing positions.

This process was repeated 2,000 times to simulate a full playoff structure, generating win probabilities and round-by-round advancement odds for each driver. Because each race uses a different track type rating — including ovals, road courses, short tracks and “restrictor plate”/drafting-style superspeedways — the model properly accounts for driver strengths and weaknesses over the upcoming schedule.

⛽ Adjusted Points+ Index

One mainstay of this system is my Adjusted Points index (Pts+) metric. Adjusted Points gives credit to drivers for finishing in each position, based on the following curve:

The average of those points per race is then scaled relative to the Cup Series norm, so that an average driver is always 100.

⛽ Track Scouting Ratings

The driver rankings then look at this across multiple seasons, in the manner of the classic “Marcel” projection system. (I used a version of this for my 2024 NASCAR season preview as well.) The following weights are applied to a driver’s total races and Pts+ from each season:

• 2025 season weight: 6.0

• 2024 season weight: 2.4

• 2023 season weight: 1.2

• 2022 season weight: 0.4

The result is a rating that uses multiple seasons’ worth of data — giving more weight to 205 — while regressing small-sample drivers toward a low Pts+ index.

For this week’s track projections, I also use a 6-3-1 weighting scheme for each driver’s performance at each track, where races at that specific track receive a weight of 6, primary similar tracks get a weight of 3 and secondary similar tracks get a weight of 1. Those are combined with the recency weighting from the overall ratings to get ratings at each track. The “Diff.” category represents the gap between weighted projected performance at that track and the driver’s overall rating. (This will tell you who does better at this track than usual.)

📊 Historical NASCAR Data Spreadsheet

Similar to in the NBA, MLB and NHL, this season paid subscribers will have access to a constantly updating spreadsheet of NASCAR (primarily Cup Series) data, including yearly registers of Adjusted Points+ index, head-to-head W-L records between drivers and their teammates/similar grid cars/car manufacturers/etc. since 1972, and average Driver Rating (split by track type, etc.) and rolling driver ratings by race since 2005.

Subscribe now

The link for this year’s sheet is below the paywall:

Read more

The following table is a list of MLB teams with their Elo ratings and their predicted wins throughout the 2025 season. Find the full list of Elo games here. See last year’s version here. Also, check out the other forecast models here.

Navigate to…

• 2025 MLB Elo ratings and win projections

• Today’s game predictions

• 2025 MLB Wins Above Replacement leaderboard

• Historical WAR data spreadsheet (✨New!✨)

⚾ 2025 MLB Elo ratings and win projections📈

Note: These ratings start the 2025 season as a regressed version of last year’s final ratings.

⚾ Today's schedule 👀

📈 2025 MLB Wins Above Replacement leaderboard ⚾

📊 Historical WAR Data Spreadsheet

Similar to in the NBA and NHL, this season paid subscribers will have access to a constantly updating spreadsheet of JEFFBAGWELL Wins Above Replacement and a variety of other advanced player metrics, plus historical team metrics and seasonal team Elo rating summaries for all of MLB history.

Subscribe now

The link for this year’s sheet is below the paywall:

Read more

This page contains a list of WNBA teams with their Elo ratings for the 2025 season. (Note: This is basic Elo, so no adjustments for trades, injuries, etc. for the moment.)

I also list the win probabilities generated by those ratings, plus a couple of sets of WNBA player rankings: One which ports over my Estimated RAPTOR ratings from the NBA — here is a full explainer of how those work — and the other which blends RAPTOR wins with Basketball-Reference’s Win Shares and Estimated Wins Added from Player Efficiency Rating (PER). Each metric is set to the same scale as Win Shares and then averaged together for a more holistic view of advanced player value.

For last year’s landing page, click here.

Table of Contents

• 2025 WNBA Elo ratings and forecast

• Win probabilities and point spreads for today’s games

• 2025 WNBA Estimated RAPTOR player ratings

• 2025 WNBA consensus player rankings

• Historical Elo and Advanced Metrics Spreadsheet

🏀 2025 WNBA Elo ratings and forecast 📈

🏀 Today's games 👀

🏀 2025 WNBA Estimated RAPTOR player ratings 📈

🏀 2025 WNBA Consensus Wins player rankings 📈

📊 Historical Elo and Advanced Metrics Spreadsheet

Similar to in the NBA, NHL, MLB and the NFL, this season paid subscribers will have access to a constantly updating spreadsheet of Estimated RAPTOR and a variety of other advanced player metrics, plus historical team metrics and seasonal team Elo rating summaries for all of WNBA history.

Subscribe now

The link for the folder is below the paywall:

Read more

The following page contains my 2024-25 season predictions for each NBA team, and Estimated RAPTOR player ratings. For more info about how it works, see below. Also check out the MLB, NFL, NHL and WNBA versions.

Navigate to…

• 2024-25 NBA forecast

• Today’s game predictions

• 2024-25 NBA Estimated RAPTOR leaderboard

• 2024-25 NBA Playoff Odds Chart

• 2024-25 NBA Elo Rankings

• Historical NBA data spreadsheet

  ---

🏀 2024-25 NBA forecast 📈

🏀 Upcoming Games 👀

🏀 2024-25 NBA Estimated RAPTOR leaderboard 🦖

🏀 2024-25 NBA Playoff Odds Chart 📈

🏀 2024-25 NBA Elo ratings and form report 📈

How the NBA forecast model works

Taking inspiration from Baseball-Reference's MLB forecast, this NBA model maintains running schedule-adjusted ratings for each team based on how they've performed on offense and defense (per 100 possessions), the locations of each game and the quality of their opponents. The ratings are then turned into win probabilities for each game on the schedule, and the rest of the season is simulated 5,000 times to track who is most likely to win the NBA championship.

One different wrinkle for the NBA (compared with MLB) is that we maintain two separate ratings for each team, which have different predictive purposes.

First, we keep a regular season rating, which is based on a team's most recent 50 games — including the end of the previous season, in the case of early season ratings — and is regressed to the mean by 15 games. This is used to create win probabilities for regular season games, and it implies that a team's recent form matters a lot during the regular season (not surprising, but still interesting to see quantified).

For each team, we also keep a playoff rating, which is used to simulate playoff games. This is based on the team's most recent 110 games, and it contains no regression-to-the-mean factor, though postseason games receive triple weight in the ratings. This means that the ratings take a much longer view of a team's performance when looking ahead to the playoffs, which also makes sense in the context of research showing that a team's previous playoff resume matters a lot relative to the regular season.

Another factor that gets baked into the simulations is uncertainty. The win probabilities of any two teams in either a regular season or playoff game vary depending on how far into the future it is, with a game tomorrow being considered more certain than a game eight months from now.

Limitations of the model

The challenge for any model is figuring out the balance between simplicity and complexity, which can mean deciding how much relevant information to include.

This forecast style is simpler than the one previously maintained by FiveThirtyEight, which was based on player ratings and accounted for trades, injuries and other player movement. Our basic model won't be able to adjust immediately to those differences. However, it will evolve over time as more of the rating sample includes the mix of players a team will use going forward.

To help mitigate some of this, the default for our model is what we call Composite Mode, which blends the pure ratings-based forecast simulations with implied betting-market probabilities from FanDuel. In theory, this should steer our forecast in the direction of more qualitative factors like injuries and trades that a statistical power rating-based system cannot readily account for.

However, if you want to see the pure stats-based version, you can toggle the interactive above to Stats-only Mode, which removes the betting-market feature.

Further reading

To learn more about how models like this work, you can read about Baseball-Reference's model here, and this is an explainer of the underlying math behind the rating method both systems use. For a history of the ratings I helped create at FiveThirtyEight, read this explainer. And for more on the logic and science behind power ratings, I would recommend the books "Who's #1?: The Science of Rating and Ranking", by Amy N. Langville and Carl D. Meyer, and "Mathletics", by Wayne L. Winston, Scott Nestler and Konstantinos Pelechrinis.

Historical NBA Data Spreadsheet

This season, paid subscribers will have access to a constantly updating spreadsheet of Estimated RAPTOR and a variety of other advanced player metrics, plus historical team metrics, going back to the 1976 NBA-ABA merger — and team Elo rating season summaries for all of NBA history.

Subscribe now

For an example of what is contained in this spreadsheet, see last year’s Estimated RAPTOR page. The link for this year’s sheet is below the paywall:

Read more

In the spirit of BracketMatrix, the following page contains a composite of probabilities to advance through — and before that, make the Round of 68 in — the 2024-25 NCAA Division I men's basketball tournament. The current component team ratings (for the bracket odds) and/or forecast models (for the mid-season forecast) are from: ESPN Analytics (BPI), Ken Pomeroy, Bart Torvik, TeamRankings, Sports-Reference/CBB, PlayoffStatus.com1 and a ✨ new addition ✨, the Dance Chance from Sports-Ratings.com.

🏀 2024-25 NCAA Tournament Odds and Probabilities 📊

How this works: Each source was compiled to create a meta-rating of point differential per 100 possessions relative to a D-I average team. Those ratings were used to create win probabilities for each NCAA tournament game based on a historical logistic regression, and the bracket was simulated 2,500 times to track the most likely winners.

🏀 2024-25 NCAA Tournament Mid-Season Meta-Forecast 📊

Filed under: College Basketball

The following is a current version of the composite preseason Major League Baseball forecasts that I introduced here last season:

Filed under: Baseball, Statgeekery

This page contains the college quarterback Points Above Replacement (PAR) metric I use for my “QB U” columns. Read about how it works below, or download the historical data here.

🏈 2024 College Football QB rankings 📈

👀 How it works

This metric builds off of the QB value stat that my colleague Maya Sweedler and I used here to rank BYU against other prolific passing programs from CFB history. I introduced it in this story about recent QB factories, and expanded on it more in this story about the most productive returning QBs of the 2024 season.

Here are the steps to calculating PAR:

• Calculate raw Points Above Average (PAA) via the following formula:
  
  PAA = Coef*(Pass Att + Rush Att)+0.408314*Completions+0.028618*Pass Yds+0.791174*Pass TDs-1.265725*INTs+0.069731*Rush Yds+1.387276*Rush TDs
  
  (These coefficients are based on a regression between standard box score QB stats and ESPN’s QBR-based value metrics.)

• As part of this, find the value for “Coef” (the multiplier on QB passing and rushing attempts) that makes the sum of all PAA across FBS QBs in a given season equal zero. (The coefficient was -0.4253 in the 2023 season.)

• Convert from raw PAA to raw Points Above Replacement (PAR) by adding 0.11334025*(Pass Att+Rush Att)

• Determine each QB’s share of team plays — i.e., his % of the team’s (Pass Att+Rush Att) by QBs.

• Calculate the schedule adjustment.

  • For complete seasons, this is done by grabbing the team’s “SOS” value from Sports-Reference/CFB’s Simple Rating System (SRS) scores, and extracting the passing-specific portion of that by multiplying by 25% (the the share of overall team Expected Points Added that is attributable to passing offense in CFB since 2005).1 Then, that number is multiplied by the QB’s share of team plays and the team’s total number of games to date, to arrive at the total SOS adjustment.

  • During the season, using actual observed SRS values early will inflate the SOS adjustments, so we taper the use of those against SOS estimates based on a prior for the opponent’s SRS, using a 3-year weighted average of their SRS from previous seasons. The blend follows this process:

• The SOS adjustment value (in points) is then added to raw PAA and PAR to obtain the schedule-adjusted versions. PAR/13 represents a QB’s prorated value per 13 team games, similar to WAR per 162 team games in my baseball JEFFBAGWELL leaderboard.

Filed under: College Football, QB U

In the spirit of BracketMatrix, the following page contains a composite of probabilities to advance through — and before that, make the Round of 68 in — the 2024-25 NCAA Division I men's basketball tournament. The current component team ratings (for the bracket odds) and/or forecast models (for the mid-season forecast) are from: ESPN Analytics (BPI), Ken Pomeroy, Bart Torvik, TeamRankings, Sports-Reference/CBB, PlayoffStatus.com1 and a ✨ new addition ✨, the Dance Chance from Sports-Ratings.com.

🏀 2024-25 NCAA Tournament Odds and Probabilities 📊

How this works: Each source was compiled to create a meta-rating of point differential per 100 possessions relative to a D-I average team. Those ratings were used to create win probabilities for each NCAA tournament game based on a historical logistic regression, and the bracket was simulated 2,500 times to track the most likely winners.

🏀 2024-25 NCAA Tournament Mid-Season Meta-Forecast 📊

Filed under: College Basketball

Throughout the 2024 NASCAR Cup Series season, I’ll be updating the following table to keep tabs on the fastest drivers in the sport, week in and week out. For more information about how it works, scroll down. If you just want the TL:DR? Every driver is assigned points based on how they finish, and I track their average points per race (relative to average) with weight also given to previous seasons. I also use a version of this (also weighted by track similarity) to predict performance at this week’s upcoming track.

🏁 2024 NASCAR Cup Series driver ratings 🚗

How it works

The basis of this system is my Adjusted Points index (Pts+) metric. Adjusted Points gives credit to drivers for finishing in each position, based on the following curve:

The average of those points per race is then scaled relative to the Cup Series norm, so that an average driver is always 100.

The driver rankings then look at this across multiple seasons, in the manner of the classic “Marcel” projection system. (I used a version of this for my 2024 NASCAR season preview as well.) The following weights are applied to a driver’s total races and Pts+ from each season:

• 2024 season weight: 6.0

• 2023 season weight: 2.4

• 2022 season weight: 1.2

• 2021 season weight: 0.4

The ratings are “padded” with about 23 races of a 53.6 Pts+ index as well, so that the following formula is used:

Rating = (6.0*2024 Races*2024 Pts Index + 2.4*2023 Races*2023 Pts Index +1.2*2022 Races*2022 Pts Index +0.4*2021 Races*2021 Pts Index + 23*53.6) / (6.0*2024 Races + 2.4*2023 Races +1.2*2022 Races +0.4*2021 Races + 23)

The result is a rating that uses multiple seasons’ worth of data — giving more weight to 2024 — while regressing small-sample drivers toward a low Pts+ index.

For this week’s track projections, I also use a 6-3-1 weighting scheme for each driver’s performance at each track, where races at that specific track receive a weight of 6, primary similar tracks get a weight of 3 and secondary similar tracks get a weight of 1. Those are combined with the recency weighting from the overall ratings to get ratings at each track. The “Diff.” category represents the gap between weighted projected performance at that track and the driver’s overall rating. (This will tell you who does better at this track than usual.)

Filed under: NASCAR

The following table is a list of MLB teams with their Elo ratings and their predicted wins throughout the 2024 season. Find the full list of Elo games here. Also, check out the NFL version here and the WNBA version here.

Navigate to…

• 2024 MLB Playoff Predictor (✨NEW!)

• Today’s game predictions

• 2024 MLB Elo ratings and win projections

• 2024 MLB Wins Above Replacement leaderboard

• WAR Injury Tracker

  ---

⚾ 2024 MLB Playoff Predictor 📈

⚾ Today's schedule 👀

⚾ 2024 MLB Elo ratings and win projections📈

📈 2024 MLB Wins Above Replacement leaderboard ⚾

🏥 Injury Tracker 🚑

(How this works)

Filed under: Baseball, Elo ratings

This page contains a list of WNBA teams with their Elo ratings for the 2024 season. (Note: This is basic Elo, so no adjustments for trades, injuries, etc. for the moment.) I also list the win probabilities generated by those ratings, plus a couple of sets of WNBA player rankings: One which ports over my Estimated RAPTOR ratings from the NBA — here is a full explainer of how those work — and the other which blends RAPTOR wins with Basketball-Reference’s Win Shares and Estimated Wins Added from Player Efficiency Rating (PER). Each metric is set to the same scale as Win Shares and then averaged together for a more holistic view of advanced player value.

Table of Contents

• 2024 WNBA Elo ratings and forecast

• Win probabilities and point spreads for today’s games

• 2024 WNBA Estimated RAPTOR player ratings (✨NEW!)

• 2024 WNBA player rankings

🏀 2024 WNBA Elo ratings and forecast 📈

🏀 Today's games 👀

🏀 2024 WNBA Estimated RAPTOR player ratings 📈

🏀 2024 WNBA Consensus Wins player rankings 📈

Filed under: WNBA, Elo ratings

One of baseball’s longest-running sources of worry has concerned its lack of household-name superstars relative to other sports like the NBA and NFL.

For years, MLB’s most famous player was Derek Jeter, and the sport faced a crisis when he retired: Its best player at the time, Mike Trout, was a budding GOAT — but God bless him, he was sort of boring. Meanwhile, Bryce Harper was too polarizing. Clayton Kershaw was too easy to take for granted. Other potential successors were either too old, not popular enough, or both. What baseball wouldn’t have given to have its own version of LeBron James back in the mid-2010s.

Fast-forward a decade, however, and the sport has not one, but two — each in their own way.

Subscribe now

Shohei Ohtani of the Dodgers keeps finding new ways to produce historic seasons. After winning MVPs with several of the best two-way campaigns by any player in forever, Ohtani was limited by injury to “only” being a hitter this year… and he became just the sixth member of the 40 HR/40 SB club when he hit a walk-off grand slam to beat the Rays last Friday.

At the same time, Aaron Judge of the Yankees keeps getting better, eclipsing 50 home runs for the second time in three seasons, setting himself up to potentially break his own American League HR record in September. And, oh by the way, he is also in the mix to win the AL Triple Crown if he can chase down Bobby Witt on batting average.

In many ways, this is the fulfillment of MLB’s dream. Judge and Ohtani are the game’s biggest stars, both having huge, historic seasons in the league’s two largest markets. Add in Juan Soto, also having a great season in the Bronx, and 2024 has been a banner year for the alignment of big names and big numbers in big cities:

In fact, it’s hard to think of a way that baseball might ever stumble across a better confluence of all these factors at once. In both Ohtani and Judge, MLB has basically maxed out its potential formula for star power + market size + elite performance. But has that mattered?

Maybe slightly. Let’s use Google Trends again to judge whether interest in Ohtani and Judge has increased relative to four mainstays of stardom in the NBA and NFL: LeBron, Patrick Mahomes, Steph Curry and Kevin Durant. We’ll look from late August one year to the next, to capture the past year and each of the previous three, looking to see how the search traffic Ohtani and Judge received compares with the big names from other sports:

Aided by his high-profile free agency decision — and subsequent massive contract — Ohtani’s online attention has risen relative to LeBron/Curry/Mahomes/Durant, though he still isn’t at their level yet. And Judge, who got a big bump for his record-setting end of the 2022 season (he broke the AL HR mark in October 2022), hasn’t gotten the same spike in interest yet this year. That may still be yet to come, if he cracks 60 dingers again and/or breaks his own record; for now, he still sits far below the non-MLB stars’ level of search traffic.

And as for each star’s respective sport, the evidence around a star-fueled rise for MLB is mixed. Conducting the same Google Trends research for teams in August-to-August windows instead of players, we see that each league has remained relatively flat over the past few years:

Google searches aren’t everything. July’s MLB All-Star Game defied predictions of declining ratings, remaining the gold standard for All-Star events across the major sports. The league has also seen rising TV ratings and attendance overall, with Ohtani’s Dodgers and Judge’s Yankees leading the way in driving national TV eyeballs.

And the postseason, of course, will be the real test of how much the sport’s big-name stars move the needle. Across the various different forecast models and the Vegas odds, there is about an 8 percent chance we get that classic Dodgers-Yankees matchup in the Fall Classic, which would make for a fitting end to MLB’s star-studded season — and provide a ratings bonanza. (As much as those still exist outside of the NFL.)

But we need look back no further than to last October for the record-low viewership of 2023’s Rangers-Diamondbacks World Series at the opposite end of the spectrum. And in the random world of the MLB playoffs, the odds are currently a coin-flip as to whether we’ll see both Judge and Ohtani flame out before the championship round, no matter what kind of rare fame and acclaim they’ve brought to their sport so far this season.

Filed under: Baseball

Do you miss the old FiveThirtyEight RAPTOR ratings? I sure do. But while those are no longer updating via Disney (I don’t know their long-term future fate), I have taken it upon myself to create the next best thing — a more basic version of RAPTOR that uses pace-adjusted box-score and on-vs-off data to estimate what the more full-fledged model (with player tracking/etc) would have said about a player.

You can find that stat — and many others — for every player not only in 2023-24, but also going back to the ABA merger (since 1976-77), here:

NBA Player and Team Stats

And here are the 2023-24 season leaders in Wins Above Replacement:

🏀2023-24 NBA Estimated RAPTOR leaderboard 🦖

For those interested, I recreated the 1997-2013 (i.e., pre-player tracking) version of RAPTOR via regression, added a mild position adjustment and a team adjustment such that each team’s efficiency gap vs. league average on offense or defense is equal to 4.5 times the minute-weighted average of its players’ ratings on either side of the ball (to account for score effects). The result is a fairly basic “statistical plus-minus” model that had a correlation of 0.89 with “full” RAPTOR from 2014-2023.

(To read more about how it works, click here.)

I’ll probably be using these throughout the season in my NBA writing, so feel free to hit me with comments/suggestions/etc. about the ratings.

Filed under: NBA

NOTE: This page is out of date. For the 2024-25 version, CLICK HERE.

The following page contains a list of NHL teams with their Elo ratings and projected wins for the 2023-24 season. It also features Goals Above Replacement data, which is described (and available for download) here. And the meta-forecast at the bottom pulls projections from the following sources: FanDuel, Hockey-Reference, Numberfire, MoneyPuck and Matt Harvey. Find the Elo data here. Also, check out the NBA, MLB and NFL versions.

Navigate to…

• 2023-24 NHL team Elo ratings and win projections

• Today’s game predictions

• 2023-24 NHL Meta-Forecast

• 2023-24 NHL Goals Above Replacement leaderboard

  ---

🏒 2023-24 NHL Elo ratings and win projections📉

🏒 Today's schedule 👀

🏒 2023-24 NHL Meta-Forecast 📊

❄️ 2023-24 NHL Goals Above Replacement leaderboard 🏒

Filed under: NHL, Elo ratings, Hockey

The following table is a list of NBA teams with their Elo ratings for the 2023-24 season. (Note: This is basic Elo, so no adjustments for trades, injuries, etc. for the moment.) Find the data here. Also, check out my more detailed NBA Forecast model here.

Table of Contents

• 2023-24 NBA Elo ratings and form report

• Win probabilities and point spreads for today’s games

🏀 2023-24 NBA Elo ratings and form report 📈

🏀 Today's games 👀

Filed under: NBA, Elo ratings

The following table shows my 2023-24 season predictions for each NBA team. For more info about how it works, see below. For the more basic, old-school Elo ratings, click here.

🏀 2023-24 NBA forecast 📈

How the NBA model works

Taking inspiration from Baseball-Reference's MLB forecast, this NBA model maintains running schedule-adjusted ratings for each team based on how they've performed on offense and defense (per 100 possessions), the locations of each game and the quality of their opponents. The ratings are then turned into win probabilities for each game on the schedule, and the rest of the season is simulated 5,000 times to track who is most likely to win the NBA championship.

One different wrinkle for the NBA (compared with MLB) is that we maintain two separate ratings for each team, which have different predictive purposes.

First, we keep a regular season rating, which is based on a team's most recent 50 games — including the end of the previous season, in the case of early season ratings — and is regressed to the mean by 15 games. This is used to create win probabilities for regular season games, and it implies that a team's recent form matters a lot during the regular season (not surprising, but still interesting to see quantified).

For each team, we also keep a playoff rating, which is used to simulate playoff games. This is based on the team's most recent 110 games, and it contains no regression-to-the-mean factor, though postseason games receive triple weight in the ratings. This means that the ratings take a much longer view of a team's performance when looking ahead to the playoffs, which also makes sense in the context of research showing that a team's previous playoff resume matters a lot relative to the regular season.

Another factor that gets baked into the simulations is uncertainty. The win probabilities of any two teams in either a regular season or playoff game vary depending on how far into the future it is, with a game tomorrow being considered more certain than a game eight months from now.

Limitations of the model

The challenge for any model is figuring out the balance between simplicity and complexity, which can mean deciding how much relevant information to include.

This forecast style is simpler than the one previously maintained by FiveThirtyEight, which was based on player ratings and accounted for trades, injuries and other player movement. Our basic model won't be able to adjust immediately to those differences. However, it will evolve over time as more of the rating sample includes the mix of players a team will use going forward.

To help mitigate some of this, the default for our model is what we call Composite Mode, which blends the pure ratings-based forecast simulations with implied betting-market probabilities from FanDuel. In theory, this should steer our forecast in the direction of more qualitative factors like injuries and trades that a statistical power rating-based system cannot readily account for.

However, if you want to see the pure stats-based version, you can toggle the interactive above to Stats-only Mode, which removes the betting-market feature.

Further reading

To learn more about how models like this work, you can read about Baseball-Reference's model here, and this is an explainer of the underlying math behind the rating method both systems use. For a history of the ratings I helped create at FiveThirtyEight, read this explainer. And for more on the logic and science behind power ratings, I would recommend the books "Who's #1?: The Science of Rating and Ranking", by Amy N. Langville and Carl D. Meyer, and "Mathletics", by Wayne L. Winston, Scott Nestler and Konstantinos Pelechrinis.

Filed under: NBA