Written By: Natalie Hansen


After a close game, it’s easy to remember big plays like walk off home runs, but it’s much harder to remember how every play contributes to a big win or a tough loss. But, what if there were a way to look at a game objectively and visualize how each play impacted the game’s outcome?

The 643 win probability model leverages our historical data to review key events in games. In a cohesive visual, users can see how each event impacts the win probability of a given game. Featured as a tool in our web application, users can select any previously played NCAA baseball or softball game dating back to 2018 and view an objective timeline of game events based on 643’s play-by-play data. Each play impacts the win probability of a given game for each team, and this is how we visualize that:

The Visualization

Win probability visual for the Evansville @ Illinois State baseball game played May 25, 2021

This visual shows how the plays over the entire game contributed to the win probability. The data points represent each individual play that occurred in the game and the blue line represents the change in win probability at these different points. Data points above this dotted line indicate that the home team (Illinois State in this case) has a higher win probability, while points under the line indicate that the away team (Evansville in this case) has a higher win probability. The top and bottom of each inning are marked along the bottom of the visual, so users can see how the win probability fluctuates between the two teams over the course of the game. In this example, we can see how the model favored Illinois State at the beginning of the game (more on this later), but at various points, events caused the model to favor Evansville. We can see how the probability swung between the two teams until Illinois State’s win in the bottom of the 12th inning of this back-and-forth game.


What does the win probability model tell us?

Our win probability model is best used to describe and reflect on a game after it occurs.

By comparing events in a given game to thousands of other college games, the model is able to objectively describe how different plays impacted a game. Since the model does not use machine learning or other predictive methods, it allows any NCAA baseball or softball team to reflect on the impact of plays based on how past occurrences impacted who won and lost the game. The model tells us what did happen in a given game situation, not what should have happened based on a predictive estimate.

This allows teams to analyze games and pinpoint how plays contributed to a win or loss through our win probability model. Rather than combing through hours of game video to pick apart the key plays in the game, coaches can use this tool to visualize entire games with a few clicks. They can hover over different portions of the game to see what play occurred at a given time and the resulting win probability calculated from the play. Viewing win probability for multiple games can help coaches and teams analyze key plays or trends for their season and objectively look at what contributed to their wins and losses.

What makes up the win probability model?

Our win probability model has two main considerations:

  1. Win probability at the beginning of a game
  2. Win probability throughout a game

How is win probability estimated before a game starts?

In order to create a visual showing win probability over a whole game, the model must have criteria that tells it who the win probability should favor before the game starts. Our model uses a conference RPI adjustment to determine this point. Home teams win more often than visitors, so one should expect home teams to be given a slight edge regardless of who’s playing before the RPI adjustment is applied. The home-field advantage is about nine percent on average.

What is conference RPI?

RPI is a measurement used to rank teams based off of their strength of schedule, and it is a weighted measure of a teams winning percentage, their opponents’ winning percentage, and their opponents’ opponents’ winning percentage.

RPI = (WP * 0.25) + (OWP * 0.50) + (OOWP * 0.25)

Our model uses the RPI for all the teams in a conference to yield a conference RPI, which allows us to measure strength of schedule for each NCAA softball and baseball conference. The conference RPI tells us how ‘strong’ each conference is, which allows us to adjust the starting win probability accordingly.

For example, if a home team is from the Summit League (Oral Roberts Baseball) and the away team is from the Big-12 (Oklahoma State Baseball), while the model would otherwise favor the home team, conference RPI is factored in, giving the away team Oklahoma State the higher starting win probability.


How is win probability estimated throughout a game?

After the initial win probability is estimated at the beginning of the game, our model then begins accounting for each play that happens. In order to determine whether a given play increases or decreases win probability, the following information is gathered:

  • Inning (top/bottom)
  • Base state
  • Out state
  • Score differential

This information creates a set of situational context that can then be compared to other situations with the same context. By leveraging our database of NCAA play-by-play data, we compare the given situation to all the similar situations to describe how frequently an event leads to an increase or decrease in win probability. We also apply our conference RPI factor at a decreasing rate throughout the plays in the game. This means that although a team from a high RPI conference may be favored at the beginning of a game’s win probability model, as plays occur in the game, the initial favoring impacts win probability less and less.

How to interact with the win probability tool

To use the win probability tool within our web application, users should navigate to the win probability tab, select baseball or softball, and then enter the team name and season. From there, users are able to see a schedule of all the games for that team’s season organized by date, allowing them to select the game that they want to view.

Once a game is selected, the win probability visual appears, and users can hover their mouse over the visual to see the win probability, score, baserunners, outs, and play-by-play narrative associated with any play in the game. The win probability visual shown here is for the March 11, 2021 softball game between Ohio State and Wisconsin.

Users can see how over the course of the game, the win probability model fluctuates between favoring Wisconsin and Ohio State. The different plays that occurred in the game causes these ‘swings’ in win probability. One way that the win probability model can be used is to determine an excitement index. The excitement index sums the changes in win probability across a given game to create an index for how exciting a game is. Close games with a lot of fluctuation in win probability result in a high excitement index, while games that favor the same team throughout result in a low excitement index.

How can the excitement index be used?

The excitement index can be used to look back on games with the most changes in win probability. We pulled data on the excitement index to determine the top 10 most exciting games of 2021 for every NCAA division of softball and baseball.

D1 Softball

Top Game:

Wisconsin vs. Ohio State (March 11, 2021)  •  Excitement Index: 1117.3

Tied 5-5 going into the 15th inning, Ashley Prange’s sacrifice bunt plated the go ahead run and Avery Clark singled through the left side to put Ohio State ahead 7-5. The Buckeyes held the Badgers in the bottom of the 15th to seal the game.

D1 Softball: Top 10 Most Exciting Games of 2021

Home Team Visitor Team Date  Excitement Index
Wisconsin Ohio State 3/11/2021 1117.3
Iowa Penn State 3/11/2021 929.2
FGCU Connecticut 2/21/2021 886.7
McNeese Arkansas 2/20/2021 872.9
Kent State Toledo 3/21/2021 825.9
Louisiana Tech Montana 3/12/2021 810.3
Purdue Iowa 3/27/2021 784.8
Seton Hall Villanova 5/2/2021 784
Idaho State New Mexico State 3/19/2021 783.2
George Mason Saint Louis 4/9/2021 773.9

D2 Softball

Top Game:

North Greenville vs. Tusculum (February 24, 2021)  •  Excitement Index: 916.6

Going into the bottom of the 12th with a 5-5 tie, North Greenville’s Josie Reed executed the suicide squeeze bunt to win the game 6-5.

Source: Conference Carolinas Digital Network  https://conferencecarolinasdn.com/?B=240120

D2 Softball: Top 10 Most Exciting Games of 2021

Home Team Visitor Team Date  Excitement Index
North Greenville Tusculum 2/24/2021 916.6
Western Oregon Northwest Nazarene 4/2/2021 865.3
Northeastern State Drury 3/6/2021 844.5
Central Oklahoma Cameron 2/4/2021 797.8
Saginaw Valley State Northwood 4/24/2021 791.1
Molloy College Bridgeport 4/3/2021 788.7
Ursuline Ferris State 2/19/2021 759.9
Black Hills State Colorado Mesa 4/9/2021 729.3
West Virginia Wesleyan Notre Dame College (Ohio) 4/3/2021 720.2
Christian Brothers UAH 3/6/2021 707

D3 Softball

Top Game:

Lebanon Valley vs. Widener (April 24, 2021)  •  Excitement Index: 1143.4

With an 8-8 tie in the bottom of the 14th, Kylie Balthaser reached first on an error by the shortstop, allowing the runner to score from third and ending the game 9-8.

Source: Lebanon Valley College https://godutchmen.com/watch/?Archive=1116&sport=12&type=Archive

D3 Softball: Top 10 Most Exciting Games of 2021

Home Team Visitor Team Date  Excitement Index
Lebanon Valley Widener 4/24/2021 1143.4
Grove City Bethany (WV) 5/1/2021 836.5
Wisconsin-Eau Claire Wisconsin-Platteville 5/11/2021 767.9
Old Westbury Mount Saint Mary (New York) 4/21/2021 763.5
Rutgers-Newark Montclair State 4/6/2021 738.4
Ohio Wesleyan Denison 4/17/2021 735.5
Greensboro North Carolina Wesleyan 4/14/2021 725.7
Suffolk Endicott 4/7/2021 710.3
Hanover Manchester 5/2/2021 706.7
St. Mary’s (MN) Augsburg 4/24/2021 705.2

D1 Baseball

Top Game:

Bryant vs. Fairleigh Dickinson (April 11, 2021)  •  Excitement Index: 1021.6

Tied in the bottom of the 12th, Sam Sinisgalli doubled down the left field line for the walk off RBI.

D1 Baseball: Top 10 Most Exciting Games of 2021

Home Team Visitor Team Date  Excitement Index
Bryant Fairleigh Dickinson 4/11/2021 1021.6
UC Santa Barbara San Francisco 3/13/2021 1006.7
Sacred Heart Wagner 5/20/2021 973.6
Iona Canisius 3/27/2021 939.3
Georgia Tech Georgia 5/18/2021 936.9
South Carolina Florida 3/26/2021 927.8
Ball State Western Michigan 3/21/2021 927.5
Presbyterian UNC Asheville 4/6/2021 921.6
Georgia Tech Louisville 5/27/2021 913.9
Illinois State Evansville 5/25/2021 900.5

D2 Baseball

Top Game:

Saginaw Valley State vs. Northwood (April 30, 2021)  •  Excitement Index: 999.2

Tied in the 12th inning, Saginaw Valley State’s Todd Paperd drove in the winning run to win 11-10 over Northwood.

D2 Baseball: Top 10 Most Exciting Games of 2021

Home Team Visitor Team Date  Excitement Index
Saginaw Valley State Northwood 4/30/2021 999.2
Lee Shorter 4/9/2021 882.7
USC Aiken Young Harris 4/18/2021 856.4
Columbus State USC Aiken 3/13/2021 847.6
Maryville Lindenwood 4/30/2021 797.5
Henderson State Augustana 5/29/2021 795.8
Delta State Valdosta State 4/3/2021 793.7
Oklahoma Christian St. Mary’s (Texas) 2/5/2021 774.3
Emporia State Rogers State 3/26/2021 764.9
Francis Marion Young Harris 2/27/2021 757.4

D3 Baseball

Top Game:

Salisbury vs. Southern Virginia (May 7, 2021) • Excitement Index: 1342.8

Tied going into the 23rd inning, Cole Campanile singled to left field bringing the score in Southern Virginia’s favor 7-6. The Knights held Salisbury in the bottom of the inning to clinch their victory for the longest game in NCAA DIII baseball history.

D3 Baseball: Top 10 Most Exciting Games of 2021

Home Team Visitor Team Date  Excitement Index
Salisbury Southern Virginia 5/7/2021 1342.8
Baldwin Wallace John Carroll 4/10/2021 934
Elmira St. John Fisher 4/28/2021 932.4
McMurry Howard Payne 4/1/2021 905.7
Knox Beloit 3/27/2021 903
Clarkson St. Lawrence 3/21/2021 902.4
Ohio Northern Heidelberg 3/13/2021 897.4
Ozarks (Arkansas) Howard Payne 3/6/2021 860.9
Rockford Elmhurst 3/13/2021 856.6
Coe Dubuque 4/24/2021 848.7

Looking at every NCAA division for baseball and softball shows us how the win probability model can be used to help determine excitement index. All the top games for 2021 featured a tie that went into extra innings.

Uses of the Win Probability Model

By looking back at the most exciting games from every division, it’s clear how the win probability tool can be useful to teams in every division of NCAA softball and baseball. It offers an objective way to describe and reflect on past games. Our win probability tool is an effective way to recap games using a single visualization connected to your team’s play-by-play data from past games. The win probability visualization can be utilized through the 643 web application.