Methodology

Bases and outs are the foundation

Every play ultimately changes, preserves, or ends a combination of two things: the position of runners on the bases and the number of outs in the inning. CBM therefore treats bases and outs as separate dimensions of the game state rather than collapsing them into a single value.

A single play can affect both dimensions at once. A sacrifice fly can create advancement that becomes part of a run while also causing an out. A caught stealing can destroy existing advancement while also consuming an out. Neither effect cancels the other; both happened.

CBM records those changes separately and assigns responsibility for each one. This allows the system to describe the actual mechanics of a play without first deciding whether the play should be classified as broadly positive or negative.

Contextual impact, causal responsibility

CBM is deliberately contextual when measuring impact, but causal when assigning responsibility.

Baseball is not played in a vacuum. The same action can have a different effect depending on the situation. A single with nobody on base may create one base of advancement. The same single with runners aboard may create several. A pitcher entering with the bases empty faces a different situation than one entering with runners already in scoring position.

CBM does not try to remove that context from the measurement. If a player creates more advancement because the situation gave them more opportunity to do so, that additional advancement still happened and still affected the game. The goal is to measure the player’s actual impact in the situation they were given.

At the same time, CBM does not assign responsibility simply because a player happened to be present when something occurred. It asks what that player directly caused.

A hitter does not receive credit for a runner already being on second base before the plate appearance, but can receive credit for advancing that runner from second to home. A relief pitcher is not responsible for the bases an inherited runner had already gained, but can be responsible for allowing that runner to advance further. Responsibility follows the specific changes in the game state, not simply the final result.

This distinction allows CBM to preserve context without confusing opportunity with causation.

Because opportunity can affect total production, CBM also preserves useful splits. On offense, Individual Bases measure advancement a player creates for themselves, while Situational Bases measure advancement the player creates for runners who were already on base.

The core principle is:

Responsibility follows advancement through time

CBM does not treat each play as an isolated event or permanently attach all value to the event label that first created it. Once advancement exists, the system follows responsibility for that advancement as the game continues.

A change in runner identity does not automatically create new advancement. For example, suppose a runner reaches first and the next batter hits into a fielder’s choice that forces that runner out while replacing them at first. The offense still has one occupied base, while an out has been added. The replacement did not create a new base of advancement simply by taking the previous runner’s place.

If that new runner later steals second, however, the steal creates a new base of advancement and that new advancement belongs to the runner who caused it.

This persistence is important because CBM is trying to account for the state of the game, not merely count isolated event labels. Responsibility can carry forward even when the player physically occupying a base changes.

The actual game becomes the next opportunity

CBM does not rewrite the game based on counterfactual outcomes. It follows the sequence that actually occurred.

Suppose a fielder drops a foul ball that should have ended an at-bat. The defender can be responsible for an Out Missed. But the game continues. If the pitcher then allows that batter to reach base, the pitcher is still responsible for the advancement they allowed. CBM does not erase that later responsibility because an earlier out could have been recorded.

This differs from systems that reconstruct an inning around errors to determine what should have happened. In CBM, once the game moves into a new state, that actual state becomes the situation the next player inherits.

The fielder is responsible for the missed out. The pitcher is responsible for what they subsequently allow. Each player is evaluated against the real opportunity they actually faced.

CBM measures what happened

CBM is designed to measure realized impact on the game. It does not estimate what a play should have been worth, what usually happens in a similar situation, or what a player’s underlying skills suggest should happen next.

If a player creates a base of advancement, causes an out, allows a runner to advance, or prevents an expected out, CBM records the change that actually occurred and assigns responsibility for it.

That distinguishes CBM from statistics built around expected outcomes, modeled run values, win values, or underlying skills such as exit velocity, launch angle, pitch movement, or spin rate. Those measurements can be useful for understanding why a player performs the way they do or predicting what may happen in the future. CBM is answering a different question: What impact did this player actually have on the games that were played?

The main exception is where probability data helps determine responsibility, particularly on defense. For example, catch probability can help distinguish an exceptional defensive play from a missed expected out. Even there, the probability is used to interpret responsibility for the actual outcome, not to replace the outcome with an expected one.

Expected statistics can help explain or predict outcomes. CBM accounts for the outcomes themselves.

Bases are a common unit of advancement

CBM treats each base of advancement as one Base because the statistic measures movement through the game state, not modeled run value.

Moving a runner from first to second and moving a runner from third to home are each one base of advancement. The latter immediately completes a run, but CBM does not need to assign it a larger theoretical weight inside the Bases statistic. Instead, the system separately records whether credited advancement ultimately became part of a run.

A run is itself the culmination of four bases of advancement: home to first, first to second, second to third, and third to home. Those four bases can be caused by different players. CBM preserves responsibility for each part rather than assigning the entire run to a single event.

Bases Scored (BS) measure the subset of a player’s credited Bases that ultimately became part of runs. Bases Unscored (BU) measure the remaining credited Bases that did not.

This allows Bases to remain a consistent unit of advancement while still preserving the distinction between advancement that ultimately contributed to scoring and advancement that did not.

Why Bases are a net stat

Bases measure net impact, not just gross production. Creating advancement matters, but so does destroying advancement that you previously created.

For example, suppose a player hits a double and is later caught trying to advance. Traditional batting statistics still record the double: the hit happened, Total Bases increased by two, and the player’s batting average improved. But from the perspective of the game state, the advancement that player created is no longer present, and the player has also caused an out.

CBM therefore records the sequence as zero net Bases and one Out Caused. The double created two Bases, but the player’s later action destroyed those two Bases before they could contribute further.

This is why CBM does not permanently lock in credit at the moment a Base is created. It follows the impact of that advancement through subsequent baserunning so that the final statistics reflect what the player actually added to—or removed from—the team’s position in the game.

Gross production is still visible through Bases Created (BC) and Bases Destroyed (BD), while the core Bases (B) statistic is net:

This distinction matters over any sample size. A player who repeatedly creates advancement and then gives some of it back through baserunning mistakes should not look identical to a player who creates the same gross advancement and preserves it.

CBM stops at direct causal impact

CBM assigns responsibility for direct changes in bases and outs. It does not attempt to trace every downstream consequence that an action might eventually influence.

A 12-pitch plate appearance may tire a pitcher. It may contribute to an earlier pitching change. That change may affect which reliever appears later. But if the reliever then allows a home run, there is no reliable way to determine how much of that home run should be attributed to the hitter who worked the earlier count.

CBM stops before that butterfly effect. If the 12-pitch plate appearance itself produces a walk, hit, advancement, or out, those direct effects are recorded. The system does not continue assigning responsibility through speculative chains of events.

This creates a deliberate causal boundary: CBM measures direct game-state impact, not every strategic, psychological, or downstream consequence of a player’s actions.

Attribution rules, evidence, and judgment

CBM uses explicit attribution rules and evidence hierarchies to determine which player caused each change in bases or outs.

In practice, many apparently complicated plays become much clearer when they are broken into individual changes in the game state: who moved, from where to where, which outs occurred, and what action caused each change. More complex sequences require more specific rules and, where available, additional evidence.

The current historical implementation reconstructs plays from historical play-by-play, including Retrosheet’s compact event notation. That record is detailed enough to resolve a large share of attribution, but it cannot contain every piece of information that would be visible to someone watching each play live. Where the historical evidence is incomplete, CBM uses consistent fallback rules rather than leaving responsibility undefined.

Modern tracking and probability data can add precision where it is available. That evidence is used to refine attribution—for example, by helping distinguish an exceptional defensive play from a missed expected out—while the observed game outcome remains the thing being measured.

Official scoring can also provide useful evidence, but CBM is not conceptually limited to official event labels such as errors. The underlying question is whether a player caused a base to be gained or lost, or an out to be created or missed. With sufficiently detailed evidence, defensive responsibility can therefore differ from the traditional scoring label attached to a play.

Some judgment will remain, especially on difficult defensive plays. The goal is not to claim that every case is perfectly objective. It is to make the rules explicit, use the best evidence available, apply those rules consistently, and reduce the number of cases that require subjective judgment.

Long term, live scoring could make attribution even stronger by recording responsibility from the play itself rather than reconstructing every decision afterward. The methodology can also continue to improve as better historical and tracking data become available.

Historical coverage and consistency

The underlying CBM framework can apply anywhere the game contains bases and outs, but the precision of attribution depends on the evidence available for that era.

Offensive, pitching, baserunning, and many defensive changes can be reconstructed from historical play-by-play without modern tracking data. More recent tracking data can provide additional evidence about defensive difficulty and responsibility that older records simply do not contain.

The current historical implementation therefore depends heavily on the availability and detail of Retrosheet data, with modern data used to refine attribution where possible. Earlier eras can still be measured through the same conceptual framework, but some categories cannot be resolved with the same precision as modern seasons.

The methodology stays the same: measure the actual changes in bases and outs and assign each change to the player who caused it. What changes across eras is the amount of evidence available to make that attribution.

What CBM adds to traditional statistics

Traditional baseball statistics capture important parts of player performance. Hits, walks, stolen bases, RBI, runs, errors, strikeouts, and other statistics each describe meaningful events in the game.

What they do not share is a single system for assigning responsibility for the underlying changes those events produce. Credit can be divided among several separate statistics, while other outcomes may be assigned entirely to one player even when multiple players contributed to them.

CBM adds a direct attribution layer. It breaks plays into their underlying changes in bases and outs and asks who directly caused each one. That allows familiar baseball outcomes to be viewed through a common framework of individual responsibility.

The goal is not to replace traditional statistics. They remain useful descriptions of what happened. CBM adds another perspective: a way to measure how much each player directly changed the state of the game and to preserve responsibility for those changes as the game continues.