backcaster 0.3.4__tar.gz → 0.3.6__tar.gz

{backcaster-0.3.4/src/backcaster.egg-info → backcaster-0.3.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: backcaster
-Version: 0.3.4
+Version: 0.3.6
 Summary: Marcel-style MLB projection CLI using local Lahman CSV data
 License-Expression: MIT
 Project-URL: Homepage, https://github.com/jeffreywevans/backcaster
@@ -36,7 +36,7 @@ Backcaster is a Python CLI that produces **Marcel-style MLB projections** from t
 It supports:
 - **Single player projections** (`player`)
 - **Single team projections** (`team`)
-- **League-wide batch projections** (`batch`)
+- **League-wide projections** (`hitters`, `pitchers`, or `batch`)
 - Output as terminal table, CSV, or HTML
 
 ---
@@ -56,10 +56,10 @@ Upgrade to the latest release available on PyPI:
 python -m pip install --upgrade backcaster
 ```
 
-To install a specific release, such as 0.3.4:
+To install a specific release, such as 0.3.6:
 
 ```powershell
-python -m pip install --upgrade backcaster==0.3.4
+python -m pip install --upgrade backcaster==0.3.6
 ```
 
 Run the command using the same Python environment where you run Backcaster.
@@ -97,19 +97,52 @@ pip install dist/backcaster-*.whl
 
 ## Quickstart CLI usage
 
-All commands require:
-- `--year` target projection year
-- `--kind` one of `batting` or `pitching`
+All commands require `--year`, the target projection year.
+The `player`, `team`, and `batch` commands also require `--kind batting|pitching`.
+The `hitters` and `pitchers` commands choose their kind automatically.
 
 Subcommand-specific required options:
 - `player` requires `--name`
 - `team` requires `--team`
-- `batch` requires no additional required option
+- `batch`, `hitters`, and `pitchers` require no player or team option
 
 ```bash
 backcaster <player|team|batch> [subcommand-options] --year <YEAR> --kind <batting|pitching> [--format cli|csv|html] [--out PATH]
+backcaster <hitters|pitchers> --year <YEAR> [--data-dir PATH] [--format cli|csv|html] [--out PATH]
 ```
 
+### Pre-offseason league projections
+
+Once the latest regular season is present in your local Lahman files, generate
+both populations without roster updates or team filters. For terminal output:
+
+```powershell
+backcaster hitters --year 2026
+backcaster pitchers --year 2026
+```
+
+To save files for comparison with Baseball Reference:
+
+```powershell
+backcaster hitters --year 2026 --format csv --out hitters_2026.csv
+backcaster pitchers --year 2026 --format csv --out pitchers_2026.csv
+```
+
+`--year 2026` projects 2026 from 2023–2025; it never reads 2026 statistics.
+Both commands support `--data-dir`, `--format cli|csv|html`, and `--out`.
+They are shortcuts for `batch --kind batting` and `batch --kind pitching`.
+For a different data directory, add `--data-dir "C:\Baseball\lahman"`.
+For HTML, use `--format html --out hitters_2026.html` (or a pitcher output filename).
+File formats require `--out`; the CLI format prints to the terminal.
+
+Players qualify for selection by meeting the existing PA/IP threshold in at least
+one of those three seasons. This is a historical eligibility list, not an active
+roster; pitchers who never reach the qualifying innings threshold (including most
+relievers) remain excluded. Qualification selects players without removing their
+other seasons from the calculation. Trades are combined into one player-season
+before qualification, and all teams' statistics contribute once. The output is
+one row per selected player, ready for comparison with reference projections.
+
 ### 1) Player projection
 
 ```bash
@@ -123,8 +156,8 @@ All projections use local Lahman CSVs. On Windows, the default directory is
 Set the `LAHMAN_DATA_DIR` environment variable to change the default, or pass
 `--data-dir` to override it for one command. No network requests or online
 fallbacks are made. The directory must contain `People.csv`, `Teams.csv`, and
-`Batting.csv` or `Pitching.csv` for the selected kind. Batter projections also
-require `Appearances.csv` and a valid `birthYear` in `People.csv`. Filenames are matched
+`Batting.csv` or `Pitching.csv` for the selected kind. Both kinds require valid
+`birthYear` values in `People.csv`; batters additionally require `Appearances.csv`. Filenames are matched
 without regard to case. The database is not bundled with Backcaster.
 
 For the default Windows directory, no additional flag is needed:
@@ -166,11 +199,12 @@ ambiguous name lookups report the matching IDs.
 
 Batter league regression and rebaselining use all non-pitcher player-seasons;
 qualification selects the displayed players. Pitching retains its qualified-only
-baseline. See [the batter model](docs/BATTER_MODEL.md) for the six-stage calculation,
+baseline, with real-age performance adjustments documented in
+[the pitcher model](docs/PITCHER_MODEL.md). See [the batter model](docs/BATTER_MODEL.md) for the six-stage calculation,
 real-age adjustment, PA formula, and explicit rebaselining choices.
-Single-player projections require qualification in all three preceding seasons.
-Team and batch projections use each player's available qualified seasons within
-that window. For team selection, a traded player's full-season totals are assigned
+Eligibility means qualifying in at least one of the three preceding seasons.
+For every selected player, projections use all available seasons in that window,
+including nonqualified seasons and seasons spent with other teams. For team selection, a traded player's full-season totals are assigned
 to the team in their latest stint for that season; this is not a current-roster forecast.
 
 ### 2) Team projection
@@ -179,7 +213,7 @@ to the team in their latest stint for that season; this is not a current-roster
 backcaster team --team LAN --year 2026 --kind pitching
 ```
 
-### 3) Batch projection (all players returned by source data)
+### 3) Batch projection (all eligible players)
 
 ```bash
 backcaster batch --year 2026 --kind batting
@@ -208,7 +242,8 @@ Example shape (columns vary by batting vs pitching):
 ```
 
 ### CSV output (`--format csv`)
-A single-row CSV for `player`; multi-row CSV for `team` and `batch`.
+A single-row CSV for `player`; multi-row CSV for `team`, `batch`, `hitters`, and `pitchers`.
+Both populations include `PlayerID` for matching records across files.
 
 ### HTML output (`--format html`)
 A basic HTML table (no custom CSS) suitable for quick sharing or embedding.
@@ -227,6 +262,7 @@ Lahman distribution from the same edition. Existing online caches are ignored.
 
 Runtime dependencies are intentionally minimal:
 - `pandas` for tabular data processing and projection output
+- `numpy` for numerical operations and rebaselining convergence checks
 
 Development dependencies (`--extra dev`) include:
 - `pytest`, `pytest-cov`

{backcaster-0.3.4 → backcaster-0.3.6}/README.md

@@ -5,7 +5,7 @@ Backcaster is a Python CLI that produces **Marcel-style MLB projections** from t
 It supports:
 - **Single player projections** (`player`)
 - **Single team projections** (`team`)
-- **League-wide batch projections** (`batch`)
+- **League-wide projections** (`hitters`, `pitchers`, or `batch`)
 - Output as terminal table, CSV, or HTML
 
 ---
@@ -25,10 +25,10 @@ Upgrade to the latest release available on PyPI:
 python -m pip install --upgrade backcaster
 ```
 
-To install a specific release, such as 0.3.4:
+To install a specific release, such as 0.3.6:
 
 ```powershell
-python -m pip install --upgrade backcaster==0.3.4
+python -m pip install --upgrade backcaster==0.3.6
 ```
 
 Run the command using the same Python environment where you run Backcaster.
@@ -66,19 +66,52 @@ pip install dist/backcaster-*.whl
 
 ## Quickstart CLI usage
 
-All commands require:
-- `--year` target projection year
-- `--kind` one of `batting` or `pitching`
+All commands require `--year`, the target projection year.
+The `player`, `team`, and `batch` commands also require `--kind batting|pitching`.
+The `hitters` and `pitchers` commands choose their kind automatically.
 
 Subcommand-specific required options:
 - `player` requires `--name`
 - `team` requires `--team`
-- `batch` requires no additional required option
+- `batch`, `hitters`, and `pitchers` require no player or team option
 
 ```bash
 backcaster <player|team|batch> [subcommand-options] --year <YEAR> --kind <batting|pitching> [--format cli|csv|html] [--out PATH]
+backcaster <hitters|pitchers> --year <YEAR> [--data-dir PATH] [--format cli|csv|html] [--out PATH]
 ```
 
+### Pre-offseason league projections
+
+Once the latest regular season is present in your local Lahman files, generate
+both populations without roster updates or team filters. For terminal output:
+
+```powershell
+backcaster hitters --year 2026
+backcaster pitchers --year 2026
+```
+
+To save files for comparison with Baseball Reference:
+
+```powershell
+backcaster hitters --year 2026 --format csv --out hitters_2026.csv
+backcaster pitchers --year 2026 --format csv --out pitchers_2026.csv
+```
+
+`--year 2026` projects 2026 from 2023–2025; it never reads 2026 statistics.
+Both commands support `--data-dir`, `--format cli|csv|html`, and `--out`.
+They are shortcuts for `batch --kind batting` and `batch --kind pitching`.
+For a different data directory, add `--data-dir "C:\Baseball\lahman"`.
+For HTML, use `--format html --out hitters_2026.html` (or a pitcher output filename).
+File formats require `--out`; the CLI format prints to the terminal.
+
+Players qualify for selection by meeting the existing PA/IP threshold in at least
+one of those three seasons. This is a historical eligibility list, not an active
+roster; pitchers who never reach the qualifying innings threshold (including most
+relievers) remain excluded. Qualification selects players without removing their
+other seasons from the calculation. Trades are combined into one player-season
+before qualification, and all teams' statistics contribute once. The output is
+one row per selected player, ready for comparison with reference projections.
+
 ### 1) Player projection
 
 ```bash
@@ -92,8 +125,8 @@ All projections use local Lahman CSVs. On Windows, the default directory is
 Set the `LAHMAN_DATA_DIR` environment variable to change the default, or pass
 `--data-dir` to override it for one command. No network requests or online
 fallbacks are made. The directory must contain `People.csv`, `Teams.csv`, and
-`Batting.csv` or `Pitching.csv` for the selected kind. Batter projections also
-require `Appearances.csv` and a valid `birthYear` in `People.csv`. Filenames are matched
+`Batting.csv` or `Pitching.csv` for the selected kind. Both kinds require valid
+`birthYear` values in `People.csv`; batters additionally require `Appearances.csv`. Filenames are matched
 without regard to case. The database is not bundled with Backcaster.
 
 For the default Windows directory, no additional flag is needed:
@@ -135,11 +168,12 @@ ambiguous name lookups report the matching IDs.
 
 Batter league regression and rebaselining use all non-pitcher player-seasons;
 qualification selects the displayed players. Pitching retains its qualified-only
-baseline. See [the batter model](docs/BATTER_MODEL.md) for the six-stage calculation,
+baseline, with real-age performance adjustments documented in
+[the pitcher model](docs/PITCHER_MODEL.md). See [the batter model](docs/BATTER_MODEL.md) for the six-stage calculation,
 real-age adjustment, PA formula, and explicit rebaselining choices.
-Single-player projections require qualification in all three preceding seasons.
-Team and batch projections use each player's available qualified seasons within
-that window. For team selection, a traded player's full-season totals are assigned
+Eligibility means qualifying in at least one of the three preceding seasons.
+For every selected player, projections use all available seasons in that window,
+including nonqualified seasons and seasons spent with other teams. For team selection, a traded player's full-season totals are assigned
 to the team in their latest stint for that season; this is not a current-roster forecast.
 
 ### 2) Team projection
@@ -148,7 +182,7 @@ to the team in their latest stint for that season; this is not a current-roster
 backcaster team --team LAN --year 2026 --kind pitching
 ```
 
-### 3) Batch projection (all players returned by source data)
+### 3) Batch projection (all eligible players)
 
 ```bash
 backcaster batch --year 2026 --kind batting
@@ -177,7 +211,8 @@ Example shape (columns vary by batting vs pitching):
 ```
 
 ### CSV output (`--format csv`)
-A single-row CSV for `player`; multi-row CSV for `team` and `batch`.
+A single-row CSV for `player`; multi-row CSV for `team`, `batch`, `hitters`, and `pitchers`.
+Both populations include `PlayerID` for matching records across files.
 
 ### HTML output (`--format html`)
 A basic HTML table (no custom CSS) suitable for quick sharing or embedding.
@@ -196,6 +231,7 @@ Lahman distribution from the same edition. Existing online caches are ignored.
 
 Runtime dependencies are intentionally minimal:
 - `pandas` for tabular data processing and projection output
+- `numpy` for numerical operations and rebaselining convergence checks
 
 Development dependencies (`--extra dev`) include:
 - `pytest`, `pytest-cov`

{backcaster-0.3.4 → backcaster-0.3.6}/THIRD_PARTY_NOTICES.md

@@ -12,11 +12,13 @@ Reference revision: `13a1c7943f5a7317c9c812a28e299c2184060046` (March 1, 2020).
 See [reference and adaptation notes](docs/MARCELR_REFERENCE.md) for the files
 reviewed, formulas examined, current differences, and attribution practices.
 
-At the time this notice was added, marcelR source had been inspected, but no R
-source had been copied or translated into Backcaster, and its datasets had not
-been bundled. Backcaster does not depend on R or marcelR. The notice is retained
-for acknowledgment and to accompany future attributed adaptations. Neither the
-review nor the acknowledgment implies endorsement or verified numerical parity.
+Backcaster's pitcher-aging equations in `src/backcaster/marcel.py` are adapted
+from the age-adjustment formulas in marcelR's `R/utilities.R` and their stat-specific
+application in `inst/scripts/marcel_data_exporter.R` at the revision above.
+The implementation is Python, with local birth-year validation and unchanged
+pitching opportunities; it is not a complete port of marcelR. No marcelR datasets
+are bundled. Backcaster does not depend on R or marcelR. Attribution does not
+imply endorsement or verified numerical parity.
 
 The following is marcelR's MIT license notice, reproduced without changes:
 

{backcaster-0.3.4 → backcaster-0.3.6}/src/backcaster/__init__.py

@@ -1,4 +1,4 @@
 """backcaster package."""
 
 __all__ = ["__version__"]
-__version__ = "0.3.4"
+__version__ = "0.3.6"

{backcaster-0.3.4 → backcaster-0.3.6}/src/backcaster/cli.py

@@ -40,9 +40,12 @@ def build_parser() -> argparse.ArgumentParser:
     p = argparse.ArgumentParser(prog="backcaster")
     sub = p.add_subparsers(dest="command", required=True)
 
-    def add_common(sp: argparse.ArgumentParser) -> None:
+    def add_common(sp: argparse.ArgumentParser, kind: str | None = None) -> None:
         sp.add_argument("--year", type=int, required=True)
-        sp.add_argument("--kind", choices=["batting", "pitching"], required=True)
+        if kind is None:
+            sp.add_argument("--kind", choices=["batting", "pitching"], required=True)
+        else:
+            sp.set_defaults(kind=kind)
         sp.add_argument("--format", choices=_OUTPUT_FORMAT_CHOICES, default="cli")
         sp.add_argument("--out")
         sp.add_argument(
@@ -61,6 +64,10 @@ def build_parser() -> argparse.ArgumentParser:
 
     bp = sub.add_parser("batch")
     add_common(bp)
+    for command, kind in [("hitters", "batting"), ("pitchers", "pitching")]:
+        league = sub.add_parser(command, help=f"Project all eligible {command} across all teams")
+        add_common(league, kind)
+        league.set_defaults(command="batch")
     return p
 
 

{backcaster-0.3.4 → backcaster-0.3.6}/src/backcaster/lahman.py

@@ -124,28 +124,33 @@ class LahmanSource:
         if kind == "batting":
             seasons, batting = self._batting_population(year)
         else:
+            self.read("People", {"playerID", "nameFirst", "nameLast", "birthYear"})
             seasons = pd.concat(
-                [self.season(y, kind) for y in (year - 1, year - 2, year - 3)], ignore_index=True
+                [self.season(y, kind, qualified=False) for y in (year - 1, year - 2, year - 3)],
+                ignore_index=True,
             )
+        eligible = seasons[seasons["Qualified"]]
         if command == "player":
-            matches = self._player_seasons(seasons, name)
+            candidates = self._player_seasons(eligible, name)
+            matches = seasons[seasons["PlayerID"] == candidates["PlayerID"].iloc[0]]
             if kind == "batting":
                 return batting[batting["PlayerID"] == matches["PlayerID"].iloc[0]].reset_index(
                     drop=True
                 )
             return project_player(
-                str(matches["Name"].iloc[0]), matches, kind, year, league_df=seasons
-            )
-        selected = seasons
+                str(matches["Name"].iloc[0]), matches, kind, year, league_df=eligible
+            ).assign(PlayerID=matches["PlayerID"].iloc[0])
+        selected = eligible
         if command == "team":
-            selected = seasons[seasons["Team"].str.casefold() == (team or "").casefold()]
+            selected = eligible[eligible["Team"].str.casefold() == (team or "").casefold()]
             if selected.empty:
                 raise ProjectionError(
                     f"No qualified players for Lahman team ID '{team}' in the prior three seasons."
                 )
         if kind == "batting":
             return batting[batting["PlayerID"].isin(selected["PlayerID"])].reset_index(drop=True)
-        return project_team(selected, kind, year, baseline_df=seasons)
+        history = seasons[seasons["PlayerID"].isin(selected["PlayerID"])]
+        return project_team(history, kind, year, baseline_df=eligible)
 
     def _batting_population(self, year: int) -> tuple[pd.DataFrame, pd.DataFrame]:
         self.read("People", {"playerID", "nameFirst", "nameLast", "birthYear"})
@@ -178,7 +183,7 @@ class LahmanSource:
         qualified = seasons[seasons["Qualified"]]
         if qualified.empty:
             raise ProjectionError("No qualified non-pitcher batters in the prior three seasons.")
-        return qualified, projected
+        return seasons, projected
 
     def _player_seasons(self, seasons: pd.DataFrame, name: str | None) -> pd.DataFrame:
         normalized = " ".join((name or "").casefold().split())
@@ -194,6 +199,4 @@ class LahmanSource:
             raise PlayerLookupError(
                 f"Ambiguous Lahman name '{name}'; matching player IDs: {', '.join(ids)}."
             )
-        if matches["Season"].nunique() != 3:
-            raise ProjectionError(f"Player '{name}' must qualify in all three prior seasons.")
         return matches

{backcaster-0.3.4 → backcaster-0.3.6}/src/backcaster/marcel.py

@@ -1,5 +1,7 @@
 from __future__ import annotations
 
+import math
+
 import pandas as pd
 
 from backcaster.normalize import safe_divide
@@ -14,9 +16,27 @@ BATTING_COMPONENTS = ["PA", "AB", "H", "2B", "3B", "HR", "BB", "SO", "HBP", "SF"
 PITCHING_COMPONENTS = ["IP", "ER", "H", "HR", "BB", "SO", "BF"]
 
 
-def _weighted_row(
-    rows: list[pd.Series], weights: tuple[float, float, float], cols: list[str]
-) -> pd.Series:
+def _pitcher_age(prior: pd.DataFrame, year: int, fallback: float) -> float:
+    age = fallback
+    if "birthYear" in prior:
+        age = year - float(pd.to_numeric(prior["birthYear"].iloc[0], errors="coerce"))
+    if not math.isfinite(age) or not 15 <= age <= 70:
+        raise ProjectionError("A valid birthYear or explicit age is required for pitcher aging.")
+    return age
+
+
+def _age_pitching(counts: pd.Series, age: float) -> pd.Series:
+    # Equations adapted from Ben Dilday's marcelR, utilities.R and
+    # inst/scripts/marcel_data_exporter.R, revision 13a1c7943f5a7317c9c812a28e299c2184060046.
+    # Copyright (c) 2017 Ben Dilday, MIT; see THIRD_PARTY_NOTICES.md.
+    benefit = 1 / (1 + 0.003 * (age - 29)) if age > 29 else 1 + 0.006 * (29 - age)
+    result = counts.copy()
+    result[["ER", "H", "HR", "BB"]] /= benefit
+    result["SO"] *= benefit
+    return result
+
+
+def _weighted_row(rows: list[pd.Series], weights: tuple[float, ...], cols: list[str]) -> pd.Series:
     total_w = sum(weights[: len(rows)])
     out = {}
     for col in cols:
@@ -124,14 +144,20 @@ def project_player(
     prior_df = _prepare_prior_df(prior_three, comps, optional_comps=optional_comps)
 
     rows = [prior_df.iloc[i] for i in range(min(3, prior_df.shape[0]))]
-    weighted = _weighted_row(rows, config.season_weights, comps)
+    weights: tuple[float, ...] = config.season_weights[: len(rows)]
+    if kind == "pitching":
+        calendar_weights: dict[float, float] = dict(
+            zip((year - 1, year - 2, year - 3), config.season_weights)
+        )
+        weights = tuple(calendar_weights.get(float(row["Season"]), 0.0) for row in rows)
+    weighted = _weighted_row(rows, weights, comps)
 
     pt_col = comps[0]
     weighted_pt = float(weighted[pt_col])
     # Regression uses weighted evidence, not the per-season average used for
     # playing time. Averaging first would multiply regression strength by the
     # sum of the season weights (12 for three seasons with weights 5/4/3).
-    evidence = weighted * sum(config.season_weights[: len(rows)])
+    evidence = weighted * sum(weights)
     evidence_pt = float(evidence[pt_col])
     reliability = safe_divide(evidence_pt, evidence_pt + reg_pt)
 
@@ -148,7 +174,7 @@ def project_player(
         regressed[c] = regressed_rate * weighted_pt
 
     growth = config.default_pa_growth if kind == "batting" else config.default_ip_growth
-    age_adj = config.age_adjustment_fn(age)
+    age_adj = config.age_adjustment_fn(age) if kind == "batting" else 1.0
     projected_pt = weighted_pt * growth * age_adj
     pt_scale = safe_divide(projected_pt, weighted_pt)
     regressed *= pt_scale
@@ -156,12 +182,16 @@ def project_player(
     if kind == "batting":
         rates = _derive_batting_rates(regressed)
     else:
+        age = _pitcher_age(prior_df, year, age)
+        regressed = _age_pitching(regressed, age)
         rates = _derive_pitching_rates(regressed)
 
     output = pd.concat([regressed, rates])
     output["Reliability"] = reliability
     output["Name"] = player_name
     output["Year"] = year
+    if kind == "pitching":
+        output["Age"] = age
     return output.to_frame().T
 
 
@@ -195,16 +225,17 @@ def project_team(
         display_name = sorted_grp["Name"].iloc[0]
         if not isinstance(display_name, str):
             raise ProjectionError("Encountered non-string player name in team input.")
-        projections.append(
-            project_player(
-                display_name,
-                sorted_grp.head(3),
-                kind,
-                year,
-                config,
-                league_rates=baseline_rates,
-            )
+        projection = project_player(
+            display_name,
+            sorted_grp.head(3),
+            kind,
+            year,
+            config,
+            league_rates=baseline_rates,
         )
+        if "PlayerID" in sorted_grp:
+            projection["PlayerID"] = sorted_grp["PlayerID"].iloc[0]
+        projections.append(projection)
     if not projections:
         raise ProjectionError("No players available to project for this team.")
     return pd.concat(projections, ignore_index=True)

{backcaster-0.3.4 → backcaster-0.3.6/src/backcaster.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: backcaster
-Version: 0.3.4
+Version: 0.3.6
 Summary: Marcel-style MLB projection CLI using local Lahman CSV data
 License-Expression: MIT
 Project-URL: Homepage, https://github.com/jeffreywevans/backcaster
@@ -36,7 +36,7 @@ Backcaster is a Python CLI that produces **Marcel-style MLB projections** from t
 It supports:
 - **Single player projections** (`player`)
 - **Single team projections** (`team`)
-- **League-wide batch projections** (`batch`)
+- **League-wide projections** (`hitters`, `pitchers`, or `batch`)
 - Output as terminal table, CSV, or HTML
 
 ---
@@ -56,10 +56,10 @@ Upgrade to the latest release available on PyPI:
 python -m pip install --upgrade backcaster
 ```
 
-To install a specific release, such as 0.3.4:
+To install a specific release, such as 0.3.6:
 
 ```powershell
-python -m pip install --upgrade backcaster==0.3.4
+python -m pip install --upgrade backcaster==0.3.6
 ```
 
 Run the command using the same Python environment where you run Backcaster.
@@ -97,19 +97,52 @@ pip install dist/backcaster-*.whl
 
 ## Quickstart CLI usage
 
-All commands require:
-- `--year` target projection year
-- `--kind` one of `batting` or `pitching`
+All commands require `--year`, the target projection year.
+The `player`, `team`, and `batch` commands also require `--kind batting|pitching`.
+The `hitters` and `pitchers` commands choose their kind automatically.
 
 Subcommand-specific required options:
 - `player` requires `--name`
 - `team` requires `--team`
-- `batch` requires no additional required option
+- `batch`, `hitters`, and `pitchers` require no player or team option
 
 ```bash
 backcaster <player|team|batch> [subcommand-options] --year <YEAR> --kind <batting|pitching> [--format cli|csv|html] [--out PATH]
+backcaster <hitters|pitchers> --year <YEAR> [--data-dir PATH] [--format cli|csv|html] [--out PATH]
 ```
 
+### Pre-offseason league projections
+
+Once the latest regular season is present in your local Lahman files, generate
+both populations without roster updates or team filters. For terminal output:
+
+```powershell
+backcaster hitters --year 2026
+backcaster pitchers --year 2026
+```
+
+To save files for comparison with Baseball Reference:
+
+```powershell
+backcaster hitters --year 2026 --format csv --out hitters_2026.csv
+backcaster pitchers --year 2026 --format csv --out pitchers_2026.csv
+```
+
+`--year 2026` projects 2026 from 2023–2025; it never reads 2026 statistics.
+Both commands support `--data-dir`, `--format cli|csv|html`, and `--out`.
+They are shortcuts for `batch --kind batting` and `batch --kind pitching`.
+For a different data directory, add `--data-dir "C:\Baseball\lahman"`.
+For HTML, use `--format html --out hitters_2026.html` (or a pitcher output filename).
+File formats require `--out`; the CLI format prints to the terminal.
+
+Players qualify for selection by meeting the existing PA/IP threshold in at least
+one of those three seasons. This is a historical eligibility list, not an active
+roster; pitchers who never reach the qualifying innings threshold (including most
+relievers) remain excluded. Qualification selects players without removing their
+other seasons from the calculation. Trades are combined into one player-season
+before qualification, and all teams' statistics contribute once. The output is
+one row per selected player, ready for comparison with reference projections.
+
 ### 1) Player projection
 
 ```bash
@@ -123,8 +156,8 @@ All projections use local Lahman CSVs. On Windows, the default directory is
 Set the `LAHMAN_DATA_DIR` environment variable to change the default, or pass
 `--data-dir` to override it for one command. No network requests or online
 fallbacks are made. The directory must contain `People.csv`, `Teams.csv`, and
-`Batting.csv` or `Pitching.csv` for the selected kind. Batter projections also
-require `Appearances.csv` and a valid `birthYear` in `People.csv`. Filenames are matched
+`Batting.csv` or `Pitching.csv` for the selected kind. Both kinds require valid
+`birthYear` values in `People.csv`; batters additionally require `Appearances.csv`. Filenames are matched
 without regard to case. The database is not bundled with Backcaster.
 
 For the default Windows directory, no additional flag is needed:
@@ -166,11 +199,12 @@ ambiguous name lookups report the matching IDs.
 
 Batter league regression and rebaselining use all non-pitcher player-seasons;
 qualification selects the displayed players. Pitching retains its qualified-only
-baseline. See [the batter model](docs/BATTER_MODEL.md) for the six-stage calculation,
+baseline, with real-age performance adjustments documented in
+[the pitcher model](docs/PITCHER_MODEL.md). See [the batter model](docs/BATTER_MODEL.md) for the six-stage calculation,
 real-age adjustment, PA formula, and explicit rebaselining choices.
-Single-player projections require qualification in all three preceding seasons.
-Team and batch projections use each player's available qualified seasons within
-that window. For team selection, a traded player's full-season totals are assigned
+Eligibility means qualifying in at least one of the three preceding seasons.
+For every selected player, projections use all available seasons in that window,
+including nonqualified seasons and seasons spent with other teams. For team selection, a traded player's full-season totals are assigned
 to the team in their latest stint for that season; this is not a current-roster forecast.
 
 ### 2) Team projection
@@ -179,7 +213,7 @@ to the team in their latest stint for that season; this is not a current-roster
 backcaster team --team LAN --year 2026 --kind pitching
 ```
 
-### 3) Batch projection (all players returned by source data)
+### 3) Batch projection (all eligible players)
 
 ```bash
 backcaster batch --year 2026 --kind batting
@@ -208,7 +242,8 @@ Example shape (columns vary by batting vs pitching):
 ```
 
 ### CSV output (`--format csv`)
-A single-row CSV for `player`; multi-row CSV for `team` and `batch`.
+A single-row CSV for `player`; multi-row CSV for `team`, `batch`, `hitters`, and `pitchers`.
+Both populations include `PlayerID` for matching records across files.
 
 ### HTML output (`--format html`)
 A basic HTML table (no custom CSS) suitable for quick sharing or embedding.
@@ -227,6 +262,7 @@ Lahman distribution from the same edition. Existing online caches are ignored.
 
 Runtime dependencies are intentionally minimal:
 - `pandas` for tabular data processing and projection output
+- `numpy` for numerical operations and rebaselining convergence checks
 
 Development dependencies (`--extra dev`) include:
 - `pytest`, `pytest-cov`

{backcaster-0.3.4 → backcaster-0.3.6}/tests/test_lahman.py

@@ -102,8 +102,8 @@ def test_short_season_threshold(lahman: Path) -> None:
     source = LahmanSource(str(lahman))
     assert "star" in set(source.season(2020, "batting").PlayerID)
     assert "star" not in set(source.season(2021, "batting").PlayerID)
-    with pytest.raises(ProjectionError, match="all three"):
-        source.project("player", 2022, "batting", name="Test Star")
+    result = source.project("player", 2022, "batting", name="Test Star")
+    assert result.loc[0, "PA"] == pytest.approx(0.5 * 190 + 0.1 * 190 + 200)
 
 
 @pytest.mark.parametrize("command", ["player", "team", "batch"])
@@ -337,3 +337,84 @@ def test_missing_birth_year_column(lahman: Path) -> None:
     source = LahmanSource(str(lahman))
     with pytest.raises(DataFetchError, match="birthYear"):
         source.project("batch", 2022, "batting")
+
+
+def test_pitcher_aging_consistent_across_commands(lahman: Path) -> None:
+    source = LahmanSource(str(lahman))
+    player = source.project("player", 2022, "pitching", name="Test Star").iloc[0]
+    team = source.project("team", 2022, "pitching", team="BOS")
+    batch = source.project("batch", 2022, "pitching")
+    assert player["Age"] == 32
+    assert player["IP"] == pytest.approx(487 / 3)
+    for result in [team, batch]:
+        row = result[result.Name == "Test Star"].iloc[0]
+        for stat in ["IP", "SO", "ERA", "WHIP", "Age"]:
+            assert row[stat] == pytest.approx(player[stat])
+
+
+def test_pitcher_requires_birth_year(lahman: Path) -> None:
+    path = lahman / "People.csv"
+    pd.read_csv(path).drop(columns="birthYear").to_csv(path, index=False)
+    source = LahmanSource(str(lahman))
+    with pytest.raises(DataFetchError, match="birthYear"):
+        source.project("batch", 2022, "pitching")
+
+
+@pytest.mark.parametrize("command, kind", [("hitters", "batting"), ("pitchers", "pitching")])
+def test_league_commands_match_batch(lahman: Path, command: str, kind: str) -> None:
+    out = lahman / f"{command}.csv"
+    batch = lahman / "batch.csv"
+    common = ["--year", "2022", "--data-dir", str(lahman), "--format", "csv"]
+    assert main([command, *common, "--out", str(out)]) == 0
+    assert main(["batch", *common, "--kind", kind, "--out", str(batch)]) == 0
+    pd.testing.assert_frame_equal(pd.read_csv(out), pd.read_csv(batch))
+    assert len(pd.read_csv(out)) == 2
+
+
+def test_pitcher_full_history_survives_team_and_qualification_selection(lahman: Path) -> None:
+    path = lahman / "Pitching.csv"
+    frame = pd.read_csv(path)
+    frame.loc[(frame.playerID == "star") & (frame.yearID == 2019), "teamID"] = "LAN"
+    middle = (frame.playerID == "star") & (frame.yearID == 2020)
+    frame.loc[middle, ["teamID", "IPouts", "ER"]] = ["LAA", 90, 20]
+    frame.to_csv(path, index=False)
+    source = LahmanSource(str(lahman))
+    player = source.project("player", 2022, "pitching", name="Test Star").iloc[0]
+    team = source.project("team", 2022, "pitching", team="LAN").iloc[0]
+    batch = source.project("batch", 2022, "pitching")
+    row = batch[batch.Name == "Test Star"].iloc[0]
+    assert player["IP"] == pytest.approx((5 * 487 / 3 + 4 * 30 + 3 * 487 / 3) / 12)
+    for candidate in [team, row]:
+        for metric in ["IP", "ER", "ERA", "WHIP", "SO", "Reliability"]:
+            assert candidate[metric] == pytest.approx(player[metric])
+
+
+def test_pitcher_trade_stints_count_once_and_qualify_together(lahman: Path) -> None:
+    original = LahmanSource(str(lahman)).project("batch", 2022, "pitching")
+    path = lahman / "Pitching.csv"
+    frame = pd.read_csv(path)
+    mask = (frame.playerID == "star") & (frame.yearID == 2021)
+    row = frame[mask].copy()
+    frame = frame[~mask]
+    for column in ["IPouts", "ER", "H", "HR", "BB", "SO", "BFP"]:
+        row[column] = row[column] / 2
+    second = row.assign(stint=2, teamID="LAN")
+    pd.concat([frame, row, second]).to_csv(path, index=False)
+    source = LahmanSource(str(lahman))
+    season = source.season(2021, "pitching")
+    assert len(season[season.PlayerID == "star"]) == 1
+    actual = source.project("batch", 2022, "pitching")
+    pd.testing.assert_frame_equal(
+        actual.sort_values("PlayerID").reset_index(drop=True),
+        original.sort_values("PlayerID").reset_index(drop=True),
+    )
+
+
+def test_pitcher_without_three_qualifying_seasons_can_be_selected(lahman: Path) -> None:
+    path = lahman / "Pitching.csv"
+    frame = pd.read_csv(path)
+    frame.loc[(frame.playerID == "star") & (frame.yearID != 2021), "IPouts"] = 90
+    frame.to_csv(path, index=False)
+    result = LahmanSource(str(lahman)).project("player", 2022, "pitching", name="Test Star")
+    assert len(result) == 1
+    assert result.loc[0, "IP"] == pytest.approx((5 * 487 / 3 + 4 * 30 + 3 * 30) / 12)

{backcaster-0.3.4 → backcaster-0.3.6}/tests/test_marcel_math.py

@@ -408,3 +408,45 @@ def test_one_season_regression_uses_only_available_weights() -> None:
     result = project_player("One Season", prior, "batting", 2025, league_rates=rates)
     assert float(result.loc[0, "HR"]) / float(result.loc[0, "PA"]) == pytest.approx(324 / 4200)
     assert float(result.loc[0, "Reliability"]) == pytest.approx(3000 / 4200)
+
+
+@pytest.mark.parametrize("age, factor", [(25, 1.024), (29, 1.0), (34, 1 / 1.015)])
+def test_pitcher_age_curve_changes_rates_not_opportunities(age: int, factor: float) -> None:
+    neutral = project_player("Pitcher", _pitching_prior(), "pitching", 2026, age=29).iloc[0]
+    result = project_player("Pitcher", _pitching_prior(), "pitching", 2026, age=age).iloc[0]
+    assert result["IP"] == pytest.approx(neutral["IP"])
+    assert result["BF"] == pytest.approx(neutral["BF"])
+    assert result["Reliability"] == pytest.approx(neutral["Reliability"])
+    for stat in ["ER", "H", "HR", "BB", "ERA", "WHIP"]:
+        assert result[stat] == pytest.approx(neutral[stat] / factor)
+    assert result["SO"] == pytest.approx(neutral["SO"] * factor)
+    assert result["Age"] == age
+
+
+def test_pitcher_birth_year_overrides_default_age() -> None:
+    prior = _pitching_prior().assign(birthYear=1992)
+    actual = project_player("Pitcher", prior, "pitching", 2026)
+    expected = project_player("Pitcher", _pitching_prior(), "pitching", 2026, age=34)
+    pd.testing.assert_frame_equal(actual, expected)
+
+
+@pytest.mark.parametrize("birth", [float("nan"), float("inf"), "unknown", 1800, 2025])
+def test_pitcher_invalid_birth_year(birth: object) -> None:
+    prior = _pitching_prior().assign(birthYear=birth)
+    with pytest.raises(ProjectionError, match="valid birthYear"):
+        project_player("Pitcher", prior, "pitching", 2026)
+
+
+def test_pitcher_missing_year_keeps_calendar_weights() -> None:
+    prior = _pitching_prior()
+    prior = prior[prior.Season != 2024]
+    result = project_player("Pitcher", prior, "pitching", 2026, age=29).iloc[0]
+    assert result["IP"] == pytest.approx((5 * 200 + 3 * 210) / 8)
+    assert result["Reliability"] == pytest.approx(1630 / 1830)
+
+
+def test_pitcher_future_season_has_no_evidence() -> None:
+    prior = _pitching_prior().iloc[[0]].assign(Season=2026)
+    result = project_player("Pitcher", prior, "pitching", 2026, age=29).iloc[0]
+    assert result["IP"] == 0
+    assert result["Reliability"] == 0