What I Built
I built a web dashboard to visualize MLB bat tracking data from Baseball Savant.
It has 5 tabs:
| Tab | What it shows |
|---|---|
| Leaderboard | Rankings by Bat Speed, Attack Angle, and more |
| Player Comparison | Radar & bar charts comparing up to 6 players |
| WBC Country Strength | Batting & pitching scores for all 20 WBC 2026 nations |
| Team Lineup Builder | Bat tracking metrics for each MLB team's 9-man lineup |
| Monthly Trend | Month-by-month bat speed trend for any player |
It also supports English/Japanese language switching.
Data Source: savant-extras
I used my own OSS library savant-extras to fetch the data.
pip install savant-extras
It's a Python library for fetching bat tracking data from Baseball Savant. I built it because the existing pybaseball library didn't support date range filtering for this endpoint.
from savant_extras import bat_tracking, bat_tracking_monthly
# Batter bat tracking data for 2025 season
df = bat_tracking(year=2025, player_type="batter")
# Monthly breakdown
df_monthly = bat_tracking_monthly(year=2025)
- GitHub: https://github.com/yasumorishima/savant-extras
- PyPI: https://pypi.org/project/savant-extras/
Why I Chose Streamlit
I wanted to share my data analysis in a visual, accessible format. Streamlit caught my eye for three reasons:
- Pure Python — no HTML, CSS, or JavaScript needed
- Free hosting on Streamlit Community Cloud (just connect your GitHub repo)
-
UI components in one line —
st.selectbox(),st.slider(), done
For example, this is all you need to add a dropdown to your sidebar:
import streamlit as st
year = st.sidebar.selectbox("Season", [2024, 2025], index=1)
player_type = st.sidebar.selectbox("Player type", ["batter", "pitcher"])
Key Implementation Points
@st.cache_data to avoid repeated API calls
Fetching data every time a user interacts with the page would be too slow. @st.cache_data caches the result so repeated calls with the same arguments return instantly.
@st.cache_data(ttl=3600)
def load_bat_data(year: int, player_type: str):
return bat_tracking(year=year, player_type=player_type)
session_state to share data across tabs
Streamlit reruns the entire script on every user interaction. To keep data loaded after pressing a button, use st.session_state:
if load_btn:
st.session_state["df_raw"] = load_bat_data(year, player_type)
if "df_raw" in st.session_state:
df = st.session_state["df_raw"]
matplotlib-fontja for Japanese font support
To render Japanese labels in matplotlib charts, I used matplotlib-fontja:
import matplotlib_fontja # noqa: F401 ← just importing this enables Japanese fonts
A Problem I Hit: japanize_matplotlib Broke on Python 3.13
After deploying to Streamlit Community Cloud, I got this error:
File "japanize_matplotlib/japanize_matplotlib.py", line 5, in <module>
from distutils.version import LooseVersion
ModuleNotFoundError
distutils was deprecated in Python 3.12 and fully removed in 3.13. I tried pinning the Python version with a runtime.txt file (python-3.11), but it didn't work — Streamlit Cloud still used 3.13.
The fix was simple: switch to matplotlib-fontja, which doesn't use distutils.
- japanize-matplotlib>=1.1
+ matplotlib-fontja
Data Caveats
A few things to keep in mind about this dashboard:
- Only MLB-rostered players are included. NPB and minor league players are not
- Name matching has limitations — spelling variations or shared names may not match correctly
- WBC 2026 scores are provisional — rosters are based on Baseball America's February 2025 projections and may differ from actual rosters
- Bat tracking accuracy depends on Baseball Savant's data quality
Please treat the results as a reference, not a definitive source.
Things I Added Later
Mobile support
The sidebar was always open on mobile and got in the way. Fixed with two changes:
st.set_page_config(
...,
initial_sidebar_state="collapsed", # start closed on mobile
)
st.markdown("""
<style>
@media (max-width: 768px) {
.block-container { padding: 1rem 0.5rem !important; }
[data-testid="column"] { min-width: 45% !important; flex: 1 1 45% !important; }
[data-testid="stHorizontalBlock"] { flex-wrap: wrap !important; }
}
</style>
""", unsafe_allow_html=True)
Also added use_container_width=True to every st.pyplot() call. Without this, charts don't resize to the container width and overflow on small screens.
Auto-load on first visit
Originally you had to open the sidebar and click "Load Data". Changed it so data loads automatically on first visit:
# Before: only load when button is clicked
if load_btn:
...
# After: also auto-load on first visit
if load_btn or "df_raw" not in st.session_state:
...
Since @st.cache_data caches the result, subsequent tab switches and filter changes use the cache — no waiting after the first load.
Summary
- Used savant-extras (my own OSS library) for data fetching
- Built a 5-tab dashboard with Streamlit
- Deployed for free on Streamlit Community Cloud
Streamlit was easier than I expected. Tabs with st.tabs(), accordions with st.expander(), column layouts with st.columns() — learning just those three got me most of the way there.
If you want to turn your data analysis into something you can actually share, Streamlit is a great option.
Links
- App: https://yasumorishima-mlb-bat-tracking.streamlit.app/
- savant-extras (PyPI): https://pypi.org/project/savant-extras/
- savant-extras (GitHub): https://github.com/yasumorishima/savant-extras
- Baseball Savant: https://baseballsavant.mlb.com/
Top comments (5)
How long did it take you to get the hang of Streamlit compared to other tools you've used?
Honestly, Streamlit was surprisingly quick to pick up — I had a basic interactive chart running within a day. The hardest part wasn't Streamlit itself but figuring out the data side (filtering, state management, etc.). If you're already comfortable with pandas, I think you can build something useful in a weekend. Definitely worth trying!
Thank you for the feedback will keep that in mind :)
Good luck with your projects! 😊
Thank you YMori!