<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[MLB Technology Blog - Medium]]></title>
        <description><![CDATA[The latest from MLB’s Technology Organization - Medium]]></description>
        <link>https://technology.mlblogs.com?source=rss----ff0f5fb80706---4</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>MLB Technology Blog - Medium</title>
            <link>https://technology.mlblogs.com?source=rss----ff0f5fb80706---4</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 00:54:39 GMT</lastBuildDate>
        <atom:link href="https://technology.mlblogs.com/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com" rel="hub"/>
        <item>
            <title><![CDATA[Scout Insights — Automating the “Whoa! Didn’t Know That” Feeling]]></title>
            <link>https://technology.mlblogs.com/scout-insights-automating-the-whoa-didnt-know-that-feeling-139df5300ff2?source=rss----ff0f5fb80706---4</link>
            <guid isPermaLink="false">https://medium.com/p/139df5300ff2</guid>
            <category><![CDATA[baseball]]></category>
            <category><![CDATA[google-gemini]]></category>
            <category><![CDATA[information-theory]]></category>
            <category><![CDATA[mlb]]></category>
            <category><![CDATA[google-cloud-platform]]></category>
            <dc:creator><![CDATA[Clint Rorick]]></dc:creator>
            <pubDate>Wed, 07 Oct 2026 01:25:00 GMT</pubDate>
            <atom:updated>2026-10-07T01:26:11.224Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*sAr_Hsv9A3LOVXAJ1tHxMQ.png" /></figure><p><em>How we measure surprising baseball facts and dynamically serve them in milliseconds, at scale.</em></p><p>On June 4, Jung Hoo Lee went 4-for-5 to extend a <a href="https://www.mlb.com/news/casey-schmitt-leadoff-homer-helps-giants-beat-brewers#:~:text=Lee%20is%20now%20riding%20a%20career%2Dhigh%2012%2Dgame%20hitting%20streak%2C">career-high 12-game hitting streak</a>.</p><p>But inside that run, Scout found a different, even more impressive achievement:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6g1JqkhHHnk62kjXSxxSGg.png" /><figcaption><em>The image above is a Scout Insight, powered by Google Cloud — </em><a href="https://www.mlb.com/news/mlb-scout-insights-gameday-powered-by-google-cloud"><em>we released this feature on Gameday on Opening Day 2026</em></a><em>.</em></figcaption></figure><p>Scout’s model estimated that only around 1 in 420,000 comparable 30-PA stretches would contain 20 or more hits.</p><p>Typical streaks are straightforward to keep track of — they either continue or they don’t. Same with many other baseball stats — MLB ingests and processes countless baseball events per season, whether it’s as old-school as a fielding error, or as new-school as a <a href="https://www.mlb.com/glossary/statcast/barrel">barrel</a>.</p><p>In this massive haystack of numbers, how can anyone find a needle?</p><p>Luckily, MLB has an army of excellent human researchers doing exactly this — but even they need a break sometimes, and baseball fans have an insatiable appetite for content.</p><p>On top of that, Gameday has an incredible amount of content for fans to take in — there’s only so much room for these insights amongst graphics, pitch trails, embedded videos — around 150 characters of room, to be precise. To paraphrase Pascal, “I would have made this shorter, but I didn’t have the time.”</p><p>But who does have time? Computers!</p><p>We built MLB Scout Insights, powered by Google Cloud, to help. You can see the results in <a href="https://www.mlb.com/gameday">Gameday</a> with a free MLB.com account — about 18 self-contained insights per game, typically highlighting the batter or pitcher currently facing off.</p><h3>The needliest of needles</h3><p>Jackson Merrill went 4-for-5 with 2 home runs, a triple and 6 RBI on September 9. Scout found the performance in the game data, estimated it as a 1 in 1,900 occurrence, and selected it to show in Gameday during the top of the ninth inning of the Padres/Giants game on September 11. MLB.com’s <a href="https://www.mlb.com/news/jackson-merrill-hits-2-home-runs-sets-season-high-in-rbis-as-padres-sweep">account of Merrill’s big afternoon</a> adds even more context: he was only the third player in franchise history with two homers and a triple in one game.</p><p>Cooper Hjerpe’s four perfect relief innings on September 6 was estimated as a roughly 1 in 2,200 occurrence, close to Merrill’s performance in rarity.</p><p>Hjerpe struck out six, retiring every batter he faced.</p><p>Scout surfaced that performance in Gameday the next time it had the opportunity, in Hjerpe’s appearance September 12 against the White Sox.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KpxoPX6goTbsafIfiwPBfA.png" /></figure><p>Jose Altuve’s better swing decisions needed a different comparison. On April 14th, his chase rate had fallen from 38.9% in 2025 to 25% in 2026. Scout compared that drop with 1,988 qualified season-to-season changes and estimated it as a 1 in 2,000 event, a similar rarity level to Merrill and Hjerpe.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*blwmUguN1R2JrhSfKEV3dQ.png" /></figure><p>A fastball does not have to be the league’s fastest to be worth noticing.</p><p>Scout can also detect when a pitcher adds a few ticks.</p><p>During his June 10 start against the Yankees, Parker Messick’s four-seamer averaged 94.8 mph — 1.4 mph above his average across his previous four starts.</p><p>Messick’s velocity jump crossed Scout’s 1-in-16 rarity threshold to qualify for display.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vm_8WUgbtD7-A4cLYbo7mw.png" /></figure><p>The increase held. Across his next 16 starts, through September 16, Messick’s four-seamer averaged about 1.4 mph above the original 4-game sample. Scout helped tell the story of Messick’s ascendant rookie season.</p><p>Dylan Cease moved in the opposite direction. On July 25, his four-seamer averaged 95.8 mph, down from 97.2 across his previous four starts. Scout estimated this as a 1 in 30 occurrence — and surfaced this insight to fans in Gameday.</p><p>During Cease’s next nine starts, his fastball averaged 95.6 mph — about 1.6 mph below the original comparison. Scout again noticed a subtle shift in a pitcher’s stuff that ended up being sticky.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*z3MTWFjee4UMViHOtfGIyA.png" /></figure><h3>Expressing surprise in bits</h3><p>“1 in 96,000” is a valid way to express a very rare event. Our silicon friends can understand it just fine.</p><p>For mortal, carbon-based humans? We prefer small numbers, that we can count on our fingers and toes.</p><p>Having a simple scale lets us precisely curate a minimum rarity threshold for each of our hundreds of insight types — said otherwise, “surface all streaks that score at least 4 bits. Surface all prior-game performances that score at least 6 bits.”</p><p>So we use a base-2 scale.</p><p>1 finger = 1 bit = 1 in 2 occurrence. 2 fingers = 1 in 4. 3 fingers = 1 in 8, and so on.</p><p>Putting this into practice: based on Scout’s database of historical streaks, an 18-game hitting streak is about a 1-in-810 event. Feels nebulous.</p><p>But taking the base-2 logarithm, 1 in 810 becomes about 10 bits of surprisal.</p><p>And we didn’t even have to take our socks off!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bLJNxfXiFtx7SodLsoVW5w.png" /></figure><h3>Which insight gets the spotlight?</h3><p>Before Freddie Freeman’s first plate appearance on September 12, we had nine different ways to highlight him: his success against right-handed pitching, a career milestone within reach, years of consistent production, and a recent slump.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Zn6d90sPEF1PXJkfdvUBEg.png" /></figure><p>Freeman was facing a right-hander. His “feasted on RHP” insight scored the highest, and was shown. Four others finished close behind, including his chance to catch Dwight Evans on the career home run list and a stretch of just one hit in his last 18 at-bats.</p><p>We give extra weight to opportunities that may soon disappear — if a stat would potentially be invalidated during the game, we serve it as early as we can. Evergreen facts are saved for later innings.</p><p>As play continues, we also favor players and subjects fans that we haven’t surfaced recently, and give a boost to categories with fewer candidates available that game. Ideally, this results in a tasty variety plate of blurbs to keep you informed.</p><h3>Giving the computer a head start</h3><p>We have about 10–15 seconds between a batter stepping to the plate in the physical world, and that batter being shown at the plate on Gameday. We give Gemini the player’s supporting facts ahead of time and save the sentence it produces. During the game, we choose among sentences already produced by Gemini using the deterministic methods described above.</p><p>(Gemini’s main job is producing a sentence that is short enough to fit on a small mobile card, but also not awkward, stilted or overly robotic. For some sentences simple templating works, but some of our generated insights are dynamic enough that we greatly benefit from an AI handling the phrasing.)</p><p>Altuve’s chase-rate insight was saved at 5:05 a.m. CDT on April 14, before his night game at Daikin Park in Houston. When Scout selected it in the sixth inning at 9:11 p.m. CDT, the wording had been ready for more than 16 hours.</p><p>Gemini is very powerful, but also needs a few seconds to complete the ask we’re making of it — “pre-baking the cake” lets us always have the insight ready when the Gameday app needs to grab it.</p><p>Across 1,273 selections in 95 games, half were saved within 1.4 seconds of the recorded game-feed update and 95% within 2.0 seconds.</p><h3>Share the work as the game unfolds</h3><p>If you and another viewer load Gameday for the same game moment, we can reuse the calculation.</p><p>If our CDN hasn’t yet received a request for a particular game moment, we check a shared origin cache before reading the saved selection from the database. Every path reuses the earlier research and Gemini wording.</p><p>Scout’s median edge cache response time was 13 ms at the median and 65 ms at the 95th percentile. <a href="https://www.mlb.com/amp/news/jacob-misiorowski-elite-velocity-extension.html?utm_source=chatgpt.com#:~:text=And%20when%20they%20are%3F%20A%20100%20mph%20fastball%20from%20Misiorowski%20will%20get%20to%20the%20plate%20in%20under%20four%2Dtenths%20of%20a%20second">A Jacob Misiorowski fastball takes roughly 350–400 milliseconds to reach the plate.</a></p><p>Scout Insights are just getting started — look for new insight types as the postseason rolls on!</p><p><a href="https://www.mlb.com/news/mlb-scout-insights-gameday-powered-by-google-cloud">MLB Scout Insights launch and product examples</a></p><p><a href="https://www.mlb.com/schedule/gameday/scout">MLB Scout Insights in Gameday</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*c6oSHVoeLRE2OJdR6OcvdQ.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=139df5300ff2" width="1" height="1" alt=""><hr><p><a href="https://technology.mlblogs.com/scout-insights-automating-the-whoa-didnt-know-that-feeling-139df5300ff2">Scout Insights — Automating the “Whoa! Didn’t Know That” Feeling</a> was originally published in <a href="https://technology.mlblogs.com">MLB Technology Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Changing the Game with Agentic AI: Highlights from MLB’s 6th Annual Hackathon, Powered by Google…]]></title>
            <link>https://technology.mlblogs.com/changing-the-game-with-agentic-ai-highlights-from-mlbs-6th-annual-hackathon-powered-by-google-54772db4d466?source=rss----ff0f5fb80706---4</link>
            <guid isPermaLink="false">https://medium.com/p/54772db4d466</guid>
            <category><![CDATA[hackathons]]></category>
            <dc:creator><![CDATA[Tani Cohen]]></dc:creator>
            <pubDate>Wed, 07 Oct 2026 00:44:13 GMT</pubDate>
            <atom:updated>2026-10-07T00:44:14.971Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>Changing the Game with Agentic AI: Highlights from MLB’s 6th Annual Hackathon, Powered by Google Cloud</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*56NBvGweZSriekME37WudQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Q7omkjVeA9HVT2PAR46PYg.jpeg" /></figure><p>This fall, Major League Baseball (MLB) wrapped its 6th Annual Hackathon, a massive initiative dedicated to driving innovation and collaboration across the organization. With <strong>Agentic AI</strong> as this year’s central theme, over 300 participants, judges, and support staff from MLB offices across New York (NY), Boulder (CO), San Francisco (CA), El Segundo (CA), Vero Beach (FL), and remote locations came together to create innovative projects.</p><p>This year’s hackathon had an almost 50% increase in participation, with 20% of participants joining from outside the Product &amp; Technology departments, embodying the “Citizen Developer” movement.</p><p>Over the course of the event, <strong>60 teams</strong> worked tirelessly on innovative projects designed to address real-world MLB challenges. The goal was clear: utilize Agentic AI to allow teams of all technology backgrounds to move from concept to working prototype in record time, developing tools that will push the boundaries of technology in baseball.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5NRWkGgV469unesFLTVN6g.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bBVcPi37DG4zjL0FO6RI7A.jpeg" /></figure><p>Teams focused their efforts across five key tracks:</p><p>1. <strong>Future of Work:</strong> This track explores how intelligent agents can reduce repetitive work, simplify complex tasks, improve collaboration, and empower people to focus on higher-value work.</p><p>2. <strong>Fan and Partner Pavilion:</strong> This track explores how Agentic AI can create more personalized, intuitive, and engaging experiences. It focuses on building solutions that make interactions more helpful, efficient, and memorable.</p><p>3. <strong>BasebALL in the Community:</strong> This track encourages solutions that use Agentic AI to grow the game, break down barriers to participation, or improve sustainability and accessibility across MLB venues and programs. Solutions must directly connect to baseball or MLB operations.</p><p>4. <strong>Imagination Engine:</strong> This track celebrates ideas that enhance storytelling, design, content creation, entertainment, learning, communication, or entirely new forms of creative expression and collaboration as it relates to baseball.</p><p>5. <strong>Wild Card:</strong> Some of the most exciting innovations don’t fit neatly into a category. If an idea crosses disciplines or explores a completely new application of Agentic AI, this is the place for it.</p><p>Projects were judged across several key criteria, with <strong>Technology</strong> and <strong>Business</strong> impact receiving special attention. In addition, a special award presented by Google was scored by an AI judge powered by Gemini. This “Judge Gemini” helped determine which team would receive the Google Award for Most Innovative GenAI.</p><p>The full slate of prototype projects, whether winners or not, were reviewed by Business, Technology, and Product leaders, and we’re now considering which projects are destined for MLB our future roadmaps.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gEoIUa2ykqCx7G_P53DLfg.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*E9Q2sBsNT8Q79ok0aK3iTw.jpeg" /></figure><p><strong>Leveraging the Power of Google Cloud and Agentic AI</strong></p><p>Now in its sixth year, the Hackathon is key to driving innovation within MLB. The 2026 Hackathon, presented by Google Cloud, the official AI provider of MLB, allowed participants to explore AI-native “re-think” opportunities with direct access and collaboration with Google as a partner.</p><p>Teams had a robust technology stack, provided by Google, at their disposal. To support Hackathoners hitting the ground running, Google Cloud offered a number of AI training sessions ahead of the event to help empower participants. These training sessions provided deep-dives into the Google Cloud Generative AI stack, and were instrumental in setting the stage for our teams to explore AI opportunities across all three key sub-tracks for Hackathon. The learning and experimentation that teams were able to apply to their projects extends well beyond Hackathon and MLB employees are now well-equipped to use these tools in their everyday project work, driving success for MLB and baseball fans alike!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZxaspftCeXA3CBaJ0FCdOg.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZmS_1njbWUytz0-mLKwDIg.jpeg" /></figure><p><strong><em>Congratulations again to all 60 teams and over 300 participants for making the 6th Annual MLB Hackathon the most innovative yet!</em></strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=54772db4d466" width="1" height="1" alt=""><hr><p><a href="https://technology.mlblogs.com/changing-the-game-with-agentic-ai-highlights-from-mlbs-6th-annual-hackathon-powered-by-google-54772db4d466">Changing the Game with Agentic AI: Highlights from MLB’s 6th Annual Hackathon, Powered by Google…</a> was originally published in <a href="https://technology.mlblogs.com">MLB Technology Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[MLB’s 5th Annual Hackathon, Powered by Google Cloud, Focused on Generative AI]]></title>
            <link>https://technology.mlblogs.com/mlbs-5th-annual-hackathon-powered-by-google-cloud-focused-on-generative-ai-761c020b68c0?source=rss----ff0f5fb80706---4</link>
            <guid isPermaLink="false">https://medium.com/p/761c020b68c0</guid>
            <category><![CDATA[hackathons]]></category>
            <dc:creator><![CDATA[Tani Cohen]]></dc:creator>
            <pubDate>Fri, 05 Dec 2025 03:00:22 GMT</pubDate>
            <atom:updated>2025-12-05T03:00:12.990Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PilaV8wVy0-zvpXfxbnN7Q.jpeg" /></figure><p>This fall, Major League Baseball (MLB) wrapped its 5th Annual Hackathon, a massive initiative dedicated to driving innovation and collaboration across the organization. With <strong>Generative AI</strong> as this year’s central theme, over <strong>250 participants, judges, and support staff</strong> from MLB came together across New York, Boulder, San Francisco, El Segundo, and remote locations to push the boundaries of technology in baseball.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GHfi8XrKoYHmkOsiLP7QGw.jpeg" /></figure><p>Over the course of the event, <strong>44 teams</strong> worked tirelessly on innovative projects designed to address real-world MLB challenges. The goal was clear: leverage Generative AI to shape the future of MLB, enhancing both the fan experience and internal operations.</p><p>Teams focused their efforts across three key sub-tracks:</p><p>1. <strong>Fan Experience:</strong> Projects aimed to reimagine how avid and casual fans connect with baseball, whether at the ballpark, at home, or on the go, focusing on personalized and interactive experiences.</p><p>2. <strong>Infrastructure and Tech Stack:</strong> This track sought to enhance MLB’s foundational technology, creating smarter, more resilient systems through backend innovation and automation, such as agentic tracking and mitigation of outages.</p><p>3. <strong>Personal Efficiencies / Automations:</strong> This track focused on streamlining day-to-day operations across all MLB departments — both technical and non-technical — to boost productivity and contribute to “operational excellence”.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZX_ZshPpoX5XI2RfHxxgvA.jpeg" /></figure><p>Projects were judged across several key criteria, with <strong>Technology</strong> and <strong>Business</strong> impact receiving special attention. In addition, a special award recognized the most innovative and groundbreaking use of Generative AI tools provided by Google Cloud.</p><p>The full slate of prototype projects, whether winners or not, will be reviewed by MLB’s Technology and Product leaders as they consider which projects are destined for MLB Technology’s development roadmaps.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OJ1AyINaLBXmFgQm8vpLJg.jpeg" /></figure><p><strong>Leveraging the Power of Google Cloud and Generative AI</strong></p><p>Now in its fifth year, the Hackathon is key to driving innovation within MLB. The 2025 Hackathon, presented by Google Cloud, a key technology partner of MLB, allowed participants to explore AI-native “re-think” opportunities with direct access and collaboration with Google as a partner.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rmeWuCI8BjB6jq_mqMbqJw.jpeg" /></figure><p>Teams had a robust technology stack at their disposal. Approved GenAI tools included <strong>Gemini for Workspace</strong> and the <strong>Gemini App</strong>, Google Cloud’s <strong>Vertex AI</strong> and <strong>Gemini API</strong>, <strong>Agentspace</strong> for building no-code agents, and the <strong>Google Agent Development Kit (ADK)</strong>. To support Hackathoners hitting the ground running, Google Cloud offered a number of GenAI training sessions ahead of the event to help empower participants. These training sessions provided deep-dives into the Google Cloud Generative AI stack, and were instrumental in setting the stage for our teams to explore AI opportunities across all three key sub-tracks for Hackathon. The learning and experimentation that teams were able to apply to their projects extends well beyond Hackathon and MLB employees are now well-equipped to use these tools in their everyday project work, driving success for MLB and baseball fans alike!</p><p><strong><em>Congratulations again to all 44 teams and over 250 participants for making the 5th Annual MLB Hackathon the most innovative yet!</em></strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=761c020b68c0" width="1" height="1" alt=""><hr><p><a href="https://technology.mlblogs.com/mlbs-5th-annual-hackathon-powered-by-google-cloud-focused-on-generative-ai-761c020b68c0">MLB’s 5th Annual Hackathon, Powered by Google Cloud, Focused on Generative AI</a> was originally published in <a href="https://technology.mlblogs.com">MLB Technology Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Automated Balls/Strike System (ABS) 2025 Spring Training Update]]></title>
            <link>https://technology.mlblogs.com/automated-balls-strike-system-abs-2025-spring-training-update-574d82c6faa2?source=rss----ff0f5fb80706---4</link>
            <guid isPermaLink="false">https://medium.com/p/574d82c6faa2</guid>
            <category><![CDATA[mlb]]></category>
            <category><![CDATA[baseball]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Graham Goldbeck]]></dc:creator>
            <pubDate>Fri, 21 Mar 2025 00:19:55 GMT</pubDate>
            <atom:updated>2025-03-21T00:38:39.532Z</atom:updated>
            <content:encoded><![CDATA[<p>| <em>Written by </em><a href="https://medium.com/@graham.goldbeck"><em>Graham Goldbeck</em></a><em>, Greg Sawers, Jimmy Heasley, Breese Sherman and Tani Cohen</em></p><p>Major League Baseball <a href="https://www.mlb.com/news/automated-ball-strike-calls-mlb-spring-games">recently announced</a> that the Automated Ball/Strike System (ABS) will be tested during 2025 Major League Spring Training. When we <a href="https://technology.mlblogs.com/developing-mlbs-automated-ball-strike-system-abs-d4f499deff31">originally posted about it</a> in 2021, ABS was still in its infancy. A lot has changed since then, such as the rollout of ABS in the Florida State League (FSL) and Triple-A level for the past several seasons, and the implementation of a challenge system. You may have seen ABS in action while attending or watching Minor League games, but with most Major League players and fans experiencing this for the first time, we thought it prudent to refresh that original post with the current state of ABS as we have launched for MLB games in Spring Training.</p><h3>Challenge Format</h3><p>There are two current versions of ABS: 1) every pitch call is made by ABS and is communicated to the umpire in real-time as outlined in the 2021 blog post, and 2) a challenge system in which pitches are called as usual but are subject to a review process. MLB has tested both systems extensively in multiple levels, and eventually chose to proceed with the challenge system exclusively during the second half of the 2024 Triple-A season. As a result, the challenge format is the version of choice for the 2025 Spring Training ABS rollout. When using the ABS challenge format, the umpire still calls balls and strikes using their own judgment as they normally do. However, if a pitcher, batter, or catcher disagrees with the ball/strike call, one of those players (and only those players) may indicate their desire to challenge the result of the pitch by giving the challenge signal. However, if the batter waits too long to challenge or the umpire observes that anyone else (including any players, coaches, fans, etc.) is influencing/compelling the player’s decision to challenge, they can deny the challenge.</p><p>If a challenge is successfully initiated, the umpire announces over the stadium speakers that the result of the previous pitch is challenged. Following the announcement, the scoreboard begins to show the challenge graphic feed, visually indicating that the pitch has been challenged. The trajectory of the previous pitch is then visually displayed from pitch release to where the ball intersects with the strike zone plane, with the graphic zooming in on the relevant quadrant of the strike zone for a prime view of any close calls. At the conclusion of the graphic simulation, which also displays the correct pitch call, the umpire announces whether the previous call was upheld or overturned.</p><p>Each Club receives two challenges a game and retains the challenge if it is successful (that is, the call is overturned).</p><p>This workflow has been tested extensively over the past few years at the Triple-A level. Here is a sample of what this challenge looked like in the 2024 season:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fwww.youtube.com%2Fembed%2FPWfQbESXnTs%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fwww.youtube.com%2Fwatch%3Fv%3DPWfQbESXnTs&amp;image=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fi.ytimg.com%2Fvi%2FPWfQbESXnTs%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/9fe4e883dddaf76adef1cafd60632216/href">https://medium.com/media/9fe4e883dddaf76adef1cafd60632216/href</a></iframe><h3>Definition of Strike Zone</h3><p>The boundaries of the strike zone that ABS uses are defined as follows:</p><ul><li>The horizontal boundaries of the strike zone are defined by the edges of the width of home plate, which are 8.5 inches in each direction, for a total of 17 inches.</li><li>The top and bottom of the zone will be set as a percentage of the batter’s measured height.</li></ul><p>To reduce the likelihood of rounding errors causing incorrect pitch calls, player measured heights are calculated in millimeters for ABS strike zone purposes. While listed heights on MLB consumer products, such as <a href="https://www.mlb.com/player/shohei-ohtani-660271">MLB.com player pages</a>, are listed in feet and inches rather than millimeters, MLB simply rounds player height measurements to the nearest inch for public display in keeping with historical precedent. Because of this, there are scenarios in which two players with the same publicly listed height may have ever-so-slightly different-sized strike zones in ABS games, though the difference would be no more than fractions of an inch.</p><h3>System Updates</h3><p>Since the original post, MLB has made some strategic technology changes to the method of communication for ABS calls. In 2019, MLB developed a custom iOS app to deliver ball/strike calls to the umpires via an iOS device attached to the umpire’s belt. As testing continued, a series of issues arose with this system of communication:</p><p>1) Wi-Fi strength is insufficient in the home plate area.<br> 2) Battery life of the device waned as they aged.<br> 3) Devices began to overheat in the hot Florida summers.</p><p>In 2022, Triple-A and FSL began using new communication devices for ABS, replacing the iOS devices previously used. These devices, leveraged around the sports world from the NFL to Formula 1, utilize a wireless protocol known as DECT and proved to be substantially more reliable in consistently delivering the ABS call to the umpire. At the same time, MLB upgraded replay communications equipment by furnishing each MLB umpire on the field with the same wireless communications devices. This enhancement allowed for public address announcements around replay reviews similar to what has historically been commonplace in other sports, and improved the speed of communications between umpires on the field and their counterparts in the Replay Operations Center in New York City. Should ABS come to MLB, it can be seamlessly integrated into the communications workflows that already exist.</p><p>Around the same time as the communication upgrades, MLB also began exploring the idea of the challenge version of ABS outlined above. In addition to the existing functionality, this change now necessitated the added capability to replay the audio of the previous pitch call on demand. To solve this problem, a new interface was developed to allow the playout of challenges at any time based on when a challenge is requested by a player and approved by the umpire. Instead of the previous setup, where a pub/sub topic was used to play audio on <em>every </em>pitch, a message could now be delivered to the communications system to trigger playback of the correct audio file at will.</p><p>The natural progression from this “audio only” state was to introduce a visual component of the challenge playback; a-la tennis in/out challenges or cricket’s Leg-Before-Wicket challenges. To accomplish this MLB has developed a server-based system for generating the challenge graphic video. Using WebGL, a technology enabling high performance graphics in the browser, MLB has been able to provide a high fidelity graphic capable of playback of the most recent pitch trajectory as soon as that pitch is tracked by Statcast. Each pitch, regardless of outcome on the field, is loaded into the graphic with the corresponding ball or strike call in anticipation of a challenge, requiring the on-site ABS Operator to simply initiate playback upon request. A timeline of how a challenge occurs operationally can be seen in the following graphic:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6zD5bXB-_hE3EGnxVmUoFA.png" /></figure><p>To ensure ABS was working smoothly in advance of 2025 Spring Training games, we conducted a test at Steinbrenner Field in Tampa using T-Mobile 5G to ensure reliability of a critical in-game feature through a trusted networking partner.</p><h3>Potential MLB Level Video Workflows</h3><p>Transitioning from Triple-A to MLB presents some new challenges (and opportunities). At the Triple-A level, the broadcast production and scoreboard production are usually one and the same. However, at the MLB level, the ABS challenge graphic needs to be provided to not only the scoreboard, but also multiple broadcasters (e.g. a Giants at Dodgers game on Tuesday night could have NBC Sports Bay Area, SportsNet LA, TBS, and NHK!). The goal is to have the challenge graphic delivered to the scoreboard and broadcasters simultaneously, so that the crowd in the park is heard reacting to the result in-time with the television audience seeing the graphic.</p><p>MLB parks already have a robust broadcast infrastructure, which supports Replay, MLB.tv, Statcast, Pitchcast, Coaching Video, and Dugout iPads. Should ABS come to the MLB level, we are able to leverage this existing infrastructure for feeding scoreboard and broadcasters the challenge graphic. The graphic is provided via a continuous feed to all partner broadcasters and the in-park scoreboard operations, allowing them to take and display the graphic just like any other video feed.</p><h3>Conclusion</h3><p>MLB is excited to introduce ABS to a wider audience of fans and players this Spring Training, and for the bright future of ABS. By using a unique combination of technologies, MLB is continuing to explore and refine the use of technology on the field in this and other areas. As we achieve our technological goals, we eagerly embrace innovations that preserve baseball’s rich history while unlocking new opportunities through state-of-the-art technology.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=574d82c6faa2" width="1" height="1" alt=""><hr><p><a href="https://technology.mlblogs.com/automated-balls-strike-system-abs-2025-spring-training-update-574d82c6faa2">Automated Balls/Strike System (ABS) 2025 Spring Training Update</a> was originally published in <a href="https://technology.mlblogs.com">MLB Technology Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Bulk Update of Fastball labelled pitches: 4-Seam Fastball <-> Sinker]]></title>
            <link>https://technology.mlblogs.com/bulk-update-of-fastball-labelled-pitches-4-seam-fastball-sinker-b7eaeeac7d7f?source=rss----ff0f5fb80706---4</link>
            <guid isPermaLink="false">https://medium.com/p/b7eaeeac7d7f</guid>
            <dc:creator><![CDATA[Tom Tango]]></dc:creator>
            <pubDate>Sun, 23 Feb 2025 15:42:19 GMT</pubDate>
            <atom:updated>2025-02-23T15:42:18.948Z</atom:updated>
            <content:encoded><![CDATA[<p>This week MLB re-classified approximately 25,000 historical pitches as four-seamers or sinkers, reversing the previous classifications between those two pitch types, based on continued upgrades to our pitch tracking technology and resulting analytic discoveries.</p><p>For background, there are two main classifications for a fastball:</p><ul><li>The most popular one is called a four-seam fastball. The behavior of the four-seamer is to give the pitch rise. The pitch is thrown as the name implies, with the fingers cutting across all four seams.</li><li>The other fastball we have classified historically as either a two-seamer or a sinker. As the two terms have been used mostly interchangeably, a few years ago we standardized the term to be sinker. The pitch is thrown with the ball orientation having the two seams under the two fingers. Hence, a two-seam fastball has the effect of a ball that drops more than a four-seam fastball.</li></ul><p>Prior to 2021, we were not able to track the seam orientation. Therefore, for pitchers who threw two fastballs, we classified each fastball based on its behavior. This worked fairly well for the large majority of pitchers and pitches.</p><p>However, following ongoing upgrades to our tracking technology, we are now actually able to directly measure the seam orientation of each pitch, rather than inferring the seam orientation as being two-seam or four-seam. As a result, we can now more accurately classify a fastball as a four-seam or a sinker (aka two-seam).</p><p>One of the most vexing of the two-fastball pitchers is Sean Manaea. Here is his movement chart:</p><p><a href="https://baseballsavant.mlb.com/player-scroll?player_id=640455#pitch_break">https://baseballsavant.mlb.com/player-scroll?player_id=640455#pitch_break</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/374/0*zZZwX9PDJHI0-8a_" /></figure><p>Note the tremendous overlap in the movement of the two fastballs. This is usually a sign that we have some of those pitches misclassified, or that it’s very difficult to classify pitches using movement without access to the seam orientation. After the reclassification for Manaea, since 2021, we will have 1120 previously-labelled 4-seamers swapped to sinkers; and similarly 859 previously-labelled sinkers swapped to 4-seamers.</p><p>Other pitchers like Calvin Faucher, reflect a pure swap: All of his sinkers will now be 4-seamers. If you go to his page, and look at the animation of his fastball, you will see a distinct 4-seam spinning action:</p><p><a href="https://baseballsavant.mlb.com/savant-player/calvin-faucher-676534?stats=statcast-r-pitching-mlb">https://baseballsavant.mlb.com/savant-player/calvin-faucher-676534?stats=statcast-r-pitching-mlb</a></p><p>In all, since 2021, we will have about 25,000 pitches moved from one to the other, with about an even split in both.</p><p>This reclassification will allow us to open up a new door in the future. Some pitchers like Josh Hader throw a distinct 2-seam fastball, but the pitch moves like a 4-seamer. It is in effect a two-seamer that doesn’t sink. Other pitchers throw 4-seamers with no ride. For purposes of labelling the pitch as a 4-seamer or 2-seamer (sinker), the behavior will no longer matter. That said, in a future effort we will have on Baseball Savant, we will be able to tag these pitches with an extra indicator. In this way, if you wanted to separate a Josh Hader 2-seamer from other pitchers who throw 2-seamers that sink, we will be able to do so.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b7eaeeac7d7f" width="1" height="1" alt=""><hr><p><a href="https://technology.mlblogs.com/bulk-update-of-fastball-labelled-pitches-4-seam-fastball-sinker-b7eaeeac7d7f">Bulk Update of Fastball labelled pitches: 4-Seam Fastball &lt;-&gt; Sinker</a> was originally published in <a href="https://technology.mlblogs.com">MLB Technology Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Deploying a Cost-Effective Kubernetes Cluster with NUC PCs and Google Distributed Cloud software…]]></title>
            <link>https://technology.mlblogs.com/deploying-a-cost-effective-kubernetes-cluster-with-nuc-pcs-and-google-distributed-cloud-software-cbb65105cdbc?source=rss----ff0f5fb80706---4</link>
            <guid isPermaLink="false">https://medium.com/p/cbb65105cdbc</guid>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[statcast]]></category>
            <category><![CDATA[google-distributed-cloud]]></category>
            <category><![CDATA[major-league-baseball]]></category>
            <dc:creator><![CDATA[Kevin Backman]]></dc:creator>
            <pubDate>Fri, 01 Nov 2024 18:56:24 GMT</pubDate>
            <atom:updated>2024-11-01T19:33:33.028Z</atom:updated>
            <content:encoded><![CDATA[<h3>Deploying a Cost-Effective Kubernetes Cluster with NUC PCs and Google Distributed Cloud software for bare metal.</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QKTIZ_Hct0M5x1u69aVKIg.jpeg" /><figcaption>Photo Credit: Kevin Backman</figcaption></figure><p>Since 2017 MLB has held the Little League Classic at Journey Bank Ballpark at historic Bowman Field in Williamsport, PA. This game is held with the goal of fostering a connection between professional players and young baseball fans from around the U.S. and across the globe while highlighting MLB’s commitment to supporting youth baseball. In previous years we have used a setup consisting of 3 Dell Poweredge XR11 servers rack mounted in a Pelican case. While powerful, these Dell servers are also expensive, heavy and power hungry. With this in mind we wanted to focus on more efficient and cost effective ways to support the current requirements of this event by piloting a cluster consisting of 8 NUC13 PCs running Google Distributed Cloud software for bare metal.</p><h4><strong>Goals of this project</strong></h4><ul><li>Support several events a year using these flypack kits including the Little League Classic, and other MLB special events such as MLB at Rickwood Field: A Tribute to the Negro Leagues, the London Series, and the Korea Series, among others.</li><li>Reduce hardware shipment to decrease hardware and shipping costs.</li><li>Reduce weight of flypack shipment to decrease required staff necessary to move and install flypack.</li><li>Reduce size of flypack shipment to provide more space in the onsite temporary data centers.</li><li>Reduce hardware requirements to decrease thermal output which is a major risk for these temporary data centers (rack servers produce a lot of heat).</li></ul><h4><strong>Flypack Infrastructure Footprint</strong></h4><p>Current Special Events Flypack Architecture:</p><ul><li>3 Dell Poweredge XR11 1U rack servers</li><li>These clusters support K8 applications such as:<br>- NTP<br>- Pitch Clock (Pendulum)<br>- Statcast Tracking System (ActiveMQ)</li><li>Total cost of the Dell flypack is over $30,000 and each unit has:<br>- A total of 24 cores and 32 GiB of memory.<br>- Max wattage draw per unit is 1120W</li><li>Lightweight Special Events Flypack Architecture:<br>- 8 NUC13anki7 small form factor pc’s<br>- These clusters support the same K8 applications as the full-sized flypack<br>- Each one of these NUC’s have 12 cores (Intel i7) and 32 GiB of memory (expandable to 64GiB of memory)<br>- Cost of building the NUC kit was around 70% cheaper than the old kit.<br>- Max wattage draw per NUC is 120W while still being able to withstand temperatures of up to 40°C.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*OV1c55WUPhXjGNbm" /><figcaption>To illustrate the difference between the two kits, you can see in this photo the new lightweight flypack on the right is substantially smaller than the old kit on the left.</figcaption></figure><h4><strong>Why Google Distributed Cloud Software For Bare Metal?</strong></h4><p>We have been running Google Distributed Cloud software for bare metal in all Major League Ballparks for the last 4 years with very few issues. When trying new architecture we want to eliminate as many unknowns as possible, as well as simplifying configuration of the new cluster and deploying corresponding applications by using existing tools to streamline the process.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*PTC1M4bf6ZoWK2Bi" /></figure><h4><strong>Deployment Process</strong></h4><ul><li>Deploying GDC to the new hardware is done using our existing Admin Workstation. We deploy a Standalone Cluster using the Edge profile detailed <a href="https://cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs/installing/creating-clusters/standalone-cluster-creation">here</a> in the Google documentation. This profile has significantly reduced system resource requirements and is recommended for edge devices with high resource constraints.</li><li>To deploy the apps supporting Player Tracking, Pitch Clock and NTP to the new cluster we use our CI/CD tool Akuity which is powered by ArgoCD. With Akuity we are able to treat this cluster like any other ballpark cluster and deploy the required apps as needed.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ioy0nAkFPmRCuKRS" /><figcaption>ArgoCD is a set of tools for Kubernetes that allow us to run workflows, manage our clusters (regardless of Kubernetes flavor), and implement the GitOps operational framework. This flexibility allows us to easily deploy our applications to new clusters.</figcaption></figure><h4><strong>Results and Benefits</strong></h4><ul><li>The lightweight flypack is able to successfully perform the tasks required of the old 3 rack server flypack while providing more computing resources, using less power and generating less heat. As shown below, the CPU and Memory requested by the applications still left more than 70% of resources free and with the increased number of nodes resiliency against hardware failures is much higher.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*jDzJQ6fnpj1QHuzV" /><figcaption>If you look at the CPU/Memory requested vs CPU/Memory allocable you can see how much headroom is left in the cluster for additional application deployments</figcaption></figure><h4>Conclusion</h4><ul><li>Overall we consider this pilot a resounding success and will look further into using this architectural setup where possible.</li><li>When looking at a single NUC there are a lot of single points of failure. However, these units are cheap enough that you can deploy large numbers in a smaller form factor and better redundancy than what one gets with a traditional rack server configuration.</li><li>Future plans or potential improvements for the new architecture.</li><li>With the success of this pilot, we believe that these clusters can be deployed to our MiLB ballparks with BMC enabled NUCs for better network resiliency.</li></ul><h4><strong>Additional Resources</strong></h4><p>If you’re curious about Akuity or GDC you can find the links with more info about them here:</p><ul><li><a href="https://akuity.io/">https://akuity.io/</a></li><li><a href="https://cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs">https://cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cbb65105cdbc" width="1" height="1" alt=""><hr><p><a href="https://technology.mlblogs.com/deploying-a-cost-effective-kubernetes-cluster-with-nuc-pcs-and-google-distributed-cloud-software-cbb65105cdbc">Deploying a Cost-Effective Kubernetes Cluster with NUC PCs and Google Distributed Cloud software…</a> was originally published in <a href="https://technology.mlblogs.com">MLB Technology Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Baseball Public Dataset]]></title>
            <link>https://technology.mlblogs.com/baseball-public-dataset-51c3aeb89322?source=rss----ff0f5fb80706---4</link>
            <guid isPermaLink="false">https://medium.com/p/51c3aeb89322</guid>
            <dc:creator><![CDATA[Tom Tango]]></dc:creator>
            <pubDate>Tue, 08 Oct 2024 19:44:19 GMT</pubDate>
            <atom:updated>2024-10-08T19:44:19.932Z</atom:updated>
            <content:encoded><![CDATA[<p>The Retrosheet event files contain extensive play-by-play data for most of MLB history. This data is now available in database form in BigQuery. All that is required of the user is to create a Google Account to access the data.</p><p>You can access the data via <a href="https://console.cloud.google.com/bigquery/analytics-hub/">Google Analytics Hub</a></p><p><em>Search the Listings for: Baseball</em></p><p>If you want a direct link, <a href="https://console.cloud.google.com/bigquery/analytics-hub/discovery/projects/257726328933/locations/us/dataExchanges/baseball_public_1922573127b/listings/baseball_public_dataset_192257d9e1f">here you go</a>:</p><p>What do we envision as the next step here? Ideally, the community will create a JSON version. You can easily see the hierarchy here, with games, plate appearances, pitches, and runner movement providing that structure. BigQuery is ready-made for JSON, which of course will make life really easy for Python coders too.</p><p>There’s nothing MLB-specific in this dataset, and there are no plans for that.</p><p><strong>The disclosures:</strong></p><p>The information used here was obtained free of charge from and is copyrighted by Retrosheet. Interested parties may contact Retrosheet at <a href="http://www.retrosheet.org./">www.retrosheet.org.</a></p><p>Data provided as a courtesy through MLB’s partnership with Google Cloud. Neither Google nor MLB has performed any validation of this data. Any issues with the data should be sent directly to the data owner, Retrosheet LLC</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=51c3aeb89322" width="1" height="1" alt=""><hr><p><a href="https://technology.mlblogs.com/baseball-public-dataset-51c3aeb89322">Baseball Public Dataset</a> was originally published in <a href="https://technology.mlblogs.com">MLB Technology Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Introducing Statcast 2023: High Frame Rate Bat and Biomechanics Tracking]]></title>
            <link>https://technology.mlblogs.com/introducing-statcast-2023-high-frame-rate-bat-and-biomechanics-tracking-3844890264a6?source=rss----ff0f5fb80706---4</link>
            <guid isPermaLink="false">https://medium.com/p/3844890264a6</guid>
            <category><![CDATA[computer-vision]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[baseball]]></category>
            <category><![CDATA[statcast]]></category>
            <category><![CDATA[biomechanics]]></category>
            <dc:creator><![CDATA[Graham Goldbeck]]></dc:creator>
            <pubDate>Mon, 27 Mar 2023 19:57:06 GMT</pubDate>
            <atom:updated>2023-03-27T19:57:06.031Z</atom:updated>
            <content:encoded><![CDATA[<p>On Opening Day 2023, MLB will introduce its upgraded Statcast platform, featuring the replacement of five of the 12 cameras in the <a href="https://technology.mlblogs.com/introducing-statcast-2020-hawk-eye-and-google-cloud-a5f5c20321b8">2020 Hawk-Eye Statcast system</a> with high frame rate cameras. This opens the door for MLB to provide fans with new bat and biomechanics tracking output, enhancing the ability to tell better stories about the game of baseball and how players are as good as they are.</p><h3>Bat Tracking</h3><p>In the 2000s, MLB used technologies like QuesTec and PITCHf/x to improve our understanding of pitching, not to mention catcher framing. In 2015, Statcast showed us the importance of pitch spin but also introduced us to hit and player tracking and created metrics such as Sprint Speed and Exit Velocity. In 2020, new data showed us how spin axis and seam orientation make certain pitches more effective.</p><p>Given all we have learned from ball and player tracking, bat tracking can be thought of as the next level to drill down in terms of understanding batter (and pitcher) performance in baseball. Bat tracking provides context around the exit velocity and launch angle datasets that have made such a large impact on the game in recent years. A ball might have an exit velocity 100 mph, but was it because the batter’s bat speed was 75 mph and they happened to just hit it right on the sweet spot, or was it closer to 90 and they happened to hit it off the end of the bat? Which batter is more likely to hit the ball hard again in the future? Who should the manager choose to pinch hit against the flamethrower with the high spin fastball up in the zone, one with an uppercut attack angle or a batter with a flatter swing?</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fwww.youtube.com%2Fembed%2FWmOHro4UBVM%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fwww.youtube.com%2Fwatch%3Fv%3DWmOHro4UBVM&amp;image=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fi.ytimg.com%2Fvi%2FWmOHro4UBVM%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/e149cee37b29f6f446bbf56ba0f0eb33/href">https://medium.com/media/e149cee37b29f6f446bbf56ba0f0eb33/href</a></iframe><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fwww.youtube.com%2Fembed%2FhXhoiUu-Ugk%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fwww.youtube.com%2Fwatch%3Fv%3DhXhoiUu-Ugk&amp;image=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fi.ytimg.com%2Fvi%2FhXhoiUu-Ugk%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/9595af4ec332fe42fe5d7e61f7ede1ce/href">https://medium.com/media/9595af4ec332fe42fe5d7e61f7ede1ce/href</a></iframe><p>These are the types of questions MLB clubs are asking themselves: 17 clubs had installed Hawk-Eye high frame rate cameras on their own by 2022. We hope to bring that insight to fans starting this season, initially with metrics like bat speed, attack angle, and contact point.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fwww.youtube.com%2Fembed%2FUUMxfpQ_7Qo%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fwww.youtube.com%2Fwatch%3Fv%3DUUMxfpQ_7Qo&amp;image=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fi.ytimg.com%2Fvi%2FUUMxfpQ_7Qo%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/ab07c470f4a92044b71c24d1d4adbc39/href">https://medium.com/media/ab07c470f4a92044b71c24d1d4adbc39/href</a></iframe><p>While the focus so far has been on the bat and ball contact interaction, many stories can also be told about swings and misses as well. Currently, the only data point we have is that there was a swinging strike on a pitch, but was that because the batter correctly identified the fastball and was just a few milliseconds late on the swing and missed contacting the ball by half an inch, or did they get a curveball when they expected a fastball and swung and missed by a foot? What if we imposed the pitcher’s average fastball trajectory instead of the curveball they actually threw, would the pitch and bat paths intersect? Bat data can also be used for pitcher evaluation too, who gets batters to swing under their four seam fastballs the most, or who fools batters the most?</p><h3>System Overview</h3><p>As a refresher, the Statcast system installed in 2020 utilizes a total of 12 cameras. Five cameras operated at 100 frames per second (fps) and were primarily dedicated to pitch and hit tracking, while an additional seven cameras were focused on tracking players and batted balls at 50 fps. As part of the upgrades for 2023, the five 100 fps cameras have been replaced with 300 fps cameras, which are circled in red below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/607/0*955BAI0G9X4kQcRX" /></figure><h3>Why Upgrade Cameras?</h3><p>The 2020–2022 Hawk-Eye Statcast system has brought a wave of new insights to fans, including data points such as <a href="https://baseballsavant.mlb.com/leaderboard/arm-strength">fielder arm strength</a>, <a href="https://baseballsavant.mlb.com/leaderboard/spin-direction-pitches">3D spin axis</a>, and seam orientation’s contribution to <a href="https://www.drivelinebaseball.com/2020/11/more-than-what-it-seams-an-introduction-to-seam-shifted-wakes-and-their-effect-on-sinkers/">seam shifted wake</a>. It also opened up the possibility of FieldVision and <a href="https://twitter.com/i/status/1586879877529178112">bringing fans a view of the game</a> only possible in virtual space. But even cameras running at 100 frames per second run into limitations when tracking incredibly fast moving objects, such as the head of a baseball bat mid swing. While 10 milliseconds (time between frames at 100 fps) may seem short, the head of a bat passing over the plate can sometimes move more than 15 inches in that timeframe, nearly enough distance to cover the entire plate between frames. With 300 frames per second video though, the cameras are able to capture an accurate swing path trajectory and contact point when paired with the hit and pitch trajectory.</p><p>In addition to the bat data, the high frame rate cameras allow for the future possibility of biomechanical analysis of pitchers and hitters and help to explain how baseball players can perform the incredible feats they do. Additionally the high frame rate cameras should provide some improved spin tracking, since there are now more frames closer to pitch release and contact point.</p><h3>Where to Find New Data</h3><p>You may already have seen some of this data featured during last year’s Home Run Derby or on Astros and Dodgers home broadcasts as the technology was being piloted in 2022. You can expect to continue to see bat tracking data featured on broadcasts across the league this year, as well as other platforms like Baseball Savant, Gameday, and other MLB properties in the months and years to come!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3844890264a6" width="1" height="1" alt=""><hr><p><a href="https://technology.mlblogs.com/introducing-statcast-2023-high-frame-rate-bat-and-biomechanics-tracking-3844890264a6">Introducing Statcast 2023: High Frame Rate Bat and Biomechanics Tracking</a> was originally published in <a href="https://technology.mlblogs.com">MLB Technology Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Weather Applied Metrics in Major League Baseball]]></title>
            <link>https://technology.mlblogs.com/weather-applied-metrics-in-major-league-baseball-aa0e556eb49f?source=rss----ff0f5fb80706---4</link>
            <guid isPermaLink="false">https://medium.com/p/aa0e556eb49f</guid>
            <category><![CDATA[meteorology]]></category>
            <category><![CDATA[statcast]]></category>
            <category><![CDATA[ballparks]]></category>
            <category><![CDATA[baseball]]></category>
            <category><![CDATA[weather]]></category>
            <dc:creator><![CDATA[Clay Nunnally]]></dc:creator>
            <pubDate>Mon, 06 Mar 2023 20:21:17 GMT</pubDate>
            <atom:updated>2023-03-06T20:21:17.130Z</atom:updated>
            <content:encoded><![CDATA[<p>On May 28th, 2021 Cubs outfielder Patrick Wisdom crushed a Sean Doolittle fastball that was destined for the left field bleachers. If you heard the announcers’ or the crowd’s live reaction, you didn’t need Statcast to tell you that balls that leave the bat at 103 mph with a 34 degree launch angle typically travel 400 feet. Cubs’ broadcaster Jon Sciambi went into his home run call: “That one drilled, left field! This one back and…”. Meanwhile, the Reds’ broadcast had a more succinct immediate reaction: “Oh boy.” Fans in the 12th row of the bleachers jumped to their feet and stretched out their arms to make a play on the ball.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fwww.youtube.com%2Fembed%2FsfUO2N7qlQg%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fwww.youtube.com%2Fwatch%3Fv%3DsfUO2N7qlQg&amp;image=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fi.ytimg.com%2Fvi%2FsfUO2N7qlQg%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/8e3b494b00ef994ea592b668092643bb/href">https://medium.com/media/8e3b494b00ef994ea592b668092643bb/href</a></iframe><p>Then, suddenly, the mood shifted. Sciambi continued, “Oh man”, having realized that Reds’ left fielder Jesse Winker was settling under the ball, not in the bleachers but just one step onto the warning track. Reds play-by-play broadcaster Tommy Thrall elaborated, “The wind is going to help out the Reds here. That ball was crushed!”</p><p>Thrall correctly pinpointed the culprit: the wind, or the weather more broadly. Sensors and advanced modeling from Weather Applied Metrics (WAM) tell us that the wind cost Wisdom 40 feet on the would-be home run on this blustery day at Wrigley Field in 2021.</p><p>After testing the technology at five parks in 2022, MLB and WAM will be providing weather data and insights across all 30 ballparks in 2023.</p><h3>Wind &amp; Park Effects</h3><p>Park sizes, orientations, shape, and weather can be considered as a group of factors in concept as a <a href="https://baseballsavant.mlb.com/leaderboard/statcast-park-factors">park factor</a>. The park-factor ingredient that has historically been most difficult to characterize is wind.</p><p>Wind at a ballpark has to negotiate large structures that surround the field. As a result, wind on-field doesn’t match what the flags are showing. The figure below illustrates how a large barrier creates a wind shadow. The height (X) and range (Y) of the wind shadow are roughly on the order of magnitude of the height of the barrier.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*bJTILuNMw5bX05lV" /><figcaption>The barrier effect and the wind shadow created over a baseball field.</figcaption></figure><p>In practice, that means, especially for large ballparks, the wind that affects ball trajectories doesn’t necessarily match the prevailing wind conditions. As a result, pitchers don’t often pitch into significant headwinds and hits can be pushed in different directions depending on where they are hit and the shape of the stadium.</p><h3>Modeling Weather Impact</h3><p>Starting in 2023 MLB will be using <a href="https://www.weatherapplied.com/">Weather Applied Metrics</a> (WAM) to quantify the impact of wind and weather effects at ballparks league wide. To do this accurately, WAM needs a higher resolution of wind and weather data than is practical using discrete weather sensors. They accomplish this using Computational Fluid Dynamic (CFD) modeling. The WAM <a href="https://www.weatherapplied.com/tech.html#howitworks">CFD modeling process</a> uses <a href="https://technology.mlblogs.com/a-lidar-validated-trajectory-for-a-long-home-run-f9c716206c8b">accurate 3D geospatial data</a> to calculate a wind-field solution which uniquely matches the park’s boundary conditions and the prevailing weather conditions measured over the stadium. WAM exhaustively computes CFD solution sets in advance so the WAM workflow can react in real time to changes in weather conditions. The image below shows a vector wind-field solution for a ballpark where swirling winds change magnitude and direction depending on location inside the park.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*7BQWANXyIpmMwwKt" /><figcaption>One view of a ballpark wind-field solution calculated by Weather Applied Metrics.</figcaption></figure><p>Along with the wind solution from their CFD model, Weather Applied Metrics uses Statcast tracking data to calculate the deviation for a hit trajectory compared to an identical hit launched in calm conditions or in an indoor environment. The “calm” deviation isolates only the wind impact on the hit, but the “indoor” deviation adjusts for all weather conditions including the elevation of the ballpark. A “calm” trajectory deviation is shown below for Matt Olson’s grand slam from April 20th, 2021.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1005/0*MEbHdmcRjPSYkk5_" /><figcaption>Wind deflection calculated for Matt Olson’s grand slam on from April 20, 2021.</figcaption></figure><p>In a similar way, the weather impact on pitch location can be calculated. Pitch deviations are typically much smaller in magnitude than hit deviations because trajectories are shorter and closer to the ground. However, in parks with consistent wind trends, pitch velocity and movement can experience slight biases related to ballpark, season, and time of day.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*XizvHeCJ5gVCSZcL" /><figcaption>Pitch trajectory deviations in 3 dimensions compared to calm or indoor conditions</figcaption></figure><p>In addition to calculating weather impact on past events, WAM can also forecast weather impact for games yet to be played. WAM forecasts use a generic home run trajectory to calculate the general expected deviation given the forecasted weather conditions for a given game. For example, the graphic below shows the change in distance (number) and deviation direction (arrow) for a generic home run trajectory.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*zdbgYJhTvgACtfWw" /><figcaption>The change in distance (number) and deviation direction (arrow) for a generic home run trajectory.</figcaption></figure><p>A collective data set from WAM can help answer the following questions:</p><ul><li>Which parks have significant wind bias that affects pitch dynamics?</li><li>In what parks does wind prevent the most home runs?</li><li>Which players would benefit the most by moving to a neutral venue?</li></ul><p>Given that WAM technology has been installed in a subset of MLB parks, we can answer another interesting question: How much <em>can </em>wind affect a hit trajectory?</p><h3>Wrigley Field and Pat Wisdom May 28, 2021</h3><p>Wrigley Field has interesting wind effects due to its proximity to Lake Michigan, but also because it’s an older, smaller park surrounded by structures that condense flow. Wrigley wind was evident on this Patrick Wisdom fly-out that was easily caught by the left fielder. Jon Sciambi’s play-by-play call reveals the story. This pitch was struck well and started on what seemed to be a home run trajectory.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fwww.youtube.com%2Fembed%2Fo-kvnEmmhGw%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fwww.youtube.com%2Fwatch%3Fv%3Do-kvnEmmhGw&amp;image=https%3A%2F%2Ffd.xuwubk.eu.org%3A443%2Fhttps%2Fi.ytimg.com%2Fvi%2Fo-kvnEmmhGw%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/cce5271be5da313dde8d774a3514a91d/href">https://medium.com/media/cce5271be5da313dde8d774a3514a91d/href</a></iframe><p>Statcast measured an exit speed of 103 mph and a launch angle of 34 degrees. Off the bat, Sciambi and Wisdom may have thought this was a home run. But this hit died on the warning track. Luckily, in this case we had WAM data and we can see how much wind affected the distance.</p><p>To give some context, we can check Statcast hit data on balls hit this hard, at this launch angle. The histogram below shows the wide variety of distance outcomes that have been tracked since 2020. You can also see this Pat Wisdom flyout has the second smallest distance of all such hits. Most balls hit this hard have traveled around 400ft, but this hit only made it to 338 ft.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*8oWIduGaaOnoFEMB" /><figcaption>Hit distance histogram for all batted balls with a 103-mph exit speed and a 34 degree launch angle since 2020.</figcaption></figure><p>What causes this hit to be a distance outlier is the wind impact. WAM used the output of the CFD wind model to calculate the wind impact throughout the trajectory. The result was found to be a 40-ft reduction in distance compared to the same hit in calm conditions. The difference in landing locations is shown below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*bNwDHH62Mxke7dwf" /><figcaption>Weather Applied Metrics change in landing location due to wind.</figcaption></figure><p>How much can wind affect a trajectory? As this example shows, wind can knock a hit back 40 ft. In this case it was the difference between a home run 9 rows back and an easy F7 on the warning track.</p><p>In cases where the hit apex is extremely high, WAM analysis has shown that the change in landing location due to wind can approach 100 ft in the most severe conditions. When weather becomes part of the story of baseball, insights from Weather Applied Metrics can be found on your team’s broadcast during the 2023 MLB season.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aa0e556eb49f" width="1" height="1" alt=""><hr><p><a href="https://technology.mlblogs.com/weather-applied-metrics-in-major-league-baseball-aa0e556eb49f">Weather Applied Metrics in Major League Baseball</a> was originally published in <a href="https://technology.mlblogs.com">MLB Technology Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Anthos on Bare Metal, how MLB Turned Ball Parks into On Prem K8 Clusters]]></title>
            <link>https://technology.mlblogs.com/anthos-on-bare-metal-how-mlb-turned-ball-parks-into-on-prem-k8-clusters-295031253a33?source=rss----ff0f5fb80706---4</link>
            <guid isPermaLink="false">https://medium.com/p/295031253a33</guid>
            <category><![CDATA[anthos]]></category>
            <category><![CDATA[anthos-on-prem]]></category>
            <category><![CDATA[major-league-baseball]]></category>
            <category><![CDATA[mlb]]></category>
            <dc:creator><![CDATA[Kevin Backman]]></dc:creator>
            <pubDate>Fri, 03 Dec 2021 20:15:43 GMT</pubDate>
            <atom:updated>2021-12-03T20:15:42.985Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>Anthos on Bare Metal: Delivering Baseball Stats on K8S in our MLB Ballparks</strong></h3><p><em>By Kevin Backman</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*U0yZPTSS2RCvWYr8HJMRgw.jpeg" /><figcaption>Credit: <a href="https://cloud.google.com/blog/topics/anthos/how-mlb-is-using-anthos">Google</a></figcaption></figure><h3>Background</h3><p>At the end of the 2019 baseball season, MLB began a hardware refresh across all our major league ballparks providing us with an opportunity to rebuild our server infrastructure using a DevOps approach.</p><h3>Goals of the Refresh</h3><ul><li>Given the patchwork nature of our environment, we wanted to simplify our infrastructure management into a single pane of glass. It was important for us to build, manage, and support 30 MLB ballparks easily from day one, and allow us to expand the approach in a repeatable fashion when the time came to refresh our minor league ballparks.</li><li>Streamline app deployments so that we could consistently and reliably deploy code using the Continuous Integration and Continuous Delivery tools currently in place.</li><li>Reduce costs and complexity by removing virtualization and running everything on Anthos bare metal, which is Googles solution for Anthos running on physical servers.</li><li>Create a resilient ballpark cluster that could handle hardware failures. In the past if a server went down, all VMs running on that server went down with it. This meant that we couldn’t fully utilize our hardware as we needed to run a minimum of two VMs for each app.</li><li>Deliver the solution within a tight schedule for the start of the 2021 MLB season with resource constraints compounded by COVID-19.</li></ul><p>Our partnership with Google provided us with a solution for having all our clusters in one place, the GCP Console. Kubernetes and Anthos on bare metal provided us with a way to do this.</p><h3>Our Ballpark Infrastructure Server Footprint</h3><p>All of our MLB ballparks have limited data center space and for each, we have a condensed and optimized footprint of four servers. These four servers are responsible for running applications that are critical to the game of baseball including our <a href="https://technology.mlblogs.com/introducing-statcast-2020-hawk-eye-and-google-cloud-a5f5c20321b8">Statcast</a> and Hawk-Eye solutions.</p><h3>Why GKE OnPrem?</h3><p>We have limited space in the ballpark datacenters and kubernetes provides us with the flexibility we need.</p><p>In the past we were running a traditional VMware (and RHEV-M in some parks) stack on physical machines. While this met our needs, we were over provisioning each VM with 8 vCPU’s and 32G of memory to run an application that only required a fraction of these resources. Using GKE was an obvious choice, allowing us to allocate the appropriate CPU and Memory resources at the pod level (e.g. cpu: 500m and memory: 1Gi). GKE also allowed us to run multiple pods across our nodes, giving us resiliency and flexibility when performing maintenance. Working with Google provided us with a wealth of knowledge that we tapped into to deploy Anthos across our parks and gave us that single pane of glass we were looking for.</p><p>With our Ballpark clusters now living in the GCP environment, we can manage them the same way we manage our GKE clusters in the Cloud and utilize our CICD tooling already in place to deploy our applications.</p><h3>How did we do it?</h3><h4>Short answer:</h4><p>Through automation and a lot of hard work.</p><h4>More detailed answer:</h4><p>Deploying new technology to 30 data centers is no easy task, without the use of automation tools and the hard work of people physically in the ballparks.</p><p>Before we could even proceed we relied on our very capable Ballpark Infrastructure and Ballpark Ops teams to actually do the hardware installation. Once that was complete (through no small feat on their part due to the pandemic) we were able to get to work on our portion of building out the new and improved infrastructure.</p><p>This was achieved by setting up and using a hardware orchestration tool called Mojo which is a software suite produced by Metify that utilizes Redfish to interact with the bare metal servers and PXE boot them. One Mojo instance can manage hundreds of bare metal servers, greatly reducing the level of effort required to manage said servers. Mojo also allows us to manage BIOS and firmware of the servers so now we are able to keep pace with patches and security hotfixes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HIeXO88K2MooBdOp6B02Zg.png" /></figure><p>With the hardware and operating system ready to go, we switch to deploying Anthos to all the servers. This process is well documented by Google; however, instead of deploying an Anthos workstation for each cluster at each ballpark, we deployed our workstations in a GCP project. This reduced and simplified managing the environment, where instead of 30 workstations across our parks, we have 3 running in the Cloud (a workstation can only deploy 1 cluster at a time).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6x6fkgf8TzacoC6EWy-MvQ.png" /></figure><p>By running Anthos on bare metal without VMWare, we have a much cleaner and simpler environment to manage. The cost savings alone are enough to make this rebuild worth the effort.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BnMR1rIzTzltVWX2RS-yIQ.png" /></figure><p>With the Anthos clusters online, it was a simple task of plugging them into our existing ArgoCD environment. Adding continuous deployments to the ballparks allows us to build a resilient environment better suited to handling hardware failures, a challenge we had when running VMWare and previous virtualization technologies.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*x8gwBYdQS59iLtgS0V2A5w.png" /><figcaption>Credit: Graeme Tower</figcaption></figure><p>The GKE solution also simplified how we handled persistent and shared storage. Using Rook Ceph within our clusters in each park allowed us to eliminate the need for a physical NAS solution and further reduce our datacenter footprint, complexities and cost.</p><h3>What did we learn?</h3><p>We learned that with automation tools, kubernetes, continuous deployment and an infrastructure as code approach, we can deploy and manage resilient systems that can be trusted to run our critical applications in the ballparks. This puts us in a position to be able to stay on the cutting edge of technology as it relates to baseball. We took what was a six hour process from start to finish and reduced it to less than an hour. In past hardware refreshes it would take a minimum of six hours for a single engineer to build the old VMware environment from start to finish. That now takes forty minutes for one park, and multiple parks can be worked on at the same time. This is a massive reduction in engineering hours leaving engineers more time to work on future improvements and projects.</p><h3>Final notes.</h3><p>As with anything, it takes a team to accomplish a task of this magnitude. With that, they are:</p><ul><li>Jennifer Ahn</li><li>Michael Durkin</li><li>Felipe Negron</li><li>Graeme Tower</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=295031253a33" width="1" height="1" alt=""><hr><p><a href="https://technology.mlblogs.com/anthos-on-bare-metal-how-mlb-turned-ball-parks-into-on-prem-k8-clusters-295031253a33">Anthos on Bare Metal, how MLB Turned Ball Parks into On Prem K8 Clusters</a> was originally published in <a href="https://technology.mlblogs.com">MLB Technology Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>