A sports analyst exists to convert the noise of competition into a decision a\ncoach, scout, or front office will actually make. Games throw off enormous\namounts of data and almost all of it is irrelevant to the next choice that has\nto be made — who to play, who to sign, how to defend a corner kick, when to pull\na starter. The analyst's reason for being is to find the small signal inside\nthat noise, attach an honest level of confidence to it, and hand it over in a\nform a coaching staff can act on under time pressure. The job is judgment about\nevidence, not the production of charts.
\n","wordCount":111},{"heading":"Core Mission","id":"core-mission","markdown":"Answer the question the coach actually needs answered — separating skill from\nluck and signal from noise well enough to change one real decision, and saying\nhow sure you are while you do it.","html":"Answer the question the coach actually needs answered — separating skill from\nluck and signal from noise well enough to change one real decision, and saying\nhow sure you are while you do it.
\n","wordCount":33},{"heading":"Primary Responsibilities","id":"primary-responsibilities","markdown":"The visible work is models, dashboards, and clips; the actual work is framing\nquestions and managing uncertainty. An analyst spends their time turning a vague\nbrief (\"are we creating enough?\") into an answerable one (\"are our open-play\nchances worth more xG per possession than last season, controlling for game\nstate?\"); building and maintaining metrics from event data and tracking data;\nscouting opponents for exploitable tendencies; evaluating players for\nrecruitment against role and budget; preparing pre-match and post-match reports;\ntagging video so a number points back to something a coach can watch; and, on\nthe broadcast or media side, explaining all of this to an audience that has\nnever heard of regression to the mean. Underneath everything sits the same\ndiscipline: deciding what a number means in *this* context before reporting it.","html":"The visible work is models, dashboards, and clips; the actual work is framing\nquestions and managing uncertainty. An analyst spends their time turning a vague\nbrief ("are we creating enough?") into an answerable one ("are our open-play\nchances worth more xG per possession than last season, controlling for game\nstate?"); building and maintaining metrics from event data and tracking data;\nscouting opponents for exploitable tendencies; evaluating players for\nrecruitment against role and budget; preparing pre-match and post-match reports;\ntagging video so a number points back to something a coach can watch; and, on\nthe broadcast or media side, explaining all of this to an audience that has\nnever heard of regression to the mean. Underneath everything sits the same\ndiscipline: deciding what a number means in this context before reporting it.
\n","wordCount":134},{"heading":"Guiding Principles","id":"guiding-principles","markdown":"- **Describe, then predict — never confuse the two.** What happened and what\n will happen are different questions. A player who shot 60% from three over ten\n games described a hot streak; he did not predict one.\n- **One actionable insight beats ten true ones.** A coach can act on a single\n clear instruction before a match. A report with forty findings gets ignored\n entirely. Pick the one that changes a decision.\n- **Sample size is the first question, not the last.** Before interpreting any\n rate, ask whether there's enough of it to mean anything. Most \"trends\" are\n small samples regressing in plain sight.\n- **A metric that drives no decision is a vanity metric.** If knowing the number\n wouldn't change selection, tactics, or recruitment, don't report it.\n- **Context is part of the number.** A 55% true-shooting wing and a 55%\n true-shooting center are not equally good shooters; role, usage, and scheme\n change what the same figure means.\n- **Communicate the uncertainty, not just the point estimate.** \"About a goal\n better, but I'm not confident\" is more useful and more honest than a false\n decimal.\n- **Watch the games.** Numbers without film drift into nonsense; film without\n numbers drifts into bias. You need both eyes open.","html":"The analyst is a translator standing between the data and the people who act on\nit. With the head coach and assistants, the work is framing and brevity — find\nthe question, deliver one usable answer in their language and their time window.\nWith sports scientists and athletic trainers, they share load and injury data\nand argue over what's signal. With recruitment and scouting, they pair models\nwith human eyes; the best calls come from the two agreeing or from understanding\nexactly why they don't. On the technical side they lean on data engineers to\nkeep pipelines clean and data scientists for heavier modeling. On broadcast,\nthe partner is the producer and the audience, and the craft is making regression\nto the mean sound like common sense. The recurring friction is trust: a number\nthat contradicts a coach's eyes is rejected unless it comes with film and\nhumility.
\n","wordCount":147},{"heading":"Ethics","id":"ethics","markdown":"Analysts increasingly hold sway over who gets paid, played, and cut, which makes\nhonest uncertainty a duty rather than a style choice. Core obligations: never\noverstate confidence to win an argument; disclose what the model ignores; resist\npressure to manufacture a number that justifies a decision already made; protect\nathletes' private biometric and medical data, which can end careers if leaked or\nmisused; and be careful that a metric doesn't entrench a bias — penalizing a\nplaying style, body type, or background the model was never built to judge. In\nbroadcast, the duty is to the audience's understanding, not to the cleanest\nstory; reporting a fluke as destiny is a small lie that compounds. The quiet\npower here is that people believe numbers more than they should, so the person\nproducing them owes them extra care.","html":"Analysts increasingly hold sway over who gets paid, played, and cut, which makes\nhonest uncertainty a duty rather than a style choice. Core obligations: never\noverstate confidence to win an argument; disclose what the model ignores; resist\npressure to manufacture a number that justifies a decision already made; protect\nathletes' private biometric and medical data, which can end careers if leaked or\nmisused; and be careful that a metric doesn't entrench a bias — penalizing a\nplaying style, body type, or background the model was never built to judge. In\nbroadcast, the duty is to the audience's understanding, not to the cleanest\nstory; reporting a fluke as destiny is a small lie that compounds. The quiet\npower here is that people believe numbers more than they should, so the person\nproducing them owes them extra care.
\n","wordCount":135},{"heading":"Scenarios","id":"scenarios","markdown":"**The team that \"can't finish.\"** The head coach, after three 1–0 losses, wants\nto drill shooting. The analyst reframes: are we creating good chances and missing\nthem, or creating bad chances? The numbers show 2.1 xG per game against 0.7\ngoals — chance quality is fine; the conversion is a small-sample cold streak that\nwill regress. Reporting that alone would feel like excuse-making, so the analyst\ngoes to the tape and finds the one real issue: the chances are coming from wide,\nlow-value cutbacks rather than central positions. The deliverable is one\ninstruction — get a runner to the penalty spot on cutbacks — plus the calm\nmessage that the finishing will return. The team isn't broken; the coach was\nabout to fix the wrong thing.\n\n**Scouting the closer.** A baseball staff wants to game-plan a relief pitcher.\nThe average line says nothing. The analyst slices by leverage and count: when\nahead, he throws the slider 70% of the time low-and-away; when behind, he\nabandons it for fastballs middle. The actionable insight is a single hitting cue\n— sit fastball when the count favors you, lay off the low slider when it doesn't.\nThe analyst states the sample (forty-one plate appearances in that split) and the\nconfidence honestly, because a tendency on a small sample can vanish, but it's\nenough to shift the approach.\n\n**The recruitment red flag.** Scouting hands up a winger with a gaudy goal tally\nin a weaker league; the model and the eye disagree. The analyst checks the base\nrate and the underlying numbers: the goals far outrun the xG, meaning he's\nfinished above expectation — a number that regresses hard, especially stepping up\nin competition. League-adjusting his output drops him from a star to a rotation\npiece. The recommendation isn't \"don't sign him\" but \"don't pay for the goals;\npay for the chance creation, which is real and translates, and expect the tally\nto fall.\" That distinction is the difference between a smart signing and an\nexpensive one.","html":"The team that "can't finish." The head coach, after three 1–0 losses, wants\nto drill shooting. The analyst reframes: are we creating good chances and missing\nthem, or creating bad chances? The numbers show 2.1 xG per game against 0.7\ngoals — chance quality is fine; the conversion is a small-sample cold streak that\nwill regress. Reporting that alone would feel like excuse-making, so the analyst\ngoes to the tape and finds the one real issue: the chances are coming from wide,\nlow-value cutbacks rather than central positions. The deliverable is one\ninstruction — get a runner to the penalty spot on cutbacks — plus the calm\nmessage that the finishing will return. The team isn't broken; the coach was\nabout to fix the wrong thing.
\nScouting the closer. A baseball staff wants to game-plan a relief pitcher.\nThe average line says nothing. The analyst slices by leverage and count: when\nahead, he throws the slider 70% of the time low-and-away; when behind, he\nabandons it for fastballs middle. The actionable insight is a single hitting cue\n— sit fastball when the count favors you, lay off the low slider when it doesn't.\nThe analyst states the sample (forty-one plate appearances in that split) and the\nconfidence honestly, because a tendency on a small sample can vanish, but it's\nenough to shift the approach.
\nThe recruitment red flag. Scouting hands up a winger with a gaudy goal tally\nin a weaker league; the model and the eye disagree. The analyst checks the base\nrate and the underlying numbers: the goals far outrun the xG, meaning he's\nfinished above expectation — a number that regresses hard, especially stepping up\nin competition. League-adjusting his output drops him from a star to a rotation\npiece. The recommendation isn't "don't sign him" but "don't pay for the goals;\npay for the chance creation, which is real and translates, and expect the tally\nto fall." That distinction is the difference between a smart signing and an\nexpensive one.
\n","wordCount":339},{"heading":"Related Occupations","id":"related-occupations","markdown":"A sports analyst shares the distributional, uncertainty-first thinking of a data\nscientist but applies it under a coach's time pressure to a single decision. The\ncoach owns the call; the analyst sharpens the evidence behind it. The athlete is\nthe subject the numbers describe and, increasingly, a consumer of them. Data\nengineers keep the pipelines that the analysis depends on. On the media side, the\nbroadcast journalist turns the same numbers into a public story, with the analyst\nguarding against the fluke-as-destiny temptation. The financial analyst is a\nsurprisingly close cousin: both separate signal from noise, fight regression to\nthe mean, and price uncertainty for a decision-maker.","html":"A sports analyst shares the distributional, uncertainty-first thinking of a data\nscientist but applies it under a coach's time pressure to a single decision. The\ncoach owns the call; the analyst sharpens the evidence behind it. The athlete is\nthe subject the numbers describe and, increasingly, a consumer of them. Data\nengineers keep the pipelines that the analysis depends on. On the media side, the\nbroadcast journalist turns the same numbers into a public story, with the analyst\nguarding against the fluke-as-destiny temptation. The financial analyst is a\nsurprisingly close cousin: both separate signal from noise, fight regression to\nthe mean, and price uncertainty for a decision-maker.
\n","wordCount":111},{"heading":"References","id":"references","markdown":"- *Moneyball* — Michael Lewis\n- *The Numbers Game* — Chris Anderson & David Sally\n- *Soccermatics* — David Sumpter\n- *Basketball on Paper* — Dean Oliver\n- *Thinking, Fast and Slow* — Daniel Kahneman\n- *The Signal and the Noise* — Nate Silver","html":"