The Cream Always Rises to the Top… Well, Not Always

“The cream always rises to the top.” You’ve heard that one your whole life. It’s the saying folks reach for when they want to believe the world works the way it ought to — that the best player makes the team, the best manager beats the market, and the best leader ends up leading. It’s a comforting thought. It’s also, I’ve learned, only half the story.

I’ve spent thirty-five years in and around places where that old saying gets tested every single day: trading floors, portfolio management, public-company boardrooms, and the long, unglamorous grind of building businesses. And here’s where I’ve landed. The saying is true a whole lot more often than the cynics will admit. But it’s true only when certain conditions are in place — and we get ourselves in trouble when we just assume those conditions are in place everywhere.

The hidden ingredient, the one tucked inside that little word “always,” is time. Cream doesn’t rise in a hurry. It rises as evidence piles up. In the short run, every competitive field is mostly noise — hot streaks, lucky bounces, one great quarter, one viral moment. Give it time, though, and the noise cancels itself out while the real signal sticks around. The flukes — the one-hit wonders, the flash-in-the-pan stories — get smaller and smaller as the record gets longer, until what’s left standing is the true measure of the person. That’s not a pep talk. That’s arithmetic. And the first part of this piece walks through why the arithmetic works.

After that, we’ll take a tour of the places where the arithmetic is allowed to do its work. Sports is the cleanest case: honest scoreboards, thousands of repetitions, and no mercy in the selection. Finance is noisier and slower, but given enough years it hands down the same verdict — the managers still standing after decades are, by and large, the ones who were actually good. Chess, surgery, music, science, and corporate life all sit somewhere along the same line, and lining them up side by side shows us the three conditions a field needs before its cream can rise at all.

Then we’ll turn to the one arena where, I’ll argue, the saying breaks down — not because the arithmetic quits working, but because the selection system points the other way. Politics doesn’t just fail to filter for excellence. In some important respects, it filters against it. The politician’s core bargain — trade your independent judgment to the tribe in exchange for keeping your seat — rewards underperformance and punishes the very traits that make people excel everywhere else. More time in politics doesn’t converge on the truth. It converges on incumbency. I’ll make that case with the same standard of evidence I’d apply to a stock, I’ll give the honest counterarguments their full say, and then you can weigh it all for yourself.

Part One: The Arithmetic of “Over Time”

Start with something simple: a coin. Flip it ten times and there’s nothing strange about getting seven heads. Roughly one set of flips in six will give you seven or more, just by chance. Now flip that same coin a thousand times, and the share of heads will almost certainly land within a few points of fifty percent. The coin didn’t change. Your luck didn’t change. What changed is the number of flips. Small samples are where flukes live. Large samples are where the truth lives. Statisticians call this the law of large numbers, and it’s the quiet engine running underneath every claim that cream rises “over time.”

Now swap the coin for a person, because a person’s results are always a blend of skill and luck — and that describes just about everybody in every competitive field you can name. In any single contest, luck can swamp skill completely. A mediocre poker player catches a big hand. A mediocre fund manager has a big year. A mediocre band stumbles onto one perfect chorus. These are what statisticians call tail events: results way out at the edge, far from the performer’s true average, produced by randomness rather than ability.

Here’s the thing about tail events, though: they can’t repeat on command. As the attempts pile up, every new result tugs the overall record back toward the performer’s real level. That tugging has a name — regression to the mean. The one-hit wonder isn’t a moral failure. It’s a statistical inevitability. One hit is exactly the sample size at which luck is best able to dress up and impersonate skill.

Trouble is, human beings are wired to misread this arithmetic. We’re storytellers before we’re statisticians, and a story needs a hero — not a bell curve. When a fund manager triples the market, the coverage explains her brilliant insight. Nobody stops to count how many equally confident managers bet the other way and quietly disappeared. When a startup founder hits it big, we study his morning habits. We don’t interview the thousand founders with the exact same habits who went under. That’s the narrative fallacy mixed with survivorship bias: the losers slip out of the sample without a sound, the winner’s luck gets relabeled as genius, and the rest of us walk away believing the fluke was destiny all along.

The practical upshot is that public reputation is a whole lot noisier than actual performance. Fame piles up on spectacular small samples. Genuine skill — which shows itself as undramatic consistency over long stretches — makes for lousy television. So if you seriously want to know whether the cream is rising somewhere, you have to start by distrusting the highlight reel and demanding the full record: every attempt, including the ones that ended quietly with nobody watching.

Michael Mauboussin, who has written about this as clearly as anyone alive, offers a simple test for how much luck a field contains: ask whether you can lose on purpose. A chess grandmaster can lose on purpose instantly, because chess is nearly all skill. A lottery player can’t lose on purpose at all, because the lottery is all luck. Most arenas sit somewhere in between — and where they sit matters enormously, because the more luck a field holds, the more repetitions you need before its results are worth believing.

That gives us the master rule for everything that follows. The honesty of any ranking comes down to three things: how clearly the outcome is measured, how many times the contest is repeated, and whether the selection system actually promotes people based on that measurement. Hold onto that rule. Every arena in this piece gets judged against it.

Part Two: Sports — The Cleanest Filter We Have

If you sat down to design a machine for making cream rise, you’d design professional sports. The outcome is measured to the decimal and posted on a scoreboard nobody can spin. The contest repeats constantly — an NBA regular season runs eighty-two games, a baseball season one hundred sixty-two, and inside every game sit hundreds of separate, recorded events. Feedback is immediate and public. And selection is bolted directly to the measurement: underperform, and you’re benched, cut, waived, or sent down. All three conditions of the master rule, satisfied at full strength.

And the results are just what the old saying promises. Short samples still produce their famous flukes. Leicester City won the Premier League at pre-season odds reported around five thousand to one. Buster Douglas knocked out Mike Tyson as a forty-two-to-one underdog. Every so often a sixteen-seed topples a one-seed in March. We remember these precisely because they’re rare — and the aftermath is always the same story: regression. Leicester spent the following seasons back in mid-table where they belonged. The Cinderella teams of a single-elimination tournament rarely survive a seven-game series — which is exactly why championship series exist. A best-of-seven is a deliberate way of enlarging the sample to shrink the role of luck.

Over career-length samples, the filter gets almost perfectly honest. The players who sustain excellence across fifteen or twenty seasons — thousands of games, multiple coaches, rule changes, and aging bodies — are, with virtually no exceptions, the genuinely great ones. And notice something else: the filter corrects its own bad predictions. Tom Brady was the one hundred ninety-ninth pick of his draft. The scouts misjudged him badly. It didn’t matter, because the system kept generating performance data, and the data kept promoting him. Michael Jordan was famously assigned to his high school’s junior varsity squad as a sophomore. The first assessment was wrong. The ten thousand measurements that came after were not.

The same machinery works just as relentlessly in the other direction. The average NFL career lasts only about three years, and rosters churn constantly as the measurements roll in. Sports is brutal. But its brutality is honest. The cream rises because nothing in the system can stop it from rising.

Sports also gives us the best-documented case of a field deliberately upgrading its own filter — and getting rewarded for it. For most of baseball’s history, talent was judged partly on measurements that turned out to be noise: batting average over on-base percentage, the look of a swing, the build of a body. The analytics movement Michael Lewis wrote about in Moneyball was, at bottom, a project to replace weaker measurements with stronger ones. The small-budget teams that adopted it first bought wins at a fraction of the going price, because the market was still paying top dollar for the noise. Within a decade, the whole league had copied them.

That’s what a healthy arena does when it discovers its scoreboard is imperfect: it sharpens the scoreboard, and the advantage flows to whoever sharpens it first. Keep that picture in mind — because further down, we’re going to meet an arena that responds to flaws in its scoreboard in precisely the opposite way.

Part Three: Finance — Noisier, Slower, Same Verdict

Markets are where I grew up professionally — I started on the retail side in 1991 and later ran quantitative portfolios — and markets are a far noisier arena than sports. One year of returns tells you next to nothing about a manager’s skill. In any given year, roughly half of active managers will beat their benchmark by plain chance, and a lucky few will beat it spectacularly. The financial industry manufactures one-hit wonders at industrial scale: the fund that returned eighty percent last year, the newsletter writer who called the crash, the trader who was up huge in one wild quarter. Every one of them is a small-sample fluke wearing the costume of genius.

Time strips the costume off. The most sobering numbers in my industry are the long-horizon scorecards: over fifteen-year periods, on the order of nine in ten actively managed U.S. large-cap funds fail to beat the S&P 500 after fees. Now, read that statistic the right way. It’s not evidence that skill doesn’t exist. It’s evidence that the filter works. Stretch the horizon out and the pretenders get removed — by redemptions, by career changes, by funds closing their doors. The market is a machine for firing underperformance, and it runs around the clock. What survives two and three decades of that machine is a small residue — and the residue looks nothing like luck.

An honest analyst has to stop here and face the standard objection head-on: survivorship bias. If thousands of managers start out and only the winners stay visible, couldn’t the survivors just be the lucky tail of one giant coin-flipping contest? For short records — absolutely. That’s the right assumption to start with. But the luck explanation has to work harder with every additional year, and at the far end of the record it flat-out collapses.

Warren Buffett has compounded at roughly twice the market’s rate for some six decades. The odds of doing that by chance aren’t small. They’re effectively zero. And Buffett himself made the deeper point in his famous “Superinvestors of Graham-and-Doddsville” argument: if the winners of a coin-flipping contest all turn out to have learned from the same teacher and used the same method, luck stops being a serious explanation. Renaissance Technologies’ Medallion Fund is reported to have returned on the order of thirty-nine percent a year, net of extraordinary fees, for roughly three decades — thousands of trading days, across every kind of market weather you can imagine. These are not one-hit wonders. This is what the far right tail of genuine skill looks like after time has cleared everybody else out of the room.

The customers, it must be said, conspire against their own interests by worshiping exactly the flukes the arithmetic warns about. Fund money chases last year’s winner with mechanical reliability: it floods in after a spectacular year — which is to say, at the very top of a small lucky sample — and floods back out after the inevitable regression, locking in the round trip. Studies of dollar-weighted returns find that the average investor earns meaningfully less than the average fund he owns — on the order of a percentage point or more a year — purely because of when he got in and out.

I watched that behavior from inside the industry for decades, and the lesson travels well beyond finance: it isn’t enough for a field to generate honest long-run data. The people doing the selecting have to have the patience to wait for it. Wherever selectors act on one-year samples, they will systematically promote luck, fire skill, and pay for the privilege of doing it. Time only works as a filter for folks willing to let it run.

So finance obeys the proverb, with two amendments. First, the required time horizon is much longer than in sports, because there’s more luck in the mix — believing any track record shorter than a full market cycle is just a category error. Second, the cream that rises is a much smaller share of the container. But the direction of the filter is right: the measurement is objective (returns are returns), the trials pile up daily, and capital — the selection mechanism — eventually flows toward performance and away from its absence. The managers who’ve been around the longest aren’t necessarily the best in any given quarter. But over time, the longest-tenured group leans strongly toward genuine, above-market skill — because tenure itself is the market’s accumulated verdict.

Part Four: A Quick Tour of Other Arenas

Line other fields up against the master rule and a clear pattern shows itself. Chess sits at the honest extreme — arguably purer than sports. Its Elo rating system turns every game into an updated measurement, luck is nearly absent, and so the rankings are close to incorruptible. Nobody argues about who the strongest players in the world are, and the same names hold the top spots for a decade at a time.

Surgery sits high on the list too. Outcomes can be counted, and the volume-outcome relationship — surgeons and hospitals that perform a procedure most often tend to get the best results — is one of the most consistently confirmed findings in health services research. Repetition builds skill, and measurement reveals it.

Music and film are luck-heavy by comparison. William Goldman’s famous verdict on Hollywood — that nobody knows anything — is really a statement about how impossible it is to predict any single outcome. The music business literally keeps a category called the one-hit wonder. But watch what time does even here: catalogs are large samples. A single hit can be luck. Twenty years of albums people still pay to hear cannot be. The artists still filling arenas three decades on are, by the only measurement that adds up honestly — sustained audience demand — the real thing.

Science is slower still, and noisier at the level of any single paper, as the replication troubles of recent years plainly show. Yet its long-run filter is famously self-correcting: fraudulent and flimsy results die when other scientists can’t reproduce them, and the Nobel committees now routinely honor work done twenty or thirty years earlier — an institutional admission that only time can certify which discoveries were cream.

Entrepreneurship and venture capital deserve their own stop on the tour, because they’re the most luck-drenched corner of business — and therefore the sternest test of the whole idea. Startup results follow a power law: in a typical venture portfolio, most investments return little or nothing, and one or two outliers pay for everything. At the level of a single company, luck — timing, a chance customer, a competitor’s stumble — can overwhelm any amount of skill, which is why one lone unicorn on a résumé proves far less than its owner usually believes.

And yet, even here, repetition slowly separates the crowd. Researchers who track founders across multiple ventures find that those who’ve succeeded before succeed again at modestly higher rates than first-timers or previous failures. And the small set of investors who beat the field across many funds and multiple technology cycles is stable enough that the industry’s limited partners fight for access to them. The signal is weaker than in chess, and the sample sizes are painfully small — a whole career might contain only a handful of companies. But the direction is unmistakable, and it’s the familiar one: even in the noisiest honest arena, more trials mean more truth. Luck sets the spread. Time still finds the average.

The hardest honest case: corporate leadership

Corporate leadership is the toughest honest case, and I want to be candid about it because it’s the world I know best. CEO results are contaminated by timing, by industry winds, by whatever the last person left behind. Plenty of executives have ridden a rising cycle up and been called geniuses for it. Attribution is genuinely hard — a company’s results are the work of thousands of hands.

But the conditions of the master rule are still there, just in weakened form. There’s a scoreboard: profit, share performance, survival. The trials repeat: every quarter, every product cycle, every downturn. And the selection mechanism, though slow and imperfect, has teeth — boards remove executives, competitors take market share, and bankruptcy is the market’s final grading pen. Executives who compound value across multiple companies, multiple cycles, and multiple decades are rare — and for the same reason long-tenured outperforming fund managers are rare. By that point, luck has largely been wrung out of the record.

So the pattern across every arena on this tour is now plain to see. Wherever outcomes are measured honestly, trials accumulate, and selection is chained to the measurement, the cream rises — quickly where luck is low, slowly where luck is high, but reliably. Which brings us to the arena where the chain is cut.

Part Five: Politics — The Filter That Points the Other Way

Now run politics through the master rule, and all three conditions fail at once.

First, there’s no scoreboard. What’s the objective, agreed-upon measurement of a legislator’s performance? Bills passed is a volume statistic, not a quality one — counting laws is like judging a chef by how many dishes leave the kitchen, without ever tasting one. Economic results lag policy by years and are shaped by forces no officeholder controls. Constituent service is invisible at scale. Voters can’t mark a politician’s performance to market, because there’s no market and no mark. There’s only narrative.

Second, the trials barely repeat. A senator faces the voters once every six years. A president, twice in a lifetime of ambition. Where a ballplayer generates thousands of measured events in a single season, a politician generates a handful of elections in a whole career — a sample size at which, as Part One established, luck and noise dominate everything.

Third — and most fundamental — the selection mechanism isn’t connected to governing performance at all. It’s connected to electoral performance. The system doesn’t promote people who govern well. It promotes people who campaign well. And those are different skills, practiced by, in many cases, different kinds of people.

The tribe trade

Now add the mechanism I believe does the deepest damage, because it doesn’t just fail the filter — it turns the filter upside down. I call it the tribe trade.

In nearly every other arena, excellence is exercised through independent judgment: the portfolio manager’s willingness to stand apart from the crowd, the scientist’s willingness to follow the evidence against the fashion, the athlete’s obligation to perform alone under the lights. In politics, independent judgment is precisely what the system prices out. A candidate doesn’t reach office — and above all doesn’t stay in office — on the strength of his own convictions. He stays by subordinating them to a party. The whip system, the primary electorate, the donor network, and his side’s media ecosystem all enforce the same contract: vote with the tribe, speak with the tribe, shelve your private judgment whenever it conflicts with the tribe’s — and in exchange, the tribe delivers the only performance metric that matters. Re-election.

That is, in the most literal sense, a paid agreement to underperform. The officeholder trades away the exercise of his own best judgment — which is the very thing excellence consists of everywhere else — and receives job security in return. In finance, we’d call that a catastrophic principal-agent problem. In politics, we call it party discipline and treat it as normal.

And the numbers say the trade clears. Public approval of Congress as an institution has spent most of recent decades in the teens and twenties — while incumbent members of the House who seek re-election win at rates routinely above ninety percent. Sit with that pairing a moment, because no honest filter could produce it: the customers overwhelmingly rate the product a failure, and the producers are overwhelmingly kept on. In any market I’ve ever worked in, a ninety percent retention rate next to a twenty percent satisfaction rate is impossible. Capital would flee. Customers would defect. The scoreboard would clear the field. In politics it’s the steady state — sustained by safe districts where the general election is a formality and the only contest that matters is a low-turnout primary decided by the most tribal slice of the electorate.

Under those conditions, time doesn’t filter. Time entrenches. Every additional term produces not a better measurement of governing skill, but greater name recognition, deeper donor relationships, and more seniority — incumbency compounding like interest, on exactly the wrong account.

The clock problem

Even if voters wanted to grade governing performance, the structure of the job frustrates the attempt — because political results ripen on a completely different clock than political careers. A pension reform, an infrastructure program, an education policy, a decision about energy capacity: these reveal their quality over ten, twenty, thirty years, long after their authors have moved on, claimed credit, or assigned blame. The officeholder sitting in the chair when the results finally arrive is rarely the one who caused them — and everybody in the system knows it.

That produces a predictable and corrosive incentive: optimize for what’s visible inside the electoral window — announcements, ribbon-cuttings, positions taken, enemies named — and discount everything whose payoff lands beyond it, however large. In finance we’d recognize that instantly as marking positions to a horizon that expires before the trade does. No serious investor would accept a track record built that way. Yet it’s the only track record politics knows how to build. The attribution problem doesn’t just add noise to the measurement. It disconnects the measurement from the work itself — so that even an infinitely patient electorate would struggle to reward the right people.

The attention amplifier

Then the modern attention economy bolts an amplifier onto the broken filter. Elections are decided by attention. Attention flows to conflict. So the system’s effective scoreboard — the one that actually moves careers between elections — is a running tally of viral moments, cable hits, and social-media engagement, every unit of which rewards heat over light. The legislator who spends a year quietly mastering a complex bill generates no clips. The one who produces a combative soundbite in a hearing raises money that same afternoon. You’d have a hard time designing a promotion metric more perfectly sideways to governing skill if you tried.

And notice what that does to the intake valve, which we’ll get to in a moment: the personality type that thrives on perpetual performative combat, and the personality type that patiently builds durable institutions, are rarely the same type. The arena is not neutral about which type it admits. It advertises for one and quietly repels the other.

The intake valve

The filter fails at the front door, too. Ask what the job actually selects for at entry — and what it costs the person entering.

Members of Congress are advised by their own party committees to spend roughly four hours a day on fundraising calls — a schedule first revealed in a leaked party orientation presentation and confirmed by numerous former members since. The pay is a fraction of what comparable talent earns in the private sector. The scrutiny extends to one’s family. And the work itself, for a junior member of a disciplined caucus, consists largely of reciting positions somebody else drafted.

Now run the thought experiment the way a recruiter would — and I’ve worn that hat many times. Describe the role honestly: the hours spent soliciting money, the prohibition on independent judgment, the pay cut, the public exposure. Then ask yourself: which candidates accept that offer?

Two economists, Francesco Caselli and Massimo Morelli, formalized the grim answer in a paper bluntly titled “Bad Politicians”: when the material and reputational rewards of office are low compared with private alternatives, the candidate pool skews toward those with the least to give up. That’s adverse selection at the point of entry — before a single vote is cast. The people best equipped to govern — the ones with demonstrated records of building, managing, and deciding under uncertainty — are, with honorable exceptions, exactly the people for whom the trade makes the least sense. They stay in the arenas with scoreboards.

The result is a political class that isn’t a fair sample of a nation’s talent, let alone its cream. It’s a sample of those willing to make a trade that talent rarely accepts.

An apprenticeship in the wrong trade

Finally, consider what the profession’s own career ladder actually trains. In every arena where cream rises, the apprenticeship consists of practicing the real work. The surgeon’s residency is surgery. The trader’s formation is trading. The point guard’s development is ten thousand possessions. The standard political ladder — campaign staffer, communications aide, party operative, candidate — is an apprenticeship in campaigning. Its curriculum is message discipline, opposition research, fundraising, and coalition maintenance. At no rung does it require building anything, managing a payroll, delivering a project on budget, or living with the long-term consequences of your own decisions.

By the time the successful climber reaches office, he may have spent twenty years becoming world-class at acquiring office — and not one day practicing its use. That’s not an accusation of bad faith. It’s an observation about training. We wouldn’t expect excellence from a surgeon whose entire residency was spent interviewing for surgical jobs, however brilliantly he interviewed. Yet that is, structurally, the résumé the political ladder produces — and then we act surprised that the institution staffed by that ladder underperforms.

Notice, please, that none of this requires cynicism about any individual, and it makes no claim about either party. It’s a claim about system design, and it would apply under any tribe and any flag. Sports would produce mediocrity too if athletes were selected by popularity contests held every few years among fans who never watched the games, and if coaches benched any player who improvised. Finance would produce mediocrity if managers were kept on for marketing skill rather than returns — indeed, in the corners of finance where exactly that is true, mediocrity is exactly what gets produced, which rather proves the point. Politics isn’t populated by worse human beings than other fields. It’s governed by a worse filter — one that measures the wrong thing, measures it rarely, and pays its participants to suppress the one trait that constitutes excellence everywhere else.

Part Six: The Honest Counterarguments

An argument this pointed deserves its strongest objections, stated fairly. Let me give them their due — because some of them are good.

“You’re judging the wrong contest”

The first objection is definitional. Democracy’s filter was never designed to select for technical excellence. It selects for consent, legitimacy, and representation — and on its own metric, it succeeds. An election answers the question “whom do the people choose?” with perfect accuracy, every single time. On this view, complaining that elections fail to identify the most capable governors is like complaining that a thermometer fails to measure weight. The instrument is doing a different job — and a valuable one, since governments that lose legitimacy fail in ways far worse than mere incompetence.

There’s real force in that, and it reframes the thesis rather than refuting it. The cream does rise in politics — cream as defined by the filter’s own metric. The best campaigners, coalition-builders, and tribal signalers reliably reach the top. The trouble is only that we then hand them a job the filter never tested them for. The proverb fails not because selection is random, but because the selected trait and the required trait are two different traits.

“But crises elevate great leaders”

The second objection is historical. Crises seem to make the cream rise even in politics. Lincoln. Churchill in 1940. When the stakes turn existential, systems somehow reach past the usual machinery and elevate genuine capacity.

I think that observation is correct — and revealing rather than contrary. What a crisis does is restore the missing conditions of the master rule. Suddenly there is a scoreboard: the war is being won or lost, visibly. Feedback arrives in weeks instead of decades. And the tribe’s tolerance for loyal mediocrity collapses, because the tribe itself is at risk. Churchill spent the peacetime 1930s exiled to the back benches — the tribal filter working exactly as I’ve described it — and was summoned only when the scoreboard lit up. The exception works by temporarily converting politics into the kind of measured arena where cream can rise. That’s less a rebuttal of the argument than a demonstration of it.

“Then why do accomplished outsiders flounder?”

A related objection comes from the other direction. If politics merely filtered out excellence, then accomplished outsiders — executives, generals, surgeons — should stride into office and outperform the career politicians. And the record shows they frequently flounder.

Fair enough, and worth taking seriously. Legislating through a body of hundreds of independent actors, none of whom can be fired, is genuinely different work from running a company — and the outsider’s private-sector scoreboard tested none of it. But look closely at what that proves. It proves political skill is real and distinct — coalition assembly, procedural mastery, managing allies who are also rivals — not that the filter selects for governing excellence. A system can require real skills and still select for the wrong ones. Indeed, that’s precisely the claim of this piece. The failure of outsiders and the mediocrity of insiders aren’t competing observations. They’re the two halves of the same diagnosis: the job requires one set of abilities, the filter tests for another, and almost no path exists that certifies both.

“Careful — ‘underperformer’ can’t just mean ‘people I disagree with’”

The third objection is a caution against my own bias, and readers should hold me to it. “Underperformer” must not become a synonym for “person whose policies I dislike.” Capable, principled people serve in public life in every generation and in every party. Some sacrifice considerably to do so, and a fair accounting salutes them. My claim is statistical, not universal: it’s about the central tendency the filter produces, not about every individual it passes.

And a fourth objection deserves standing room: maybe the comparison is simply unfair, because governing is harder to measure than shooting a basketball, through no fault of the system’s designers. Partly true — but only partly. Because measurement in politics isn’t merely difficult. It’s actively resisted. Institutions that would sharpen the scoreboard — independent scoring of legislation, sunset clauses that force a retrospective look, nonpartisan redistricting that would make general elections competitive again — are proposed perennially and blocked perennially, most vigorously by incumbents. A field can be forgiven for having a noisy scoreboard. It’s harder to forgive a field for dismantling the scoreboard on purpose.

Conclusion: You Get What You Select For

So the proverb, corrected, reads something like this: cream rises wherever time is allowed to multiply honest measurements — and only there.

In sports, the conditions are nearly perfect, and the cream rises fast. In finance, the noise is heavier and the rise takes decades, but the long-tenured survivors of the market’s continuous culling are the real thing — time shrinks the flukes, retires the one-hit wonders, and lets the true measurement come through. Chess, surgery, science, music, and business each confirm the same law at their own speed. Across all of them, the pattern isn’t that virtue triumphs. It’s that filters deliver exactly what they select for — no more, and no less.

Politics delivers exactly what it selects for, too — and that’s the indictment. It selects for electability, tribal reliability, and the willingness to trade your own judgment for your own seat. And so that’s what floats to the top: not our best and brightest, but our most durable campaigners, sustained by a system in which twenty percent approval coexists comfortably with ninety percent re-election. The one-hit wonders of markets get washed out by time. The incumbents of politics get fortified by it. Same variable, opposite sign — because the underlying selection points the opposite way.

The practical conclusion follows directly, and it’s more hopeful than the diagnosis sounds. If the problem were the people, we’d be stuck, because moral lectures have never improved a candidate pool. But the problem is the filter — and filters can be redesigned. Anything that restores the three conditions moves the system toward honesty: make more elections genuinely competitive, so the trials mean something. Build scoreboards — retrospective review, transparent scoring, published results — so performance actually gets measured. And raise the rewards for independent judgment relative to tribal loyalty, so talented people find the trade worth making.

None of that is easy, and incumbents will resist all of it, for reasons this piece has made plain. But the direction isn’t mysterious. Every arena that gets selection right teaches the same lesson, and it’s the lesson I’d leave you with: stop asking why the cream isn’t rising, and start asking what the filter is actually skimming. The cream always rises to the top — in systems built to let it. Build the system, and time, the most honest judge we have, will do the rest.

Where to read more

If you’d like to dig into the ideas behind this piece, here are some good next stops. I’ve stuck to reference works, the original research, and scorekeepers rather than opinion pages — you’ll find a range of conclusions in them, and that’s rather the point.

• The Law of Large Numbers — Wikipedia: the plain-language reference for why averages settle down and tell the truth as the number of trials grows.
• Regression Toward the Mean — Wikipedia: why a spectacular performance is usually followed by a more ordinary one, with everyday examples from classrooms to baseball.
• Survivorship Bias — Wikipedia: how looking only at the winners still standing — funds, founders, or returning aircraft — quietly distorts what we think we know.
• Passive Management and the SPIVA Scorecards — Wikipedia: an overview of index investing that walks through the S&P Dow Jones Indices SPIVA results behind the nine-in-ten funds figure used above.
• “Skill or Luck?” — strategy+business: a review of Michael Mauboussin’s The Success Equation, including his “can you lose on purpose?” test for telling skill from luck.
• “Bad Politicians” — Caselli & Morelli, NBER Working Paper 8532: the original economics paper on why low rewards for office can skew the candidate pool (later published in the Journal of Public Economics, 2004).