A Portfolio of Almost Is Still Exposure

Many positions that nearly became wise, which is a beautiful way to be broke #

By February the portfolio contains a kind of pleasant fog. The MVP ticket lost in the last week of November, but cleanly; the division future lost a coin flip on tiebreaker; the team total over on the high-pace team was a stadium-roof game; the receiving-yards prop on the same offense was beaten by a third-quarter injury. Each loss has a small story attached, and each story argues, quietly, that the bettor was right about something. The arguments aggregate into a feeling of intelligence undermined by luck. The ledger, asked to confirm this feeling, declines.

A portfolio of almost is, in my experience, the most flattering shape a losing season can take. It preserves the bettor s sense of competence by parceling out the disappointment into discrete near-misses, each defensible on its own terms. The structural problem is that the discrete near-misses are not independent. They are usually expressions of the same hypothesis wearing different uniforms, and the same hypothesis, if wrong, takes down the whole rack. Almost is not a settlement category. The portfolio either managed its exposure or it did not, and the manager s feelings about the matter are admissible only after the arithmetic.

The market mechanic of disguised exposure #

The sportsbook is happy to offer the bettor a wide range of markets that all settle in approximately the same direction. A QB MVP ticket, a team season over, and a wide-receiver receiving-yards prop on the same offense are three markets whose outcomes depend on a shared substrate: the offense functions, the quarterback stays healthy, the schedule cooperates. The book is not engineering the trap. The book is presenting the menu honestly. The bettor who reads the menu as a menu — three items, three independent decisions — has misread the document. The items rhyme.

The realized exposure on a portfolio of rhyming positions is much larger than the stake count implies. A bettor with twelve tickets distributed across four independent theses owns a portfolio with four positions, not twelve. The stake exposure on those four positions, when correctly aggregated, often consumes most of the discretionary risk the bettor believes she has allocated across the season. The portfolio looked diversified at the time of placement. It became concentrated the moment the first thesis broke, because the broken thesis broke three tickets at once.

The thesis review and what it actually replaces #

The exercise I have come to depend on, before any portfolio-level grading at the end of a season, is a thesis review. I write down each ticket s thesis in a single sentence, without team names, without market labels. I then group the sentences. The number of distinct sentence-groups is the number of bets the portfolio actually placed; the number of tickets is irrelevant. The exercise rarely flatters the placement record. Twelve tickets typically reduce to between three and five distinct theses, and the largest cluster usually contains the position the bettor was most confident about — which is also the cluster most likely to have failed in a correlated way.

The thesis review is the only audit I have found that survives the postseason. It replaces the season-end ritual of grading each ticket against its result with a structural question: which theses did the portfolio fund, and at what cluster-adjusted stake. The answer is usually different from the answer the slip suggests. The bettor who treats the slip as authoritative will continue to believe, season after season, that she was diversified. The thesis review will tell her where the exposure actually lived, which is the only basis on which the next season s position sizing can be reset honestly.

What near-misses are allowed to teach and what they are not #

Near-misses contain real information. A futures ticket that finished one game outside a hedge window may have been correct in thesis and slightly wrong in price; a prop that lost on an injury was correct in design and unlucky in realization. Neither of these is a verdict on the bettor s process. What near-misses are not allowed to do is rescue a portfolio whose cluster-adjusted exposure was too large. The verdict on the portfolio is the cluster-adjusted P&L, not the count of tickets that were live in Week 17. Almost is not a settlement category, and the ledger does not permit it to be one.

I try to keep two columns in the season-end review: the per-ticket P&L, which is the result of placement, and the per-thesis P&L, which is the result of portfolio construction. The per-thesis column is the column that informs next season. If the same thesis appears in next year s portfolio, it will appear at a stake size determined by the cluster-adjusted result, not the per-ticket result. This is a small, mechanical adjustment. It is the only way the portfolio learns from a year in which the bettor was right about several things at the same time and lost meaningfully anyway.

Almost can teach, but it cannot be allowed to flatter. The portfolio either managed exposure or it did not. Memory is welcome to testify after the ledger speaks first.

The discipline of grouping before grading #

The procedural discipline that makes the thesis review possible is to group bets at placement, not at settlement. Each ticket is added to the portfolio with a thesis tag — the offense functions, the defense regresses, the rookie lives up to draft capital, the coach loses the locker room — and the tag travels with the ticket through the season. At any moment, the portfolio s cluster-adjusted exposure can be read off the tag column rather than reconstructed from memory in February. The bettor who tags at placement is the bettor who can size correctly. The bettor who reconstructs at settlement is the bettor who will continue to declare twelve tickets and own four positions.

The tag column has a second benefit, which is to make the in-season hedge decision tractable. A bettor who can see that three tickets share a thesis can hedge the thesis once, at the cluster level, rather than three times at the ticket level. The hedge is cleaner, the price is better, and the residual exposure is more honest. The slip view of the portfolio cannot produce this hedge; the cluster view can. The discipline is to maintain both views, with the cluster view treated as the operative one whenever a decision is being made.

The closing argument #

The portfolio is not what the slip says it is. The slip is a placement artifact; the portfolio is the cluster-adjusted exposure that survives once the rhyming positions have been grouped honestly. A season-end review that grades each ticket on its own terms is a review that protects the bettor s feelings at the cost of the bettor s next sizing decision. The grouping is the work. The grading follows. Without the grouping, every season produces the same fog: a long list of near-misses, a feeling of intelligence undeployed, a ledger that refuses to confirm the feeling.

I find that a portfolio constructed with the tag column at placement and reviewed with the thesis report at settlement produces a different kind of season summary. The summary is shorter. The summary is harder to argue with. The summary names the theses the bettor was right about, the theses she was wrong about, and the cluster-adjusted stake she should consider on each next year. None of this prevents being wrong. It prevents being wrong four times in the same direction without noticing. Almost is welcome at the end of the year; it is not allowed to write the budget.