What the Market Is Actually Saying

Closing line value is the strongest early signal a quantitative system can produce. It does not ask you to trust a backtest. It asks a simpler question: did the entry age well as the market absorbed more information?

How to think about CLV in plain language

Suppose ArcVest enters a position at a price that implies a lower probability than the closing market eventually implies. That means the entry was better than the final consensus price. Put differently, ArcVest entered the position before the market fully repriced it.

That is the sports-market equivalent of entering a trade before the rest of the tape acknowledges the information. The final outcome still matters for realized return, but the price move itself tells you whether the process saw value before the market corrected.

This is why Price Capture Alpha is the strongest single trust metric on the platform. It does not ask you to believe a backtest. It does not ask you to trust selective screenshots. It asks a simpler question: did the entry age well as the market absorbed more information?

Why line shopping still matters

CLV is only meaningful if the entry is real. That means finding the best available price matters. Two users can hold the same opinion on a side and still end up with different long-run results because one consistently enters at a stronger number.

This is where line shopping becomes an execution discipline rather than a convenience. A small difference in entry price compounds across a portfolio. The objective is not to collect more apps. The objective is to reduce friction between model signal and actual entry quality.

Why ArcVest makes it public

Most products discuss edge in a way that cannot be audited. ArcVest treats that as a credibility failure. If a model claims skill, the public should be able to inspect whether its entries actually beat the close over time.

That is why Price Capture Alpha is surfaced in the ArcVest record and why the Performance Log exists. The ledger is not there as decoration. It exists so readers can inspect sealed positions, compare entry and closing prices, and evaluate whether the process improved relative to the market.

That is the difference between narrative confidence and measurable confidence. One is copy. The other is evidence.

What CLV can and cannot do

CLV is the best early proof of process quality, but it is not the only metric that matters. It does not replace calibration, allocation discipline, or realized portfolio management. It also does not make every individual position correct in hindsight.

What it does do is answer the most important pre-result question available to a public-facing quantitative system: did the model get into the market at a number that the market itself later validated? If the answer is consistently yes, that is the strongest available sign that the process is operating on real informational edge rather than on luck.

For informational purposes only. Not financial or betting advice. Past model performance does not guarantee future results.

Start with the right question

A losing stretch feels personal because the result is immediate and the signal is emotional. But a drawdown does not begin as a moral event. It begins as a diagnostic event.

The useful question is not, "Did the portfolio lose this week?" The useful question is, "What part of the process moved?" ArcVest answers that by separating drawdowns into three components: Price Capture Alpha, Selection Alpha, and Outcome Variance.

This matters because not every losing stretch means the model is broken. Some losing stretches are standard noise. Some reflect unstable conditions. A small number indicate that the process itself deserves investigation.

The three-part decomposition

Price Capture Alpha asks whether the portfolio entered at prices that the market later improved through. If this remains healthy during a drawdown, the model may still be seeing the market well even if outcomes are temporarily unfavorable.

Selection Alpha asks whether the model's probability estimate was superior to the market's implied probability at entry. This is the valuation layer. If Price Capture Alpha is steady but Selection Alpha is weakening, the signal may still be reaching the market efficiently while the underlying event assessment softens.

Outcome Variance is the luck layer. It measures the gap between realized return and expected return. In plain language, it asks how much of the recent pain came from events resolving worse than the underlying prices implied.

The variance-status ladder

ArcVest translates drawdown states into a simple diagnostic ladder. Standard Noise means recent results are uncomfortable but still sit within the portfolio's expected distribution. Elevated Noise means recent outcomes have become meaningfully more unstable and deserve closer review. Critical Sigma means the system is outside normal expectations and should be investigated with more care.

This framing matters because it gives the user a vocabulary that is calmer than panic and more useful than denial. "We are down" is a fact. "We are in an Elevated Noise state" is a fact plus context.

That is the difference between a reactive interface and a disciplined one. A good interface does not hide the red number. It explains what the red number is likely saying.

What to do during a bad week

The first discipline is not to abandon line quality. Drawdowns often tempt users to rush, overtrade, or ignore execution quality. That usually compounds the problem. During unstable stretches, the need for good entry prices becomes more important, not less, which is why line shopping across market makers remains essential.

The second discipline is to keep process and outcome separate. If Price Capture Alpha remains healthy and Outcome Variance is doing most of the damage, the correct response is usually restraint and review, not emotional intervention.

The third discipline is to use the public record. The Scorecard exists so that a bad stretch can be interpreted in context rather than in isolation.

When a drawdown becomes more than noise

There are losing stretches that deserve concern. If Price Capture Alpha weakens, Selection Alpha deteriorates, and variance status escalates together, the system is no longer just absorbing noise. It may be losing its pricing advantage, misreading the market, or operating in conditions where the prior assumptions no longer hold.

This is exactly why a disciplined platform should show more than record and ROI. Record and ROI describe what happened. The decomposition starts to explain why.

The mistake most readers make

The most common error is to treat every losing streak as proof that the process has failed. That interpretation is emotionally understandable and analytically weak. A short drawdown can emerge from a run of unfavorable realizations even while the process is still entering the market well.

The second most common error is the opposite one: assuming every drawdown is harmless. Sometimes the market is signaling that the edge has narrowed or that the environment has changed. The point of the decomposition is not to excuse losses. The point is to diagnose them honestly.

Serious users do not need reassurance. They need a framework that tells them which layer of the process is carrying the stress.

For informational purposes only. Not financial or betting advice. Past model performance does not guarantee future results.

Sizing is where discipline becomes real

Most readers spend too much time deciding what they like and too little time deciding how much to allocate. That is backwards. In a portfolio system, sizing is not a footnote to the signal. It is the mechanism that determines whether the signal compounds or self-destructs.

ArcVest uses fractional Kelly because it converts estimated edge into a position size in a way that is mathematically coherent. It does not ask how confident the user feels. It asks how large the edge is relative to the price and how much capital the portfolio can responsibly expose.

This is why the platform sizes positions with a formula, not a feeling. Feelings do not scale across a portfolio. Rules do.

What Kelly is doing in plain language

The Kelly framework starts with a simple idea: when the market price is favorable relative to the estimated probability, the position deserves capital. When the edge is larger, the position deserves more capital. When there is no edge, the position deserves nothing.

The result is a fraction of bankroll. A Kelly fraction of 0.04 means 4% of current bankroll on this position. That fraction shrinks when the edge is thin and grows when the edge is wide. It naturally sizes down as the bankroll falls and sizes up as it grows.

Why fractional, not full

Full Kelly maximizes long-run expected portfolio growth under certain assumptions. In practice, those assumptions are rarely met precisely. Win probability estimates carry uncertainty. Prices move between model run and execution. The assumed independence of positions does not fully hold across a multi-sport slate.

Running full Kelly under realistic uncertainty produces higher variance and deeper drawdowns than the math on paper suggests. ArcVest uses quarter Kelly as a deliberate conservative scaling. The expected growth rate is lower, but the drawdown profile is substantially more controlled. For a platform with a public track record that builds trust over time, drawdown control is as important as growth rate.

The role of line shopping

Sizing and price quality are inseparable. The first determines how much the portfolio commits. The second determines whether that commitment was made efficiently. A small difference in entry price compounds across a portfolio. Readers who want to improve execution quality should always compare multiple market makers before committing capital.

What the minimum-allocation floor is for

ArcVest's minimum-allocation floor exists because not every mathematically positive position deserves a practically tiny allocation. In a pure model environment, extremely small position sizes can emerge from very modest edges. In a real operating environment, those positions may be too small to execute efficiently or too small to matter after friction.

The floor solves that by enforcing a minimum operational threshold. It is not there to exaggerate conviction. It is there to keep the portfolio executable. This is an important distinction. A minimum floor is an execution rule, not a confidence signal.

How the daily allocation works

Each day's card should be understood as a portfolio, not as a stack of isolated opinions. That means total allocation matters in addition to individual stake sizes. A disciplined system manages both the size of each position and the aggregate exposure of the daily set.

This is how ArcVest turns model output into an operating framework. Edge identifies which positions qualify. Fractional Kelly scales them. Portfolio caps control total exposure. The result is a repeatable allocation process instead of a collection of guesses.

For informational purposes only. Not financial or betting advice. Past model performance does not guarantee future results.

Playoff strength isn't full-season average or hot streak — it's a calibrated blend. ArcVest uses 75% full-season Pythagorean strength and 25% "Last 15" games for NBA playoffs, with 85/15 for NHL. This isn't arbitrary. It's tuned to playoff evidence while guarding against recency bias.

Why Full Season Stays Dominant

Full-season strength remains dominant because playoffs reward sustained quality. The NBA Pythagorean exponent (14.3) and NHL's PythagenPuck coefficient (0.458) derive win expectancy from points and goals scored and allowed across 82 games. That captures roster depth, coaching quality, and matchup-adjusted performance better than any short window. But late-season noise — load management, tanking decisions, clinch rest — distorts the full-season average. Enter the Last 15 overlay.

The Last 15 as Turbocharger

The Last 15 acts as a turbocharger, not a replacement. Recent Pythagorean strength is computed from the team's own most recent 15 completed games. For NBA, where load management inflates regular-season variance, 25% recent weight acknowledges momentum without overreacting. NHL's cleaner late-season incentives justify just 15% — hockey's physical toll shows in shorter meaningful streaks.

The ±8% Circuit Breaker

Unchecked, a 15-game heater could swing strength 20% or more — chasing noise rather than signal. The ±8% shift cap is the circuit breaker. Compute the deviation δ = R_last15 − R_season, clamp it to [−0.08, +0.08], then apply the weighted blend to the clamped value. This bounds the turbocharger while letting genuine signal through.

Tunable by Design

All parameters live in Config as named constants — no recoding required to adjust the blend:

This bounded approach embodies ArcVest's philosophy: math over narrative. Full season anchors reality. Recent form turbocharges relevance. The cap enforces discipline. Playoff edges emerge from that balance — not from chasing streaks.

For informational purposes only. Not financial or betting advice. Past model performance does not guarantee future results.

A team can be favored in the series right in front of it and still be a long shot to win the title. That is not a contradiction. It is how path-dependent probability works.

If a team has a 60% chance to win its current series, that sounds strong in isolation. But a championship future is not one position. It is a chain of conditional events. To lift the trophy, that same team might need to survive three or four consecutive rounds against increasingly strong opponents. Even before we account for matchup quality, injuries, rest asymmetry, or home-ice/home-court dynamics, the math compounds quickly.

The Compounding Problem

Here is the intuition. Suppose a team has a 60% chance in Round 1, then 45% in Round 2, 35% in Round 3, and 20% in the Final. Its title probability is not 60%. It is the product of the whole path.

That is why single-series odds are incomplete. They tell you the probability of surviving the next gate. They do not tell you how difficult the road becomes after that gate opens.

What Bracket Simulation Solves

Bracket simulation solves the compounding problem by mapping every plausible route through the field and aggregating the probability of each route. Instead of asking, "Can this team win tonight?" the model asks, "What is the full probability distribution of surviving the entire campaign?"

This matters because the market often prices the next series more efficiently than the entire path. A club with a favorable opening matchup may look attractive in series odds, but if the likely semifinal and final opponents are elite, its title path can still be thin.

Path-Pricing vs. Headline-Pricing

ArcVest's view is that futures should be path-priced, not headline-priced. The right question is never just whether a team is favored now. The right question is whether the market has fully accounted for the sequence of opponents it would need to beat next.

By running tens of thousands of simulated tournaments and tracking which teams survive each round, the model generates a full championship probability distribution — not a single series price, but a complete map of every path to the title and its likelihood.

For informational purposes only. Not financial or betting advice. Past model performance does not guarantee future results.

Every position on ArcVest passes through two independent filters before it's logged. Most candidates fail. The ones that survive are genuine edges — not just low odds or high-probability favorites.

Gate 1: Expected Value ≥ 5%

Expected value (EV) measures how much you expect to return per dollar risked, on average, if you made the same position thousands of times.

The model win probability comes from Pythagorean expectation — a formula that uses points scored and points allowed to estimate true team strength, independently of the win-loss record.

Gate 2: Positive Edge vs. No-Vig Probability

Even a position with positive EV might just be exploiting overpriced odds — not a genuine model edge. The second gate compares the model probability directly against the no-vig (fair) implied probability.

This ensures the model actually disagrees with the market's fair probability — not just the vig-inflated line.

Why Both Gates?

EV alone can be gamed by taking huge underdogs — technically positive EV but high variance. Edge alone doesn't account for the actual return magnitude. Together, they ensure every logged position is both mathematically positive in expectation and based on genuine model disagreement with the market.

What Comes Next: Kelly Sizing

After filtering, each position receives a quarter Kelly stake — a conservative position size derived from the EV and odds. All positions are then scaled uniformly so total daily risk stays within portfolio caps. No single position dominates the portfolio.

For informational purposes only. Not financial or betting advice. Past model performance does not guarantee future results.

Odds formats aren't just display preferences — they shape how you think about value. American odds dominate U.S. market makers because they're compact and intuitive for execution. Decimal odds rule in modeling because they make expected value math direct and repeatable. Neither is universally better, but one will serve your workflow more effectively depending on whether you're scanning lines or sizing positions.

American Odds: The Execution Standard

American odds are the default at FanDuel, DraftKings, and every U.S. market maker for a reason. The +/− sign tells you at a glance whether it's a favorite (−150 means lay $150 to win $100) or a longshot (+200 means win $200 on $100 risked). No mental math needed to classify the side.

Strengths: Native to U.S. execution — matches exactly what you see when placing the position. Compact for dense displays. Psychologically clear: positive feels like upside, negative feels like cost.

Weaknesses: EV requires a conversion step. For a quick scan, American wins. Open a spreadsheet for Kelly sizing and the conversions start adding up.

Decimal Odds: The Modeler's Language

Decimal odds treat every price as a total return multiple. A 2.50 decimal means $1 risked returns $2.50 total — $1.50 profit. Implied probability is always 1 ÷ decimal odds. Expected value simplifies to one formula regardless of whether you're looking at a favorite or a longshot.

Strengths: Universal payout formula, no sign-based branching. No-vig and Kelly sizing flow without reformatting. Consistent across sports, markets, and position types.

ArcVest's engine computes entirely in decimal internally — that's why the model stores model_odds as 1.91, not −110. It's the natural format for the two-gate filter, fractional Kelly, and CLV calculation.

Head-to-Head: Five Real Positions

Same five positions, same model probabilities — the only difference is which format you're reading.

The Trader Workflow

Scan in American. Your market maker shows −135, not 1.74. American matches the execution environment.

Model in decimal. Once you're computing EV, Kelly fraction, or CLV, decimal removes all the conversion friction. One formula, every market, every sport.

The hybrid approach is what disciplined traders actually use. American for the glance, decimal for the math.

For informational purposes only. Not financial or betting advice. Past model performance does not guarantee future results.