Back to cover

SportsLogicOS Framework

MODEL INPUTS · MARKET EVIDENCE · SELECTION & BANKROLL RULES

Operational Philosophy

Beyond Narrative Betting

SportsLogicOS is a decision-support tool for evaluating candidate wagers. It combines team and market estimates, current prices, and risk controls; it does not guarantee outcomes or profitability.

A useful way to describe the workflow is: estimate outcomes, evaluate the available price, then manage exposure. Inputs and checks vary by sport and market.

The Mathematical Reality

Break-even at −11052.38% break-even
Break-even at other pricesDepends on the odds
Win rate by itselfDoes not establish profit
Accuracy claimsNeed market and sample context
Overall Grade Key — A+ through C

Evaluated candidates receive a conviction score and a server-stamped overall grade from A+ through C. The grade is mapped from the score; candidates rejected by an eligibility gate may be shown as C in the discarded-candidate view. Passing a conviction floor alone does not qualify a pick: EV, sport- and market-specific checks, mode rules, and exposure limits also apply. The slate is not built to fill a fixed number of daily picks.

A+Conviction score ≥ 0.78
AConviction score ≥ 0.66
A-Conviction score ≥ 0.55
B+Conviction score ≥ 0.44
BConviction score ≥ 0.33
B-Conviction score ≥ 0.22
CConviction score < 0.22

Default conviction floors: Full Kelly 0.35 · Half Kelly 0.42 · Quarter Kelly 0.50. These settings are configurable; passing the floor is necessary, not sufficient.

Pillar I
Quantitative Handicapping

The quantitative pipeline combines a probability estimate with bounded inputs from the relevant sport and market. Inputs are scoped by availability, freshness, overlap, and sport-specific rules.

·Team ratings and market-derived ratings
·Sport-specific performance indicators, bounded by model rules
·Fresh injury and availability adjustments where data is available
·Rest and travel interactions in supported sports
·MLB starter, bullpen, park, and umpire adjustments
·Weather and scoring context in supported totals models
System Implementation
Auto-Slate · Sport models · Injury context · MLB pitcher/bullpen inputs · Rest and travel · Market-derived ratings
Pillar II
Market Evidence & Review

Market evidence and post-wager evaluation have separate roles. Selected line-movement signals require verified Pinnacle history; public split data is retained for context, and closing-line value is evaluated after the market closes.

·RLM and steam use verified Pinnacle line or price movement
·Overlapping movement signals are deduplicated
·Other-book breadth may confirm a move but cannot originate a Pinnacle signal
·Public ticket and money splits are audit/display context only
·CLV is retrospective and excluded from live conviction ranking
·Derivative-market support is limited and sport-specific
System Implementation
Pinnacle movement checks · Market context · CLV tracking and calibration
Pillar III
Strategic Execution

A candidate is not selected on a model estimate alone. The engine applies price, EV, sport- and market-specific checks, conviction rules, and the selected bankroll mode before sizing a position.

·Expected value uses the available price; the base directional EV floor is +0.5%
·Additional sport- and market-specific checks can apply
·Conviction floors vary by bankroll mode and are configurable
·NFL key-number checks apply where supported
·Auto-slate base sizing uses EV and conviction, not rank position
·Mode-specific tier brackets and daily exposure caps can reduce or remove positions
System Implementation
EV checks · Key-number checks · Signal grade · Eligibility gates · Kelly-derived sizing
Capital Allocation · Auto-Slate Sizing

Auto-slate base units are calculated from EV and conviction, not assigned by slate rank. When the sharp score is explicitly zero, the base size is multiplied by 0.75. The selected mode then applies its tier-specific bracket and daily portfolio cap. Public ticket/money splits do not affect confirmation or sizing. The settlement-aware Kelly calculation is not enabled for live sizing.

Sizing Order
EV + conviction base calculation → sharp-score adjustment → 0.5–4u base bounds → mode/tier bracket → daily cap
1u represents 1% of the bankroll used for sizing
Conviction Promotion Floor
Candidates below their mode's conviction floor are not promoted. Passing the floor does not bypass EV, sport/market eligibility, tier, or exposure checks. Default floors are configurable.
Auto-Slate Stake Calculation

The current base calculation is driven by EV and conviction. The selected bankroll mode applies an additional tier bracket and daily portfolio cap; the values below describe the base calculation and default limits, not a rank ladder.

Base formulaEV% ÷ 4.4 × (0.70 + 0.60 × conviction)A 25%-Kelly-inspired approximation at −110 when inputs are present; missing/nonpositive EV or missing conviction currently falls back to 1u.
No sharp score× 0.75Applied when the sharp score is explicitly zero.
Per-pick bounds0.5–4.0uRounded to the nearest 0.5u, then clamped before mode-specific limits.
Bankroll unit1u = 1%One unit equals 1% of the bankroll used for sizing.
Portfolio capMode-dependentDefault daily caps are 2.0%, 1.5%, and 1.0%; settings may override them within the configured hard limit.
Bankroll Mode Rules — Selection & Exposure

The selected mode applies mode-specific conviction floors, tier brackets, and daily exposure limits to the user-facing slate. Full, Half, and Quarter Kelly are mode labels; the current auto-slate base size is calculated from EV and conviction, while mode rules can cap or remove it. Modes do not switch automatically after losses.

Full Kelly
StakingWidest mode-specific tier brackets
Per-pick cap4u maximum
Daily cap2.0% default (~2u at 1% per unit)
Tier CExcluded from every official slate
Live CNot suppressed by this mode; other gates apply
Conv. floor0.35 default; configurable
Half Kelly
StakingHalf-Kelly tier ceilings
Per-pick cap2u maximum
Daily cap1.5% default (~1.5u at 1% per unit)
Tier rulesTier C and Live C suppressed
Conv. floor0.42 default; configurable
Quarter Kelly
StakingQuarter-Kelly tier ceilings
Per-pick cap1u maximum
Daily cap1.0% default (~1u at 1% per unit)
Tier rulesTier C and Live C suppressed
Conv. floor0.50 default; configurable

The exact-settlement Kelly sizing path is currently disabled.Live auto-slate units use the published EV/conviction calculation, then mode-specific tier ceilings and the configured daily cap. The daily caps shown above are defaults and may be changed in settings. Review-All displays qualifying candidates before the mode-specific filters and skips daily portfolio trimming; any shown units are suggestions, not an enforced exposure limit.

Selected Model Inputs, Market Evidence & Controls
Model Inputs
Team and market-derived ratings
Sport-specific performance indicators
Fresh injury and availability adjustments where supported
Rest, schedule, and travel interactions in supported sports
Baseball starter, bullpen, park, and umpire context
Weather and scoring inputs for supported totals markets
Market Evidence & Review
Pinnacle-sourced line movement with provenance checks
RLM and steam signals are deduplicated when they overlap
Pinnacle price comparison where available
Public ticket and money splits are audit context, not scoring inputs
Closing-line value is tracked after the wager for evaluation
Selection & Sizing
Expected value uses the available price
Sport- and market-specific eligibility checks
Conviction score and server-stamped overall grade
Mode-dependent conviction floors and tier rules
EV- and conviction-based auto-slate sizing
Daily bankroll exposure limits
Model & Evaluation Principles
Do Not Chase Losses
A losing run does not make the next wager more likely to win. Keep decisions within a pre-set bankroll plan.
Evaluate the Available Price
Expected value depends on both the estimated outcome probability and the price actually available. Comparing accessible books can identify a better price; no timing rule guarantees one.
Verify Market Evidence
RLM and steam rely on verified Pinnacle line or price movement. Public ticket and money percentages are retained for audit context and do not create a sharp signal.
Win Rate Is Not Profitability
A win percentage has no standalone meaning without the odds and the sample behind it. CLV is a separate, retrospective measure of price quality.
Do Not Assume Signals Are Independent
Some inputs overlap. The scoring pipeline applies deduplication and correlation controls where implemented; the displayed categories are not a count of independent models.
A Smaller Slate Is Valid
Candidates must clear the applicable EV, conviction, sport, market, and bankroll-mode rules. The system does not need to fill a fixed number of daily picks.