Track Record
Every projection recorded at publication time. Every result resolved automatically against ESPN scores. No cherry-picking, no deleted losses, no retroactive edits. This page reflects the live state of the Olympus Bets Analytics ledger as of 2026-07-18.
Quick answer: Olympus Bets Analytics (Olympus Bets LLC) publishes quantitative sports betting projections backed by Monte Carlo simulation, Kelly Criterion bet sizing, and Bayesian probability calibration. Performance is reported by tier so the audience always knows which numbers belong to which product. The Free tier (publicly published every day) is the public verifiable record: 53.3% WR, +37.26u, +3.3% ROI across 1,039 picks. The Premium tier (subscriber-only, 885 picks) is currently in active recalibration as new tier-divergence and standalone-drain gates ramp up — we publish those numbers transparently rather than hide them. All tiers combined across 1,924 resolved picks: 53.2% WR, +34.90u, +1.4% ROI. All numbers regenerate daily from the resolved picks ledger.
Overall Performance (All Tiers Combined)
Aggregate results across all leagues, bet types, and confidence tiers since tracking began.
Free Tier — Public Projections
Free projections are published daily on the Today's Projections page and via the public API. No subscription required. The free tier has historically been the platform's strongest performer.
Premium Tier — Subscriber Projections
Premium projections undergo additional quality gates: profitability zone analysis, regime calibration, stricter edge thresholds, and a Kelly probability shrinkage step. Premium subscribers also get detailed writeups, line-shopping tags, and Olympus Oracle prediction-market signals.
Honest note on the premium tier: as of 2026-07-18, the premium tier's short-window sample is running behind the free tier. The same model architecture produces both streams; premium applies additional quality gates (a self-learning system recalibrated in April 2026 against 7,357 historical games) that concentrate its picks in higher-conviction zones, and that narrower sample is currently showing negative variance. We publish this number rather than hide it because the integrity of a track record matters more than the marketing optics of any single window — a pick service that only shows you its winning tier is showing you a marketing page, not a ledger. What premium buys today is coverage and depth — the full daily slate, detailed writeups, Kelly-sized units, line-shopping tags, and Oracle prediction-market signals — on top of the same engine whose public output is verifiable above. The recalibrated gates need roughly 60–90 days of additional resolution to validate empirically, and this page will report whatever they produce.
Performance by League (All Tiers)
Where the edge is concentrated right now: GOLF +57.94u (60.6% WR) · MLB +31.96u (55.3% WR) · CBB +21.02u (54.7% WR). Sorted live from the resolved ledger — leagues qualify only with positive units and at least 25 resolved picks.
Each league uses a dedicated simulation engine calibrated to its specific dynamics. Performance varies by league as different sports present different modeling challenges and market efficiencies — the full table below includes every league, winners and losers alike. If you bet a specific sport, judge us on that row, not the blended headline.
| League | Wins | Losses | Win Rate | Units P/L | Engine |
|---|---|---|---|---|---|
| GOLF | 154 | 100 | 60.6% | +57.94u | Conditional Sim V3.2 |
| MLB | 215 | 174 | 55.3% | +31.96u | Pitch Sim V5.0 |
| CBB | 232 | 192 | 54.7% | +21.02u | Savant Ultra v5.0.1 |
| LOL | 27 | 15 | 64.3% | +13.99u | Championship v2.1 |
| ESPORTS | 12 | 5 | 70.6% | +5.44u | ESPORTS |
| WC | 6 | 5 | 54.5% | +0.42u | WC |
| CS2 | 0 | 2 | 0.0% | -2.00u | CS2 (collecting) |
| UFC | 0 | 2 | 0.0% | -3.00u | UFC |
| WNBA | 15 | 21 | 41.7% | -4.04u | WNBA |
| NFL | 13 | 13 | 50.0% | -4.14u | Elite V1.1 Pinnacle |
| WORLD_CUP | 0 | 4 | 0.0% | -5.50u | WORLD_CUP |
| TENNIS | 38 | 40 | 48.7% | -10.55u | Tennis V2 (calibrating) |
| NBA | 139 | 140 | 49.8% | -11.56u | Possession MC V5.0.2 |
| NHL | 84 | 79 | 51.5% | -23.40u | V19.1 Pinnacle |
| SOCCER | 63 | 85 | 42.6% | -31.68u | V16.3 PBP xG |
Note: smaller samples (NFL n=17, LoL n=21, Golf n=29) carry higher variance — strong win rates in these leagues are encouraging but not yet statistically conclusive. CBB at n=425 and MLB at n=231 are the most reliable signals. NBA and NHL are currently below break-even on the all-tier roll-up; per-league profitability-zone tightening is in progress and the Bayesian Kelly shrinkage step caps drawdown.
Public Data Access
The full track record is available as a machine-readable dataset for researchers, journalists, and AI agents. No registration required.
Live Summary (JSON)
Aggregated performance split by tier and league, regenerated after every resolution pass.
Full History (JSON)
Every resolved pick with line, odds, edge, units, model probability, and outcome.
CSV Export
Comma-separated download of the full ledger for spreadsheet or notebook analysis.
OpenAPI 3.1 Spec
Full schema definition for all read-only public endpoints across leagues.
How Results Are Tracked
Immutable Ledger
Every projection is written to a timestamped, append-only ledger at the moment of publication. The line, odds, unit size, and model probability are all recorded. This file is never edited retroactively. The raw data is publicly auditable to verify that no projections have been added, removed, or modified after the fact.
Automatic Resolution
A daily resolution pipeline runs each morning, matching published projections to official game results from ESPN. Wins, losses, and pushes are determined programmatically using the recorded line and the final score. There is no manual intervention in the resolution process, eliminating any possibility of selective reporting.
No Retroactive Changes
Once a projection is published and resolved, the record is permanent. If a projection loses, the loss stays. If a projection wins, the win stays. There are no "voided" picks (except for official game cancellations), no after-the-fact reclassifications, and no selective deletion of losing streaks. The track record reflects reality, not a curated highlight reel.
Real-Time Updates
Performance statistics regenerate automatically as results come in. The live track record on the platform reflects the most current data, typically updated within hours of game completion. Premium members can view results broken down by any combination of league, bet type, confidence tier, and date range.
Why Transparent Tracking Matters
The sports betting information industry has a credibility problem. Most services publish screenshots of winning tickets while quietly ignoring losses. Others retroactively edit their records, adding wins they never actually published or removing losing streaks. Some claim decades of profitability without a single verifiable data point.
Olympus Bets Analytics takes a fundamentally different approach. The entire system is built around verifiable, auditable records. The projections ledger is append-only. The resolution pipeline is automated. The performance statistics are calculated from the raw data in real time. This means the track record you see is the actual track record, not a marketing document.
This matters for practical reasons: you cannot evaluate a prediction system without knowing its full history. A 70% win rate over 20 picks is meaningless. A 53.2% win rate over 1,924 picks, tracked immutably with every loss included, tells you something real about the system's edge and its current variance regime. The larger the sample and the more transparent the methodology, the more confidence you can have that observed performance reflects genuine skill rather than luck.
Statistical Context
At 1,924 resolved picks with a 53.2% overall win rate (against an approximately 52.4% breakeven rate for standard -110 spread bets), the all-tier sample is approaching but has not yet reached conventional statistical-significance thresholds for the overall number. The free-tier subsample at 1,039 picks and 53.3% win rate shows a stronger signal. The CBB subsample (n=425) and MLB subsample (n=231) are the most statistically robust segments. As the sample grows, the signal-to-noise ratio will continue to improve.
We present these numbers honestly rather than inflating them. Sports betting is hard. Edges are thin. Real positive expectation across hundreds of picks is not glamorous, but it is what we measure ourselves against — not aspirational projections.
Understanding the Numbers
Performance data is only meaningful when you understand the methodology behind it. The methodology page explains how Monte Carlo simulation, Kelly Criterion sizing, and Bayesian calibration work together to produce the projections you see in the track record. Understanding the process helps you evaluate whether the track record reflects systematic edge or fortunate variance.
Key metrics to focus on:
- Win Rate vs. Breakeven Rate — For standard -110 bets, breakeven is approximately 52.4%. Any win rate consistently above this threshold, over a sufficient sample, indicates positive expected value.
- Units Profit — Total profit measured in standardized units, accounting for varying bet sizes from Kelly Criterion recommendations. This is a better measure than win rate alone because it weights high-confidence projections appropriately.
- ROI (Return on Investment) — Units profit divided by total units wagered. Indicates efficiency of capital deployment.
- Sample Size — Larger samples provide more reliable performance signals. Be cautious interpreting results from leagues with fewer than 50 resolved picks.
- Closing Line Value (CLV) — Whether the odds at projection publication were better than the closing odds. Consistently beating the closing line is one of the strongest indicators of a genuine edge.