Open paper-trading research lab for value betting methodology. We publish the math, the database, and every loss — because trust is built through transparency, not secret formulas.
Our north star: a fully open research lab where every formula, every paper bet, and every miss is publicly verifiable.
We're building an open paper-trading research lab for value betting methodology. The mission is narrow but uncompromising: trust through transparency, not secret formulas.
Industry standard right now is black-box scanners (Allbestbets, RebelBetting, Pikkit, Trademate) that publish ROI claims without methodology, without paper-trade history, and without CLV transparency. We publish all three: full math, live paper-trade DB, and CLV beat-rate as our primary KPI.
Target audience: educated retail value bettors with €500–€5k bankrolls, learning the math, not professional arbers. We're not building for sharp syndicates or live-betting bots — we're building for the literate retail bettor who got burned by ROI claims and wants to verify the methodology before risking a euro.
A quick scan of paid value-betting tools shows a consistent pattern — claims without verification, methodology hidden behind paywalls, no public paper-trade DB.
Black-box ROI claims dominate the industry (~70% of paid tools). Sites advertise "10% monthly ROI" or "120% annual returns" without disclosing sample size, time window, bookmaker mix, or the inevitable variance bands. Without N, the number is marketing — not data.
Methodology obscurity is the second pattern. Even premium players like RebelBetting publish a "How it works" video without the actual math — fair-line construction, EV thresholds, and Kelly fractions are abstracted away. You're asked to trust the engine, not understand it.
No public paper-trade verification. ROI claims are unverifiable when the underlying bet log is private. There is no industry standard for publishing every paper bet — wins, losses, and voids — with timestamps and CLV. We're trying to change that.
Our answer: open methodology + open paper-trade DB + CLV-first framing. See our head-to-head comparisons: Σ vs. paid tools, vs. Allbestbets, vs. RebelBetting, vs. Pikkit, vs. Trademate, vs. OddsPortal.
We're pre-real-money. Every signal is a paper bet logged with timestamps, fair-line snapshot, and CLV after settlement. Real money is gated by a quantitative test, not a calendar date.
Stage: paper-trading research, pre-real-money validation. The engine runs production cron jobs, posts every value signal to Telegram and Bluesky, and writes outcomes to the paper-trade DB after settlement.
Coverage: multi-sport — tennis (ATP+WTA), soccer (EPL/La Liga/Bundesliga/Serie A/Ligue 1/UCL/RPL/MLS and 30+ leagues), hockey (NHL/KHL/Liiga), basketball (NBA/Euroleague), baseball (MLB), American football (NFL), MMA UFC, boxing, AFL, NRL. The list grows as fair-line consensus becomes reliable for a new league.
Engine: v1.9 + M297 anti-bleed guard, live since 13 May 2026. After the audit phase identified soft-soccer-league bleed, we paused 4 minor leagues and reset baseline stats. Read the deep-dive: Engine v1.9 + M297 anti-bleed methodology.
Sample size: N is accumulating from the 13 May reset. Currently small post-reset — see live numbers at /track/. Pre-reset archive contains 280 bets used to calibrate the engine. Real-money phase is locked until N=200+ paper bets show consistent positive CLV (statistically significant beat-rate above 50%).
Timeline of milestones — from initial conception in April 2026, through audit and baseline reset, to the multi-day megasprints that produced the current production stack.
Boring, well-tested infrastructure. SQLite where Postgres would be overkill. Custom in-house engine where no library fits. CI runs on every commit.
SQLite in WAL mode handles 262 MB of paper-trade history, sharp-book snapshots, and Pulse engagement metrics. WAL mode allows concurrent read while cron tasks write. Postgres migration is reserved for after the N=200 milestone if necessary.
Median-of-3 sharps consensus (Pinnacle + Betfair Exchange + Matchbook). ADR-008 LOCK filter (k∈[1.30,3.00], EV≥3%, sharp_spread≤8pp). M297 anti-bleed guard (league blacklist + draw-outcome skip for soft soccer). News-flag down-weighting from RSS corroboration.
Day-schedule polling resolves 90%+ of paper bets within 6 hours of event close. Walkover detection from Sofa day-schedule sufficient. Stuck-bet auto-void after 48 hours for niche leagues without resolver coverage.
python-telegram-bot drives @value_sniper_bot DM flow (26 commands, RU+EN parity). atproto cross-posts to Bluesky in English. Resend handles email digests and Premium subscriber notifications (M346–M349).
Ubuntu 24.04 on Ruthenium DE. Cloudflare Pages serves the lander, Workers handle API endpoints, R2 holds daily off-site DB backups (30-day retention). Old VPS decommission across 3 service tiers.
69 pytest tests cover engine filter logic, fair-line median computation, CLV derivation, and bilingual DM templates. GitHub Actions CI runs on every push since M322. Pre-commit guard blocks files >50 MB from polluting the repo.
Open methodology: all formulas at /methodology/, paper-trade DB live at /track/. Future: Engine v2.0 design doc references — line-movement signal incorporation post-N=200, CLV-anchored re-weighting of fair-line consensus.
Small on purpose. One operator end-to-end so methodology stays consistent. Not hiring.
Founder: Сано (Alexander). Solo founder. Background in software engineering and quantitative methods. Built the engine, the lander, the bot, and the paper-trade DB from scratch starting 26 April 2026. Anonymous-preferred default unless press explicitly requests a profile interview.
Looking for: educated retail beta users to stress-test signals and report edge cases. Not hiring currently — solo operator is intentional through the N=200 milestone to keep methodology consistent. Post-N=200 the team picture may change.
Contact: hello@valuesniper.app for general inquiries. TG admin via @valuesniper_ru public channel. Press resources at /press/.
Six principles that drive every product decision — from how we publish a missed bet to whether to launch a feature without backtest.
Explicit scope. We say no to a lot, on purpose. Every line below is a feature request we have already declined.
Public channels and a direct inbox. For press, see the press kit. For general questions, email or DM the bot.