How the Confluence Score Works
Every number on Form55 is derived from public Canadian government data. Here is exactly how scores are calculated and what the backtest shows.
Read the backtest top-down. The score's job is to concentrate signal: the 435 signals scoring 75 or above returned +9.8% on average with a 63% hit rate, and the top 90–100 tier hit 75%. The broad overall figure (+6.7% avg, 44% hit) is diluted by thousands of low-scoring TSXV micro-caps the platform tracks but does not surface — that dilution is what the score is for, not a flaw in it. Hit rate climbs cleanly with the score:
6,706 signals backtested across TSX & TSXV companies, 2020–2026. Return is measured over the 90 calendar days following each signal date, equal-weighted. The tier table below is unfiltered — every score band, winners and losers.
| Score Tier | Signals | Signal Avg | Hit Rate |
|---|---|---|---|
| 90 – 100 Exceptional | 73 | +10.7% | 75% |
| 75 – 89 Strong | 362 | +9.7% | 60% |
| 60 – 74 Moderate | 1,310 | +5.2% | 49% |
| 40 – 59 Weak | 4,424 | +7.1% | 41% |
| 20 – 39 Low | 537 | +4.5% | 40% |
| Overall | 6,706 | +6.7% | 44% |
Methodology & caveats: Returns are equal-weighted — each signal counts once regardless of company size, measured over the 90 calendar days after the signal date. Survivorship: the 2020–2024 history was rebuilt from a deeper TSXV data feed that covers currently-listed companies, so names delisted before today are largely absent. That gap is concentrated in the low tiers and, if anything, flatters them — failed micro-caps that vanished aren't counted — so the 20–59 numbers shown are an optimistic read of a cohort the score already de-emphasizes. Delisted names are included where price data was available at the time and excluded where it was not (a disclosed omission). No transaction costs or slippage assumed. Past performance is not indicative of future results. This is not investment advice.
Confluence signals (insider buying + active lobbying in the same 90-day window) are markedly more consistent than standard insider-only signals — a 58% hit rate versus 41%. Average returns are similar (+6.8% vs +6.7%), so confluence isn't about bigger wins — it's about far fewer losers.
| Signal Type | Signals | Avg Return | Hit Rate | Median |
|---|---|---|---|---|
| Confluence Insider + Lobbying | 959 | +6.8% | 58% | +2.7% |
| Standard Insider only | 5747 | +6.7% | 41% | -0.4% |
Confluence is a consistency filter, not a magnitude amplifier. A 58% hit rate with a +2.7% median means fewer surprises — versus a roughly flat -0.4% median for standard signals — suitable for systematic positioning. Insider-only signals carry more variance: when they work, they tend to work bigger; when they fail, the drawdown is steeper. The Confluence badge on the leaderboard flags the more reliable cohort.
The single strongest predictor of magnitude in the model is the number of unique insiders buying. A lone insider (+8.0% avg) beats a crowd of 20+ (+2.8% avg) — large-cap, high-volume insider patterns (banks, pipelines) are mostly noise. The trade-off is variance: lone-buyer signals actually hit positive less often (41% vs 51%), they just pay more when they land. High conviction means bigger average returns with a wider spread, not more frequent wins.
| Unique Buyers | Avg 90-Day Return | Interpretation |
|---|---|---|
| 1 | +8.0% | Lone insider with conviction — highest signal |
| 2 – 4 | +5.6% | Small cluster, still meaningful |
| 5 – 19 | +4.2% | Diluted signal, possible plan participation |
| 20+ | +2.8% | Likely institutional or DRIP noise |
| Component | Max Points | Logic |
|---|---|---|
| Conviction | 30 | 1 unique buyer = 30 pts · 2–4 = 22 pts · 5–9 = 15 pts · 10–19 = 8 pts · 20+ = 3 pts |
| Buy/Sell Ratio | 20 | Ratio of buy value to sell value × 20. Pure buying = 20 pts. |
| Dollar Value | 25 | $1M–$10M = 25 pts (sweet spot) · $100K–$1M = 20 pts · $10M–$50M = 15 pts · $50M+ = 8 pts · <$100K = 5 pts |
| Lobbying Sweet Spot | 15 | 51–200 active regs = 15 pts (engaged, not ubiquitous) · 11–50 = 12 pts · 1–10 = 10 pts · 200+ = 5 pts (mega-lobbyists are noise) |
| Contract Wins | 10 | $10M+ federal or provincial contracts = 10 pts · $1M+ = 7 pts · 3+ wins = 4 pts · 1+ win = 2 pts |
| Private Placement Conviction | 8 | Credible investors writing cheques in a private placement is pre-public conviction. $1M+ deal = 8 pts · $100K+ = 6 pts · any deal = 4 pts. Sourced from Globe Newswire press releases and SEDI acquisition filings. |
| Insider Track Record | 10 | Credibility score of buying insiders based on their historical trade accuracy. ≥70% credibility = 10 pts · ≥50% = 6 pts · ≥30% = 3 pts |
| Confluence Bonus | 10 | Insider buying is the anchor. Insider buying + active lobbying + contract wins in the same 90-day window = 10 pts · insider buying + any one of lobbying, contracts or a private placement = 5 pts. A name with lobbying or contracts but no insider buying earns no confluence bonus — the bonus corroborates an insider signal, it doesn't replace one. |
| Cluster Timing (Velocity) | up to +10 | How tightly clustered open-market insider buying is. 2+ insiders buying within 1-3 days = 10 pts · same-day or within 4-14 days = 4-7 pts · single buyer or no cluster = 0. |
| Buy/Sell Reversal Pattern | up to +8 | Insiders who were net sellers in the prior 180 days and have switched to open-market buying — a pattern change is a stronger signal than consistent buying. 3+ reversing insiders = 8 pts · 2 = 5 pts · 1 = 3 pts. |
Component totals can exceed 100 before bonuses — the final score is capped at 100. A signal scoring 100 has maxed the conviction, value, and confluence components with velocity or reversal confirmation on top.
SEDI (System for Electronic Disclosure by Insiders)
Canadian Securities Administrators. All TSX & TSXV companies. 231,000+ transactions from 2020 to present.
Office of the Commissioner of Lobbying (OCL)
172,000+ registrations from 1996 to present. Matched to TSX tickers by company name.
Federal procurement databases
1.26M+ federal contracts. 41,300 matched to TSX issuers.
QC, ON, BC, AB, NS, NWT open data portals
542,000+ provincial & territorial contracts. 8,699 matched to TSX issuers. Paid tier only.
Globe Newswire press releases & SEDI acquisition filings
PP announcements detected via newswire RSS feeds and SEDI transaction code 22 (acquisition by private placement). Globe Newswire releases give the announced deal size; SEDI filings capture an insider's participation in the placement. Coverage is growing as the parser matures.
Elections Canada contribution records
Individual political donations (corporate donations banned federally since 2007), matched to insiders by name. Surfaced on company & insider pages as a factual political footprint — display-only, NOT a Confluence Score input.
Beyond the four scored inputs above, company pages surface additional public-data layers as context. These are shown for research — they are NOT Signal Score inputs (unless explicitly noted). Each is sourced from a public registry or feed.
Provincial mining-title registries — BC & Ontario
DataBC Mineral Titles Online and Ontario MLAS, mapped to TSX/TSXV tickers. New claims and land-package expansion (claims staked adjacent to ground a company already holds) — a quiet pre-announcement signal in mining.
Insider / director interlock
People who file as insiders at this company and at other TSX/TSXV companies, derived from SEDI filings (funds and 10%+ holders excluded). A well-connected insider taking a seat at a new junior can be an early tell.
Outstanding warrant blocks (analyst tier)
Strike, expiry and moneyness vs the current price, compiled from public filings & placement disclosures. In-the-money blocks near expiry signal likely dilution; deep out-of-the-money near expiry can flag a market-vs-insider divergence.
Canadian Impact Assessment Registry (IAAC)
Where a company's major project sits in the federal impact-assessment process — entering the impact-assessment / decision window is a permitting catalyst.
First Nations consultation / IBA status (mining)
Per-project Indigenous engagement status — Impact Benefit Agreements signed, consultation ongoing, or active opposition / legal challenge — compiled from public filings, registries and news (each row sourced). Often the gating factor for a Canadian mining project's viability.
CIRO Consolidated Short Position Report
Aggregate short position and net change per security, published twice monthly. Rising short interest against a bullish insider signal is a notable tension setup.
ClinicalTrials.gov (health / biotech)
Trials where a tracked company is the lead sponsor. Phase advances, completions and terminations are material catalysts for clinical-stage names.
SEDAR+ alerts & mining technical reports
Material change reports and other SEDAR+ filings, plus NI 43-101 technical-report detection for miners — recent filing activity surfaced on the company page.
Wikipedia pageview trends
A no-key public proxy for retail attention — a spike in a company's Wikipedia pageviews can accompany a news catalyst.
Curated roster on the SEDI register
A hand-curated roster of notable Canadian resource financiers & company-builders, flagged where they appear on a company's public SEDI insider register. When a marquee name is on the register, the involvement can be a conviction tell. Coverage reflects what SEDI captures, so a name surfaces mainly at their flagship vehicles. See all →
Federal grants & contributions
Federal transfer payments (SIF, Western Diversification, SDTC, NRCan and others) from the Government of Canada's Proactive Disclosure of Grants & Contributions, matched to a company by legal name. Government funding can be a pre-revenue validation and capital signal.
Every signal generated since January 2025 is tracked in real time at /track-record — individual named signals with actual TSX entry prices, 90-day exit prices, and benchmark comparison against XIU.TO over the identical window. Returns are verifiable against any TSX data source.
Closed signals (90-day window elapsed) show confirmed returns. No cherry-picking — every signal with score ≥ 40, one per ticker per month. 16 of 928 signals are quarantined for data-quality reasons before any statistic is computed. Headline figures use the tradable-universe weighting (entry price ≥ $1.00); a sizable share of the alpha is concentrated in mining-sector and less-liquid names — see the full methodology & validation breakdown for weighting schemes, sector attribution, and score-bucket statistics.
The 0–100 Confluence Score measures insider conviction on the buy side — the scale the backtest validates. Insider selling is a different, noisier signal, so we keep it on a separate categorical axis rather than folding it into the score. Conflating the two would degrade the validated buy-side number. A holding can therefore read, e.g., “71 ⚠ Watch”. The warning axis has three states: none, watch (one detector), and warning (two detectors).
| Detector | What it is |
|---|---|
| Bearish reversal | Sustained insider buying (3+ open-market buys over the prior 180 days) followed by an open-market sell cluster (2+ unique sellers in the trailing 30 days) — a turn in the same names that were being accumulated. |
| Score collapse | The Confluence Score fell 25+ points over 30 days on new insider selling (buy/sell-ratio deterioration). A score that decays only because past buying aged out of the 90-day window is treated as a neutral signal-expiry event, not a warning. |
This is factual pattern detection, not a recommendation. A warning describes what insiders did — it is never advice to sell. Warning states apply from launch forward and are not back-applied to historical track-record cohorts. Detectors that did not survive validation (cluster-distribution selling, lone credible-insider exits) were dropped.
Internal note (hidden in production): §5.3 historical validation completed; summary pending review before publication. Bearish-reversal is shipped as “developing” given small historical sample.