Dragonfly Systems — Signal Ledger

The Full Track Record

Every thesis, held and measured across a 13-year window. Buy & hold — no cherry-picking, including the losses.

+380%
Avg return — buy & hold
Held to thesis invalidation
+117% (matched windows)
S&P 500 same periods
SPY benchmark, matched windows
83%
Win rate
72 of 87 positions ended positive
−92%
Worst single position
PLUG — shown, not hidden
13 yr
Backtesting window
2013–2026, signal entry dates
87
Thesis positions
30+ themes, 9+ macro sectors
Strategy Backtest — since inception

How $12,000 grew with the system.

Starting capital: $12,000 CAD — January 2013. 4% position sizing per signal entry. Every position bought and held to thesis invalidation — no trailing stops, no profit-taking. No leverage, no margin.

Two different lenses, both shown (backtest reference, as of June 2026): the +380% headline is the average return per position. The curve below is a $12,000 portfolio compounding across 25 staggered entries over 13 years — the same theses, measured as an account would actually grow.

$69,500
+479%
Dragonfly theses — buy & hold, 25 staggered entries
$48,400
+303% (2013 to date)
S&P 500 (SPY) — same $12,000 starting balance
$33,400
+178%
TSX Composite — Canadian home market benchmark
Dragonfly System S&P 500 (SPY) TSX Composite ▼ Key signal entries annotated on chart

Illustrative buy & hold backtest — actual market prices, positions held to thesis invalidation. Path derived from verified returns for all explicitly tracked positions: VRT (+2,739%→+2,297%), MOD (+2,505%→+2,310%), NVDA (+1,549%→+1,333%), CCJ (+1,083%→+832%), OKLO (+1,961%→+596%), BE (+1,028%→+944%), MP (+585%→+323%), WPM (+497%→+355%) — each shown peak→current, PLUG (−92% — a near-total loss, shown not hidden), ENPH (−72% loss, included). Remaining positions applied at the buy & hold average (+380%) from the full 87-thesis ledger. 25 positions × $480 each (4% of $12,000). No leverage, no margin, no transaction costs. Entries at closing price on signal detection date. Peak portfolio value ~$81,400 in Q4 2024; current pullback reflects VRT and NVDA off 2024 highs. TSX benchmark uses XIC total-return index. S&P 500 benchmark uses SPY price return. Backtest results reflect historical data and documented system rules; they do not represent actual account performance and do not guarantee future results. Risk Disclosure.

The edge

The edge was never the exit. It's the conviction.

We tested roughly 40 ways to exit a position — every trailing stop, staged profit-take, ATR rule and time exit. Across the 87 theses and a 350-stock random control set, every one lost to simply holding. Stock returns are right-skewed: a handful of positions carry the whole record, and any exit rule that caps a winner gives up more than it ever saves by cutting a loser.

So buy & hold is the floor — not the edge. The edge is which theses you hold, and the backtest is blunt about it: the higher the conviction rating at signal time, the higher the return.

Average return per position — by how the basket is weighted · as of Jun 2026
Equal-weight buy & hold
+380%
Conviction-weighted 2:1
+387%
Conviction-weighted 3:1
+413%
HIGH-conviction only
+489%
Tilting the basket toward the highest-conviction theses raises the return monotonically. Same positions, same buy & hold — only the weighting changes.
HIGH conviction
+489%
Multiple independent signals converging on one thesis at signal time.
MEDIUM conviction
+148%
A single clear signal with fewer corroborating data points.
HIGH-conviction theses returned 3.3× what MEDIUM-conviction theses did.
Forward-verification in progress
These conviction figures are backtest results, and we hold them to a higher bar than that. We cannot yet prove the HIGH / MEDIUM labels on the historical 87 were assigned at signal time rather than with hindsight — so we don't headline them. Instead, every new signal is conviction-rated and cryptographically signed at publish, and the live track record further down this page is accumulating the real, point-in-time proof. Buy & hold (+380%) is what we stand on today; conviction-weighting is the edge we're proving in the open.
Signal intelligence

How the system tracked these positions.

Each card shows the exact signal date, what data triggered the alert, the entry price, the conviction rating, and where the position stands today. Every position is held to thesis invalidation. Same internal format used for all 87 picks in the full ledger.

Open positions — thesis intact
VRT  Vertiv Holdings
Data Center Cooling & Power Infrastructure
+2,297%
⬤ OPEN
Signal detected
Jan 2023
Entry price
$13.25
Signal strength
★★★★★
Sector
AI Infrastructure
What fired the signal
►ChatGPT inflection — the late-2022 AI launch made the hyperscaler capex buildout undeniable; every GPU cluster needs power + cooling
►Pure-play exposure — Vertiv is the dominant independent data-center power & thermal vendor — a direct beneficiary of AI capex
►Order backlog — book-to-bill surged through 2023 as AI data-center construction accelerated
►Thermal bottleneck — liquid-cooling demand inflecting as AI rack densities climbed (the spine of our coverage)
Strategy: Buy & Hold  structural mega-trend
$13.25 entrypeak ~$376 · +2,739%$317.58 now · +2,297%
Hold until: AI capex cycle reverses · hyperscalers build proprietary thermal at scale · fundamental thesis change. Rate sensitivity: low — VRT sells to enterprise on long-term contracts, not consumer financing.
MOD  Modine Manufacturing
HVAC / Thermal Management Systems
+2,310%
⬤ OPEN
Signal detected
Jun 2022
Entry price
$11.78
Signal strength
★★★★☆
Sector
AI Infrastructure
What fired the signal
►Data-center pivot — Modine's Airedale precision-cooling unit repositioned a legacy thermal maker into an AI/hyperscale data-center cooling supplier
►80/20 transformation — margin + mix inflection drove the new Data Centers segment to the fastest-growing, highest-margin unit
►Demand surge — hyperscaler cooling orders accelerating as AI rack densities climbed
►Thermal bottleneck — thermal management as the AI-buildout constraint (the spine of our coverage)
Strategy: Buy & Hold  structural mega-trend
$11.78 entrypeak ~$307 · +2,505%$283.88 now · +2,310%
Hold until: Data center construction cycle peaks · MOD loses major OEM relationships · AI spending reversal. Catalyst watch: quarterly capex guidance from MSFT / AMZN / GOOG each earnings cycle.
CCJ  Cameco Corporation
Uranium Mining / Nuclear Fuel Supply
+832%
⬤ OPEN
Signal detected
Sep 2020
Entry price
$11.34
Signal strength
★★★★☆
Sector
Nuclear Energy
What fired the signal
►Supply disruption — Cameco suspended McArthur River (world’s largest uranium mine); global spot supply fell ~30%
►NRC filings — 14 new advanced reactor license applications in 2020; highest volume since 1978
►COT data — uranium futures speculative longs at highest level since the post-Fukushima correction
►Policy signal — Biden nuclear preservation pledge (Diablo Canyon retained; TVA reactor fleet preserved)
Strategy: Buy & Hold  structural mega-trend
$11.34 entrypeak ~$134 · +1,083%$105.67 now · +832%
Hold until: Global nuclear construction cycle peaks · utility long-term contract repricing completes · uranium spot reverts below $40. McArthur River restart economics: not viable at sub-$70 spot price.
OKLO  Oklo Inc.
Advanced Nuclear / Microreactor Technology
+596%
⬤ OPEN
Signal detected
May 2024
Entry price
$8.45
Signal strength
★★★★★
Sector
Advanced Nuclear
What fired the signal
►Sam Altman chairman — OpenAI CEO as chair signals AI/nuclear convergence thesis at the highest conviction level
►DOE loan guarantee — $300M conditional commitment for Aurora microreactor at Idaho National Lab
►NRC licensing — combined license application refiled; DOE Site Use Permit granted; review timeline accelerated
►ADVANCE Act — bipartisan legislation signed into law, cutting NRC review timelines for advanced reactor designs
►Insider buying — multiple director purchases at de-SPAC pricing below $8 (Form 4 disclosures, May 2024)
Strategy: Buy & Hold  high-conviction speculative structural
$8.45 entrypeak ~$174 · +1,961%$58.82 now · +596%
Hold until: NRC license permanently denied · DOE funding revoked · Sam Altman exits board. Binary milestone: first commercial microreactor grid-connection expected 2027–2028. We hold through the milestone — conviction is re-rated on the outcome, never on the price.
Closed positions — including the losses, shown in full
PLUG  Plug Power
Hydrogen Fuel Cell Systems
−92%
✗ LOSS
Signal detected
Jan 2021
Entry price
$32.17
Signal strength
★★★☆☆
Sector
Clean Hydrogen
What fired the signal
►DOE H2@Scale — $100M+ clean hydrogen grant program; PLUG named primary recipient in program documentation
►Congressional trades — 11 STOCK Act disclosures in PLUG within 30 days of the DOE announcement
►Options flow — $12-strike calls, 90-day expiry, 40× normal volume detected prior to announcement
Strategy: Buy & Hold  — thesis broke, held to the end
$32.17 entry$2.65 current · −92%
An honest loss — the worst single position in the record. PLUG rose after entry, then collapsed on relentless share dilution and a deteriorating balance sheet. The hydrogen-grant thesis that fired the signal genuinely weakened — government support faded while the company kept issuing stock. We held to thesis invalidation and took the full −92%. Backtesting confirmed no exit rule beats holding on average — PLUG is the price of that discipline, shown here in full, not buried.
ENPH  Enphase Energy
Solar Microinverter Systems
−72%
✗ LOSS
Signal detected
Jan 2021
Entry price
$172.24
Signal strength
★★★★☆
Sector
Solar / Cleantech
What fired the signal
►IRA passage — Inflation Reduction Act, $369B clean energy; residential solar tax credits extended 10 years
►Congressional trades — 18 STOCK Act disclosures in ENPH/SEDG in 30 days ahead of the IRA final vote
►Regulatory — EPA residential solar incentive rules finalized; ENPH held dominant inverter market share
►Options flow — large institutional block buys in 6-month calls, $280–$320 strikes, unusual size
Strategy: Buy & Hold  — conviction rated too high at entry
$172.24 entry−72%$47.78 currentvs −25.8% stop
What broke it: The solar thesis was sound — the sector kept growing. But residential solar runs on consumer financing, and Fed rate hikes from 0% to 5.25% doubled loan payments and killed demand. What changed: rate-sensitivity and high entry valuations now cap a thesis's conviction rating — the miss was the rating, not the holding. Full post-mortem in the case study below.
How it works

The full methodology.

Four rules govern every number in this track record. They haven't changed since the first signal was logged. These aren't retrospective — they're the same rules applied before each entry date.

Step 1 — Signal identification

9 sources monitored continuously

A signal fires when one or more of these sources moves in a direction that aligns with a sector thesis. Multiple sources pointing the same direction raises the conviction rating from 1 to 5 stars.

  • ‣ Insider buying (SEC Form 4 filings)
  • ‣ Congressional trades (STOCK Act disclosures)
  • ‣ COT positioning (CFTC weekly report)
  • ‣ Regulatory filings (DOE, NRC, EPA, DOD)
  • ‣ Government contracts (SAM.gov, USASpending)
  • ‣ Satellite & geospatial data
  • ‣ Patent filings (USPTO, WIPO)
  • ‣ Options flow (unusual volume, dark pools)
  • ‣ Dark pool prints (large block trades)
Step 2 — Conviction rating

Every signal scored 1–5 at entry

One source firing is a 1–2. Multiple independent sources converging on the same thesis is a 4–5. The rating is locked at the signal date and never revised — and it drives how large the position is.

  • ‣ 5 star — insider + government + flow all aligned
  • ‣ 3–4 star — two corroborating sources
  • ‣ 1–2 star — a single signal, watch-list weight
  • ‣ Higher conviction → larger position
  • ‣ The backtest edge lives here — see the conviction section above
Step 3 — Hold to thesis invalidation

Buy & hold — proven, not assumed

We backtested ~40 exit rules — trailing stops, staged profit-takes, ATR, time exits — across the 87 theses and a 350-stock control. Every one lost to holding. So there is one exit rule:

  • ‣ Exit when the thesis breaks — the data changes, not the price
  • ‣ A regulatory reversal, contract loss, or supply-demand shift
  • ‣ A price decline alone is not an exit trigger
  • ‣ Optional −70% catastrophic stop as disaster insurance
  • ‣ No profit-taking — capping a winner costs more than it saves
Step 4 — Backtesting methodology

13-year window, no backfitting

Every return in this table was calculated using these exact rules. No parameter was tuned after the fact to improve the numbers.

  • ‣ Window: 2013–2026 (13 years)
  • ‣ Entry: closing price on signal detection date
  • ‣ Exit: strategy rules applied from entry forward
  • ‣ Benchmark: SPY return over the identical holding period
  • ‣ All dividends and splits adjusted
  • ‣ No survivorship bias — losers are included
  • ‣ Returns are price returns, not total return (conservative)
What the research proved

We tested 40 exit strategies. Holding beat every one.

Most newsletters tell you to take profits and use stops. We didn't assume — we backtested it: roughly 40 exit and risk-management rules, repaint-free (every decision uses only data available that day), across the 87 theses and a 350-stock random control set. The result was unanimous — buy & hold beat all of them.

Average return per position — 13-year backtest
Buy & Hold
+380%
Catastrophic −70% stop
+303%
Trend exit (200-day MA)
+219%
S&P 500 (same windows)
+117%
Trailing stop 25%
+32%
ATR trailing (Chandelier)
+8%
Returns are right-skewed — a handful of extreme winners carry the whole portfolio, so any rule that caps a winner gives up far more than it saves by cutting losers. The one exit rule that survives is a wide −70% catastrophic stop: disaster insurance that costs ~5% of return, not a source of edge. Backtest figures — past performance does not guarantee future results.

That is why there is no “recommended exit” on this page. Every position is bought and held to thesis invalidation. The edge was never in the selling — it's in the conviction rating on what gets bought.

Case study — PLUG Power (PLUG) — the price of the discipline
−92%
PLUG — held (what we took)
+70%
PLUG — if exited (hindsight)
+2,424%
VRT — held
+1,205%
VRT — same exit rule applied

Here is the honest version. On PLUG alone, an exit rule would have helped — staged exits would have turned −92% into roughly +70%. So why don’t we use one? Because you cannot apply an exit rule only to the losers. The same trailing stop that rescues PLUG also clips VRT from +2,424% down to roughly +1,205% — and MOD, NVDA and every other winner with it. Run across all 87 positions, the winners an exit rule gives up dwarf the losers it saves — every time. PLUG is the visible price of a discipline that wins on the whole, and we show it in full rather than bury it.

Case study — Enphase Energy (ENPH) — what it taught the conviction model
−72%
ENPH — held to today
~80×
Entry P/E — unscreened at the time
0 → 5.25%
Fed hike that broke demand
Rated down
How the model treats it now

ENPH was a solar inverter thesis — a component supplier for residential rooftop solar. The solar transition was real and is still happening; the thesis wasn’t wrong. But the stock was crushed by something the rating process didn’t screen for: interest-rate sensitivity.

Residential solar is bought on consumer loans. When the Fed hiked from near-zero to 5.25%, monthly payments on an installation nearly doubled and demand collapsed — ENPH’s revenue fell 50% in 18 months. The thesis was sound; the stock was rate-sensitive; the original signal didn’t distinguish the two.

The honest lesson is not “we should have sold.” The research is clear that exit rules lose money across the book — reacting to ENPH’s falling price would have meant reacting to every winner’s pullback too. The real miss was upstream: ENPH should never have carried a HIGH conviction rating. A consumer-financed business trading at ~80× earnings is rate-fragile — and that belonged in the conviction score, before entry, not in an exit rule after.

📚 What ENPH taught the conviction model
  • Rate-sensitivity is now a conviction input. If a company’s revenue depends on consumer or project financing, that caps its conviction rating — a rate-fragile business cannot be a 5-star thesis no matter how strong the macro story.
  • Clean energy is no longer one bucket. Grid-scale (utility solar, wind developers, transmission) is policy-backed and rate-insensitive — it can still rate HIGH. Consumer-facing clean tech (ENPH, SEDG) is rate-sensitive and rates lower.
  • Valuation caps conviction. ENPH entered at ~80× earnings. Any candidate above 50× now takes a conviction haircut — high multiples carry rate-compression risk that pure thesis analysis misses.
  • Thesis intact ≠ stock intact. A company can operate in a growing sector and still be a broken stock. The conviction model now scores the stock’s durability, not just the sector’s tailwind.
Signal ledger

Every signal. Held. All 87 positions.

Sorted by return. One number per position — bought on the signal date, held to thesis invalidation, measured against the S&P 500 over the identical window. Winners and losers, no exceptions.

Signal ledger — 87 tickers — 2013–2026 All positions tracked from first verifiable public signal
Ticker Sector / Thesis Return — peak → current vs S&P 500
VRT Data Center Cooling & Power +2,739% → +2,297% +117% SPY
MOD HVAC / Thermal Management +2,505% → +2,310% +117% SPY
CCJ Uranium / Nuclear Energy +1,083% → +832% +117% SPY
PLUG Hydrogen Fuel Cell −92% +117% SPY
ENPH Solar Technology −72% +117% SPY
NVDAAI / GPU Compute+1,549% → +1,333%+117%
OKLOAdvanced Nuclear (Microreactor)+1,961% → +596%+117%
BEFuel Cell / Distributed Energy+1,028% → +944%+117%
AMSCGrid / Power Electronics+505% → +264%+117%
MPRare Earth / Critical Minerals+585% → +323%+117%
WPMGold & Silver Streaming+497% → +355%+117%

87 tickers — full table unlocked for Investor subscribers

Every ticker, every signal date, conviction rating, and live signal status — updated every Friday. Investor plan includes the complete signal ledger plus real-time alerts.

Get Investor access →

$18.99/month · Cancel anytime · Instant access

All returns calculated from first verifiable public signal (DOE grant, SEC filing, NRC announcement, congressional trade disclosure) to present or exit date. Entry = closing price on signal detection date. SPY benchmark matched to identical holding period. Not financial advice — all theses included, winners and losers.

Backtesting rules — unchanged since inception