The Brown Bear Report

The most promising signals the research has produced, the names each one is firing on now, and — for every one of them — what the evidence actually is. Names caught by a validated avoid flag are marked wherever they appear.

Signals as of 2026-08-13 · All prices are the previous close. Nothing on this page is a live quote.

The imported signals are 13 days old.

Brown Bear Core, the score drops, FCF/assets, V0 repair and all three avoid flags are built off-server and were last refreshed 2026-08-14. Treat the lists below as of that date, not today.

Where to look first

Two tiers. Convergent names are fired by two or more signals and caught by none of the avoid flags. Clean fires are fired by exactly one signal and caught by none. Anything carrying an avoid flag is excluded from both and listed separately underneath.

Convergent 3

  • AMAT Applied Materials, Inc. Brown Bear Core + RPS and PHS drops
  • PANW Palo Alto Networks, Inc. Brown Bear Core + V0 repair
  • RELY RPS and PHS drops + V0 repair

Convergence is not a tested rule. No study here has ever shown that two signals firing together beats the better of the two alone — and the standing lesson from the quality work is the opposite, that composites come out weaker than their best component. This list is a convenience for finding overlaps, not a ranking, and the order carries no meaning beyond the number of signals.

Clean fires 66

Fired by one signal, caught by no avoid flag. They are listed in each signal's own card below, where the evidence for that signal sits beside them — which is the only place the name means anything.

Fired but flagged 64

  • JNJ Johnson & Johnson Brown Bear Core + RPS and PHS drops Low volatility, Close at range high
  • MCD McDonald's Corporation Quality oversold repair + RPS and PHS drops Low volatility
  • MTL.TO RPS and PHS drops + V0 repair Low volatility
  • VRSN VeriSign, Inc. FCF / assets + RPS and PHS drops Low volatility
  • ABNB Airbnb, Inc. RPS and PHS drops Close at range high
  • AIZ Assurant, Inc. RPS and PHS drops Low volatility

Signals

Brown Bear Core

Imported from research as of 2026-08-1312 firing

The programme's flagship list: established, cash-generative companies that are leading the market, accumulated gradually while other investors are fearful and released while they are calm. The rank is the list's own order of preference. How the list is built is not published.

Stock CloseRank Avoid flags
VICR 226.64 1.00 clear
KRYS 332.86 2.00 clear
AMAT Applied Materials, Inc. 534.54 3.00 clear
LRCX Lam Research Corporation 337.01 4.00 clear
FTI 69.62 5.00 clear
CAT Caterpillar Inc. 854.60 6.00 clear
FTNT Fortinet, Inc. 165.44 7.00 Low volatility
FIX Comfort Systems USA, Inc. 1,727.64 8.00 clear

Screened 199 names · 12 entries recorded so far · headline 3,018% total (23.1%/yr, Sharpe 0.757) gross, no costs applied

One of the inputs behind this list was last updated 27 days ago. The names are current; treat anything time-sensitive about them with caution.

Tested out-of-sample on an independent platform — no live record yet

Backtest
3,018% total return from January 2010 to August 2026, about 23.1% a year, with a Sharpe ratio of 0.757 and a worst peak-to-trough fall of 42%. Measured on QuantConnect, an independent backtesting platform using its own market data and its own universe of companies — not on this project's data. That matters: where the two have disagreed before, the independent platform has been right and the local data flattering.
Live tracker
A live ledger started 2026-08-14 and records the list every session. No live track record yet — everything above is a simulation, and a simulation is a hypothesis about the future, not evidence from it.

Read this too: Four things this headline does not say. (1) Most of the money was made by being invested in the stock market generally; the list's own contribution above a plain S&P 500 tracker is roughly 8.7 percentage points a year, and that is the number to judge it on. (2) No trading costs are deducted anywhere — realistically 0.7 to 1.3 percentage points a year would come off. (3) A 42% drawdown is worse than the index suffered over the same period, and the worst of it came during a relatively mild market decline, so the ride is rougher than the return implies. (4) The simulation ended holding far more positions than it was designed to hold, because of how it sells; the entry list shown here is sound, but the exit behaviour above is not fully characterised.

Quality oversold repair

Computed on this server as of 2026-08-265 firing

The regime narrative for a name with sound fundamentals whose technicals have been beaten down and are starting to mend. Computed on this server from last night's S&P 500 run, so this list is the freshest on the page.

Stock ConfidenceBusiness convictionPosition healthClose Avoid flags
VRSK Verisk Analytics, Inc. 100.00 90.50 68.90 187.87 Low volatility
INTU Intuit Inc. 94.70 86.80 55.90 345.88 Close at range high
PODD Insulet Corporation 82.40 77.90 21.40 143.38 clear
MCD McDonald's Corporation 75.70 60.20 19.70 266.93 Low volatility
DECK Deckers Outdoor Corporation 75.20 74.30 21.60 89.47 clear

Live, measured against SPY over 133 episodes since 2026-06-23: 3d +0.48% hit 48% · n=117 5d +1.21% hit 54% · n=114 10d +2.72% hit 58% · n=104 20d +4.57% hit 70% · n=81

Live tracker only — no backtest of its own

Backtest
None. The regime narratives were hand-specified, not fitted, so there is no in-sample number to quote — which is why the live tracker is the whole of the evidence here.
Live tracker
One of only three narratives out of fifteen whose live forward returns carry the sign the narrative claims, and it strengthens with horizon. The figures on the card are measured against the S&P 500, not raw.

Read this too: Read this against its own family, which is mostly not working: the two highest-volume regime narratives are INVERTED live — support-resistance bounce and rotation-leadership upgrade both go the wrong way over more than a thousand matured episodes each. That this one reads correctly is genuine, but it sits inside a set where being correct is the exception, and the sample starts in June 2026. Treat it as a narrative that has not yet been falsified rather than one that has been confirmed.

RPS and PHS drops

Imported from research as of 2026-08-1399 firing

Names whose Rotation Progress Score or Position Health Score fell by 10 points or more from one run to the next. Counter-intuitively this is a BUY list, not a sell list — see the evidence.

PHS drops 431

After a drop, average raw return: 3d +0.95% n=2131 5d +1.53% n=1973 10d +1.88% n=1528

Showing the 60 most recent of 431.

Stock DroppedPoints Avoid flags
AMGN Amgen Inc. 2026-08-13 -10.30 Low volatility
TDY Teledyne Technologies Incorporated 2026-08-13 -10.40 Close at range high
AZO AutoZone, Inc. 2026-08-13 -10.90 clear
COST Costco Wholesale Corporation 2026-08-13 -11.00 Low volatility Close at range high
LULU lululemon athletica inc. 2026-08-13 -11.10 Close at range high
AMAT Applied Materials, Inc. 2026-08-13 -11.20 clear
IEX IDEX Corporation 2026-08-13 -11.20 clear
IFF International Flavors & Fragrances Inc. 2026-08-13 -11.30 Low volatility

RPS drops 360

After a drop, average raw return: 3d +1.26% n=1181 5d +1.75% n=1082 10d +3.40% n=714

Showing the 60 most recent of 360.

Stock DroppedPoints Avoid flags
HMR 2026-08-13 -10.00 clear
TMO Thermo Fisher Scientific Inc. 2026-08-13 -10.00 Low volatility Close at range high
HCA HCA Healthcare, Inc. 2026-08-13 -10.20 Low volatility
JCI Johnson Controls International plc 2026-08-13 -10.20 clear
MTD Mettler-Toledo International Inc. 2026-08-13 -10.20 Low volatility Close at range high
VLTO Veralto Corporation 2026-08-13 -10.50 clear
ALB Albemarle Corporation 2026-08-13 -10.60 clear
HLT Hilton Worldwide Holdings Inc. 2026-08-13 -10.80 clear

Live event study — well powered, and unflattering to our own scores

Backtest
None by design. Business conviction is roughly 90% fundamentals and cannot be reconstructed historically without point-in-time data, so this could only ever be answered forward, on real scores as they accrue. That is the right way to have built it.
Live tracker
Across thousands of matured events, a drop of 10 points or more is followed by POSITIVE returns and a jump by negative ones — consistently, for both RPS and PHS, at 3, 5 and 10 days. The most likely reading is that the scores lag price: they rise after a move has happened and fall after a pullback, so buying the drop is buying the dip.

Read this too: This is a real, live, well-powered finding and it is unflattering to Brown Bear's own scoring system — that is exactly why it is on this page. Two limits. The returns are raw, not measured against the S&P 500, so some of the gain is simply the market going up. And the sample starts in June 2026, which is a single market regime. Note also that the BCS leg behaves differently and is deliberately not shown: a large move in business conviction is followed by negative returns in EITHER direction.

FCF / assets

Imported from research as of 2026-07-2912 firing

The names ranking highest on free cash flow as a share of total assets. This is the measure that survived testing — not FCF as a share of revenue, which is the one shown elsewhere on this site.

Stock FCF / assets Avoid flags
WINA 1.80 Low volatility
VRSN VeriSign, Inc. 0.81 Low volatility
BCRX 0.67 clear
DAVE 0.59 clear
SEZL 0.52 clear
OPFI 0.51 clear
NVDA NVIDIA Corporation 0.47 clear
MANH 0.45 clear

1,063 names ranked · 1 cohorts registered, 0 matured · 20 needed before this can be read at all

Real, but not ours — a published factor, replicated here

Backtest
FCF/assets and operating-cash-flow/assets both survived on the Brown Bear universe, while every accrual-based measure died. That reopened an axis three earlier studies had closed.
Live tracker
A live shadow race is running with the ranked cohort recorded in full. No cohort has matured yet, and by design that will take years.

Read this too: Two things must be said with this one. First, it is a replication of a published academic result (Ball and others, 2016), not a Brown Bear discovery — presenting it as proprietary would be a misrepresentation. Second, do NOT build a quality score out of it: the standing lesson from this work is that the composite comes out weaker than its best component. The unit of inference is the cohort date, roughly four independent periods a year, so a positive number two years from now is not confirmation.

V0 repair

Imported from research as of 2026-08-1312 firing

The book held by the repair arm of the V0 live race. The names marked "differs" are the ones the repair arm holds that the plain baseline arm does not — that difference IS the hypothesis being tested.

Stock Differs from baseline Avoid flags
BNY The Bank of New York Mellon Cor no Low volatility
FIVN no clear
MPC Marathon Petroleum Corporation no Low volatility
MTL.TO no Low volatility
OOMA no clear
PANW Palo Alto Networks, Inc. no clear
PSX Phillips 66 yes Low volatility Close at range high
RELY no clear

baseline +10.74% over 21 sessions repair +11.98% over 21 sessions

Accruing — far too early to read

Backtest
The factor set this arm modifies rests on a five-year, single-bear panel and the leave-one-out test that justified its shape has never been re-run on the deeper spine. That gap is known and recorded.
Live tracker
A frozen two-arm race: baseline against repair, marked daily. The design is sound — the runtime discriminates between two live hypotheses rather than between a signal and nothing.

Read this too: Weeks of data, not years. The return figures shown are for the whole arm over a handful of weeks and are dominated by market direction; annualised or risk-adjusted versions of them would be meaningless at this sample size and are deliberately not shown. Nothing here should be read as evidence in either direction yet.

Avoid flags

These describe conditions that have been found to precede under-performance. A name on one of these lists is marked wherever it appears in a buy signal above. They are not sell signals: the research separately found that a failed position is positive-expectancy to keep, and none of these flags has ever been tested as an exit rule.

Low volatility

Imported from research as of 2026-08-13 266 flagged

Names that are consistently quiet — low realised volatility, low beta, low correlation to the market and small overnight gaps. Meeting three of those four flags the name.

Matured 60-day excess return since flagged: seed -9.40% median · hit 74% · n=275 out of sample no matured episodes yet

Showing the 60 most extreme of 266. The cross-reference against the buy signals above uses all 266.

Stock VolatilityBetaSince flagged Also flagged
WINA 0.41 0.60 -15.73 —
EBAY eBay Inc. 0.33 0.74 -14.35 Close at range high
DOC Healthpeak Properties, Inc. 0.42 0.58 -13.80 —
NI NiSource Inc. 0.20 0.06 -12.69 —
MO Altria Group, Inc. 0.28 -0.48 -12.57 —
LIN Linde plc 0.20 0.08 -12.49 —
FIS Fidelity National Information Services, Inc. 0.39 0.22 -12.36 —
EXEL 0.38 0.60 -12.20 —

The best-evidenced avoid in the system

Backtest
The broadest effect in the whole dataset, era-stable and found five separate ways: low beta −4.1pp over 60 days, low correlation to the market −5.7pp (identical in both halves of the sample), low average true range −5.0pp, low volatility −4.4pp, small gaps −3.6pp. Combined with weak 12-month returns it reaches −8.0pp.
Live tracker
By far the best-populated flag on the page, and the live shape matches the backtest: the penalty deepens monotonically with horizon, exactly as predicted.

Read this too: The study's own qualification: the penalty concentrates in rising markets and mildly REVERSES in 2022 and 2024. In a genuine bear year, low volatility defends. This is a bull-leaning tilt, not a timing signal — and the panel it was found on contains only one mild bear, so that caveat has never actually been tested.

Heavy buying, weak RSI

Imported from research as of 2026-08-13 1 flagged

Price absorbing heavy buying without going up: up-volume running at nearly twice down-volume while the 14-day RSI stays at or below 45.

Matured 60-day excess return since flagged: seed -6.93% median · hit 62% · n=8 out of sample no matured episodes yet

Stock Up/down volumeRSISince flagged Also flagged
ESS Essex Property Trust, Inc. 2.20 36.30 -2.14 Low volatility

Strong backtest, no live power

Backtest
−7.9pp over 60 days, and −13.2pp once weak 12-month returns are added — the single deepest negative cell found anywhere in the search, and negative in every year from 2022 to 2025. The study calls it a genuinely novel negative.
Live tracker
Almost nothing. The flag fires roughly a dozen times in six months, so there is no live power to speak of and there will not be for a long time.

Read this too: At this sample size the live mean and median disagree violently — a strongly positive mean sitting on a strongly negative median, which is the signature of one name dominating the average. Read the backtest, treat the live column as decorative, and check the underlying names before quoting any figure from it.

Close at range high

Imported from research as of 2026-08-13 137 flagged

Names closing near the top of their daily range persistently over 20 days — the top fifth of the cross-section. Intrabar exhaustion.

Matured 60-day excess return since flagged: seed no matured episodes out of sample no matured episodes yet

Showing the 60 most extreme of 137. The cross-reference against the buy signals above uses all 137.

Stock Close in rangeSince flagged Also flagged
TPR Tapestry, Inc. 0.55 -19.87 —
EBAY eBay Inc. 0.59 -15.95 Low volatility
CINF Cincinnati Financial Corporation 0.55 -11.41 Low volatility
CMP 0.53 -11.33 —
SYK Stryker Corporation 0.57 -9.81 Low volatility
CDW CDW Corporation 0.57 -8.03 —
OTIS Otis Worldwide Corporation 0.55 -7.09 Low volatility
PG The Procter & Gamble Company 0.54 -5.82 Low volatility

Instrumented but unmeasured — weakest item on the page

Backtest
−1.92pp over 60 days, in both halves of the sample and within-ticker. The research ledger rates this low confidence: it is the weakest of the avoid set by a wide margin, it is in-sample, it is bull-heavy, and it emerged from a three-stage arc in which the original hypothesis was killed and the family was shown not to stack.
Live tracker
None whatsoever — zero matured episodes in both the seed and live columns. It is wired and recording, and that is all.

Read this too: This flag currently has NO data behind it in either column. It is on the page so that a name it catches is disclosed rather than hidden, but it should not on its own change a decision, and it is the one item here that a hostile reader would be right to discount entirely.

How to read this page

Every card carries its own source and date. Computed on this server means the list was rebuilt from last night's run. Imported from research means it was built where the research data lives and shipped here — those signals cannot run on this machine, and rebuilding approximations of them under the same names would be worse than importing them late.

Nothing on this page is wired into any decision. No signal here sizes a position, fills a slot, or triggers a trade. Several of them are recorded specifically because they have not yet been shown to work, and the live trackers exist to find that out. Where a signal has no matured out-of-sample data, the page says so rather than showing an in-sample number in its place.

Returns described as excess are measured against the S&P 500; returns described as raw are not, and in a rising market a positive raw number may be no achievement at all. No figure anywhere on this page has trading costs deducted. On Brown Bear Core that is worth roughly 0.7 to 1.3 percentage points a year — set against an edge over the index of about 8.7 points a year, it is a real bite, not a rounding error.