HIGH CONVICTION PATTERNS

High Conviction Pattern Families

Historical shape is light gray; future projection is aqua for bullish families and fuchsia for bearish families.

Select a pattern family to view details The selected family and its variants will appear here.
METHODOLOGY

Cross-Market Pattern Discovery Framework

The pattern library was generated from approximately 2,500 equities using three years of hourly price history. Rolling price segments were normalized and compared using Pearson correlation across both intra-ticker and cross-ticker histories. Qualifying matches were evaluated over defined forward horizons to measure directional frequency, return magnitude, and statistical uncertainty. The complete search involved trillions of correlation calculations.

SEARCH ARCHITECTURE One rolling candidate is evaluated against the broader historical segment universe.
01 · INPUT UNIVERSE
~2,500 equities 3 years · hourly bars
→
02 · Rolling segmentation
Variable-length windows Every eligible historical position
→
03 · Cross-ticker Pearson search
Candidate ↔ historical segment universe Same ticker + the other securities in the dataset
candidate
→
04 · Forward outcome evaluation
Pattern-specific horizon Return and direction recorded
→
05 · Directional statistics
Frequency + magnitude Wilson 95% confidence bound
Directional frequency
Return magnitude
Wilson 95 lower bound
01

Search universe and rolling comparison

The search starts with the historical dataset itself rather than with a predefined list of named chart formations. Rolling windows move through the full hourly history of each security. At each eligible position, a candidate segment can be compared with other segments from the same ticker and with segments from the other securities in the universe.

With approximately 2,500 tickers and 3 years of hourly history, the number of pairwise segment comparisons becomes extremely large. The full discovery run therefore required trillions of correlation calculations.

Universe~2,500 tickers
Historical depth3 years hourly
Search scaleTrillions of comparisons
02

Normalization and Pearson similarity

Price segments are normalized before comparison so that similarity is driven by path shape rather than absolute share price. Pearson correlation is then used to quantify how closely two normalized sequences move together.

1.00Near-identical normalized movement
0.95Extremely similar shape
0.90Strong positive similarity
0.00No linear similarity

In the current published bank, mean match correlations are 0.964–0.980.

03

Minimum historical support

A high directional percentage is not treated as meaningful when it is supported by only a few examples. Candidate structures require enough historical occurrences for their forward statistics to be statistically informative. The current retained variants contain 95–638 evaluated historical occurrences, with a median of 218.

Many of the strongest patterns therefore have hundreds of historical occurrences found and compared across the dataset. Larger samples reduce the uncertainty around the estimated directional rate and make it harder for a pattern to rank highly because of a short lucky run.

Minimum retained95
Median retained218
Maximum retained638
Minimum retained95
Median retained218
Maximum retained638
04

Forward outcome measurement

After a qualifying historical match is identified, the realized return over that pattern's defined forward horizon is recorded. Each pattern uses its own forward evaluation horizon. Repeating this for every qualifying historical occurrence creates the outcome distribution used for direction and magnitude statistics.

Directional confidence same-direction outcomes ÷ total evaluated occurrences
Magnitude distribution of realized forward returns
Forward horizonPattern-specific
Outcome basisAll qualifying historical occurrences
05

Direction and expected magnitude

Direction is determined from the observed forward outcomes rather than from a subjective interpretation of the chart. For an UP pattern, positive forward outcomes contribute to directional confidence; for a DOWN pattern, negative outcomes do.

Magnitude is summarized separately. Median final return reports the middle historical outcome and is less sensitive to extreme observations, while mean final return reports the arithmetic average. Showing both makes it easier to see whether the typical move and the average move tell a similar story.

Typical emphasisMedian final return
Supplementary viewMean final return
06

What Wilson 95 means

Directional confidence tells us how often historical matches moved in the expected direction. But the same percentage can carry very different weight depending on how many observations produced it. Wilson 95 adjusts that directional rate for sample-size uncertainty.

Same observed rate. Different confidence. More observations → less uncertainty → the Wilson lower bound moves closer to the observed rate.
SMALLER SAMPLE 10 observations
Observed rate 90.0%

9 of 10 historical matches moved in the expected direction.

50%75%100%
59.6%Wilson 95 lower bound 90.0%Observed rate
LARGER SAMPLE 300 observations
Observed rate 90.0%

270 of 300 historical matches moved in the expected direction.

50%75%100%
86.1%Wilson 95 lower bound 90.0%Observed rate
How to read this:

Both samples produced the same 90% observed directional rate. With only 10 observations, however, there is much more uncertainty, so the conservative Wilson lower bound falls to 59.6%. With 300 observations, the result is much better supported and the lower bound rises to 86.1%.

Example from the current retained pattern bank
262 observationshistorical matches evaluated
96.95%observed directional confidence
94.09%Wilson 95 lower bound

In this example, the pattern moved in the expected direction in about 97% of its historical observations. Even after accounting for sample-size uncertainty, the conservative lower end of the 95% Wilson interval remains about 94%.

Wilson 95 is not a p-value. It is a confidence interval around an observed proportion and is used here to account for sample-size uncertainty when evaluating directional consistency.

INTERPRETATION

Interpretation of the resulting statistics

The objective is to identify recurring price structures whose subsequent behavior has been both directionally consistent and supported by a substantial historical sample. Similarity identifies the recurring structure; forward returns measure what followed; directional frequency measures consistency; return statistics measure magnitude; and the Wilson lower bound quantifies uncertainty in the directional estimate.

The resulting values describe historical conditional behavior. They do not imply that a future occurrence will produce the same result, and they do not remove the effects of changing volatility, liquidity, market regime, or other conditions.

GAMESTOP

Chart Data

This section is reserved for GameStop chart data and market-structure research.

GAMESTOP

Company Information

SEC filing history and insider transaction data, organized for direct filtering and comparison.

Visible0
Static preload for the current build. Filing links open the SEC EDGAR company page; a later sniffer can replace this embedded dataset without changing the UI.
FiledFormFiler / subjectDescriptionSource
Transactions0
Net shares0
Acquired0
Disposed0
Date range—
Daily net shares Hover, click, or tap a bar for details
1,179 transactions. Magnitude is shown as reported share quantity; transaction price is not included in this dataset.
DatePersonTitleCodeTransactionDirectionSharesDerivative
GAMESTOP

Options Flow

This section is reserved for GameStop options-flow data and analysis.

RULEMAKING ACTIVITY

SEC

563 records. Every displayed record links to its source document.

Visible0
PublishedTypeTitleEffectiveDocument #CitationLink
Page 1
RULEMAKING ACTIVITY

FINRA

8,560 records. Every displayed record links to its source document.

Visible0
PublishedTypeTitleEffectiveDocument #CitationLink
Page 1
RULEMAKING ACTIVITY

CFTC

278 records. Every displayed record links to its source document.

Visible0
PublishedTypeTitleEffectiveDocument #CitationLink
Page 1
RULEMAKING ACTIVITY

DTCC

This section is reserved for DTCC, NSCC, DTC, and FICC material.

RESEARCH

Due Diligence

This section is reserved for original due-diligence work and long-form research.

SUPPORT

Support the Site

theUltimator5 research and data tools are freely available. If you find them useful and want to help support continued development and data costs, you can contribute through Cash App.

QR code for Cash App handle $theultimator5
Scan with your phone camera
CASH APP
$theultimator5

Use the QR code or open the Cash App profile directly.

Open Cash App ↗
Before sending, please verify the recipient is $theultimator5.