Research Report · Twitter Backtest · Market Color

@pennycheck 585-Day Backtest: Only 36 of 165 Tweets Carry Extractable Cashtags - the Real Signal Is Outside the Regex

Quantitative backtest of @pennycheck (61,226 followers, 50,091 lifetime tweets, bio: “finding asymmetric opportunities in any and every mkt.”). 165 original tweets sampled across 585 days, but only 36 contain extractable $TICKER cashtags. The author refers to own positions with a penguin emoji rather than $TICKER — meaning the regex sentiment classifier under-counts. The recent (Apr-May 2026) pivot from $POET bull to $POET bear is well-documented and was a meaningful inflection.
Published 2026-05-25 Backtest Window: Oct 15, 2024 -> May 24, 2026 (585 days) Benchmarks: SPY (broad) / XLK (tech sector) Entry: T+1 open · Exits: 5d / 14d / 30d / 90d close
VERDICT: CHATTER / LOW-SIGNAL · DO NOT FOLLOW MECHANICALLY
36 / 165
Ticker-mention tweets out of original tweets
10 / 2 / 24
BULL / BEAR / NEUTRAL classified
+5.98%
BULL mean / 5d (n=14, WR 79%)
+11.37%
BULL mean / 14d (n=5, WR 100%)
61,226
Followers · account since 2013-08-30
50,091
Lifetime tweets · very high chatter ratio
6
$POET mentions · recent BULL -> BEAR pivot
585d
Wall-clock window · signals concentrate late
Prefer the print version? Download the 3-page PDF: profile dossier, full sentiment-by-horizon table, most-mentioned tickers, methodology, limitations.
Download PDF ↓

Profile & Persona Who pennycheck is and why the regex backtest under-reads the signal.

The author. 61,226 followers, account opened 2013-08-30, 50,091 lifetime tweets (~12 / day average), bio: “finding asymmetric opportunities in any and every mkt.” Style is asymmetric / niche micro-cap and semi positioning, heavy emoji use, very short cryptic posts (avg ~50 characters / tweet).

The measurement problem. Most own-pick references use a penguin emoji rather than the $TICKER cashtag — meaning a standard cashtag regex misses the directional posts the author actually makes about own positions. Of 165 original tweets sampled, only 36 contained any extractable cashtag. The reported BULL count (n=10) is therefore a heavy under-count of the true directional surface.

The well-documented inflection. Across April-May 2026 the author publicly pivoted from $POET bull to $POET bear. This was a tape-changing call for the stock and is visible in the 6 $POET mentions inside the sample. Mechanical sentiment classification splits these inconsistently because the language is short and emoji-heavy; reading the actual tweets gives the inflection clearly while the regex does not.

Forward Returns by Sentiment & Horizon Entry: next trading day open after tweet date. Exit: close at +5/+14/+30/+90 trading days. Alpha = signal return minus same-window benchmark return.

SentimentHorizonnMean RetMedianWin RateBestWorstAlpha SPYAlpha XLK
BULL5d14+5.98%+5.62%78.57%+18.45%-12.11%+4.70%+1.94%
BULL14d5+11.37%+12.82%100.00%+17.70%+0.16%+7.65%+0.42%
NEUTRAL5d20+2.36%+1.01%55.00%+24.59%-17.31%+0.89%-2.40%
NEUTRAL14d5-3.24%-0.21%40.00%+15.62%-25.02%-5.86%-11.81%
NEUTRAL30d1+176.77%+176.77%100.00%+176.77%+176.77%+173.48%+175.82%
NEUTRAL90d1+461.62%+461.62%100.00%+461.62%+461.62%+460.71%+464.58%

The headline read. BULL 5d (n=14, WR 79%) is directionally fine but under-powered; BULL 14d (n=5, WR 100%) is too small a sample to read seriously. The two NEUTRAL 30d and 90d rows are single-signal outliers (a one-off micro-cap that ripped); they should be treated as anecdotes, not as evidence of a 460% mean edge.

The most honest read of this table: the BULL 5d alpha vs SPY is real (+4.70%) but inside the noise floor against XLK (+1.94%), and the entire dataset is too thin (36 ticker mentions across 585 calendar days) to support mechanical mirroring.

Most-Mentioned Tickers Top 20 by mention count. The skew is heavy — $POET dominates and then everything else is 1-3 mentions.

#TickerMentions#TickerMentions
1$POET611$MU1
2$NVDA312$IBM1
3$INTC313$GOOG1
4$PENG214$INFN1
5$NOK215$SHMD1
6$AOSL216$LAW1
7$AMPX217$CRBS1
8$SMCI218$GLXY1
9$AMD219$ADEA1
10$FORM120$AVGO1

Five Questions, Five Answers

Q1. Is there a real edge in the data?

BULL 5d alpha vs SPY is +4.70% on n=14 (WR 79%) - directionally positive, statistically thin. Against the tech-sector benchmark (XLK) the alpha collapses to +1.94%, which is inside the noise floor for this sample size. The 14d / 30d / 90d rows resolve on too few signals to read.

Q2. Why is the sample so thin?

The author uses a penguin emoji to refer to own positions rather than $TICKER cashtags. A standard cashtag regex therefore captures only the cases where a stock is named explicitly — 36 tweets out of 165 sampled. The true directional surface is much larger; the published numbers are a heavy under-read.

Q3. What did the $POET pivot look like?

Across April-May 2026 the author moved publicly from $POET bull to $POET bear. This was a tape-changing call for the stock. Followers who read the actual posts (rather than relying on aggregated sentiment) saw the inflection in real time. The regex backtest does not capture this kind of qualitative pivot well.

Q4. What are the NEUTRAL 30d / 90d rows?

Single-signal outliers (one ticker that ripped +176% / +461% in the relevant window). They are present for transparency but should not be read as “NEUTRAL signals produce 460% returns.” n=1 is anecdote, not data.

Q5. How should I use this feed?

As a market-color feed. Read the posts to absorb the asymmetric / micro-cap mindset and to catch qualitative pivots (like the $POET reversal). Track when the “own-pick” penguin wording resolves to a specific name — that is the actionable subset. Mechanical follow of the cashtag subset is not justified by the data.

Methodology

Data source. User-timeline endpoint, original tweets only (replies and pure retweets excluded). 165 tweets across 585 calendar days — the maximum the timeline endpoint surfaces for this account given its very high posting cadence (account-lifetime ~12 tweets / day).

Ticker extraction. $TICKER cashtag regex with a curated non-ticker stop-list. The penguin-emoji own-pick convention is invisible to this regex by design.

Sentiment classifier. Regex keyword lexicon (BULL: bullish, buy, long, added, breakout, ATH, target, runner; BEAR: bearish, short, sell, sold, dump, dilution, scam). BULL/BEAR if count strictly dominates; otherwise NEUTRAL. Short cryptic posts are likely under-classified into NEUTRAL.

Backtest mechanics. Entry next trading day open (T+1). Exits at +5d / +14d / +30d / +90d close. No transaction costs / slippage.

Benchmarks. SPY (broad market) and XLK (tech sector).

Limitations & Honest Caveats

  1. Short-tweet undercount. Cryptic posts (avg ~50 chars / tweet) plus penguin-emoji own-pick convention means a regex captures only a small fraction of the true directional surface. The classified BULL n=10 is a heavy under-count.
  2. Sample thinness. Only 36 ticker-mention tweets across 585 days — the 14d / 30d / 90d rows resolve on 1-5 signals each. Read 5d as the only bucket with enough n to read directionally.
  3. Single-signal outliers in NEUTRAL. The 30d (+176%) and 90d (+461%) NEUTRAL rows are n=1 each — a single micro-cap winner inflating the mean. Treat as anecdote.
  4. Wall-clock window does not mean coverage. 585 days end-to-end but the timeline endpoint surfaces the most recent ~3,200 tweets only; for a 50,091-lifetime-tweet account, dense sampling concentrates in the last 6-12 months.
  5. Qualitative pivots invisible to regex. The $POET BULL -> BEAR reversal across Apr-May 2026 was a meaningful inflection that the sentiment classifier does not encode cleanly. Read the actual posts.
  6. No risk-adjustment. Sharpe / Sortino not computed. Position-sizing, drawdowns, and correlation across signals are not modeled.

Final Verdict

! Tier: CHATTER / LOW-SIGNAL

BULL 5d (n=14, WR 79%) is directionally fine but underpowered; against XLK the alpha is inside the noise floor. The 14d / 30d / 90d buckets are too thin to read. Mechanical follow is not justified by the data.

! Measurement caveat

Penguin-emoji own-pick convention and very short cryptic posts mean a cashtag regex captures only 36 of 165 tweets. The published numbers are a heavy under-read of the true directional surface.

+ How to use

Market-color feed. Read posts to catch qualitative pivots (the $POET BULL -> BEAR reversal was a real call). Track when penguin-emoji own-pick wording resolves to a specific name — that subset is the actionable read.

Full 3-page PDF report Profile dossier, full sentiment-by-horizon table, top-20 most-mentioned tickers, methodology and limitations.
Download PDF ↓

Report generated 2026-05-25 by the fxcryptobots research desk. Source: @pennycheck public X timeline (user-timeline endpoint, original tweets only); price data from Yahoo Finance daily bars. Sentiment classifier is regex-based and explicitly under-counts emoji-driven and short-tweet styles. Per-name returns are total return including dividends via auto-adjusted close. This is research and educational analysis, not investment advice; see our risk disclaimer.