The author. 55,358 followers, account opened 2009-07-08 (16-year account), 19,662 lifetime tweets, bio: “Trading as a hobby. All tweets personal opinions.” Style is macro + semis trader: CPU / HBM / photonics rotation, geopolitical hedges ($OXY, $NTR, oil), bond options ($TLT), frequent retweets of other analysts. Top names across the 42-day window: $NVDA, $LITE, $NOK, $AAOI, $AMD, $MRVL, $SMTC, $CBRS, $WYFI, $SANM, $FLEX, $GLW, $UCTT, $FSLY.
What the backtest does not see. Roughly 30% of the feed is retweets (excluded from this run since the methodology only considers original posts). Rapid trade-flow updates are common (own positions get added / trimmed within the same day). The 14d / 30d horizons therefore probably mis-measure the actual hold periods — the author often takes profit inside the first 1-3 days.
| Sentiment | Horizon | n | Mean Ret | Median | Win Rate | Best | Worst | Alpha SPY | Alpha XLK |
|---|---|---|---|---|---|---|---|---|---|
| BULL | 5d | 13 | +2.75% | +8.32% | 76.92% | +19.98% | -39.76% | +0.78% | -2.17% |
| BULL | 14d | 7 | +6.78% | +10.73% | 71.43% | +21.57% | -14.74% | +3.08% | -3.52% |
| BEAR | 5d | 4 | +9.86% | +8.26% | 100.00% | +15.76% | +7.16% | +8.04% | +4.98% |
| BEAR | 14d | 3 | +28.02% | +28.90% | 100.00% | +48.71% | +6.43% | +24.19% | +17.08% |
| NEUTRAL | 5d | 36 | +0.44% | -2.03% | 44.44% | +52.24% | -28.75% | -1.24% | -4.01% |
| NEUTRAL | 14d | 19 | +7.54% | +3.23% | 63.16% | +66.88% | -18.81% | +4.36% | -2.71% |
The headline read. BULL signals look directionally positive in raw return (5d +2.75%, 14d +6.78%) and the win rate is 71-77%, but the alpha vs the tech sector ETF (XLK) is NEGATIVE at both horizons (-2.17% / 5d, -3.52% / 14d). The picks underperform a passive tech-sector hold inside the same window.
The BEAR rows look spectacular (5d +9.86% WR 100%, 14d +28.02% WR 100%) but the samples are tiny (n=3-4); a single signal in either direction would flip the read. NEUTRAL 5d is essentially flat with negative XLK alpha — the broad ticker-mention surface does not produce equity alpha in this window.
Where the value actually is. Inside the wider timeline (not the equity table) the consistent strength is macro overlays — $TLT and US10Y options around FOMC / NFP, oil hedges around geopolitical shocks, SPX timing around macro prints. These tweets do not always carry equity cashtags so they are not in the table, but they are the part of the feed that resolves with edge.
| # | Ticker | Mentions | # | Ticker | Mentions |
|---|---|---|---|---|---|
| 1 | $NVDA | 6 | 11 | $FLEX | 2 |
| 2 | $LITE | 5 | 12 | $GLW | 2 |
| 3 | $NOK | 3 | 13 | $UCTT | 2 |
| 4 | $AAOI | 3 | 14 | $FSLY | 2 |
| 5 | $AMD | 3 | 15 | $CRDO | 1 |
| 6 | $MRVL | 2 | 16 | $MXL | 1 |
| 7 | $SMTC | 2 | 17 | $SPX | 1 |
| 8 | $CBRS | 2 | 18 | $SMH | 1 |
| 9 | $WYFI | 2 | 19 | $MU | 1 |
| 10 | $SANM | 2 | 20 | $DGXX | 1 |
No, not against the tech-sector benchmark. BULL alpha vs XLK is -2.17% / 5d and -3.52% / 14d. Raw returns are positive but the picks underperform a passive XLK hold inside the same window. Against SPY the alpha is small-positive (+0.78% / 5d, +3.08% / 14d) but not large enough to justify single-name execution risk.
Spectacular on paper (5d +9.86% WR 100%, 14d +28.02% WR 100%) but n=3-4 each. Statistically uninterpretable. One signal in either direction would flip the read. Do not size off this row.
Macro overlays — $TLT and US10Y options around FOMC / NFP, oil hedges around geopolitical shocks, SPX timing around macro prints. These tweets do not always carry cashtags so they are invisible to the equity backtest, but they are the part of the feed that consistently resolves with edge.
Two reasons. First, the author often takes profit inside the first 1-3 days; 14d / 30d exits mis-measure the actual hold. Second, the universe leans heavily into mid-cap component suppliers ($LITE, $AAOI, $MRVL, $SANM, $UCTT, $FLEX) which were generally outpaced by mega-cap during this specific 42-day tape.
As a macro / event-timing context feed. Read for NFP, FOMC, oil shock framing and bond-options posture. Do not mirror the equity picks — the data shows they underperform a passive tech-sector hold inside this window. Treat retweeted analyst threads as research input, not signal.
Data source. User-timeline endpoint, original tweets only (replies and pure retweets excluded for signal extraction). 232 tweets across 42 days — the maximum the timeline endpoint surfaces for this account given its mixed retweet / original cadence.
Ticker extraction. $TICKER cashtag regex with a curated non-ticker stop-list. Macro / event-timing posts often lack cashtags and are therefore not in the equity return table.
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-tweet styles under-classify into NEUTRAL.
Backtest mechanics. Entry next trading day open (T+1). Exits at +5d / +14d / +30d / +90d close. No transaction costs / slippage. Multiple mentions of the same ticker on the same day produce separate signals (no deduplication).
Benchmarks. SPY (broad market) and XLK (tech sector). Most signals are in semis / tech, so XLK is the demanding benchmark.
BULL 14d alpha vs XLK is -3.52% on n=7 (WR 71%). Raw returns are positive but the picks underperform a passive tech-sector hold. Against SPY the alpha is small-positive but not large enough to justify execution risk.
$TLT / US10Y options around FOMC / NFP, oil hedges around geopolitical shocks, SPX timing around macro prints. These posts are largely cashtag-free so they are not in the equity table, but they are the part of the feed that resolves with edge.
Macro / event-timing context feed. Read for FOMC, NFP, oil shock framing and bond-options posture. Do not mirror the equity picks. Treat retweeted analyst threads as research input, not signal.
Report generated 2026-05-25 by the fxcryptobots research desk. Source: @labubu_trader public X timeline (user-timeline endpoint, original tweets only); price data from Yahoo Finance daily bars. Sentiment classifier is regex-based. 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.