Research Report · Twitter Alpha Backtest · AI Infrastructure

@Ren_aramb 60-Day Backtest: BULL Signals Returned +25.22% Mean at 14 Days, Win Rate 87%, +13.67% Alpha vs XLK

Quantitative backtest of @Ren_aramb (Ren, 24,432 followers, bio: “Investing into the AI buildout. Head of AI / 10y Product Manager”). 376 original tweets across the 60-day window, 255 ticker-mention tweets, sentiment mix BULL 139 / BEAR 30 / NEUTRAL 86. Universe: AI infrastructure, photonics, NeoClouds, semiconductors.
Published 2026-05-25 Backtest Window: Mar 25 -> May 25, 2026 (60 days) Benchmarks: SPY (broad) / XLK (tech sector) Entry: T+1 open · Exits: 5d / 14d / 30d / 90d close
VERDICT: STRONG ALPHA · FOLLOW (semi-niche, time-bounded sizing)
+25.22%
BULL mean return / 14d (n=125)
87.20%
BULL win rate / 14d
+13.67%
BULL alpha vs XLK / 14d
+21.16%
BULL alpha vs SPY / 14d
376
Original tweets sampled (60d)
255
Tickered tweets · 139 BULL
24,432
Followers · 3,291 lifetime tweets
60d
Sample window inside AI rally (regime caveat)
Prefer the print version? Download the 3-page PDF: profile dossier, full sentiment-by-horizon table, most-mentioned tickers, best/worst 30-day BULL calls with tweet previews, methodology, limitations.
Download PDF ↓

Profile & Persona Who Ren is and what kind of signal his timeline produces.

The author. 24,432 followers, account opened 2020-11-12, 3,291 lifetime tweets, bio: “Investing into the AI buildout. Head of AI / 10y Product Manager.” Long-form fundamental theses on AI infrastructure, photonics, NeoClouds and semiconductors. Top thesis names across the 60-day window: $SIVE / $SIVEF (Sivers Semiconductors AB, repeatedly thesis-built), $LITE (Lumentum), $NBIS (Nebius), $POET, $AAOI, $INTC, $SNDK, $AXTI, $MRVL, $AMD, $MU, $IQE, $ARM, $AEHR.

Signal shape. Author is openly long most names tweeted. The edge is directional thesis sourcing inside an AI-infrastructure universe, not a stock-picking arbitrage that would survive a regime change. Treat the feed as research input, not signal output.

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
BULL5d193+5.54%+4.13%66.32%+84.59%-39.76%+4.24%+2.13%
BULL14d125+25.22%+18.64%87.20%+93.76%-24.77%+21.16%+13.67%
BULL30d5+19.67%+12.62%80.00%+57.44%-7.56%+7.56%-7.73%
BEAR5d55+11.19%+9.28%70.91%+52.01%-42.76%+9.08%+5.81%
BEAR14d49+24.12%+20.83%85.71%+183.08%-5.89%+19.86%+12.74%
BEAR30d11+91.07%+88.75%100.00%+270.77%+5.76%+79.97%+63.40%
NEUTRAL5d89+9.90%+6.48%73.03%+72.86%-55.06%+8.09%+5.27%
NEUTRAL14d75+24.79%+19.20%88.00%+200.00%-62.98%+20.41%+12.90%
NEUTRAL30d8+110.16%+107.60%100.00%+325.59%+14.35%+97.26%+80.58%

The headline read. The 14-day horizon is the cleanest sample (n=125 BULL signals). Mean return +25.22%, median +18.64%, win rate 87.2%, alpha vs the tech sector ETF (XLK) +13.67% and vs SPY +21.16%. This is best-in-class for AI-infrastructure thematic depth inside the sample window.

Note that the BEAR and NEUTRAL buckets also produced strong forward returns: most are ticker-mention tweets where the author was not making an explicit directional call. Inside an AI rally, just being early on the right names was enough.

Most-Mentioned Tickers (60-Day Sample) Where the conviction concentrated. Top 20 by mention count.

#TickerMentions#TickerMentions
1$SIVE8811$LPKF18
2$SIVEF3112$LPK17
3$LITE2813$AMD16
4$INTC2714$MU16
5$SNDK2615$IQE16
6$NBIS2516$ARM14
7$POET2517$AEHR14
8$AAOI2418$TRT14
9$AXTI2019$COHR10
10$MRVL1820$SOI10

The top thesis is $SIVE / $SIVEF (Sivers Semiconductors AB, Swedish indium-phosphide laser play) with 119 combined mentions over 60 days; this dominates the conviction surface but is excluded from the return table because yfinance does not resolve the cashtag for the Stockholm listing. The published edge therefore understates the thesis-quality signal — the most-repeated bull call is not in the numbers.

BULL Calls With Resolved 30-Day Returns Only 5 BULL signals had a full 30-day forward window inside the sample. Small sample, presented as-is.

DateTickerRet 30dTweet preview
2026-04-02$AAOI+57.44%“Lightcounting is projecting $100B optical interconnects by 2030. That's 5x from here...”
2026-04-02$TSEM+25.80%“Lightcounting is projecting $100B optical interconnects by 2030...”
2026-04-03$SPY+12.62%“For the first time in the year Oil and $SPY are up...”
2026-04-08$SPY+10.06%“$SPY Seems like tomorrow will hold and go up...”
2026-04-05$ASTS-7.56%“$ASTS I've been following this stock for a while, solid fundamentals, Space X IPO as a catalyst...”

Five Questions, Five Answers

Q1. Is there a real edge inside the sample window?

Yes. 14-day BULL alpha vs XLK is +13.67% on n=125 (the largest, cleanest bucket). Win rate 87% is materially above coin-flip. The edge concentrates in AI-infrastructure / photonics / semis where the author writes long-form thesis posts.

Q2. Will the edge survive a regime change?

Unknown. The entire 60-day sample falls inside the March-May 2026 AI rally (SPY +6%, XLK +12%). No bear-tape data in this run. Treat the +13.67% XLK alpha as an upper bound, not a forward expectation.

Q3. Is the 30-day horizon reliable?

Not at this n. Only 5 BULL signals had a full 30-day forward window inside the sample; the BEAR 30d bucket (n=11, +91% mean) is a survivorship artefact of a few large winners. Use the 5-day and 14-day rows for sizing decisions; treat 30d as anecdotal.

Q4. What is the highest-conviction name in the dataset?

$SIVE / $SIVEF (Sivers Semiconductors AB) with 119 combined mentions over 60 days. Excluded from the return table because the Stockholm listing does not resolve in the US pricing source — meaning the published alpha is conservative.

Q5. How should I use this feed?

As a research feed, not a signal feed. Use Ren's thesis posts to discover names in AI infrastructure / photonics / semis you would otherwise miss, then do your own work. Sizing should be moderate, time-bounded (≤30d), and risk-controlled against the regime caveat.

Methodology

Data source. User-timeline endpoint, original tweets only (replies and pure retweets excluded). 376 tweets across 60 days — the maximum the timeline endpoint surfaces for an account at this posting cadence.

Ticker extraction. $TICKER cashtag regex with a curated non-ticker stop-list (e.g. $AI, $US, $YTD removed).

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.

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) — this reflects a literal “follow every tweet” baseline.

Benchmarks. SPY (broad market) and XLK (tech sector). XLK is the more demanding benchmark since virtually all signals are in AI / semis. Alpha = signal return minus same-window benchmark return.

Limitations & Honest Caveats

  1. Sample window is 60 days. The user-timeline endpoint surfaces only the most recent ~3,200 tweets per account; for an active poster this maps to ~2 months of coverage. The published edge is a snapshot, not a multi-cycle backtest.
  2. Regime bias. The entire sample falls inside the March-May 2026 AI rally. SPY +6%, XLK +12% across the window. Stock-picking edge is inseparable from beta on a tape this strong.
  3. 30d / 90d samples are tiny. Most signals fall in the last 60-90 days of the lookback, so the longer horizons resolve on only 5-11 data points per row. The 5d and 14d rows are the only buckets with enough n to read seriously.
  4. Foreign / delisted tickers excluded. $SIVE / $SIVEF (Sivers Semiconductors AB, Stockholm) is mentioned 119 times but does not resolve in the US pricing source; the highest-conviction name in the dataset is absent from the return table.
  5. Author is openly long. Most tweeted names are positions, not arbitrage candidates. The edge measured is “buy what an early AI-infrastructure long is buying” — useful, but not market-neutral.
  6. No risk-adjustment. Sharpe / Sortino not computed. Position-sizing, drawdowns, and correlation across signals are not modeled.

Final Verdict

+ Tier: STRONG ALPHA · FOLLOW

BULL signals returned +25.22% mean / 14d (n=125, WR 87%) with +13.67% alpha vs XLK and +21.16% vs SPY. Best-in-class for AI-infrastructure thematic depth inside the sample window.

! Caveat: AI-rally regime

The 60-day window is a single bullish AI tape. No bear-regime data. Treat the published alpha as an upper bound — size for the possibility that the regime change cuts the edge in half (or worse).

+ How to use

Research feed, not signal feed. Discover names through Ren's thesis posts, then size moderate and time-bounded (≤30d). The highest-conviction name ($SIVE) is excluded from the numbers — the published edge is conservative.

Full 3-page PDF report Profile dossier, full sentiment-by-horizon table, top-20 most-mentioned tickers, best/worst 30-day BULL calls with tweet previews, methodology and limitations.
Download PDF ↓

Report generated 2026-05-25 by the fxcryptobots research desk. Source: @Ren_aramb public X timeline (user-timeline endpoint, original tweets only); price data from Yahoo Finance daily bars. Sentiment classifier is regex-based and may under-count cryptic / 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.