How It Works

The scoring system behind the FinTwit Leaderboard — what we track, how we measure it, and why alpha matters.

What is Alpha?

Alpha (α) measures how much a stock pick outperforms the broader market. Instead of just asking “did the stock go up?”, we ask “did it beat what you'd get from simply buying the S&P 500?”

+3.0α
Strong outperformance
Picks beat SPY by 3% per call
0.0α
Market-performing
Picks match the S&P 500
-2.0α
Underperforming
Losing to the index by 2%

How Scoring Works

1
Capture
We monitor stock calls from tracked FinTwit accounts — tweets with tickers and a clear directional view.
2
Record Entry Price
We log the stock’s closing price on the day of the tweet. This is the entry price — no retroactive adjustments.
3
Track Performance
We check the stock’s performance at 7, 30, 90, and 180 days after the call.
4
Calculate Alpha
For each horizon: α = stock return − S&P 500 return over the same period. This isolates the caller’s skill from general market movement.
5
Score the Call
A call is a HIT if α > 0 (the pick beat the market). A MISS if α < 0 (the market did better).

What Counts as a Call?

The litmus test: if an average person scrolling Twitter came across this tweet, would they read it as a stock recommendation they could act on? If yes, it's a call. If it's just vibes, news, or commentary — it's not.

COUNTS
Has a ticker symbol ($NVDA, $TSLA, etc.)
Expresses a clear directional view — buy, sell, price target, or explicit position
Tweets adding to an existing position count as new calls at the current price
DOES NOT COUNT
Generic sentiment — “congrats NVDA holders” is not a call
News reporting or analyst commentary without a position
Retweets of other people’s calls
Promotional tweets or paid content

No Revisionist History

Deleted tweets are tracked and still count toward a caller's record. If someone deletes a bad call, it stays on the leaderboard. No one gets to erase their misses — that's the point.

Classification & Accuracy

The classification of whether a tweet is a stock call or not is a judgment made by a trained NLP (natural language processing) model. Like any automated system, it is entirely subject to making mistakes — a tweet may be incorrectly flagged as a call, or a legitimate call may be missed entirely. We review and improve the model regularly, but errors are inevitable.

If you believe a call has been misclassified, or have any other feedback, please submit a request using the feedback form in the bottom-right corner of the page.