XAnalytics · Working papers · @bubbleboi Live

@bubbleboi — every call, examined

All 233 bets we found, with the real outcome of each — filter, sort, and watch the charts update live.

How to read this page

The benchmark. Not just “did it go up?” — we report how much each call beat (or lost to) a fair benchmark, the extra return it added on top.

vs. its sector = vs other stocks in its industry (the fair skill test). vs. the S&P 500 = vs the broad market. Toggle to switch every chart.

The running total. One step per bet: the stock's return after the call, minus the benchmark's return over the same window, summed in posting order.

The luck test. The same bets replayed 1,000 times with each direction chosen at random. Where the real record finishes among the reruns shows how much chance could explain.

Time frames. Each call is tracked 1 day → 6 months after the post. The Horizon control picks which one drives the charts (1 month by default).

Hit rate = share of bets that beat the benchmark (50% = coin-flip). Conviction = how hard they pushed; direction = long (rise) or short (fall).

Horizon

Compare to

Bet

Conviction

Type

Post type

Their horizon

The running total — if you'd copied every call

The luck test — could a coin flip have done this?

Where the edge shows up

How individual calls landed

Is it carried by a few big wins?

Is the edge holding up?

Every call

Full statistical profile

Computed over the full record. The filters above don't apply here.

Evidence of skill · t-value 2.4

could be luck t = 2 almost certain

Fewer than 1 in 20 luck-only records reach this level.

Independent bets · 171 of 203 84%

all one trade every bet independent

Same-week calls in the same sector move together and prove less than separate wins. These 203 bets carry the evidence of 171 independent ones.

Consistency · per-bet Sharpe 0.18

0 0.1 solid 0.3 excellent

Whether the edge arrives steadily or in a few jackpots. For single stock calls, 0.1 to 0.2 is solid; above 0.3 is excellent.

Directional skill · IC +0.135

0 0.03 real 0.10 elite

Direction only: long calls should precede gains against the sector, short calls losses. Professional managers typically sit near 0.05.

Shape of the wins · skew +1.4

a few big losers balanced a few big winners

A few large winners carry the average.

Edge over time · 1st vs 2nd half +4.1% → +3.8%

older half +4.1%

newer half +3.8%

Softer in the newer half, but most of the edge remains.

How each number is computed

t-value2.4: The average beat divided by its own uncertainty. It rewards both size and sample: a modest edge proven over hundreds of bets scores higher than a large edge shown a handful of times. Above 2, fewer than 1 in 20 luck-only records reach this level; above 3, almost none do.
avg beat ÷ its standard error = x̄ ⁄ (s ⁄ √effective-bets)
Effective independent bets171 / 203: Each call enters the average, but calls in the same sector and the same week move together, so they overlap as evidence. The standard error is widened to account for this (cluster-robust by sector and calendar week), and the result is expressed as the number of fully independent bets that would carry the same proof.
bets ÷ design effect, design effect = (clustered error ⁄ naïve error)²
Per-bet Sharpe0.18: The average beat divided by the spread of individual results. Two accounts can both average +2%: one with steady small wins scores high, one with a single +60% winner among many losers scores low. The steadier record is more likely to repeat. For single stock calls, 0.1–0.2 is solid and 0.3+ excellent.
avg beat ÷ std-dev of the beats = x̄ ⁄ s
Information coefficient (IC)+0.135: The correlation between the called direction and the stock's subsequent return relative to its sector, ignoring magnitudes. Markets are noisy, so absolute values run low: 0.03 indicates real signal, 0.05 is strong, 0.10 elite.
correlation( call direction ±1 , stock return vs its sector )
Return skew (γ₃)+1.4: Whether results lean on a long tail. Strongly positive means a few large winners support the average, which is fragile if they were one-offs. Strongly negative means a few large losses drag down an otherwise better record. Values near zero indicate a balanced, more repeatable shape.
average of (each beat − x̄)³ ÷ s³
Edge stability · 1st → 2nd half+4.1% → +3.8%: The record is split in half by date and each half is scored separately. A durable method keeps working; a hot streak or a single market regime usually does not. Similar or rising halves indicate persistence; a steep drop means the early numbers flatter the present.
avg beat of the older half vs the newer half (split by date)

All measures are equal-weight on the deduplicated bets, with closed positions measured only to their exit — a read on the signal, not an investable portfolio.