Fair Value screener track record

Live forward results of published picks · 118 trading days recorded since 2026-06-02 · prices through 2026-11-19

This is the forward record of what the Fair Value screener actually published: each trading day’s Active picks are frozen at that day’s close and then measured against the S&P 500 over the next 21, 42 and 63 trading days. Returns are gross of trading costs, and every figure is reported per model version — results are never pooled across a model change.

Algorithmic ranking, informational only. Past performance is not indicative of future results. See Terms §17.

Current model: v12, live since 2026-11-09

The current model’s forward record is only 9 trading days old, and none of its picks has completed a 21-day window yet. Its first 21-day result lands on or about 2026-12-08; 63-day results follow about two months after that.

Results for earlier model versions are shown below for context, each labelled with its version. They are not this model’s record.

Results by model version

Version Live Days Picks L / S 21d long vs S&P 21d hit 63d long vs S&P 21d short vs S&P 21d long−short
v12 current
Split-adjusted share counts, repaired live inputs
2026-11-09 – 2026-11-19 9 6 / 100 pending — pending — —
v11.3
Regularized models, calibrated expected return
2026-10-02 – 2026-11-06 26 96 / 30 −22.6%n=47 · 7 names 2%0/7 names pending −14.4%n=3 · 2 names −31.1%2 days
v11.2
Long/short side set by the forward-return model
2026-07-08 – 2026-10-01 60 207 / 26 +4.8%n=140 · 15 names 65%7/15 names −0.3%n=77 · 7 names +4.4%n=26 · 11 names +5.8%11 days
v11
Hybrid return + rank model, conviction score
2026-07-01 – 2026-07-07 3 13 / 0 −0.1%n=7 · 7 names 57%4/7 names +5.3%n=11 · 8 names — —
v9
Live tracking begins
2026-06-02 – 2026-06-30 20 88 / 4 +2.0%n=48 · 30 names 63%16/30 names +6.5%n=58 · 36 names +5.2%n=4 · 4 names +11.1%1 day

“n” is the number of pick-days with a completed window and “names” the distinct stocks behind them: a stock held on consecutive days counts once per day, so n overstates the number of independent bets. A version’s figures cover only days it was live; cohorts whose window is still open are excluded until it closes.

Cumulative alpha: following the screener daily

Each day’s 21-day Active picks held from that close to the next, equal weight, summed day by day versus the S&P 500. A day with no picks on a side counts as flat. Dashed lines mark model-version changes. Gross of costs.

Monthly cohorts

Month Version Days Picks L / S 21d long vs S&P 21d hit 42d long vs S&P 63d long vs S&P 21d short vs S&P
Nov 2026 v12 9 6 / 100 pending — pending pending —
Nov 2026 v11.3 5 11 / 14 pending — pending pending —
Oct 2026 v11.3 21 85 / 16 −22.6%n=47 · 7 names partial 2%0/7 names pending pending −14.4%n=3 · 2 names
Oct 2026 v11.2 1 1 / 0 — — pending pending —
Sep 2026 v11.2 20 78 / 4 −3.2%n=39 · 10 names 38%3/10 names −19.6%n=16 · 6 names partial pending +8.0%n=4 · 2 names
Aug 2026 v11.2 21 62 / 0 +12.2%n=53 · 6 names 92%6/6 names +13.1%n=42 · 4 names −15.2%n=30 · 4 names partial —
Jul 2026 v11.2 18 66 / 22 +3.1%n=48 · 8 names 56%6/8 names +17.5%n=30 · 5 names +9.1%n=47 · 6 names +3.8%n=22 · 10 names
Jul 2026 v11 3 13 / 0 −0.1%n=7 · 7 names 57%4/7 names +3.0%n=8 · 8 names +5.3%n=11 · 8 names —
Jun 2026 v9 20 88 / 4 +2.0%n=48 · 30 names 63%16/30 names +9.5%n=50 · 27 names +6.5%n=58 · 36 names +5.2%n=4 · 4 names

Each month pools the picks published that month (one row per model version when a change fell mid-month). partial = some days are still inside their window; pending = none has completed it yet. Short picks vs S&P is positive when the short worked.

Most recent cohort: 2026-11-18

v12 · 1 session held
Symbol Side Active at Conviction Entry close Return so far vs S&P
FLEX LONG 63d 83% $74.84 −0.8% +1.1%
SBUX SHORT 21d, 42d, 63d 86% $109.11 −4.3% +2.3%
YUM SHORT 21d, 42d, 63d 84% $160.80 −0.9% −1.0%
AFL SHORT 21d, 42d, 63d 38% $125.27 −0.4% −1.6%
AON SHORT 21d, 42d, 63d 80% $332.47 −0.4% −1.5%
MDLZ SHORT 21d, 42d, 63d 79% $66.56 −1.3% −0.6%
BDX SHORT 21d, 42d, 63d 70% $216.08 −1.6% −0.4%
ERIE SHORT 21d, 42d, 63d 84% $276.18 −2.6% +0.7%
TSCO SHORT 21d, 42d, 63d 91% $39.01 −0.7% −1.3%
LVS SHORT 21d, 42d, 63d 46% $39.27 +0.2% −2.1%
MCD SHORT 21d, 42d, 63d 83% $274.03 −0.6% −1.3%
CPRT SHORT 21d, 42d, 63d 70% $31.46 −1.9% −0.1%
GE SHORT 21d, 42d, 63d 49% $328.10 −2.8% +0.8%
AKAM SHORT 21d, 42d, 63d 49% $91.70 +0.0% −2.0%

Picks published at the 2026-11-18 close, marked to the 2026-11-19 close. “vs S&P” is signed so positive means the pick is working (for a short, the stock lagging the index). The 2026-11-19 cohort (0 long, 11 short) entered at that day’s close and starts counting tomorrow.

How this record works

Every trading day the Fair Value screener ranks each stock’s forward-return outlook and flags the strongest names on each side as Active. This page records those picks the moment the trading day closes, freezes them, and measures what happened next. It is a forward record: nothing is re-run or re-fitted after the fact, and a pick cannot be removed once published.

What counts as a pick

A pick is a stock the screener showed as Active in its default view: an S&P 500 member ranked in the strongest 5% on its side at a given horizon after the risk filters, with model conviction of at least 30% (below that the quote page reads “neutral”). A pick is judged at the horizon it was active for — a 21-day pick on its 21-day return, a 63-day pick on its 63-day return.

How returns are measured

Entry is the close of the publication day; exit is the close 21, 42 or 63 trading days later. Each return is compared with the S&P 500 (SPY) over the identical window. For a long pick, alpha is its return minus the index; for a short pick, it is the index minus the stock, so a positive number always means the call worked. The hit rate is the share of picks with positive alpha. The long−short spread is the equal-weight long return minus the equal-weight short return, on days that had both.

What “n” counts

Every figure’s n is a count of pick-days, not of stocks. The screener re-ranks daily, and a stock that stays in the strongest 5% is published again each day, so one name held for a week contributes five picks whose windows overlap almost entirely. The name count next to n is the number of distinct stocks behind it; when it is small, the result is a few bets counted many times, and one stock’s move dominates the average.

Why so few short picks

Short picks are only published when the broad market is trading below its 50-day trend, and the short-side filters remove high-quality, high-beta, lightly shorted and strong-momentum names. In a rising market most days have no short picks at all.

Model versions

The model behind the screener is retrained and occasionally redesigned. Results from an older version say little about the current one, so every figure is reported per version and the chart marks each change. The current version’s record starts the day it went live; until its first 21-day window closes it has no completed results, and this page says so rather than borrowing older numbers.

Costs and the backtest

All returns are gross — no commissions, spreads, slippage or borrow fees — so real results would be lower, especially for shorts and for daily rebalancing. The screener’s walk-forward backtest replays the model over history; this page counts only picks after they were published live, so it cannot benefit from hindsight. It is a much shorter record, which is why every number carries its sample size.

Frequently asked questions

What does the Fair Value track record show?

The forward record of the picks the screener actually published. Each day’s Active picks are recorded at the close, frozen, and measured over the next 21, 42 and 63 trading days against the S&P 500 over the same window.

How is a pick defined?

A stock the screener showed as Active in its default view: an S&P 500 member in the strongest 5% on its side at a horizon after the risk filters, with conviction of at least 30%. It is judged at the horizon it was active for.

Why are there so few short picks?

Shorts are only published when the market is below its 50-day trend, and the short-side filters remove quality, high-beta, lightly shorted and strong-momentum names. The long−short spread is computed only on days with both sides.

Why is n bigger than the number of stocks?

n counts pick-days. A stock that stays Active is published again every day, so one name held for a week is five picks with overlapping windows. The name count beside n says how many distinct stocks are behind it.

Why are results split by model version?

Results from an older model say little about the current one, so every figure is per version and the chart marks each change. The current version has no completed results until its first 21-day window closes.

Are these returns after trading costs?

No. They assume entry at the publication-day close and exit at the final close, with no commissions, spreads, slippage or borrow fees.

How is this different from the screener’s backtest?

A backtest replays a model over history; this page only counts picks after they were published live, so it cannot benefit from hindsight — at the cost of a much shorter record.