MoonshotScore / Track record
Over the full three years, the higher the score, the better the stocks did in our simulation.
We rebuilt the score month by month back to November 2023 and followed every scored US stock, the ones that later delisted included. Over the full three years the order held, though not in every stretch. Against the S&P 500 it did not win. From the first close after its rules were made public the record runs live, on rules frozen in advance.
Stock Expert AI research desk · Published 14 Nov 2026 · Data to the 13 Nov 2026 closeSimulation, not client results
Total return, 13 Nov 2023 to 13 Nov 2026
Hypothetical results from a simulation, not real trading. No client money was managed. Past performance does not guarantee future results.
What $10,000 became in each score group
Weekly values, dividends reinvestedHypothetical, simulated results with the benefit of hindsight. Not an account, not a forecast.
Simulation. Score groups re-formed at every month-end from 31 Oct 2023 to 30 Oct 2026, equal weight, 10 bp per side; values rebased to $10,000 at the 13 Nov 2023 close (80+ $17,248, 60–79 $15,976, 40–59 $13,219, below 40 $11,874, S&P 500 $17,770). The S&P 500 line is SPY with dividends. Source: Stock Expert AI backtest of MoonshotScore V2, prices and filings from Financial Modeling Prep.
A sorting tool, not a market-beating machine
The cleanest thing the score did over the full three years was put stocks in the right order. Stocks scoring 80 and above went on to do better than those at 60 to 79, which did better than 40 to 59, which did better than the rest.
The gap between top and bottom is large: +72.5% against +18.7%. It is also the part we trust most, because it does not depend on a benchmark choice. The 80+ group also finished ahead of an equal-weight basket of every US company above $2 billion, 19.9% a year against 15.5%, with a little less volatility.
It did not hold in every stretch. In the first two months of the test (Nov–Dec 2023) the lowest group led, and in 2025 the 40 to 59 group edged out 60 to 79.
| Period | 80+ | 60–79 | 40–59 | Below 40 |
|---|---|---|---|---|
| Nov–Dec 2023 | +18.1% | +18.3% | +15.1% | +33.4% |
| 2024 | +13.0% | +10.7% | +5.4% | −0.7% |
| 2025 | +15.8% | +8.7% | +9.1% | +8.1% |
| 2026 to 13 Nov | +15.1% | +15.2% | +1.2% | −8.5% |
| Three years | +72.5% | +59.8% | +32.2% | +18.7% |
Calendar periods run close to close. "Three years" is 13 Nov 2023 to 13 Nov 2026. Simulation.
Inside the Nasdaq-100, a small edge every year
The fairest test is like against like: Nasdaq-100 members that scored 80 or more, held in equal amounts, against all Nasdaq-100 members held the same way. Same stocks to choose from, same weighting, same costs.
The 80+ half came out ahead in every period, by three to six points, or about 3.7 points a year over the full three years. That is consistent. It is also small, and with 37 months of data it is not statistically conclusive (t-statistic 1.34; the usual bar is about 2).
Simulation. Index membership rebuilt for each month-end from 40 dated Nasdaq-100 changes since Oct 2023. Three-year row shown as a yearly rate so all rows share one scale (+76.4% against +60.6% in total). The 80+ basket was ahead in 22 of 36 months; average monthly edge +0.31%, t = 1.34.
Against the S&P 500, it fell short over three years
The 80+ group returned +72.5% over the three years. The S&P 500 returned +77.7%. Most of that gap opened in 2024, when a small number of very large technology companies carried the index and an equal-weight group of a thousand stocks could not keep up by construction.
The last twelve months went the other way, +19.1% against +13.5%. We are not leaning on that: one good year after one bad one is exactly what noise looks like. Across all 36 months the difference from the S&P 500 is statistically zero (ahead in 17 of 36 months, t = -0.3).
| Period | 80 and above | S&P 500 (SPY) | Difference |
|---|---|---|---|
| 2024 | +13.0% | +24.9% | −11.9 pts |
| 2025 | +15.8% | +17.7% | −1.9 pts |
| 2026 to 13 Nov | +15.1% | +11.5% | +3.6 pts |
| Last 12 months | +19.1% | +13.5% | +5.6 pts |
| Three years | +72.5% | +77.7% | −5.2 pts |
Weight in Apple and Nvidia alone, in the one version that looked strongest: the 90+ stocks held by company size. Its five largest holdings averaged half the basket. That is a bet on a handful of giants during their rally, not proof of stock-picking skill, and it trailed the S&P 500 in 2026 (+8.5% against +11.5%). So we do not present it as a result. We are tracking it live instead.
Simulation. Market-cap weighting splits each company's value across its share classes so it counts once. Largest holdings at the 30 Oct 2026 rebalance: Apple 15.8%, Nvidia 15.6%, TSMC 6.6%, Alphabet (two share classes) 10.8%.
- Business quality90
- Financial safety75
- Valuation33
- Growth durability78
- Momentum92
The same number you see on every stock page
MoonshotScore V2 rates a company from 0 to 100 on five things: business quality, financial safety, valuation, growth durability and momentum. Of the 20,000+ US stocks we cover, about 6,000 carry a score.
For this study we ran the site's own scoring code, unchanged, on the data that was public at each month-end. The full method is on the methodology page.
In plain words, including the weak spots
- Only what was known thenEach month's score uses only filings the SEC had accepted before that date, and prices up to that day. No later information leaks in, apart from occasional restatements of old numbers.
- The losers stay inOf the 8,234 stocks that passed through the study, 2,196 later delisted or went inactive. All are counted, 348 of which once scored 80+. A stock that stopped trading is sold at its last price.
- Monthly, with costsGroups re-formed at each month-end, 37 times. Every trade pays 10 basis points each way, which is generous to the smaller names. Taxes are ignored.
- Checked against the real thingOur live score archive starts on 23 Aug 2026. On five dates since, the rebuilt scores matched the site's published ones closely: correlation 0.96 to 0.97, an average gap of about 4 points, 84–87% in the same group.
- Hindsight, the biggest caveatV2 was designed in August and September 2026, after the period it is tested on. Results from a formula built with the past in view usually flatter it. Expect the future to be weaker.
- It is a simulationNo one held these baskets. Each one owns about a thousand stocks and turns over most of them in a year. It was never available as a product.
Every version we tried, reported whether it helped or not
- All US stocks 80+, equal weight
- The same, without the $1 / $100M size filter
- 80+ among companies above $2 billion
- Nasdaq-listed stocks 80+
- Nasdaq-100 members 80+, equal weight
- All US stocks 85+, equal weight
- All US stocks 90+, equal weight
- Nasdaq-100 85+, equal weight
- Nasdaq-100 90+, equal weight
- All US stocks 80+, by company size
- All US stocks 90+, by company size
- Nasdaq-100 80+, by company size
- Nasdaq-100 90+, by company size
Thirteen versions of the top-score basket, plus the three lower score groups, measured against ten benchmarks (SPY, QQQ, QQEW and equal-weight and size-weighted rebuilds of the market, the Nasdaq-100 and the S&P 500). When you test this many versions, the best-looking one is the most flattered by luck. That is why the headline on this page is the score ordering over the full three years, and not the best single line.
We expected to write a different page. A score is hard to judge fairly when you built it, and the first question we asked was the one everyone asks: did it beat the market? Over three years, no. We could have kept slicing until some version said yes. One did, and it turned out to be two stocks.
What we found instead is quieter and, we think, more useful. Stocks with high scores did better, as a group, than stocks with low ones. That is what we built it to do: help you look at a company clearly. It is not a recommendation on any stock.
The past here is rebuilt. The future will not be. From the first close after its rules were made public the record is kept live, on rules we wrote down and made public before knowing the outcome.
The Stock Expert AI research desk
Live from the the first close after the rules are publicly timestamped, on rules frozen in advance
These three rules were frozen on 14 Nov 2026 and the file was then made public, so anyone can check it was not changed later. The record starts at the first market close after that public timestamp. The baskets use the scores the site archived after the last close before the start, so the scores are always known before the trade. Each month-end we add one line, using only scores the site actually archived. Nothing gets re-fitted. If we ever change a rule, a new record starts with its own date, and this one keeps running beside it.
The score ladder
Every investable US stock with a score, split into 80+, 60–79, 40–59 and below 40, equal weight inside each group.
Passes a review if the groups finish in score order, all three steps.
| Month | 80+ | 60–79 | 40–59 | <40 |
|---|---|---|---|---|
| Start | Groups formed at the start. Value 100. | |||
Nasdaq-100, scoring 80+
Nasdaq-100 members on each rebalance date that score 80 or more, equal weight.
Against all Nasdaq-100 members (Nasdaq's own published list, saved each month), equal weight. "Ahead" = higher return since the start; called meaningful only at t ≥ 2.
| Month | 80+ | All members | Difference |
|---|---|---|---|
| Start | Basket formed at the start. Value 100. | ||
90+, held by company size
Every investable US stock scoring 90 or more, weighted by market value; a company with two share classes still counts once. The version that looked best in the simulation, now tested where hindsight cannot help it.
Against the S&P 500 (SPY, dividends included). Same test as B; top-five weight published every month.
| Month | 90+ | S&P 500 | Difference |
|---|---|---|---|
| Start | Basket formed at the start. Value 100. | ||
533fcdf7d77d793685f36536068ac9f81f22fe05b1f56b8de883b80c0702bd53
Starting holdings: published here with their SHA-256 once the baskets are formed
Read this before using any number above
Every result for Nov 2023 to Nov 2026 is hypothetical. It was produced by running today's scoring formula on historical data, not by managing money. Simulated results have built-in limits: they are prepared with the benefit of hindsight, they do not involve real trading, and they cannot fully account for costs, liquidity or the decisions a real investor would face. No client money was managed. Actual results would have differed, perhaps materially. The live record is also a model portfolio, not an account.
This is educational content, not investment advice. Past performance does not guarantee future results.
Method and score definition: MoonshotScore methodology.