Better odds are not a guaranteed outcome
A positive historical relationship, when present, changes probabilities rather than guaranteeing an outcome. A cheap stock can become cheaper, and an expensive stock can become more expensive.
Do our valuations
Every valuation on Alpha Spread is scored against what happened afterwards, across thousands of stocks, over years of history.
Valuation Backtest
What $10,000 became over 5 years
Loading historical results…
Historical results aren’t available yet.
This chart will appear once six months of complete historical data is available.
The chart couldn’t be loaded.
Please refresh the page to try again.
This chart compares two simulated $10,000 portfolios with the S&P 500 price index. The benchmark excludes dividends and trading costs. Historical results do not predict future returns.
For professionals
Methodology, in full
How the historical sample is built, when information becomes eligible, and how every signal, return, and statistic is calculated.
The calculation uses today’s eligible company stocks with a primary, active listing and current market capitalization of at least $100 million. A stock also needs a valid historical price and valuation for the observation month.
- Funds, secondary listings, and other non-company securities are excluded.
- Using today’s active universe and classifications introduces survivorship and look-ahead bias into historical results.
- Observations without a usable ending price or historical exchange rate are excluded from the relevant metric and sample size.
For each month-end, the calculation selects the latest available price, valuation, and exchange rate dated on or before that cutoff. It never fills a missing historical value with a later one.
- DCF and multiples history come from the historical sources currently available in Alpha Spread.
- Those sources are not complete archived data vintages, so a backtest rerun can reflect restatements or model-history corrections.
The valuation gap compares the model estimate with the market price available at the snapshot. Positive gaps indicate undervaluation; negative gaps indicate overvaluation.
- For group comparisons, eligible stocks are ranked from most undervalued to most overvalued. Q1 contains the most undervalued stocks and Q5 the most overvalued. Every observation is assigned using floor(zero-based rank index × 5 / sample size), so any remainder is distributed across the groups and group sizes differ by no more than one.
- Observations within ±5% of fair value are neutral and excluded from directional hit rate.
Each historical signal is matched with the adjusted-close return of that stock over the following 12 months. Prices are translated to U.S. dollars using historical exchange rates.
- An undervalued call is correct when the forward return is positive; an overvalued call is correct when the forward return is negative.
- Value spread is the median return of the most undervalued group minus the median return of the most overvalued group.
Every metric has a predefined neutral reference. Directional accuracy is read against 50%, value spread and rank correlation against zero, and the undervalued portfolio against an overvalued portfolio built from the same dates and universe.
- Control portfolios use the same eligible stocks, rebalance schedule, and weighting method so the valuation signal is the intended difference.
- Market-wide gains or losses are not treated on their own as evidence that the valuation signal worked.
The chart normalizes two simulated portfolio NAV series and the S&P 500 benchmark to $10,000. One portfolio holds the most undervalued quintile and the control portfolio holds the most overvalued quintile.
- Both long-only portfolios are equal weighted and rebalanced monthly using a valuation from the signal month.
- The signal and simulated execution use the same adjusted month-end close, which is a simplifying assumption and can introduce look-ahead bias.
- Adjusted close keeps the portfolio on a consistent split-adjusted, dividend-reinvested total-return basis.
- The displayed series deducts estimated 5 bps commission, 10 bps one-way slippage, and a 5 bps FX spread on non-USD trades.
- The S&P 500 benchmark uses the month-end USD price index and excludes dividends and trading costs.
- The result is shown before investor-specific taxes.
Medians and ranks reduce the influence of a small number of extreme winners or losers. VIC 12M is the cross-sectional Spearman rank correlation between valuation gaps and subsequent returns.
- Accuracy, spread, and VIC are reported only when the relevant sample contains at least 30 observations.
- The displayed sample size changes with the selected metric because each metric has its own eligibility rules.
A new result is calculated after each monthly valuation refresh. The latest month is the row with the most recent published date.
- The date shown on this page identifies the latest completed monthly result.
- When the methodology changes, the affected history is recalculated so every point uses the same rules.
Use it correctly
What these numbers don’t mean
These aggregate results describe how the valuation rules behaved historically. They do not promise what will happen to any single stock.
Valuation does not tell you when the gap will close
A valuation gap can persist for months or years. The signal identifies a difference between price and estimated value; it is not a catalyst, an entry point, or a short-term price target.
A valuation is an estimate, not an exact future price
It depends on current financial data and assumptions about the business. New information can change the estimate itself, even when the market price has not moved.
A backtest is not a forecast
Backtests describe how the rules behaved historically. These results include estimated trading costs but remain before personal taxes; markets, companies, and model relationships can change.
