A five-year stock forecast can show two apparently reasonable averages that imply different ending values. One averages the forecasts’ annualized rates. The other averages their projected cumulative gains. Applying the same compounding formula to those summaries does not make them interchangeable.
This distinction matters before anyone compares forecast rankings or discusses predictive performance. A reader needs to know which calculation produced the percentage. More decimal places cannot repair an unclear calculation method.
A simple example with a measurable difference
Consider two hypothetical forecasts with equal weights and a starting index value of 100. One projects 0% annual price growth for five years. The other projects 20% annually over the same period.
The first ends at 100. The second ends at 248.832. Averaging those ending values gives 174.416, a five-year gain of approximately 74.42%.
The simple average of the annual rates is 10%. Compounding that rate for five years gives 161.051, a cumulative gain of approximately 61.05%. The difference is 13.37 percentage points, even though neither forecast nor its weight changed.
The reason is that compounding is nonlinear: averaging rates before compounding generally differs from compounding each rate and then averaging the resulting ending values. For rates above −100% and a horizon longer than one year, differences between the forecast rates raise the average compounded ending value relative to the result of compounding the average rate.
There is also a third valid summary: the annual equivalent of the average ending value. The constant rate connecting 100 to 174.416 over five years is about 11.77%. It describes that shared ending value; it is not the simple average of the two annual forecasts.
These are alternative projections, not two independent investments. Equal weighting is a modeling choice. An average of opinions is not automatically an expected return weighted by each outcome’s probability, and this calculation does not identify which opinion deserves more confidence.
A check using archived forecasts
I checked this distinction against an archived extract from iPulse AI’s original Batch 7 forecast records. The selection included all active equity records at one-year and five-year horizons, with each forecast given equal weight. There was no filtering by bullishness, ranking, projected return or the size of the difference between the two calculations.
The extract contained 766 unique records: 383 equities at each horizon. It used a September 20, 2026 snapshot, a September 18 reference closing price and scores calculated on September 21. The data was retrieved on October 2. All four return fields used in the original comparison contained values.
For each five-year record, I compared its stored average cumulative gain, excluding dividends, with the gain obtained by compounding its stored average annualized rate, also excluding dividends, for five years. The stored average cumulative gain was larger in all 383 records. The gap exceeded one percentage point in 289 records, approximately 75.5% of the group. At the one-year horizon, the two stored figures matched.
That is evidence about how stored forecast summaries should be interpreted. It is not evidence that the projections will occur. The input values were rounded for storage, and the calculation did not reconstruct every underlying quarterly forecast path. It does not establish whether the forecasts accurately reflect the likelihood of outcomes, whether the equities represent the broader market or whether the forecasts predict returns accurately.
The original research and calculation context describe the archived dataset and distinguish these stored summaries from forecast views subsequently combined or recalculated using different methods. A historical record should retain its own calculation and date context.
Match what the returns include before converting the horizon
For a cumulative gain G and a horizon of Y years, with returns expressed as decimals, the annual equivalent is:
Annual equivalent = (1 + G)^(1/Y) − 1
For an annual compound rate r, the corresponding cumulative gain is:
Cumulative gain = (1 + r)^Y − 1
Those conversions put returns onto the same time horizon. They do not resolve differences in what the returns include. A price-only forecast and a dividend-inclusive forecast measure different things even when both are annualized. Dividend reinvestment, taxes, fees, currency and the starting price can also change what a displayed return means.
Compound annual growth rate, or CAGR, expresses the constant annual rate connecting a starting value to an ending value. It does not describe drawdowns, when an investor might need access to their money, the probability of reaching the ending value or the sequence of returns along the way. Two paths can have the same annual equivalent and very different risks.
A practical reading test
Before interpreting an average forecast return, record its starting price and date, the latest information included, the forecast horizon, what the return includes and how the forecasts were combined. Ask whether the platform averages annual rates, ending values or returns at each step along a forecast path.
When converting a displayed cumulative gain, label the result “annual equivalent of displayed endpoint.” That wording makes clear that the converted rate describes the displayed ending value and is not necessarily the platform’s original average annualized forecast rate.
Then examine the underlying assumptions. What business outcome supports the projection? What valuation translates that outcome into a price? Which observable development would weaken the investment case? Comparable arithmetic makes those questions easier to ask, but cannot answer them on its own.
Finally, keep a forecast separate from a track record. Performance evaluation requires the original dated projection, later observations and disclosed rules for dividends and other income, corporate actions such as stock splits, missing data and benchmarks. A portfolio simulation requires selection rules, investment weights, costs and a clear separation between information available at the time and information learned later.
The first useful check is modest: identify exactly what the percentage measures before asking whether the forecast is good.
Comments
Log in or sign up to join the conversation.